In contrast to time-series analysis, cross-sectional studies usually ...... unions and employers' organisations in Europe (Eclipse Group Ltd, London) ..... bondsgroei vooral gestimuleerd door prijs- en loonstijgingen, ook wel aangeduid als.
TRADE UNION GROWTH AND DECLINE IN THE NETHERLANDS
The book is no. 101 of the Tinbergen Institute Research Series. This series is established through cooperation between Thesis Publishers and the Tinbergen Institute. A list of books which already appeared in the series can be found in the back.
Trade union growth and decline in the Netherlands
ACADEMISCH PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam, op gezag van de Rector Magnificus prof. dr P.W.M. de Meijer ten overstaan van een door het college van dekanen ingestelde commissie in het openbaar te verdedigen in de Aula der Universiteit op woensdag 29 november 1995 te 13.30 uur
door
Jeannette Erica van den Berg geboren te Leiden
Promotiecommissie Promotor:Prof. dr J. Hartog Overige leden:Prof. dr W.L. Buitelaar Prof. dr L. Noordegraaf Prof. dr J. Theeuwes Dr K.G. Tijdens Dr J. Visser Prof. drs. P.J. Vos
PREFACE
At last this thesis has come to an end, and although it took considerably longer than expected, I have no regrets about this undertaking. Apart from the moments that nothing seemed to work, this research has been very fulfilling, for which I owe a lot to many people at work and at home. My family has been a great support throughout these past seven years, and I thank Richard Cörvers for encouraging me to take on this research in a period in which I had no scientific aspirations whatsoever. Today I could not think of any other kind of work that would suit me, so I am very happy with the opportunities offered to me at my new job, at the Economic Institute of Utrecht University. In my former jobs I have always been surrounded with helpful colleagues, a few of whom I would like to mention separately. In the first place I want to thank my supervisor Joop Hartog, who has always given me valuable comments that really improved my work. Many times he surprised me with his reading speed, which was never at the cost of accuracy. In this last year he managed to speed me up as well, which helped me to round off this research. During my stay at the department of Microeconomics of the University of Amsterdam I had the pleasure of working together with Wim Groot on two articles, which are now integrated in this thesis. In Amsterdam I also received back-up from several fellow aio's of the Tinbergen Institute, especially Andre Voskamp, Lourens Broersma, Robin de Vilder, and Inge van den Doel, whose expertises formed a very useful contribution. My warmest thanks go to my colleagues at the department of Economics at Leiden University, who have been most helpful and provided me with a very stimulating working environment. Especially Marcel Kerkhofs, Ron Linssen, Peter van Wijck, and Jan Kees Winters spent a lot of their time in answering my questions, commenting on my estimation results and reading several parts of my book. I am very grateful for all their efforts. Finally I would like to thank Roger Barton for reading my whole thesis and for correcting the English. Leiden, September 1995.
Annette van den Berg
CONTENTS 1.Introduction 1 1.1Object of this research 1 1.2Theoretical framework 2 1.3Outline of the book 5 2.Union growth and decline: economic theories and empiricism 2.1Introduction 7 2.2Time-series studies 7 2.2.1The first theories 7 2.2.2Theoretical foundation of time-series models 9 2.2.3Models of union growth: empirical results 11 2.3Cross-sectional studies 17 2.3.1Personal characteristics 17 2.3.2Occupational and industrial characteristics 19 2.3.3Wage related variables 23 2.3.4Regional characteristics 24 2.3.5The socio-legal environment 25 2.3.6Union leadership & policies 26 2.4Concluding remarks 27 3.Short history of the Dutch trade union movement
29
3.1Introduction 29 3.2The formation of the first trade unions, between 1860 and 1900 29 3.3Growth of the trade union movement, 1900-1940 36 3.3.1Emergence of the tripartition in the labour movement 36 3.3.2Fluctuating union membership in the interwar period 39 3.3.3Development of the institutional framework 44 3.4The interlude of the second World War 52 3.5Trade unions in the welfare state, 1945-1990 54 3.5.11945-1959: centralization and co-operation 54 3.5.21959-1975: towards more decentralization and conflicts 61 3.5.31975-1990: shift of power, deterioration and some recovery 67 3.6Conclusions 75 Appendix: list of abbreviations 83
7
4.Time-series analysis on trade union growth 4.1Introduction 4.2Short review of the dependent and independent variables 4.3The Dutch data and their peculiarities 4.4The models of Ashenfelter & Pencavel and Bain & Elsheikh 4.5New modelling 4.5.1Building a model covering the whole period 4.5.2The pre-war model 101 4.5.3The post-war model 104 4.6Conclusions 109 Appendix: Tables 4.10 - 4.13 112
85 85 85 89 94 97 97
5.Cross-sectional analyses 119 5.1Introduction 119 5.2Econometric framework 120 5.3The first data set 122 5.4Empirical results 126 5.5The second data set 134 5.6Econometric framework 140 5.7Empirical results 141 5.8Conclusions 147 Appendix: Tables 5.10 - 5.11 150 6.Supplementary lines of approach from the social sciences 155 6.1Introduction 155 6.2Sociologic and psychologic theories of union membership 155 6.3Empirical findings of social scientists 157 6.3.1Questionnaires: results and discussion 157 6.3.2Statistical modelling: results and discussion 161 6.4Incorporation of sociologic notions into economic modelling 164 6.5Conclusion 173 Appendix 6.I: Data used in section 6.4 174 Appendix 6.II: Simultaneous modelling: derivation of the restrictions from the system of four equations 184 7.Summary and conclusions
187
References 195 Samenvatting (summary in Dutch) 203
1.INTRODUCTION 1.1 Object of this research Within the field of labour economics, the research on trade unionism constitutes a considerable share. The majority of studies on trade unions deals with bargaining theories, in which unions are assumed to strive after certain goals to benefit their members. Unfortunately, because the bargaining models easily become too complicated, the formulated union goals are usually confined to reaching an as high as possible level of wages and employment. In reality, there is of course a much wider variety of bargaining issues. A further drawback of many of these union models lies in the fact that the level of trade union membership is fixed, which means that the membership rate is unaffected by the union's bargaining performance. This is not a very realistic supposition, as the union's bargaining results surely influence the people's decision whether or not to join or to leave the union. Alternatively, the size of the union movement constitutes an important determinant of the outcome of the bargaining process, because unions derive their power for a large part from the number of their adherents. Hence, trade union membership should be taken up as an endogenous variable, which in fact has been done in more recent, dynamic bargaining models. However, here again for simplicity's sake the membership level is usually only determined by the results of the wage negotiations and their consequent effect on the employment outcome. The motives of workers to join or to leave the union are of course much more diverse, and a wide range of literature has developed on this specific issue. Whereas union bargaining theories are mainly of interest to labour economists, union membership theories are the object of research of various disciplines. As each has its own lines of approach and research methods, often different aspects of the unionization decision are accentuated. By their very nature, economists emphasize costs and benefits motives, while sociologists lay more stress on social control motives. Historians contribute to the subject by analyzing societal developments and events, which affect the rate of change of trade union membership. This does not necessarily imply that each discipline comes up with a completely different set of explanations for the growth and decline of unions. Still, by putting emphasis on diverging determinants of the unionization decision, each approach tends to underestimate or omit certain elements of the explanation that are considered of high importance in another field of science. For that reason, this study intends to follow a multidisciplinary approach to the explanation of fluctuations in trade union membership in the Netherlands, from its early history onwards.
1
Very schematically, we could look at unionization in the following way: ┌─── supply of union services ──┐ ┌──┴─────────┐ ┌───────────┴──┐ ┌─────────── │(potential) │ │ trade unions │ ───────┐ │ │ members │ │ │ │ │ └────────────┘ └──────────────┘ effect cost-benefits or effect upon other deliberations upon individual union decisiondemand for union services policies making ┌────────────────────────────┐ │ │ │ environmental factors │ │ └──--────────────┤ (institutional, cultural, ├─────────-──┘ │industrial,etc.etc) └────────────────────────────┘
This outline takes the economic lines of approach as a starting point, but at the same time incorporates determinants of unionization that are put forward by other disciplines. Theoretically, this will be elaborated below. 1.2 Theoretical framework The point of departure will be the framework set out by Hirsch & Addison in their book The economic analysis of unions (1986), in which the demand for and the supply of unionism is analyzed. According to them, 'formal development of a model of the demand for unionism can help in understanding the large differences in union representation over time and across industries, workers and geographic areas. However, because the costs of organizing and providing union services are likely to vary substantially across jobs, the supply of unionism will also be an important determinant of union status' (p. 29). The demand function expresses the demand of all workers for union representation and services, while the supply function reflects the supply of union services. The underlying basic assumption of the demand - and supply functions of union representation is that demand as well as supply are determined by costs-benefits considerations. Very briefly, the two functions can be described in the following way.1) The demand function is given by Ud = f(p,y,diff,s,z,t) The price p represents the costs of membership (initiation fees and dues) relative to the price of other assets and goods. Wealth or permanent income, y, has a certain impact upon the decision to unionize. If union services are a normal good, then the demand for union representation should increase with income. But since this is not always the case, the relationship between unionism and wages must be estimated empirically.
1)In chapter 2 a more detailed explanation of all the variables involved can be found.
2
The next determinant, diff, stands for the relative union-nonunion compensation differential. It is widely accepted that union representation is most likely where the union wage effect is largest. However, 'because the relative wage differential cannot in general be measured directly, studies typically examine the relationship of unionism with personal and industry characteristics, which serve as proxies for the expected benefits of union representation' (p. 31). Good examples of such industry characteristics are the degree of concentration and capital intensity, both indicators of lower demand elasticities, which are favourable to union wage gains. Similarly, several personal characteristics such as age, sex and skill level have a certain impact upon union compensation gains. The variable s represents the costs of substitute services. It is argued that, for instance, the increase of social welfare benefits lessens the demand for union services. Next, nonpecuniary benefits from a unionized work environment, z, are incorporated in the demand function. This variable relates especially to the improvement of working conditions and is indirectly measured through such industry characteristics as firm size, capital intensity and accident risk. It is assumed that 'unions are more likely to arise in those firms where formalization of the work environment provides the largest benefits' (p. 32). And finally, allowance is made for the individual's taste for unionism, t. Differences in attitudes towards unionism are believed to stem mainly from differences in the expected wage and nonwage benefits of unionism, but are also in part captured by variables such as region and urbanization. The supply function is given by Us = g(p,CO,CS,G) The price, or revenue of union services, p, is supposed to have some positive relationship with Us, although in the case of unions there is no real question of profit maximization. The costs of supplying union services can be subdivided into organizing and servicing costs. Concerning the costs of organizing, CO, it can be said that this involves a substantial fixed-costs component and exhibits sizeable economies of scale. The fixed costs comprise the setting up of the machinery, procedures and expertise to organize new members. It is predicted that union coverage is greater in large firms and in more concentrated industries, since marginal organizing costs are lower there. By the way, 'much of the cost of organizing new workers is borne by existing members' (p. 33). Servicing costs, CS, are also likely to have a large fixed-costs component, so again this leads to the expectation that unionism is less likely in small bargaining units, unless such a unit is covered by the collective bargaining outcomes in other units. The costs of organizing and servicing will also be affected by the employer's attitude towards collective bargaining (which is usually, but not necessarily, dictated by economic considerations), and by the legal structure in which unions must operate. The last variable in the supply function, G, stands for union goals and is in fact the determinant of p, CO and CS. Dependent on the union's goal, for instance, maximizing wages or membership, the level of p, CO and CS is affected. The equilibrium level of unionism is determined by
3
U = Ud = Us In a reduced form, U and p are functions of all other variables within the system, the unionism equation being U = h(y,diff,s,z,t,CO,CS,G). Quoting Hirsch & Addison, 'Empirical studies generally estimate some variant of the reduced-form equation. We know of no study that specifies and estimates structural demand and supply equations. The absence of such a study reflects the paucity of data measuring directly the factors specified in Ud and Us. In addition, the quantity of union services, U, is not directly observed. However, if the level of union services is proportional to the level of unionization across jobs, direct measures of union membership status or collective coverage can be used to estimate U' (p. 35). The measurement of U constitutes a problem in this study. The demand for unionism expresses itself in the number of individuals who are, or want to become a member, while the supply of unionism becomes manifest in the number of existing trade unions. In other words, Ud and Us can be measured, but differently. Alternatively, the demand for and supply of union services deal with the same variables all right, but these cannot be measured (quantified) properly. In the empirical part of this study U will be put on a par with the number of union members. This level of unionization is however not just determined by the variables formulated in the demand function, but also by variables connected with the supply side. Thus far it has been assumed that the unionization decision is determined by costs and benefits considerations. This assumption detracts from less materialistic inclined motives of unionization, which are stressed by other disciplines. Here it is the aim to incorporate these determinants of union membership into the demand and supply framework as well. Without yet going into details2), this concerns factors such as ideological motives, social pressure and custom, and commitment to the union on the demand side, with on the supply side the impact of organizational structure (centralized versus decentralized) and goals of unions on the commitment of (potential) members. Recalling the demand - and supply functions, these are given by Ud = f(p,y,diff,s,z,t) Us = g(p,CO,CS,G) As for the demand side, the added factors can be classed in the variable t (attitudes towards unionism), which is rather underappreciated in the purely economic approach. And concerning the supply side, the additions can appositely be classed in the variable G, union goals. It should be kept in mind, that on both sides the variables are in fact influenced by various environmental factors - stemming from societal developments -, which do not enter the demand and supply function themselves; we have seen this in the original functions too, where, for example, the impact of the employer's attitude and the legal structure, though not visible, are taken into account. 2)Chapter 6 discusses the most important social-scientific lines of approach.
4
At this stage it should be made clear, that it will not be possible to estimate all the determinants of union membership empirically, and certainly not by means of only one model. The reason for this drawback lies in the availability of the data, which are incomplete, sometimes inconsistent and often measured at different levels (aggregate versus micro data). Moreover, except for the problem of quantifying correctly, several important factors are hard to quantify at all, such as - gradual - changes in mentality and legal status. This does not imply that the theoretical framework outlined above has to be abandoned, just because several determinants of unionization cannot be captured by means of econometric modelling. The scheme of demand and supply should be elaborated by formulating all the questions that need to be answered, in order to get an, as complete as possible, insight into the determinants of Dutch trade union membership through time. Part of the answers can be given with the aid of econometric modelling, another part can be found by sound reasoning: relevant events and developments should be mapped out and interpreted in the light of the various scientific theories on this subject. The basic questions concerning the demand for union representation and services run as follows: what determines the chance whether or not to become or remain a union member, and what kind of services are demanded? And, subsequently: how and why do these matters change through time? Underneath, a lot more questions can be unearthed like different pieces of one puzzle. To name just a few: in what way do personal characteristics play a role, what is the impact of the work environment, and how important is the effect of the institutional setting and of prevailing opinions? The basic question concerning the supply of unionism can be split into two main questions: which factors determine the supply of unions, and what kind of services are offered? And, again: how and why do these matters change through time? Consequent subquestions are: when, and in which industries do unions arise; how are they organized; how do they present themselves and what kind of activities do they develop to attract and retain members? Finally, the theoretical notion of a market where demand for and supply of unionism meet, can be translated into the following question: does the supply of union services link up with the demand; if so, does this lead to an increase of trade union membership, and if not, does it result in decline?
5
1.3 Outline of the book The main task of this study is to establish which factors carry the most weight in the determination of trade union membership in the Netherlands. In order to answer all the relevant questions in a systematic way, the structure of the book is as follows. Chapter 2, which gives a survey of the (empirical) literature in the field of economics, demonstrates that a great many determinants of union membership have been taken into consideration; these are either direct measures of or proxies for the variables incorporated in the original demand and supply functions as described in section 1.2. This overview will serve as guidance for the selection of data to be used in the empirical part of this study. Before the actual estimations can take place, in chapter 3 attention will be paid to the historical development of Dutch trade unions. Each country has its own specific, historically grown context, in which unions operate. A short look at the history of the trade union movement in the Netherlands may enable us to formulate several theoretical assumptions, in addition to the general theories on trade union growth and decline. Dependent on the type of variable and the data, the quantifiable variables enter either a time-series model or a cross-sectional model. Both types of research will be taken up in this study. Chapter 4 is dedicated to time-series analysis, while chapter 5 deals with cross-section data. Finally, in chapter 6 supplementary lines of approach from the social sciences will be discussed, followed by a small empirical exercise. The overall conclusions will be recapitulated once more in chapter 7.
6
2.UNION GROWTH AND DECLINE: ECONOMIC THEORIES AND EMPIRICISM 2.1Introduction In this chapter an overview will be given of the different explanations of union growth, the way they have been advanced in the economic literature from the beginning of this century until most recently. In economic theory, attention is focused on the costs-benefits approach of union membership: the costs of joining are weighed against the benefits. A thorough study of the literature teaches, that there does not really exist one coherent theory that can explain why people join or leave the union. It already makes a difference whether union membership is studied at the aggregate or at the individual level. Moreover, the economic theories are strongly linked to the applied research method. Broadly speaking, two different approaches to empirical union membership analysis can be distinguished: time-series models and cross-sectional models. In time-series studies, a model is estimated to explain annual fluctuations in (aggregate) membership growth rates over a long time span. A detrimental fact is the relative small number of explanatory variables at disposal, due to problems of poor availability of long series on the one hand, and of changes in the definitions of series on the other hand. In contrast to time-series analysis, cross-sectional studies usually employ individual data, and lay more stress on the structural determinants of union growth, such as the (changing) structure of the labour force and of the social and legal environment. The drawback of cross-section data is, that the data set mostly covers only one point in time, which prohibits the measurement of changes through time. The advantage lies in the magnitude and variety of the explanatory variables, which enable the researcher to test the theories to a much larger extent. Quite recently several researchers (Hirsch & Addison 1986, Ellwood & Fine 1987) have suggested to integrate these two approaches, as far as the necessary data are available over a substantial period of time. To us this seems the most fruitful approach. In the next sections however, a discussion follows of the two separate categories of literature, in which theory and empiricism go hand in hand. 2.2 Time-series studies
2.2.1 The first theories
The first theorists of union growth (Barnett 1916, Weyforth 1917, Commons 1918, Wolman 1924) were convinced that the size of labour unions was determined by the business cycle. They saw a strong connection between fluctuations in union growth and changes in business conditions. In years of prosperity, when rising prices and higher profits swelled the demand for labour, the costs of living and the hours of labour also increased. This would cause growing dissatisfaction among workers, who would then demand higher wages and shorter working hours. Under those circumstances unions could win concessions from employers and gain membership. However, during a downturn of the business cycle employers were able to withstand union pressure and the labour movement would decline again. This prosperity thesis was advocated until it 'failed to work' during the flourishing nineteen twenties: in spite of the cyclical upturn union membership in the United States declined (see Wolman 1936).
7
Researchers were then forced to revise the business cycle theory. Davis (1941) was the first to make a good attempt. He did not reject the theory, rather he made a distinction between business prosperity and rising prices. He argued that these two indicators do not necessarily converge and that union growth was not so much influenced by the former, but the more so by the latter. He stressed the fact that an upturn of the business cycle is primarily restricted to the capital industries, whereas a rise in consumer prices affects all workers. As wages tend to lag behind price increases, the fear of a fall in the standard of living would impel employees to join unions. And because increases in labour costs can more easily be passed along to consumers in times of rapidly rising prices, employers would be more willing to negotiate with unions. The 'prosperity theory' was therefore replaced by a 'prices theory'. Nevertheless, Davis' contribution could only be regarded as a modification of the early business cycle theory, for the prosperity theory and the prices theory employ virtually identical arguments. They both showed that membership variations are at least partly determined by fluctuations in economic growth, which was supported with empirical evidence especially by Davis. Although all these early theorists did acknowledge the importance of several other factors in determining union growth, none of them made an integrated, systematic multi-causal approach. Dunlop (1949) and Bernstein (1954) were the first writers to attempt this. They both distinguished between long-term and short-term determinants of union growth. First, they rightly argued that several long-term factors have a considerable influence upon unionism. Among other things, they mention such factors as occupational, institutional and social circumstances. The short-term determinants put forward by Dunlop and Bernstein have been subjected to a large amount of criticism. The two theorists both advanced a sort of 'disaster' approach to union growth, claiming that fluctuations in the size of unions were caused by wars and social unrest. But they failed in substantiating their theory with convincing empirical evidence. In fact several contemporaries (Shister 1953, Rezler 1961, Blum 1968) point out that these ad hoc 'disasters' are not the real causes of union growth, they merely stimulate the more fundamental and systematic determinants, such as rising prices and union recognition. As a matter of fact, Dunlop himself argued that during periods of wars and so-cial unrest the intensity of the business cycle was greatest. He even related fluctuations in union growth to the fluctuations of the long business cycle waves, or Kondratieff cycles. Bernstein also had to admit that the business cycle was closely connected to short-term variations in membership in 'specific historical contexts'. In short, research into explaining short-term union growth still relied heavily on the business cycle theory. Although numerous determinants had been put forward, no one had ever attempted to integrate at least some of these determinants in a systematic quantitative framework in order to estimate their relative importance. This matter was first taken up in the nineteen sixties, when researchers developed econometric models to establish a more exact relationship between the explanatory variables and union growth.
8
2.2.2 Theoretical foundation of time-series models The theories on union growth and decline are usually embedded in a conventional economic framework: union services are being supplied and demanded at a given price (membership fee). In research, emphasis clearly lies on the demand side, where workers compare the costs and benefits of union membership. The supply side is seldomly elaborated, probably due to a lack of data; this pursues such matters as union recruiting activities and determination of the membership subscription level. In an influential study by Bain & Elsheikh (1976, hereafter referred to as B & E), a more neutral framework is applied: they explain changes in union membership in terms of changes in the propensity and the opportunity to unionize. The theories are then formulated around the explanatory variables, the way they show up in the econometric models. For that reason the next paragraphs are devoted to an enumeration of the relevant variables, combined with their theoretical foundation. Inflation. Price increases can have a stimulative effect on union membership in two ways (B & E, pp. 62-63). First, they radiate a so-called 'threat effect' on workers, inducing them to join or remain in the union, which is expected to defend their standard of living. Second, rising prices might be a proxy for an upward movement in the business cycle. In those circumstances employers are more inclined to concede union demands, which in turn attracts more members; this is referred to as the 'prosperity effect' of inflation. Wages. Wage increases are also expected to influence union membership in a positive way: when workers attribute a pay-rise to the efforts of the union, it may encourage them to become or stay a member, in hopes that this reinforcement of the union will improve their future income as well. This is called the 'credit effect' of wage rises (B & E, pp. 63-64). Unemployment. The impact of unemployment on union membership can either be positive or negative (see Visser 1988, p. 146; B & E, pp. 65-67). Especially at the start of an unemployment period, people can decide to join or hold on to the union, hoping that the latter can protect their jobs. Unemployed workers may choose to remain a member for several reasons. Mostly their dues are reduced or waived, so there are little or no costs involved; the gains of continued membership can be found in ideological motives, or practical motives: through the union they keep in touch with the labour market, which might improve their chance on finding a job. Moreover, if the trade union movement plays a role in administering unemployment benefits of a country, it is argued that unemployment, through this institutional arrangement, has a positive effect on membership rates. On the other hand, there are several arguments why unemployment decreases un-ion membership in the longer run. In particular when an unemployment spell is long and severe, the union movement is forced on to the defence, implying that it is less able to serve the workers' interests. This position reduces the attraction for workers to join or stay in the union, while unemployed members will loose their faith. Those who are still employed sometimes reject union membership so as to avoid the possible retaliation by their employer, who is then in a much more powerful position. Union density rate, which stands for the proportion of union members in the dependent labour force (which includes all wage-earners and those looking for a job). B & E (pp. 67-68) have argued that when union density starts to rise from a low level, at first nonmembers are incited to join as well; this is called the 'enforcement effect'.
9
However, after a certain level a further extension of the union movement becomes ever more difficult, which is referred to as the 'saturation effect'. Institutional variables. The impact of politics on union growth and decline has been taken into account by several theorists, who distinguish the effect of political parties on the one hand, and of laws on the other. In the influential study by Ashenfelter & Pencavel (1969, p. 439; hereafter referred to as A & P) it is argued that the percentage of Democrats in the House of Representatives proxies the degree of prolabour sentiment in society. More generally, the proportion of leftist parties in Parliament correlates positively with trade union membership, as those parties are expected to support the goals of the labour movement and hence facilitate the recruitment power of unions. Alternatively, the effect of labour laws is not a negligible factor, for these can either promote or obstruct unionization. Several country studies have supported this view (A & P, p. 438; B & E, pp. 87-98). Likewise, the social security system and its subsequent social laws can play a prominent part (Neumann & Rissman 1984, pp. 176177, 182-186). Unions used to have an important role in providing social benefit payments to their members. With the rise of the welfare state the national governments took over that function to a greater or lesser degree, thereby taking the wind out of the unions' sails; that is why we call this the 'substitution effect'. Strikes. Industrial conflicts often induce workers to join a union, not least in order to receive strike pay. If these actions are successful, this attracts new members as well; in union circles this flow is known as 'warprofits'. This is a very appropriate term, as these profits can easily be lost again after the immediate cause of the labour dispute has been solved. Subscription fee. A rise in the level of the union dues, being a cost factor, is expected to have a negative impact on membership rates. Although this variable is supposed to be influential on theoretical grounds, it is almost never incorporated in economic models, for time series on dues are hardly available. Exceptions can be found in Schnabel (1989b), who does not detect a significant impact of dues on German union membership, and in Pencavel (1971, p. 187), who finds a small significantly negative effect of dues on British union membership. Union organizing expenditures. It can be argued that the degree in which unions exert themselves to recruit new members, also influences the overall membership rates. One way to proxy these union efforts is by looking at the amounts of money they spend on organizing nonunion workers (see Voos 1984, Hirsch & Addison 1986, pp. 55-56). After this short exposition of the most relevant theories on union growth and decline, the next section will discuss the most interesting empirical results in time-series analyses.
10
2.2.3 Models of union growth: empirical results The earliest econometric model of union growth was developed by Hines (1964) on British unionization at the aggregate level during the period 1893-1961. Hines chose the rate of change3) of union density (ΔD) for the dependent variable. He then set out to examine which explanatory variables had to be put on the righthand side of the equation. After running some tests on the relation between ΔD and the state of demand for labour as measured by the level of unemployment and/or the rate of change of unemployment, Hines found that these variables were not significant. This same conclusion was drawn with respect to the rate of change of money wage rates. After these eliminations Hines arrived at the following model: ΔDt = c0 - c1Dt-1 + c2ΔPt-½ + c3Zt-½ + εt where ε is a random disturbance term. The level of unionization (D) was included to test the saturation effect, while the rate of change of retail prices (ΔP) was incorporated to capture the threat effect. Lastly Hines included the level of real profits (Z) as an indicator of the prosperity of industry, which influences the attitude of the labour movement during the process of collective bargaining as well as the ability of the employers to meet the workers' requirements. The estimation of the above equation led to the conclusion that ΔDt was only significantly affected by Dt-1 and ΔPt-½. Hence, according to Hines just two of the original five variables had explanatory power. Unfortunately, since it has been shown that the Hines model has many empirical, theoretical and methodological shortcomings (see for instance Purdy & Zis 1973, Thomas & Stoney 1970, Bain & Elsheikh 1976), it is impossible to indicate the significance of its results. The second econometric model of union growth that will be stated here refers to American unionization between 1900 and 1960 and was put forward by Ashenfelter and Pencavel (A & P) in 1969. Although their model has not gone uncontested either, it has nevertheless made a valuable contribution to the quantitative study on union growth and stimulated further research into this area. A & P defined their explanatory variables in such a way, that these could fit into the framework of the conventional costs-benefits analysis. However, they took a broad view of the assessment of expected gains and costs of union membership, for they also incorporated the impact of social and political forces into their model. The first two variables are purely economic, namely the percentage rate of change of prices (ΔP) and of employment in the highly unionized sectors of the economy (ΔE). A & P suggest that ΔP can function as a proxy of changes in real wages. When consumer prices increase rapidly, real wages are being endangered, at which point the expected benefits of joining a union are highest. Likewise, ΔE also reflects some costs and benefits of unionization. A & P argue that 'a major cost of joining a union, employer retaliation, is likely to be lowest when the labor market is tightening' (p. 436). Moreover, employment growth is favourable to successful membership drives: the organizing funds are larger and employees are more receptive to organizing 3)Henceforth the rate of change of any variable X is defined as: ΔX = (X - X )/X t
t
t-1
t-1
11
efforts. And finally 'with the advent of union security agreements in more recent years, increases in employment often lead automatically to upturns in union membership'. The third variable in the model is socio-economic by nature. Unions are not just regarded as organizations that can improve the conditions of employment, they can also be looked upon as 'agencies of protest' against the overall living conditions of the labour force. A & P contend that 'worker discontent is a function of the amount of unemployment in the preceding trough of the business cycle' and of the years elapsed since then (g(U,t-θ); where t is the current year and θ is the year of the preceding trough). Next, A & P introduce the variable (M/E)t-1, which they adopt from Hines. It stands for the proportion of workers that were previously members in the more unionized sectors and is included in the model to test the saturationist thesis. Finally, A & P take into account the effects of legislation. They realize that this legislation is 'to a large extent a reflection of the general climate of opinion', but since these attitudes towards trade unions can not be measured they settle for an indirect variable Dmt, which stands for the percentage of Democrats in the House of Representatives. It is assumed that most legislation in favour of unions is enacted by the Democratic Party and that most union members vote for Democrats. Having defined these explanatory variables, A & P incorporate them into the following equation: ΔMt = β0 + β1ΔPt + Σ=0 β2iΔEt-i + β3g(U,t-θ) - β4(M/E)t-1 + β5Dmt + εt where ΔMt is the annual percentage change in trade union membership. About 75 percent of the variance in union growth is accounted for by the explanatory variables; the empirical results quite strongly support the hypotheses advanced by A & P. For instance, they conclude that the significant coefficients on M/E indicate that 'further growth of unions is hampered by their own size' (p. 444). They also reaffirm Davis' findings that price changes have a strong effect on unionization (about 0.65% point). The coefficients on previous peak unemployment and on Democratic representation do have the right sign, but only considerable changes in these variables affect the size of the labour movement. Lastly, A & P find that union growth responds fairly strong to ΔE: in a period of three years an employment change of 1% leads to almost a 0.4% point change in union membership. After examining four historical subperiods, A & P come to the conclusion that they do not have to amend their model for any of these specific periods, which leads to their statement that 'a single behavioral relationship can explain the progress of the American labor movement in the twentieth century' (p. 434). The A & P model has led to many reactions; the most important criticisms of the choice of the variables will be mentioned below. Bain & Elsheikh (B & E 1976) bring up several arguments to adjust some of the explanatory variables. Regarding ΔE, they argue that there are three instead of one separate employment effects, which can all be of different magnitude. 'First, there is the effect of changes in the rate of growth of potential membership which, other things being equal, might be expected to lead to equal proportionate changes in actual membership. Second, there is the effect of changes in the excess demand for labour which, in A & P's view, determines the ability of employers to oppose unions and the ease with which unions can launch successful membership drives. Third,
12
there is the effect of changes in the distribution of employment between high and low density sectors of the labour force' (p. 38). Because these different effects can not be distinguished, B & E object to the use of ΔE, since this variable only measures employment in the highly unionized industries. In any case they prefer to introduce the unemployment rate, which refers to all sectors of employment. This variable would certainly do a better job in measuring the second employment effect mentioned above. A & P do use an unemployment variable (UP) themselves, but for quite another purpose, namely to capture the effect of labour's grievances on unionization. If these grievances mainly stem from dissatisfaction with rising prices and falling employment, which both are already included in the A & P model, then there is really no need to include UP also. And if there are other factors causing workers' discontent, these should be spelt out more precisely, according to Bain & Elsheikh. Their next comment relates to the variable M/E. In imitation of Mancke (1971) they question whether or not the relationship between M/E and ΔM should always be negative, as A & P imply. If the level of union density is quite low, employment growth might very well stimulate unionization; a saturation effect would then certainly not occur. Further, the correctness of the variable Dm comes up for debate. In the first place, not all Democrats are pro-labour and they have not promoted unionization constantly during a 60 year period. Secondly, other institutions such as the senate, the presidency and the judiciary are at least equally important political factors. And thirdly, in the United States much legislation affecting unions differs considerably from state to state. A final objection to the A & P model concerns the claim that it is capable of covering the whole period. Several students of labour economics (among others, Sheflin, Troy & Koeller 1981) have demonstrated theoretically as well as empirically (by re-estimation) that there really was a significant break in the structure of the model around 1937, shortly after passage of the Wagner Act. This law granted many rights and protection to unions and their members, thereby greatly facilitating union growth. Despite all the criticism of the A & P model it is still considered to be of fundamental importance, for it incited many researchers to elaborate and adjust the model and to apply it to other countries and/or later periods. Two more time-series studies influenced by Ashenfelter & Pencavel will therefore be touched upon. In 1971 Sharpe applied a similar model to Australia; he slightly adjusted the model both on theoretical grounds and because of specific Australian legislation concerning unions. The variables M/E and ΔE are maintained, but Sharpe uses U 'to measure more directly the effect of the excess demand for labour' and Δ(W/P) 'to measure more directly the effect of real wages'. Lastly, he introduces a dummy variable C to obviate the effect of special labour laws upon unionization during the early years of this century; these are assumed to have exerted a positive impact on union growth between 1907 and 1913. Although Sharpe has made a noteworthy attempt to explain Australian union membership growth he did not succeed in doing so, because other students (especially Bain & Elsheikh) have shown that his work contains too many theoretical, statistical and methodological weaknesses. Another application of the A & P model stems from a more recent date. In 1984 Neumann & Rissman published a time-series analysis concerning American union
13
membership during the period 1904-1980 'in the spirit of Ashenfelter and Pencavel's influential study' (p. 177). Like A & P, they also allow for the effects of variations in the economic as well as the legal and institutional environment. In addition, they consider the effects of alternative sources of supply of union services, in the shape of government expenditures on social welfare. The economic variables in their equation are captured by the rate of inflation, levels of unemployment and changes in the highly unionized sectors. These are used to re-estimate the A & P model, in which Neumann & Rissman at first only make one modification: the dependent variable, the rate of growth of union membership, is replaced by the percentage of the work force organized. The results of their new test are quite consistent with those of the original model, as far as the three economic variables are concerned. The other two original variables perform less well, which becomes especially clear when the old A & P model is used to forecast the unionization rates for the new period, 1961-1980. It turns out that the actual and predicted percentages diverge in the nineteen seventies. In order to adapt the A & P model, Neumann & Rissman do not omit any variables but they add two new ones. The first represents variations in the fraction of representation elections won by unions (%WONt), which 'can be interpreted as the direct manifestation of workers' preferences for union representation' (p. 182). The second explanatory variable, SWt, stands for government social welfare expenditures expressed as a fraction of GNP, included to test the substitution hypothesis mentioned in section 2.2.2. Estimation of the adjusted model leads Neumann & Rissman to the conclusion that the variable %WONt does not have much predictive power, but that the inclusion of SWt improves the efficacy of the model significantly. Hence, the substitution effect is substantiated with empirical evidence. We shall now end our overview of time-series studies by discussing the model developed by Bain & Elsheikh (B & E) in 1976 on union growth in the United Kingdom between 1893 and 1970. Although this model contains only four explanatory variables, it is very much worth discussing, since it certainly is one of the best reflected studies on unionization we have encountered. B & E have taken into consideration all the weaknesses of the earlier models before building an alternative one. Their model has however not been constructed to capture all the different determinants of union growth but rather to identify 'those of a strategic nature' (p. 116), which cause changes in union membership at relatively short intervals. The general form of their model is:ΔM = f(ΔP, ΔW, U, D, ε) The dependent variable, the proportional rate of change of actual union membership (ΔM), is preferred to union density (D) for various reasons. Firstly, in longitudinal studies annual variations in D are quite small. Secondly, 'the potential membership component of union density is generally characterized by considerable measurement error, and it is well known that such error in the dependent variable tends to produce autocorrelation' (p. 59). And finally, there are reasons to place D on the right-hand side of the equation, which will be shown below. The rate of change of retail prices (ΔP) is included to test the threat effect as well as the prosperity effect. Secondly, the rate of change of money wages (ΔW) is included because B & E argue that, since ΔP and ΔW do not necessarily have the same magnitude, they should both be incorporated. Furthermore, 'wages may be important not only in relation to prices but also in their own right' (p. 64), because employees
14
may want to join unions in order to raise their wages regardless whether there is inflation or not (credit effect). The third determinant relates to the level and/or the rate of change of unemployment (U, ΔU). It is expected that the total effect of unemployment on unionization is negative and lagged by one or two years. Lastly, as mentioned before, the prevailing level of union density (Dt-1) is included. The reason why D affects ΔM negatively is explained by the saturation effect. Yet B & E realize that there can also be an 'enforcement effect'. As union density increases, non-organized employees are incited to join as well 'either because of social coercion or because of union security provisions' (p. 68). B & E therefore take full account of the possibility that 'union density will have a positive impact up to some critical level of density after which its impact will become negative' (p. 68). This implies that D enters the equation in a linear as well as a quadratic form ((Dt-1)2), while the expected signs are positive and negative, respectively. After testing their model by means of Ordinary Least Squares it turns out that around 70% of the variance in the annual growth rates is explained by the four variables. The model 'is clearly satisfactory when judged by the criterion of overall goodness of fit as well as in terms of signs, magnitudes and significance of the estimated regression coefficients' (p. 79). In order not to undergo the same criticism that they themselves pass on against the earlier models, B & E subject the model to additional tests on predictive ability and on a possible structural break in the relationship which the model describes. They find the model's predictive power reasonably satisfactory and application of the Chow test indicates a structural stability. Next, they also experiment with the inclusion of a political variable similar to that used by A & P, and two dummies regarding public attitudes towards unions so that their model does not only consist of pure economic variables. However, neither of the alternative variables produces a good result. Since B & E think that socio-political factors can influence unionism they assume that these determinants 'may have an indirect impact upon union growth which is captured by the economic variables which are included in the model' (p. 86). Ultimately, B & E apply their basic model to three other countries with broadly similar industrial relations systems. In the cases of the United States and Australia, allowance is made for the effects of pro-union legislation upon union growth: in both models B & E introduce a dummy variable, to cover the effects of the Wagner Act in the U.S. between 1937 and 1947, and the effects of a compulsory arbitration system in Australia between 1907 and 1913. As for the third state, Sweden, B & E cannot find any statistical evidence that pro-union legislation has affected unionization, hence the original model is applied. After testing this model it turns out that the centralization of the bargaining system in Sweden since 1939 is the cause of a significant break in the structure of the model; however, B & E fail to estimate a preand post 1939 model. The insignificant sign of ΔW is attributed to this institutional change. It confirms the hypothesis that the more centralized the bargaining system, the smaller the credit effect of wage increases on unionization. All in all, B & E are satisfied with the results of these international comparisons: despite a few deviations, they ascertain that their four chosen variables, in some cases supplemented with a political variable, are important determinants of the rate of change of union membership.
15
Like all preceding empirical work, the book of B & E also received some critical comments. Pedersen (1978, p. 376) claims that 'union growth in the post-war years must be explained by a different structure from the one prevailing in the interwar years', while Richardson (1977, 1978) finds their theoretical analysis shallow and ad hoc, and criticizes their omission of compositional or structural variables. In a reply to this last criticism, Elsheikh & Bain (1978, p. 99) contend that 'such variables as the industrial and occupational composition of the labour force exhibit little annual variation because they are dominated by their time trends, and hence they are irrelevant in a rate of change model such as ours'. Even though some of the above models have been able to explain the annual fluctuations in union growth quite well, all researchers agree that time-series studies can never provide a conclusive explanation for membership growth and decline, since they only take short-term determinants into consideration. The rate of unionization is however also affected by a long list of structural factors, several of which are accounted for in the cross-sectional studies discussed below. The great merit of time-series models is the fact that they indicate to what extent membership fluctuations can be explained by a limited number of (mostly economic) key variables. As Bain and Elsheikh note, 'for a multiplicity of determinants constitutes a threat to explanation and understanding' (p. 117). But theorists like Hirsch & Addison (1986) object to such a confined explanation, especially because according to them the variables in time-series models only very indirectly measure the costs and benefits of union membership: 'possibly a more fruitful approach in the time-series analyses would be to use models that incorporate the personal and industry characteristics variables utilized in cross-sectional studies, variables believed to systematically proxy the benefits and costs of unionism. [...] it is possible that such factors could explain some significant portion of the change in union membership over time. (p. 57)' For this reason the next section will be devoted to an exposition of cross-sectional analysis concerning unionization.
16
2.3 Cross-sectional studies To account for changes in unionization, cross-sectional models include explanatory variables that represent several characteristics of individuals, occupations, industries and regions in particular. These features play a role in the decision to join a union but in view of their nature they do not exert much influence on the short-term fluctuations in union membership, but the more so on the long-term rate of trade union growth. Cross-sectional models can be used to serve a double purpose. The first and most frequently pursued objective is to determine the relative weight of each of the variables incorporated at one point in time by means of a multivariate analysis. The next aim is to determine how shifts in these structural variables have affected the pattern of union growth, by repeating the analysis at different points of time. This second objective conforms to the suggestion by Hirsch & Addison mentioned above and has already been tested on a few variables in the Dutch case by Visser (1987). Since most theorists have mentioned many characteristics in a greater or lesser degree, it does not seem meaningful to discuss all these studies separately. Instead, six groups of determinants are distinguished, which cover almost all the variables put forward by others. In succession, we shall expound personal, occupational and industry characteristics, wage-related variables, regional characteristics, the socio-legal environment, and trade union leadership & policies.
2.3.1 Personal characteristics Because joining a union is primarily an individual choice4), personal characteristics play an important role in determining the union status of workers. We shall discuss most of these characteristics mentioned in the literature. Sex. One of the most striking differences between groups of employees as for their union status, concerns the level of unionization among females set against that among men. The much lower rate of organization of women has traditionally been explained by the nature of female employment: women often work only temporarily in the labour force because of marriage and motherhood, they frequently occupy part-time jobs and most of the time their incomes are secondary to that of their husbands. Under these circumstances the degree of attachment to trade unions is low, and the costs of membership for women clearly outweigh the benefits. (see: Shister 1953, Moore & Newman 1975, Hirsch & Addison 1986) In addition to this argument, many authors (e.g. Bain & Price 1983, Freeman & Medoff 1984, Visser 1987) argue that the occupational distribution of females plays a more important role in explaining their low rate of unionization: employed women tend to concentrate in the less-organized sectors, especially in white-collar jobs and smallscale firms. We shall embark on these job and industry related characteristics later on in this section. At this point it is sufficient to observe that in most studies more stress is laid on the uneven occupational distribution than on any 'intrinsic characteristics peculiar to women'. 4)Economic theory is usually rooted in rational-choice, in which individuals compare the costs and benefits of any decision. With respect to unionization, an exception to this individual decisionmaking can be found in a 'closed-shop' system, in which union membership is compulsory. Furthermore, in sociologically oriented literature stress is often laid on the impact of social coercion, implying that union membership is not (always) the result of individual choice.
17
Age or years of experience. Shister (1953, pp. 421-422) theorized that younger workers show a greater propensity to unionize than older workers on several grounds. Among others, youngsters do not yet have so much feelings of loyalty towards their firm, their job mobility is higher, and they have not yet built up many seniority rights or other firm related benefits. This implies that employer retaliation for being a union member does not cost that much for younger workers compared to older workers. However, since then empirical work has not clearly confirmed Shisters arguments. The results are quite ambiguous, as is shown by Fiorito & Greer (1982) and Hirsch & Addison (1986). In fact many empirical studies have indicated that older workers are more likely to be a union member than younger workers. Bain & Price (1983) explain this result by reasoning that older employees have more need for a union to promote their interests, since they have fewer opportunities to improve themselves through job turnover, and because in many jobs productivity tends to decline with age. Schippers (1986) also finds a positive relationship between membership and years of experience; he ascribes this to the fact that the longer one works at one single firm, the more 'firm-specific' knowledge is amassed, which increases the worker's dependence on his employer, and by that also the need of union protection. Finally, several authors postulate a concave union-experience relationship. For instance Hirsch & Berger (1984, p. 675) find that 'the likelihood of union membership increases at decreasing rate, peaking at approximately 35 years of experience'. Due to all these different results, several age variables are constructed in empirical studies and no particular sign is predicted for their regression coefficients. Education. Most researchers presume a negative relationship between membership and schooling. It is assumed that higher educated employees have greater individual bargaining power and that they sometimes identify more with management than with the labour movement (Bain & Elias 1985, Fiorito & Greer 1982, Hamermesh & Rees 1988). In addition, Schippers (1986) argues that more schooling is most of the time accompanied by higher earnings, which in turn is presumed to exert a negative effect on union membership. Empirical work sometimes supports these assumptions (Scoville 1971, Farber & Saks 1980, Bain & Elias 1985), but in other cases no clear relationship is found at all (Hirsch & Berger 1984, Booth 1986). Recently, Hudley (1988, p. 195) has expounded that 'estimates of the relationship between union membership and schooling depend on the degree to which investigators control for occupational differences in propensity to unionize. When occupational factors are excluded, the relationship is negative; when included, it is positive'. Hence, like the sex variable, the education variable seems to affect union membership also via occupational factors, which are controlled for separately. The following variables will be mentioned only briefly, since they are included in only a few reports and because their impact on unionization has been found to be relatively modest. Race. This variable is only included in American studies, the majority of which reports a positive relationship between being non-white and union membership (Scoville 1971, Moore & Newman 1975, Fiorito & Greer 1982, Hamermesh & Rees 1988). It is reasoned that in spite of union racial discrimination minority groups still choose for collective protection by that union. Freeman & Medoff (1984) and Hirsch & Addison (1986) also point out that many studies find the union wage effect to be generally larger for non-whites than for whites. Yet, it is highly probable that the race 18
variable does not perform very well in other countries because of their different history with regard to discrimination practices. Family composition. It has been suggested by some authors (e.g., Freeman 1976) that marital status and number of dependents influence the propensity to become a union member, but only a few researchers have tested this hypothesis. Studies by Scoville (1971), Duncan & Leigh (1980) and Schippers (1986) indicate that workers with greater family responsibilities are more likely to join unions. On the other hand, Bain & Elias (1985) and Booth (1986) find the variable family commitments insignificant for men and women in Britain. Political preference or activity. Little attention has been paid to the influence of political party preferences or membership on unionization. Kornhauser (1961) establishes a positive relationship between a preference for the Democratic party and union membership in the U.S.A., but claims that this correlation is a consequence rather than a determinant of union membership. More recently, Schippers (1986) finds a significantly positive relationship between membership of a political party and union membership in the Netherlands. Unfortunately, this is all the evidence we have found on this variable until now. Religion. Again, only Kornhauser and Schippers include this variable in their research and once more their findings do not correspond. The first author finds no significant connection between religious involvement and union membership, while the second shows that in the case of the Netherlands there is quite a remarkable difference between the religious persuasions with respect to unionization. For instance, Humanists have a positive propensity to unionize, whereas the Roman Catholic variable has a significantly negative coefficient.
2.3.2 Occupational and industrial characteristics It is rightly argued by many authors (e.g., Bain & Price 1983, Hirsch & Berger 1984) that the personal characteristics influencing unionization are often interrelated with occupational factors, and 'affected by the way in which employment is distributed across and within industries'. Therefore we shall now turn to a survey of the most important occupational factors subdivided into job related, company related, and industry related characteristics. For the sake of brevity we have in places reduced some factors to the same denominator. Blue-collar versus white-collar workers. The literature is almost unanimous when it comes to assessing the propensity to unionize of manual and non-manual employees. Researchers have found blue-collar workers more likely to be union members for a great number of reasons. In contrast to white-collar workers, blue-collar workers have relatively more homogeneous preferences and working conditions, often characterized by team production, repetitive and machine-dominated work. These circumstances obstruct identification with management and make them easier to organize (see Hirsch & Addison 1986). Most studies discuss more extensively why non-manual employees are less inclined to join a union (e.g., Bain & Price 1983, Freeman & Medoff 1984, Visser 1987). Traditionally, students stressed the social position and status-consciousness as the main obstacles to unionization. More emphasis is nowadays laid on the fact that white-collar workers are less in need of union representation because they are better paid and higher educated. Beside that, employers have often been more opposed to the unionization of their non-manual employees. Finally, some recent studies ascribe
19
the low density of white-collar union members to the fact that these workers are very often employed in smaller firms and in smaller numbers. This effect of employment concentration on union growth will be discussed in more detail later on. One last aspect of the blue- versus white-collar issue concerns the so-called 'demanualization of the labour force'; the steady growth of non-manual employment during the last twenty-odd years has led to a preponderance of white-collar occupations in many Western countries, the effect on the rate of unionization will be commented on below. Industrial composition of the labour force. A lot of attention is paid to the industrial distribution of workers. Traditionally, some employment sectors have always had a low union density (especially agriculture), whereas others have had a high density level for a long time. Some authors distinguish a long list of different industries (see for instance Bain & Elsheikh 1979), here we shall follow Visser's (1987) distinction of just the three main sectors, namely agriculture, the industrial sector, and services. Agricultural workers have always had those characteristics that are unfavourable to unionization: they usually work isolated in small groups, alongside the employer and their jobs are often temporary. In most Western countries the share of agricultural employment has rapidly declined ever since the nineteen twenties. Today, agricultural workers account for no more than two per cent of the total labour force, which implies that their low propensity to unionize does not constitute a problem to the labour movement any more. From early days, the manufacture formed the stronghold of unionism. A few of the reasons have already been mentioned under the heading blue-collar workers. Further, industrial production is often characterized by a high concentration of workers, which makes them easier to organize. Also, many industries are capital intensive, implying that strikes are very costly for employers; this in turn means that unions in those sectors have a strong position, which attracts members. The industrial sector expanded everywhere until about the 1960's, but next industrial employment started to decline increasingly as a consequence of structural developments. For a long time this did not affect the union density rate negatively, but in many European countries union membership in manufacturing sank down after 1979; in the U.S.A. this already started after 1973 (Troy 1986). The activities within the third major sector, services, cannot really be lumped together because of their great heterogeneity. After Visser (1987), we can distinguish three subsectors, namely transport and communication; 'producer-oriented services' (commerce, banking and finance); and 'consumer-oriented services' (public service sector, hotel- and catering industry, real estate services and many others), all but the first one are dominated by white-collar jobs. As far as Europe is concerned, we can make the following remarks. The first subsector employs a stable 6 to 8% of the work force in most countries, and is highest organized, often even higher than the industrial sector. This also holds for the U.S.A. On the contrary, employees in the so-called 'producer-oriented services' have a very low propensity to join unions, while their share in total employment has grown steadily up to circa 15% in most countries. This development clearly causes a threat to the overall union density rate. Lastly, the 'consumer-oriented services' sector nowadays employs the majority of the labour force in quite a number of countries (e.g., Scandinavia, Great Britain and the Netherlands), mainly due to the strong expansion of the public sector since the nineteen sixties. Here the fairly high union density rate comes almost solely because of the public services, to which we will pay attention
20
more closely below. As for the U.S.A., Troy (1986) discusses services with the exclusion of the public sector. Here too services have the highest employment since 1982, but almost the lowest union density. Private versus public sector. If we compare membership developments in Europe between the private and the public sector, it becomes clear that unionization levels are generally much higher in the latter (Visser 1987). Several explanations have been put forward to account for this substantial difference. Much stress has been laid on the bureaucratic characteristics of the government sector, in which unions are expected to protect and improve especially the fringe benefits and other rights of public employees, since they have not so much control over wage bargaining (see Visser 1987 for references). Another important aspect is mentioned by Bain (1970), namely the fact that governments themselves often have been the first employers who recognized and facilitated the activities of trade unions. Furthermore, Visser claims that 'public employees enjoy a greater amount of job security and turnover rates are lower [...], [which] should favour union membership, as it favours both the chance of recruitment and membership retention (p. 48)'. Finally, Schippers (1986) ascribes the high density rate to the fact that public employees have little opportunities to find another employer in case of a conflict, which makes the need of union protection very great. Strangely enough, the union density rate of public employees in the U.S.A. has developed quite differently. For a long time unions in the public sector were heavily opposed by governmental agencies and hindered by unfavourable legislation. So, in American studies the percentage of state employment mostly has a negative effect on unionization (Scoville 1971, Moore & Newman 1976). Only recently (see Freeman & Medoff 1984) authors seem to have an eye for the fact that public sector unionization has expanded tremendously since the nineteen sixties, thanks to new public sector labour laws, reflecting a more kindly disposed attitude of the government towards the organizing of public employees. Industry concentration, establishment size and capital intensity. In this subsection we discuss some industry- and company-related characteristics, closely bound up with each other as is set out by Hirsch & Berger (1984). Employees in firms situated in highly concentrated industries are more likely to be unionized for a number of reasons. Firstly, an industry with a high concentration of sales will probably also have a high concentration of workers, which reduces the organizing costs for unions. Secondly, such an industry is likely to have a lower labour demand elasticity, which increases 'the union's ability to organize and to acquire compensation gains' (p. 667). And lastly, a highly concentrated industry usually yields high profits, thereby attracting unions who attempt to cream off part of these profits in favour of their members. All arguments mentioned above also apply to establishment size: larger firms are usually positively related to unionization. Another explanation for this higher union density can be found in the more impersonal relationship between workers and management, reflected by bureaucratic rules. However, it is also argued that in the case of very large companies, management might very well try to keep unions away by offering the same sort of services that are normally supplied by unions exclusively (Freeman & Medoff 1984, Cornfield 1986). Another company related aspect concerns capital intensity, to which also the same arguments more or less apply. Again, Hirsch & Berger (1984) point at the more inelastic labour demand, this time caused by a higher capital to labour ratio, implying
21
greater fixed costs and hence less firm mobility. This enables unions more easily to realize their demands. Furthermore, capital intensive firms are expected to sustain short-run economic profits, which stimulates union organizing activities. Finally, Hamermesh & Rees (1988) argue that these firms usually have invested much capital in expensive equipment that has to be protected by means of strict working rules. Machine-dominated work by itself already is not very gratifying, but when it is accompanied by close supervision, these unpleasant working conditions encourage labourers to join a union. This brings us to the next aspect, concerning firm ownership, especially with regard to supervision. Fiorito & Greer (1982) and Bain & Price (1983) pay attention to the employer's impact on unionization. Regardless of establishment size, it makes a difference whether the firm forms part of a large enterprise or exists on its own. In the first case opportunities for union growth are much greater because such a workplace is very likely to be characterized by impersonal bureaucratic rules, contrary to a single independent company. Moreover, workers in a multi-establishment enterprise have a sense of common interests induced by the so-called 'proximity influence' (Shister 1953), which makes unionization easier. And lastly, unions deal in these enterprises with professional managers, who are more inclined to recognize unions than the paternalistic managers or owners of independent firms will be. One final factor of importance may be foreign ownership, which is found to have a negative influence on unionization. Other job related characteristics. The remaining characteristics are quite divergent. The first one is in a way associated with the preceding factor, supervision. It concerns the occupational rank of an employee. Kornhauser (1961) and Schippers (1986) address this subject, but they interpret this concept in a different way. The former author talks about criteria of rank in terms of education, income and prestige. She finds that non-manual workers with a low rank are more likely to join a union; this result is linked up with the fact that this group of lower white-collar employees also have specific working conditions that favour unionization. Schippers operates the concept of rank in terms of position in the company; more precisely, he examines the relationship between union membership on the one hand and being in charge of a number of employees on the other. He finds that membership at first increases and later on decreases with the number of subordinates, but the variable turns out to be insignificant. The next characteristics are part-time employment and labour turnover, both exerting a negative impact upon unionization (Fiorito & Greer 1982, Bain & Price 1983, Schippers 1986). Part-time employment is closely linked up with female employment. Less working hours lead to a lower labour force attachment, and hence to a lower propensity to join a union. Moreover, although membership dues are often income dependent, it usually still is relatively more expensive to become a member. Research by Schippers confirms that people working less than 15 hours per week have a much lower union density than others. Bain & Price point at the important fact that part-time employment and other job and industry related characteristics that are unfavourable to unionism reinforce each other. A similar argument holds for job turnover. Since 'many of the advantages of unionisation are only forthcoming after several years of continuous pressure upon employers' (Bain & Price, p. 25), it does not pay to join if one often changes jobs. Again, this variable is closely associated with other unfavourable industry characteristics such as small firms and private ownership.
22
The very last job related characteristic has to do with risk due to injury or illness, otherwise called job hazards. Among others, Hirsch & Berger (1984) find a positive relationship between dangerous jobs and union membership. The need for union protection is greater in such jobs. Working conditions in general have an impact on the unionization rate. It is estimated that not only machine-pacing, but also high effort levels and low autonomy influence union membership positively. Beside that, several other dissatisfaction measures have been studied (see Fiorito & Greer 1982 for references). At this point we close this overview of occupational determinants and turn to the next group of characteristics.
2.3.3 Wage related variables Although one could think of reasons to classify this group of characteristics in one of the two preceding sections, most writers discuss it separately. Much attention has been paid to the effect of unionization on earnings, but just as well several researchers argue that earnings also have a positive impact on unionization. According to Ashenfelter & Johnson (1972) the effect of wages on unionism is statistically even more significant than the effect of unions on wage rates. Here we shall not pursue this question of causality any further, but focus on the impact of income on the level of unionization. Wages and earnings. Fiorito & Greer (1982) as well as Hirsch & Addison (1986) mention that many American studies find a positive impact of income on unionization, which they ascribe to the fact that union services are viewed as a normal good. Yet both writer-couples are cautious in accepting these results. Fiorito & Greer mention that 'none of these studies has controlled for likely simultaneity effects' (p. 11), while Hirsch & Addison argue 'that union membership is particularly unlikely for high-income workers, that is has decreased during a period of generally rising income, and that the positive relationship may be restricted to production workers in manufacturing' (p. 68). Bearing in mind these difficulties, Bain & Elias (1985) try to determine the relationship between earnings and unionization for Great Britain. They treat the earnings variables as exogenous and include them only on the right-hand side of the unionization equation. They find a non-linear, parabolic relationship: the probability of unionization increases with earnings but at a decreasing rate, and after a certain income level it decreases absolutely. A similar result is found by Booth (1986), but she also makes a distinction between manual and non-manual. The latter group deviates from the general pattern since their unionization probability only rises with an increase of the lowest incomes. Bain & Elias (1985, p. 84) interpret their findings as follows. They argue that 'this relationship may result from the way that the costs and benefits of union membership vary with the level of earnings'. One cost of membership are union dues, which tend to decrease as a proportion of income as the level of earnings rises. The second cost factor is employer retaliation, which usually increases with earnings from a certain income level, because employers are often more opposed to membership as the employee has a higher rank in the firm, and hierarchical position and level of earnings are positively related. The benefits of membership as a proportion of earnings also tend to decrease as wages rise, because unions often attempt to reduce the dispersion of earnings, thereby favouring the workers at the bottom of the income distribution, but injuring the workers at the top.
23
The union-nonunion wage differential. Hirsch & Addison (1984) are in favour of
including the potential relative union wage differential, because in their view it 'may measure much of the benefit associated with representation' (pp. 68-69). Unfortunately this variable is very difficult to observe, but several researchers (Lee 1978, Duncan & Leigh 1980, Farber 1983a, Hirsch & Berger 1984) have made an attempt and have found a positive, though sometimes only weak relationship between membership and the union wage differential. A correct measurement of this variable is hindered by at least two opposing effects. Firstly, students have to deal with the problem that the observed union wage differential is not identical with the expected potential differential at the time the decision to join a union is made. And secondly, as the differential increases, so does the opposition of employers against unionization. Somewhat more comprising than the wage differential, Fiorito & Greer (1982) mention wage inequity variables, which not only consist of wage differentials due to unions, but also of measures of perceived underpayment. They refer to several authors who 'all find similar pay inequity or dissatisfaction measures to be predictors of union membership' (p. 11).
2.3.4 Regional characteristics Regional differences are believed to be largely caused by differences in personal characteristics and in industry structure, already discussed in the above sections. Yet another part of these regional variations stem from different characteristics across areas, two of which will be reviewed below. Degree of urbanization. As becomes clear from the overview by Fiorito & Greer (1982), almost all American cross-sectional studies with urbanization measures, find a positive impact upon unionization. The most extensive explanation is given by Kornhauser (1961, pp. 53-54). She argues that large cities, especially if they are located in urbanized regions, are very likely to contain big industries, carrying with them a high concentration of the work force. This in turn leads to three positive effects on unionization. First, economies of scale result in lower organizing costs for unions. Second, the concentration of so many employees 'favors the development of a common outlook', thereby increasing the propensity to unionize (cf. Shister's proximity influence). And third, 'the competitiveness of the urban labor market and the proliferation of interest groups combine to lessen the power of employers in the large city as compared to small towns'. Curiously enough, most British studies do not pay any attention to the impact of urbanization. Only Bain & Elsheikh (1979) initially include the variable conurbations in their model, but omit it again after establishing that it does not significantly differ from zero. They expected this result more or less, because 'once the industrial distribution of industry is controlled for, then region [or urbanization] does not have a significant impact upon the degree of unionisation' (p. 152). Interarea differences. Unionism can greatly differ between the various regions of one state. American as well as British studies report a marked difference between the southern and northern part of the country, originating from all kinds of causes. Hirsch & Addison (1984, p. 64) guess that 'most interarea differences result from differences in personal, industry, or labor market characteristics, but that regional differences in attitudes do in fact make some difference'. These divergent attitudes are historically determined, and probably closely connected with industrial
24
developments. In the south of the U.S. agriculture has predominated for a very long time, which, as we have seen, does not form a fertile soil for unionism. On the contrary, rapid industrialization in the North favoured unionization there. Strong antiunion sentiments still prevail in the South, especially on the part of the employers: Farber (1983, p. 1432) states that 'the length of the queue for union jobs relative to demand is much longer in the South', indicating 'supply constraints on union jobs which may be due to a social and legal climate (typified by Right-to-Work laws common in the South)'. These laws will be discussed in the next section. A similar explanation holds good for Great Britain. This country can also be roughly divided into two distinct regions, the North and the South, the former having a much higher level of unionization. Research by Booth (1986, p. 52) 'may lend support to the view that membership customs are well established in these regions'. Blackburn (1967) amplified this statement by writing 'it seems clear that in the North there is more manufacturing industry and less money, while the people tend to be in lower social-status groups and more likely to join unions'.
2.3.5 The socio-legal environment As the social climate and the legal framework are not regionally determined per se, we will go into this matter in more detail by itself. However, before we shall discuss two different variables, it is good to mark an important passage out of Shisters work (1953). In this, emphasis is laid on the fact that although the legal framework does affect unionism, it should be remembered that this framework in turn is a lagged derivative from the prevailing climate of opinion, which again in turn is strongly influenced by the current work environment. Laws and public policies. Many American cross-sectional studies examine the impact of right-to-work (RTW) laws upon unionization. The Taft-Hartley Act of 1947 gave states the right to pass RTW laws that forbid a 'closed-shop' system. This means that in RTW states union membership is not compulsory, whereas in non-RTW states workers usually must join a union if they are hired by an employer in a union covered firm. Since 1947, twenty states, primarily in the south-west, have RTW laws, and on average unionization is much lower there. The issue is whether or not these low membership rates are directly caused by RTW laws. According to Hirsch & Addison (1986, p. 65), 'apart from a couple of recent studies, most estimates indicate that the effects of RTW laws on union membership are not statistically significant'. Most writers conclude that RTW laws mirror preexisting preferences, hence there is a big possibility of a reverse causality, leading from anti-union attitudes to RTW laws to low levels of unionization. One exception is a study by Ellwood & Fine (1987), who examine the effect of RTW laws on union organizing. They find a significantly negative effect of RTW by itself, because union organizing decreased only after the passage of these laws. A second government variable mentioned by Fiorito & Greer concerns the National Labor Relations Board (NLRB), which is a council in the U.S.A. that oversees representation elections and determines what are unfair labour practices by employers. Among others, the impact of NLRB political composition and procedures upon unionization has been examined. If we turn to the British situation, it becomes clear that in this country laws and government action have also significantly influenced unionization (see Bain & Price 1983). Union recognition and growth have been promoted with ups-and-downs not
25
only by favourable legislation, mainly when the Labour Party was in power, but also by the attitude of the subsequent governments, encouraging employers to adopt a more professional approach to industrial relations and to accept collective bargaining. Hence, we round off this discussion of the legal structure by quoting Hirsch & Addison (1986, p. 67): 'legal rules, or, more generally, any factors significantly affecting the marginal costs of union organizing, are important determinants of unionism'. We shall pay some more attention to union organizing in the next section, where we discuss union policies. Leftist governments. Although this variable has been included in several time-series models, it has also been used in cross-sectional studies in which unionization is compared between different countries. A recent example of this is an article by Wallerstein (1989), in which he analyzes union density in advanced industrial societies. He finds the so-called cumulative impact of leftist governments upon unionization levels very significant. A similar exercise has been undertaken by Visser (1987), who also examines the political influence of Labour. He finds a very high correlation between union membership rates and left party control, but he is hesitant to pronounce upon the direction of causation or upon the relative weight of this variable.
2.3.6 Union leadership & policies Shister (1953) distinguishes three main determinants of union growth. Besides the work environment and the sociologic framework, union leadership plays a very important role. According to him, the ability of union leaders to pursue the correct organizing techniques, to set up the appropriate structural and administrative machinery, and to create good bargaining relationships with employers and governments all contribute to the unionization level, within the limits set by the economic and institutional environment. Bain & Price (1983, p. 31) criticize this point of view: 'only a very small amount of union growth can be unambiguously attributed to the independent influence of union leaders and policies'. Thereupon they quote Bain & Elsheikh (1976, p. 23) who argue that 'union leadership is dependent upon, and constraint by, the same socioeconomic forces which motivate or enable workers to join trade unions' and, 'as such, it is very much a secondary and derivative determinant of aggregate union growth'. Still, Fiorito & Greer (1982) report a lot of studies that include measures for union instrumentality beliefs or measures for perceptions about union effectiveness. These can be conceived as perceptions about union policies, in which case this factor does play a role in determining the volume of the labour movement. A similar case is made by Albeda (1971), who claims that the high turnover in union membership can be viewed for a large part as a reaction to union policies and bargaining results. One final point of discussion relates to union efforts to fight the free-rider problem and to reduce high turnover rates in membership. The degree of success in handling these problems undoubtedly influences the unionization level in the long term. A closed shop system, in which union membership is compulsory, would be most effective, but is forbidden in many states and countries. Alternatively, unions can try to attract new members and to maintain the old ones by offering selective benefits and services only available to members (see Visser 1987 for references).
26
2.4 Concluding remarks The survey of economic literature on trade union membership makes clear that a great variety of explanatory variables has been brought forward. The majority of the theories adduced above can be applied to any given western country, as far as the availability of data permits. Still, allowance has to be made for specific circumstances in the country at issue, in this case the Netherlands. A study of the evolution of the Dutch labour movement, against the background of historical developments in the society at large, may lead to additional insights into the unionization in the Netherlands, which can not be explained by the general theories. Therefore the next chapter will be dedicated to the history of the Dutch trade union movement, with special reference to membership developments.
27
28
3.SHORT HISTORY OF THE DUTCH TRADE UNION MOVEMENT 3.1Introduction In this chapter an overview will be given of the emergence and further developments of trade unions in the Netherlands. The history of the Dutch labour movement has already been described in detail by several authors. Among those are Brugmans (1925), who described the early period, and Harmsen & Reinalda (1975) and Windmuller et al. (1987), who have analyzed the whole history of Dutch trade unionism. The aim of this treatise is to give a deeper understanding of the specific features of the Dutch union movement by discussing the economic, social and institutional climate of the Netherlands, from the period in which the first unions emerged until now. Against this background the development of the union membership figures will be brought into prominence. Furthermore, the findings will be interpreted in terms of the supply and demand framework as set out in chapter 1. It must be noted however, that information about the number of union members in the 19th century is very scarce. This conclusion had already been drawn in 1904 by Hudig, writer of the first thesis on the Dutch trade union movement (pp. 245-248). This lack of data implies that the description of the early period will largely be qualitative by nature. 3.2The formation of the first trade unions, between 1860 and 1900 The emergence of the first unions in the Netherlands coincided with a certain acceleration of economic growth between 1860 and 1870 (Windmuller, p. 15; Van Vuurde, pp. 101-102). These workers' associations did not arise among industrial labourers, but among the more educated craftsmen. The explanation of this phenomenon is sought in the socio-economic conditions in which the Dutch wageearners lived in the middle of the 19th century (Brugmans, pp. 188-192; Harmsen, pp. 3-7). Large industries had not yet evolved, and consequently large concentrations of the labour force did not exist. Even more important, there was as yet no question of any class-consciousness among wage-earners. Due to a scarcity of work they lived in such straitened circumstances (bad housing, bad health, little or no education) that they completely resigned themselves to their fate. Many people were either dependent on the poor relief of the churches and government, or committed to the goodwill of the entrepreneurs, who were in a way regarded as philanthropists. Acquiescence and ignorance are the keywords that every historian uses in describing the labour force of those days. In explaining why the mass of the workers became unionized relatively late compared to countries like Great Britain, the United States and Germany (Eickhof, pp. 89-94), Brugmans concludes by saying that "in order to initiate the corporate life it is necessary to reach a certain level of education, as well as a certain level of well-being, however small" (translation, p.189). This last remark makes it understandable why the first workers to organize were educated craftsmen, whose market position was much better than that of the underdeveloped masses. There already existed several social clubs and mutual relief
29
funds among typographers and diamond cutters. These associations created a fertile soil for the first real trade unions, which came into being in a period characterized by fast rising prices, causing a decline in real income (Brugmans, pp. 258-261; Lammers, p. 16). Also in the other countries the educated artisans organized in unions well before the un-trained workers, for the same reasons (Eickhof, pp. 76-86). The first trade union5) was founded in 1861 by typographers in Amsterdam, and was called the Mutual Relief fund 'Typography' (Onderling Hulpfonds 'Boekdrukkunst'). Its main object was to distribute allowances to unemployed members at such a high level, that this would prevent workers from accepting pay below the market wage. Although this clearly was a defensive objective, from 1865 onwards the union also engaged in campaigns for higher wages (Brugmans, p. 259; Giele, pp. 36-37). Not long after this foundation many other crafts followed, in the beginning mainly in Amsterdam. Other centres of some importance were Arnhem, the province of Friesland, and the larger cities in the West (Brugmans, p. 261; Giele, p. 38; Hudig, p. 245). The dominance of Amsterdam in the early period is explained by Brugmans (pp. 261-262). To begin with, unionism arose first among craftsmen, who were still preponderant in Amsterdam. Next, these trade unions gave the example to other groups of workers in the city, but not so much outside of the city, since inter-urban communication was not very far developed yet, especially not among the working class. Finally, Brugmans attaches great value to personal influences. The emergence of the so-called secularists or freethinkers movement in Amsterdam had an important impact on the young labour movement, because several union leaders emanated from this group of democrats.6) Already in 1866, the typographers took a step farther by forming the first national union, the General Netherlands Typographers Union (ANTB)7). Objectives were support for unemployed members and campaigns for higher wages. The ANTB was in favour of defensive strikes, in case the terms of employment were threatened. This was very remarkable, since the prohibition to strike was still in force (this law was abolished in 1872). The employers reacted with a frontal attack, by firing all union members and putting them on a black list. The subsequent large appeal to the union's insufficient unemployment fund caused the membership to drop from 755 in 1866 to just 274 in 1869 (Giele, pp. 37-38; Harmsen, p. 8). The next move of the unions was an attempt to join forces beyond the bounds of occupation. However, differences of opinion about union policies and attitude existed from the very beginning, which impeded a united front. One group strove after class struggle to reach the socialistic ideal state. They joined the International Workingmen's Association, and founded a Dutch branch in 1869: the First InternaA trade union is defined as an organization which has as its primary aim improving the terms of employment of its members, if necessary by putting up a struggle.
5)
This movement's objectives were to dissociate themselves in their thinking from the authority of church, tradition or convention, and to "search instead for the truth, guided by nature and reason", for the benefit of the society.
6)
In the appendix of this chapter a list is taken up of all abbreviations and their Dutch and English names in full.
7)
30
tional (Giele, pp. 39-43; Van Vuurde, pp. 103-104; Brugmans, pp. 263-264; Harmsen, pp. 10-13). They aimed at becoming an umbrella organization for all trade unions, but although a few hundred individuals joined, hardly any union did. This was probably due to a lack of information about the International and its goals, its radical and anticlerical image, and the fact that union members had to pay extra subscription fee. The International was granted only a short life8), but it did have an important impact on the development of the labour movement. It gave the initial impetus to the consciousness-raising of the workers, by propagating a class-feeling and by vindicating political rights, notably universal suffrage. For the first time manual labourers were addressed, which induced a number of new unions for several groups of unskilled workers. In 1871 a second attempt was made to set up a trade union federation: the General Netherlands Workingmen's Association (ANWV), which attracted much more members9) because of its moderate attitude (Brugmans, pp. 265-268; Giele, pp. 4344; Van Vuurde, pp. 104-105; Lammers, p. 17). Most Dutch workers were not interested in international solidarity, nor in class-struggle; they preferred to reach an understanding with their boss through co-operation. These traditional views on the relationship between employer and employee were reflected in the program of the ANWV, displaying a stand against socialism, revolution and violence; as strikes were to be avoided, the ANWV did not find it necessary to build up a strike fund. Cooperation with the employers was considered of high importance. The ANWV aimed at raising the welfare of the working class by making an effort for the advancement of education, the establishment of shop co-operatives and universal suffrage. It was determined that only wage earners could join (for fear of patronizing), but the Liberal Party still had a lot of influence on the federation's policies, as a result of which the ANWV in fact not really served the workers' interests (Hudig, pp. 77-78). Not all members supported the moderation of the ANWV. Criticism came from the right as well as the left wing, leading to two separate secessions (Van Vuurde, pp. 122126; Giele, pp. 44-47; Harmsen & Reinalda, hereafter referred to as H & R, pp. 5153, 57-60; Windmuller, pp. 21-22). On the right, Calvinist members objected to the "unchristian" policies of the ANWV, such as the refusal to repudiate strikes, and the assembling on Sundays. Immediate cause of the separation was the official decision of the ANWV to devote itself to the advancement of neutral education, which induced Christian members to found their own organization: Patrimonium, in 1877. This establishment was really a step back in the development of the union movement, because Patrimonium threw its membership open to workers as well as employers. Its main objective was to act After a split in the international association between the followers of Marx and those of Bakoenin in 1872, the Dutch section shortly afterwards ceased to exist.
8)
At its formation, 5 local unions joined, representing 740 members. Within one month, this number increased rapidly to 10 branches with 1,800 members in total. Mid-1872 the membership figure had already risen up to 3,400 (within 16 unions), while the ANWV reached its peak in 1876, comprising about 5,500 members belonging to 56 unions (cf. Heldt, pp. 4, 5, 23).
9)
31
according to the Christian principles, and to strive after class reconciliation instead of class struggle. It rejected any form of government intervention in labour relations. Therefore, Patrimonium cannot be considered a real trade union, which only slowly changed in the 1890's. The left wing of the ANWV consisted of ex-members of the former International. They tried to incite the ANWV to more action and to take a progressive stand. The first clash occurred when the ex-Internationalists wanted nonworkers to be able to join as well. Half a year later, in June 1878, the split became an accomplished fact, as the leftist members had in vain proposed to adopt the social-democratic program of the German Labour Party. After this failure they founded their own society in Amsterdam, which became the Social-Democratic League (SDB) in 1881. Hence, from the late 1870's the Dutch trade union movement was characterized by three mainstreams: a liberal-moderate, a Protestant-Christian, and a socialistic movement. Due to the emergence of the last two directions, the first one (the ANWV) lost its ascendancy during the course of the 1880's. Its membership gradually declined from about 4,000 members in 1880 to a little over 2,000 members in 1893 (Heldt, pp. 41, 117). A lot of its branches were dissolved or went over to the SDB. The ANWV probably did not lose much of its members to Patrimonium, which mainly attracted unorganized people who considered both the SDB and the ANWV too radical. Its membership grew quite rapidly from approximately 1,000 in 1880 to 9,765 in 1892 (Kruithof, p. 371), but not all of these people were wage-earners10). The same holds for the SDB, which really was a socialist political movement with trade unions at its basis. Although the SDB instigated the formation of many labour organizations, its own membership figures were never that spectacular. The radical SDB was in such a bad odor with the employers, that the threat of discharge kept many workers and unions from joining (Giele, p. 49 f). Van Horssen & Rietveld (pp. 18-20, 28-33, 50-52, 58) have assessed that the SDB started with about 200 members in 1881, and at first only slowly increased with a few hundreds each year. The economic hardships in the middle of the 1880's made the workers susceptible of the political agitation of the SDB11), which led to a sudden increase of members to 4,000 in 1886, especially in the cities in the West. However, many of these new members left again soon afterwards, when it became clear that their objectives would not be realized at short notice. A second period of bloom occurred at the start of the 1890's. Due to a severe agricultural crisis many country labourers joined the SDB, which reached its peak in 1893 with about 5,000 members. Thereafter the organization declined owing to a rift in 1894 between the anarchistic and the parliamentarist movement, whereupon the last group founded the Social Democratic Labour Party (SDAP). The SDB, or
Hagoort estimates the number of nonworkers around 8% of the members (p. 55).
10)
The SDB pleaded ardently in favour of universal suffrage, which was thought to bring about a socialistic society soon, thanks to the majority of the labourers' votes (Verberne, pp. 110-112; Van Horssen & Rietveld, pp. 42-43).
11)
32
actually its successor12), even fell back to just 600 members in 1898, after its charismatic leader (Domela Nieuwenhuis) had left, and two years later ceased to exist. In the last decade of the 19th century the Dutch labour movement was reinforced by several developments (Verberne, pp. 118-121, 138-140). The number and intensity of labour disputes rose from 1888 onwards, often leading to satisfactory results for the strikers. This had more appeal for the workers than politics, especially since the efforts of the SDB to acquire universal suffrage had fallen short of expectations. As it became clear that the SDB was not capable of organizing the majority of the workers, an international socialistic initiative led in 1893 to the formation of a 'national labour secretariat': the NAS (H & R, pp. 66-73; Verberne, p. 141; Lammers, pp. 20-22). It was meant to become the centre for all labour organizations, unions as well as political parties. In its initial pursuit of neutrality it also invited the ANWV and the confessional organizations to enter; that failed, but almost all the other unions and the SDB did join, which showed the need of such a co-ordinating body. After the above mentioned split within the SDB in 1894, the SDAP also joined the NAS, but in that way the divergence of political views was taken along. This situation caused the union movement much harm 13), and so it was decided to drop the political parties from the NAS, which consequently became a pure trade union federation in 1896. However, it soon became clear that the leaders of the NAS adhered to the syndicalistic ideas of the SDB. This boiled down to a line of policy in which the union leadership was guided in most matters by the opinion of its members. Its main objective was to stimulate revolutionary strikes as much as possible, but its weakness lay in the absence of strike funds. Financial support for the strikers was supposed to come from spontaneous solidarity, but that often failed to occur. The resulting defeats demoralized the workers, keeping the NAS from growing (Lammers, pp. 2224). Meanwhile, the ANWV experienced a new bloom after its low point in 1893, due to an orientation towards the 'liberal democracy'. This led to an increase of members to 5,000 in 1902. However, the ANWV dealt more with politics than with real action on behalf of the labourers, and therefore it slowly withered away until its end in 1921 (Verberne, p. 108). The close of the century also showed the first cautious move in the direction of unionism among Christian workers, Protestants as well as Roman Catholics. In the course of the 1890's, the Protestant leaders gradually became receptive to the cry from below for more workers' involvement, prompted by agricultural crises and a breakthrough of the industrialization process in the Netherlands. In 1891 a Christian Social Congress endorsed social legislation and even favoured the right to strike, within certain limits. This stimulated the formation of a few unions, but these remained formally dependent on Patrimonium until 1905. Still, the mental attitude towards real trade unions slowly ripened, all the more since the view, that workers In 1894 the SDB was banned, but was almost immediately replaced by the Socialistenbond (SB).
12)
In its first year, the NAS counted 15,728 members, which climbed to 18,700 in 1895, but fell back again to 12,700 members in 1896 (H & R, p. 430).
13)
33
should be enabled to stand up for their own rights, had gained the upperhand (Windmuller, pp. 29-31; Kruithof, pp. 352-360). The first Catholic associations emerged in 1888, but were not real trade unions either. They were led or at least advised by priests, and their main objectives were social gathering, mutual support and, above all, protection against the advancing socialism. This submission to the Church had its origins in the age-long isolation and discrimination of the Roman Catholics by fellow countrymen who were predominantly militant Calvinists. An important stimulus for the emancipation of the Catholic workers was the encyclical Rerum Novarum of 1891, which spoke openly in favour of separate labourer's associations, and of social legislation. Socialism was powerfully condemned. Thereupon, more Catholic trade unions emerged, but the preponderant influence of the clergy remained. This was perfectly illustrated after the foundation of 'Unitas', an interdenominational union for the textiles industry, in 1895. After several years of debate, the Dutch bishops decided in 1906 that Roman Catholic workers should be organized exclusively in Catholic organizations, and this was done accordingly (Verberne, pp. 132-137; Windmuller, pp. 31-33; Roes, pp. 3032). One last development should be mentioned, before this section on the 19th century is closed. This concerns the foundation of the General Netherlands Diamond Workers Union (ANDB) in 1894, which is considered to have been a watershed in the history of the Dutch union movement (Windmuller, pp. 28-29; H & R, pp. 7477). The leader of this democratic socialist union, Henri Polak, had worked in England for four years where he had become enthusiastic about the organizational principles of the 'New Unionism', which would come to serve as a model for his own union. In short, these principles implied a centralist administration, unconditional discipline, full-time paid leadership, and high dues to finance large strike and insurance funds to be allocated to members only. This approach bore fruit from the very start, turning the ANDB into a strong, efficient union that was soon able to make the membership mandatory for all diamond workers.14) From this it can be deduced that Polak was diametrically opposed to the policies of the syndicalist NAS, which was accentuated by the fact that the ANDB called itself the 'modern trade union movement' as distinguished from the 'old-fashioned' principles of the NAS. It would lead to a decisive confrontation in the first decade of the 20th century. Even though so little is known about the overall number of union members before 1900, scholars agree that the union density rate (the number of union members as a percentage of the dependent labour force) must have been very low (Hudig, pp. 245248; van Arnhem & Knegtmans, p. 47). According to Giele (pp. 48-49), until 1878 union members could only be found among skilled workers, who constituted one third of the total working population at the most. He roughly estimates that in 1878 The success of the ANDB was also partly due to the favourable environment: the highly skilled diamond workers were almost entirely concentrated in Amsterdam, which made them a strong collective power, if organized.
14)
34
no more than 10% of these skilled workers were members of a real trade union. Lammers (p. 21) rates the total membership figure in 1896 at 30 to 35 thousand, which would imply a density rate between just 1.5% and 1.8%. The causes of this slow start can also be explained in terms of supply and demand. The supply of unions was never really hindered by the law, as the right of free association already existed since 1855 (Verberne, p. 66). Moreover, political groupings (socialists as well as liberals) tried to stimulate Dutch unionism, but without many results. The relative late industrialization of the Netherlands, and the large dispersion of economic activity delayed the development of a labour movement. The demand for union organizations therefore first arose in the artisan trades among the skilled workers, who had some experience in assembling and whose market position enabled them to make a moderate stand against the employers when their standard of living was being endangered. Still, most associations had to deal with strong resistance from the employers and with economic downturns, both leading to big fluctuations in the trade union membership. Within the scope of this, the 19th century showed a fast emergence and disappearance of unions, all characteristics for a "birth era": the supply of, and the demand for unions were still in disequilibrium. These phenomenons could also be detected in the surrounding countries (Eickhof, pp. 100, 105-107). Just as elsewhere (see Eickhof, p. 170 f) Dutch unions soon supplied social insurances against illness, decease and unemployment, to retain members (Hudig, pp. 252-253), but only the ANDB had a strike fund as well. This union was also the only one that had a closed shop, while this was already much more common in Great Britain and the United States, where it could lead to membership growth (Eickhof, pp. 179-183). The rise of Dutch unionism was further hampered by the prevailing attitude of the labourers - submissiveness and willingness to co-operate with the employers -and by their religious beliefs which preached this obedience. In addition, the Protestant and the Roman Catholic leaders stimulated the formation of labour organizations under the wing of the church, to restrain the christian workers from joining socialistic unions. These actions also obstructed unity in the union movement. In fact, they were the beginning of a long-lasting tripartite division of the whole Dutch society, better known by the terms 'pacification', 'fractionalization', or 'pillarization'.
35
3.3 Growth of the trade union movement, 1900-1940
3.3.1 Emergence of the tripartition in the labour movement
From the beginning, the labour associations were divided among one another with respect to objectives and ideology. Roughly, four directions could be distinguished at the turn of the century, namely Protestants, Roman Catholics, liberals (ANWV), and the progressives (predominantly NAS and ANDB). The Christian groupings were both growing stronger, while the ANWV had lost its prominence. Within the leftist camp, the NAS was becoming ever more syndicalistic, reducing the possibility of an overture to the social democratic ANDB. All these differences of opinion became painfully manifested during the railroad strikes of 1903 (see Rüter for a full description). In January 1903 there was a strike in the port of Amsterdam, which spread to the railroads when the railroad men refused to move freight from and to the striking dockareas. Since their working conditions were bad as well, they could easily identify with the dockworkers. This led to a very sudden and complete collapse of the rail transport. The railroad companies were taken by surprise and quickly met all the demands of the strikers. This victory convinced the syndicalist unions that the time was ripe to increase the existing revolutionary sentiments. But it also induced the Calvinist government to introduce bills that would prohibit all strikes in public services and the railroads, which prompted the socialists and syndicalists to jointly form a Resistance Committee so as to organize a general strike. Protestant and Catholic organizations reacted to that move by forming their own committees, to counteract revolutionary agitation. By the time the proposed bills were to be discussed in Parliament, the Resistance Committee had become internally divided on the question whether to strike or not. The syndicalists wanted to proceed, but the socialists feared that alert military forces would torpedo their attempt. Still, it was decided to call the railroadmen out on strike on April 5, while a general strike was proclaimed on April 8. However, the walkout lasted only a few days and was far from successful, due to a bad organization, military counter-moves, Christian strike-breakers and the denouncing public opinion. The defeat was at once followed by the dismissal of thousands of labourers, mainly railroad men. This disastrous second strike is viewed by Windmuller as "one of the most important turning points in the country's social history" (p. 36). Regarding the labour movement, the already existing differences of opinion between socialists and syndicalists, as well as between socialist and denominational organizations, had sharpened. The socialists accused the NAS of pursuing an irresponsible policy, and many members concurred: after the first strike, the NAS membership rose from 10,500 to 17,00015), but due to the following great defeat this figure declined again to just 8,000 at the end of 1903, and even fell back to circa 3,500 in the second half of that decade (H & R, p. 430). Conversely, the confessional unions had gained members, which was seen as a justification of their attitude during the strikes (Verberne, p. 168 f; Windmuller, p. This number is mentioned in the union periodical of the NAS, 'Het Correspondentieblad', in volume 1902/03.
15)
36
37). The socialist unions had lost members, except for the ANDB, which had been against striking, but had participated nevertheless, out of solidarity. This stand and its resulting stable membership had much impressed the other socialist unions, and would come to serve as an example. The growing antagonism between syndicalists and modernists impelled the NAS in 1904 to challenge the critics to found a second federation. The gauntlet was taken up by several union leaders, who asked Henri Polak of the ANDB to take the first step in forming a new centre. This was done accordingly, and so the Netherlands Federation of Trade Unions (NVV) came into being on January 1, 1906 (Verberne, pp. 164-167; H&R, pp. 88-95; Windmuller, pp. 39-40; Lammers, pp. 26-27). The policies and organizational structure of the NVV were modelled after the ANDB. This implied a strong central leadership with full-time paid administrators, disciplined behaviour and workers' education. Strike action would only be taken when necessary, but always well organized and financed out of centralized strike funds. However, to improve the working conditions preference was given to collective agreements and protective labour legislation. Finally, the federation built up benefit funds in the event of illness, accidents, unemployment and decease. By means of these social insurances the NVV tried to hold on to its members. Originally, the NVV wanted to remain independent of the SDAP, to attract nonsocialist workers as well. This soon appeared to be untenable, since the NVV strived after labour legislation, for which it needed the parliamentary support of the SDAP. From 1908 the two socialist organizations worked closely together, initially to realize universal suffrage. Eleven unions had joined the NVV in 1906, representing 18,960 members. The number of affiliated unions and members rose quickly and three years later the NVV had already doubled (see table 3.1 on page 48-49). This rapid growth accelerated the formation of a Roman Catholic and a Protestant federation. At the outset of the 20th century, a few labour associations could be distinguished within the Protestant movement (Kruithof, pp. 357-365, 372-373). Next to Patrimonium, which mainly represented the working-class circles, the Christian National Workmen's Union (CNWB) emerged in 1890, which recruited its members predominantly from the lower middle-class. Patrimonium did not really co-operate with this very conservative CNWB, which had about 8,000 members in 1903. Only in this year the two organizations shortly joined forces, during the railroad strikes. Within Patrimonium a difference of opinion existed whether to found independent unions or not. When a majority was in favour of this, it led to the formation of the Christian Labour Secretariat (CAS) in 1900. Only in 1905 the CAS became fully independent from Patrimonium. Its membership remained very modest, due to a lack of decisiveness and coherence. At its height in 1904, not more than 3,000 workers were members, while this number declined to just 786 in 1909. The emergence of the strong NVV finally urged the Protestants to create a united front against the advancing socialism in the labour movement. For that reason the Christian National Trade Union Federation (CNV) was founded on July 1, 1909. The term 'Christian' was deliberately chosen, because this centre intended to bear an interconfessional signature. But it soon became apparent that the federation would turn into an exclusive Protestant organization, because of a strong resistance from the
37
Catholic Church as well as the own ranks. The CNV started with 5,000 members, and increased to 11,000 just before the first World War in 1914. Although the Catholics shared the opinion of the CNV about the 'pernicious socialism', they dissented from interconfessional co-operation, at least the church authorities did. Out of self-preservation the Church was forced to permit the formation of a national union centre. Hence, in July 1909 the Bureau for Roman Catholic Trade Unions (Bureau voor de Rooms-Katholieke Vakorganisatie) was installed, counting about 9,000 members at its start; in 1914, this number had already increased to 29,000 (Windmuller, pp. 42-45; Roes, p. 33 f; Perry, p. 295 f). The foundation of the Bureau regenerated the controversy between advocates of the so-called estate organizations (standsorganisaties) and those who were in favour of national trade unions. Since 1906 the former group was organized in the Federation of Diocesan People's and Workers' Leagues (Federatie van Diocesane Volks- en Werkliedenbonden). Even though religion was the first matter in both Catholic centres, the Federation put more emphasis on the satisfaction of spiritual needs, while the Bureau strove after improving the terms of employment. The workers themselves felt more need for the promotion of their material well-being, which caused a threat to the survival of the estate organizations. As the Catholic Church wanted to retain her ideological influence on the labourers, from 1916 onwards all Roman Catholic union members, except for white-collar workers, were obliged to become a member of a diocesan workers' league as well. This dual structure within the Catholic labour movement would last until the liquidation of the estate organizations in 1963.16) Summarizing, at the beginning of the 20th century three federations came into being, which would dominate the union movement until the 1970's. These union centres were each an exponent of one of the three main ideological segments in Dutch society. The Protestants and the Catholics had been diametrically opposed to each other ever since the 17th century. The rise of socialism created the third ideological block, as the leaders17) of both religious blocks kept their ranks firmly closed against any socialistic influence. Although the fourth movement of the syndicalists continued to exist until World War II, its size was too small to play any role of importance. Graph 3.1 illustrates the development of all four federations. Furthermore, a few other centres came into being in the first half of the century, the membership of which remained relatively insignificant18). Despite this multitude of movements we still usually speak of a tripartite division of the labour movement. The obligatory double membership, accompanied by double subscription, lasted until 1945 (Perry, p. 301). Several authors argue with forceful arguments that the emergence of the pacification was imposed authoritatively by the church leaders; the initiatives certainly did not come from below (see H & R, p. 102; Kruithof, pp. 366-367). This concerns the Liberal Netherlands Federation (ANV), founded in 1912, which merged in 1929 with the Federation of White-collar Workers (VVH, dating from 1925) into the liberal Netherlands Trade Union Centre (NVC). Also, a few break-aways could be detected, of which the Netherlands Syndicalist Federation (NSV; a split-off from the NAS since 1923), was the largest one. Furthermore, there existed two centres especially for civil servants (the CRP from 1916, and the CMHA from 1917). See also table 3.1.
16)
17)
18)
38
Graph 3.1Union membership of the NAS, NVV, CNV and RKWV* 1899-1939 (end-of-year data, in thousands)
*
Including its predecessor, The Bureau for Roman Catholic Trade Unions.
Source: see below table 3.1
In the United States, Great Britain and in Scandinavia, trade unions were not divided along political and religious lines, unlike several Western and Central European countries. In Belgium and France, the main division existed between leftist and Christian (i.e., Roman Catholic) unions, while in Germany and Switzerland there was also a distinction between Catholics and Protestants at first, but this soon disappeared (see European Industrial Relations Review; Eickhof, pp. 129-132, 140-147)
3.3.2 Fluctuating union membership in the interwar period Although the Netherlands remained neutral in the first World War, the Dutch economy suffered a great blow. This was mainly due to the heavy dependence on international trade, that was severely disrupted because of the war. Exports as well as imports were greatly reduced, causing a rise in unemployment and in many prices. Especially the growing unemployment caused the trade unions a lot of difficulties, as their insurance funds were much too small to satisfy all the demands made on them by members who had lost their jobs. Because state insurance schemes did not exist in those days, unemployed non-members had even more problems. The government however, quickly recognized the very serious state of affairs and established the Royal National Emergency Relief Commission (Koninklijk Nationaal Steuncomité), in which the unions were asked to participate. The task of this Commission was to administer unemployment relief, rationing of scarce commodities,
39
rent and price controls; as this was done on the national as well as the local level, it contributed greatly to the social integration of the unions (see Windmuller, pp. 4850). A further recognition of the union movement could be found in the government financial assistance to the union unemployment funds, which were at first replenished by municipal subsidies. From 1917 a more structural decree was passed, in which the government committed herself to subsidize the unemployment funds with an amount that equalled the deposits made by the union members. This became known as the 'Ghent system', and was retained until World War II. Although half of the contributions came from the state, the unions were allowed to administer the funds by themselves; an important effect of this was, that it induced more people to join a union because they received benefit payments from their union rather than from private or public charity. In Britain, this had worked out the same way (Windmuller, p. 50; H & R, p. 125; Hueting et al., pp. 58-59). After 1915, part of the economy revived, bringing about a growing demand for labour, while the increased profits created possibilities to claim higher wages. Trade unions were ever more able to reach labour agreements, without the need to strike for it first. This implied that unions were increasingly recognized officially at the bargaining table. The number of collective contracts, which had only been 87 (covering 23,000 workers) in 1911, rose quickly to 932 (118,000 workers) in 1918. All this strengthened the position of unions, which in turn attracted new members (Hueting et al., pp. 63-64; Van Zanden & Griffiths, pp. 88-89; Windmuller, p. 51). The economic growth was not equally distributed; unemployment remained high in several industries. Moreover, the people had to cope with high prices and rationing. This gave rise to increasing social unrest, leading to membership growth of all union federations. Despite the growing discontent with government policies, the first national elections with universal suffrage (for men) in 1917 did not bring an eclatant victory for the SDAP, although 22 seats out of the 100 meant quite an improvement. With the end of the war, several Central European governments were overthrown by socialistic revolutions. Fired with these successes, the NVV and SDAP demanded social reforms, and even the confessional unions asked for some changes in favour of the working class, although they took a firm stand against any revolutionary undertaking19). The government had to give in partially, to soothe the masses. Several measures were taken, such as women's right to vote (1919), an eight-hour day and a 45-hour week (1919), and improved provisions for industrial accidents, disability, and old age. Furthermore, the Ministry of Social Affairs became established in 1919, the year in which also the High Council of Labour (Hoge Raad van Arbeid) was installed. This council would come to serve as an important advisory board on social policy, and was made up by representatives of the three major trade union federations, the four In Parliament, the leader of the SDAP, Troelstra, had proclaimed a socialistic revolution on November 11, 1918. This turned out to be a huge mistake, because not only the CatholicProtestant coalition was against an overthrow of the capitalistic system, but the majority of the socialist party as well. To support the government, the confessional unions organized massive pro-Orangist demonstrations, showing that a large part of the labour force was against socialism (Windmuller, p. 57; H & R, p. 130).
19)
40
principal employers associations, independent experts, and top-ranking civil servants (Windmuller, pp. 56-58; Hueting et al., pp. 66-68, H & R, pp. 126-134). While a large part of Europe had to deal with social upheavals and economic chaos, the Dutch economy experienced a boom period with fast rising employment until the end of 1920. Together with the reforms in favour of the working class, this led to a huge rise in union membership of all pillars between 1918 and 1920. The NVV remained the largest by far, with 259,700 members at the start of 1920; but the CNV and the Roman Catholic Bureau had relatively grown even quicker, both more than doubling their membership in only two years. Even the NAS revived temporarily. By 1920, the union movement had already been centralized to a large degree: less than one fifth of all members was not affiliated to one of the federations, against two third in 1910. (Van Zanden & Griffiths, hereafter referred to as Z & G, pp. 88-89, 109; Windmuller, p. 58). In the course of 1920 a world economic depression set in that would last until approximately 1924 (Z & G, pp. 109-112). While all prices dropped, nominal wages tended to remain stable, causing profits to decline rapidly, which in turn led to fast rising unemployment (from 4.7% of the dependent labour force in 1920 to 8.9% in 1923; deducted from Den Bakker & Van Sorge, p. 229, 231). In this situation, the employers used their power to reinstate the 48-hour week in 1922; also, numerous attempts were taken to cut back wages by up to 10 or even 15%. The union movement tried very hard to prevent these deteriorations by organizing a large number of long lasting strikes; while the annual average number of working days lost due to industrial conflicts had been 490,000 between 1913 and 1918, this average rose to 1,715,000 between 1919 and 1924, with a maximum of 3,156,000 days in 1924. This number would never be reached again (Windmuller, pp. 59-60; Hueting et al., pp. 70-74; H & R, pp. 140-142, 149-152). Despite these union efforts, most strikes were lost. Usually, the confessional workers were the first to give in (H & R, pp. 141-142, 149-152; Windmuller, pp. 58-59). Many members resigned, especially from the leftist unions; compared to their top at the end of 1919, the NAS declined by almost 75% (38,700 members) between 1920 and 1923, while the NVV lost 30% (76,400 members). The Bureau and the CNV only slowly started losing members from 1921, but their decline developed further and would not reverse before 1926 (see table 3.1). This pattern of union growth and decline during and right after World War I, could be found in most OECDcountries: increasing employment and inflation caused a rise, while the depression led to a fall in membership (Neumann et al., p. 4). The main cause of the Dutch decreasing membership rates was the worsening of the working conditions, but there was an additional reason. During the 'revolutionary' years 1918 and 1919 many workers had joined the (socialist) union hoping that this would contribute to lasting improvements for the working class. When this did not materialize, members of the progressive unions were the first to resign. The confessional unions did not de-cline so soon, which could be ascribed to the fact that their new members had joined partly to support the pro-Orangist, anti-socialist movement. Afterwards, the drop in Christian membership lasted longer, which might be related to the lack of fighting spirit of the confessional unions, which could have disappointed many workers (Boomgaard, pp. 363-364; H & R, p. 152).
41
Still, the size of the Christian union movement in the Netherlands was much larger than in most other countries, indicating the great influence of the Dutch churches. At the beginning of 1922, the NVV had almost exactly as many members as both confessional federations together, namely 224,000. In Germany at that time, this ratio was 8 to 1 million members in favour of the socialist federation. Co-operation between the different denominations was very difficult, although it would have strengthened their defence against the employers. However, a merger between the NVV and a Christian federation was unthinkable, and in 1921 the NVV made an unsuccessful attempt to join forces with the NAS and the liberal ANV (see note 14)(H & R, pp. 152-154; Hueting et al., pp. 77-78). When the American economy was picking up after 1923, and Germany had solved its hyperinflation, Dutch trade and industry revived as well; scholars often speak of the 'golden years' between 1923/24 and 1929/30. Real national income increased with 5% annually, and especially the industrial sector bloomed (Z & G, p. 115 f). In this period, the trade unions were able to improve the terms of employment much more gradually. Union membership increased again, but not at a very spectacular pace. Only in 1930 the NVV surpassed its former maximum of 1919, while the union movement as a whole surpassed it in 193120). But this time it was a stable growth, not instigated by political sentiments (Lammers, pp. 46-47). The Roman Catholic union movement managed to fortify in 1925 by merging the estate organizations with the trade unions into one federation, the Catholic Workers Federation of the Netherlands (RKWV). Within the RKWV, the 5 diocesan leagues and the 25 national unions kept their own governing bodies, each with an equal position, but gradually the unions gained ascendancy (Windmuller, pp. 62-63). For the Netherlands, the consequences of the economic crash of 1929 became really apparent after 1930. All economic activities deteriorated until the trough was reached in 1936, and then the government finally abandoned the gold standard. Together with the international upturn, this devaluation immediately promoted Dutch trade again. While 1938 was another year of decline, 1939 showed a strong revival (Z & G, p. 129 f). The depression of the 1930's hit the Dutch population very hard. In 1930, 173,000 people had been jobless, and this number would rise to 648,000 in 1936. In percentages of the dependent labour force, this came down to an increase from 7.0 to 24.2%. Moreover, statistics of early 1937 show that half of all registered unemployed were jobless for over two years. In 1939, still 426,000 workers were unemployed.21) (Goudriaan, pp. 21-22). In the 1930's, a compulsory unemployment insurance did not exist yet, so that very many families were reduced to poverty. The union members who had voluntarily insured themselves through their unions, did receive unemployment benefits, but, The high union density rate of 1919, 38.6%, would first be bettered after World War II; in the 1930's the maximum was 37.1%, in 1932. See table 3.2 on page 50.
20)
Unemployment figures before World War II are estimates, and the several existing statistics differ significantly from each other. Here the most recent estimates are used, derived from Den Bakker & Van Sorge, pp. 229, 231, 238.
21)
42
depending on their occupational group, only for 6 or 13 weeks. After that, they too had to fell back on a local relief scheme, consisting of a basic amount per adult and per child up to a maximum of 65% of the former wage. However, nonmarried (adult) children still living at home were excluded, just like unemployed women; unemployed union members received weekly one guilder more than nonmembers. (Levenbach, p. 550; Goudriaan, pp. 26-28; H & R, pp. 172-173). Contrary to the large unemployment spell of the early 1920's, the Great Depression did not lead to a real exodus of union members. In fact, membership figures kept rising until 1932/33, after which this trend was reversed. The NVV lost 16% (55,000) of its members, the RKWV 13.6% (27,000) and the CNV only 6.8% (8,000). When the economy improved in 1937, all unions (except the NAS) increased again. Membership rates rose until the second World War put an abrupt end to that. At the beginning of 1940, the trade union movement counted 809,000 members, which implied a density rate of 32.5%. The NVV with 322,300 members was still the largest federation by far, followed by the RKWV (191,700 members) and the CNV (120,300 members). Membership developments in other countries were very diverse; it would seem that unionization was mostly influenced by the business cycle on the one hand, but just as much by political developments (pro- or anti-union governments and laws) on the other (cf. Owen Smith). In explaining why the union movement did not lose so much ground during the depression, commentators point at two things. First, the unemployment insurances assured the unions of many members. As these benefits only lasted 6 or 13 weeks, there was always the risk of postponed loss of members. To prevent this from happening, the NVV founded so-called 'unemployment unions' in 1933, especially for members who could not draw unemployment benefit anymore22). Second, it is argued that gradually union membership had become an accepted phenomenon, and for many workers even a 'moral duty'23), out of solidarity with one's own class. The workers owed much to the union movement, which had greatly contributed to their welfare. Unions had become a strong countervailing power against the employers, and had also given the initial impetus to many social services24). Not The members of the unemployment unions didn't have to pay any dues. To be admitted, they had to declare that they would become members of their appropriate union after finding employment (Windmuller, p. 205 of the original English version). These unions turned out to be rather successful, as their number increased to 98 with 10,528 members in 1938 (Hueting et al., p. 84, 111).
22)
For a few workers it was an actual duty: closed shops hardly existed outside the printing industry, where it had become a usance since 1913 (Van der Wal, pp. 228-229). The highest number of workers covered by a collective agreement for whom membership was mandatory, came to 58,000 in 1922, only 3% of the dependent labour force. This would decline to just 23,000 (0.9%) in 1940 (CBS, Collectieve arbeidsovereenkomsten in Nederland op 1 juni 1941, pp. 10-11).
23)
To improve social legislation, the three federations maintained relations with their allied political parties. The closest bonds existed between NVV and SDAP, in which the former really had a say in matters concerning the labour market; NVV leaders held one fifth of the SDAP seats in parliament. The RKWV also consulted permanently with the Roman Catholic State Party, in which a few prominent union leaders took a seat. The CNV had no formal ties with its two
24)
43
only the workers' material well-being had improved, but also their spiritual development, by means of all kinds of educational and recreational facilities set up by unions. Moreover, according to Windmuller, deep-seated religious commitments had made the formation possible of viable Catholic and Protestant unions, which 'probably meant a net addition to the ranks of the organized' (Windmuller, pp. 77-81; Lammers, pp. 56-61; Boomgaard, pp. 365, 371-372).
3.3.3 Development of the institutional framework With the ascent of trade unions, the employers felt the need to organize as well, to be able to offer resistance against the growing power of the labour force. In 1899 the (non-religious) Association of Dutch Employers (VNW) became established. The immediate cause was the intention of the government to introduce social legislation, more specifically a law on industrial accidents (enacted in 1901), which would oblige the employers to compensate injured workers for their loss of income. During the first two decades, the VNW revealed itself as a public spokesman for employers on a very wide range of issues; in 1920, it founded a separate agency for labour issues, the Central Board for Consultation in Labour Relations (Centraal Overleg in Arbeidszaken). The Board would become an important institution of its own, thereby inducing a dual structure that would last until 1968 (Andriessen et al., pp. 7374; Windmuller, pp. 52-53). In view of the arisen pacification, the emergence of separate employers' associations of Protestant and Catholic signature was bound to happen. Already in 1892, the Protestants had established 'Boaz', consisting of industrial employers, small businessmen and farmers. Because of this heterogeneous composition Boaz lacked effectiveness, which led in 1918 to the foundation of the Federation of Protestant Employers (VPCW), but it would remain much smaller than its neutral counterpart. The federation of Catholic employers was also less influential, and at the same time hampered by the same dual structure of estate organizations and genuine trade associations that characterized the Catholic workers' organizations. Due to their small power, both religious employers' associations followed the lead of the VNW. So, there was much less disunity among the employers than among the trade unions of different persuasions, which was also expressed in the initiation of a co-ordinating council (Kring van Werkgeverscentralen) in 1921 (Windmuller, pp. 54-55; Andriessen et al., p. 75). Where individual employers used to have difficulties to withstand the union demands, their organizations could take joint action against trade unions. This especially yielded results during the slump of the early 1920's. Another effect of the emergence of the employers' associations was the initial fast growth of the number of collective labour agreements, and the number of workers involved. This rise continued until 1920, after which it slowed down until about 1926. Thereafter a revival occurred, which was reversed again after 1931; finally, from 1937 a new growth set in. This is shown in table 3.3. In view of the business cycle, this implied related political parties, but there was frequent consultation (Lammers. pp. 60, 196-200; Hueting et al., pp. 106-109; Windmuller, pp. 62-66).
44
that during an economic upturn many collective contracts were concluded. In such a period, often characterized by rising prices, employers were very much more willing to fix the wages for a longer time, since the principle of pay indexing did not exist yet. For the same reason, during a recession the employers were not eager to negotiate collective agreements, as wages tended to fall (Hueting et al., p. 91; Boomgaard, pp. 346-347). The huge expansion of collective contracts between 1913 and 1920 caused problems in supervising the application of the agreements. This induced the government to effect a statutory regulation, the Law on Collective Bargaining Agreements, which came first into force in 1927, after very long negotiations. The government had proposed to extend a contract to an entire industry if the employees of firms subscribing to the agreement made up a substantial portion of the total workforce in that sector. In that way, not only wage competition among employers could be avoided, but also the risk that organized workers would be replaced by cheaper labour. However, most employers opposed extension to a whole sector because they were against too much government interference. At the insistence of the union movement, only the clause which obliged an employer to apply the collective agreement to all his personnel, was kept in the law of 1927. Unions realized that this could also work against them: as the results of their bargaining efforts would automatically apply to all workers in a firm, the danger of free riding would emerge (Windmuller, pp. 74-75). The depression made the employers change their mind about state intervention. A new law on extending collective labour agreements to a whole branch of industry could provide two kinds of benefits: it would prevent unfair wage competition and it could lead to increasing purchasing power, which could stimulate the economy again. Hence, in 1937 the Act on the Extension or Nullification of Provisions of Collective Labour Agreements was passed. As the name already indicates, the government had the right to invalidate (part of) a collective labour contract if its effects were thought to be conflicting with the public interest (Windmuller, pp. 75-77). Although the total number of collective labour agreements rather fluctuated in the interwar period, the number of firms subscribing to these agreements rose almost uninterruptedly from 17,000 to 70,000 between 1928 and 1940. This indicated that the collective labour agreements obtained a more national character, probably owing to the laws of 1927 and 1937. At the same time, ever more nonpecuniary arrangements were incorporated in these contracts, especially after the 1937 law on extension. This concerned matters such as the apprentice system, arbitration, the prohibition of strikes as well as lock-outs, and sometimes personnel representation within the firm (Hueting et al., pp. 91-92). The encouragement of more staff participation (medezeggenschap) in the firm's management was one of the most important goals of the unions, after the achievement of political democracy through universal suffrage. Most employers were against any co-determination of employees in their own company. Some works councils became established, but they only had very limited consultative rights. Therefore, the union movement put all emphasis on reaching worker participation at the industry level. Except for the liberals, who saw a threat to the employers' right to manage, all political and labour organizations favoured the establishment of industrial boards, in
45
which employers and employees would jointly decide over industry policies and terms of employment. However, when this scheme was finally passed as the Law on Industry Boards (Bedrijfsradenwet) in 1933, its terms were mitigated due to a lot of compromises. All the law provided was the right - not the duty - to form joint boards for organized consultation on a broad range of socio-economic issues. Until the second World War the new system hardly got off the ground, as only 21 boards came into existence, none of them being really successful (Windmuller, pp. 70-73; Hueting et al., pp. 92-93, 108). At the national level, there was more joint consultation between employers and trade unions, namely in the High Council of Labour. Windmuller (pp. 64-66) looks on the existence of this board as a valuable contribution to a better climate of industrial relations, lessening some of the existing social tensions. Its significance can further be found in the field of social legislation, where about 80% of the 104 official recommendations of the Council were followed in whole or in part by the government between 1919 and 1939. Gradually at first, and then precipitated by the Great Depression, the strive after more co-operation between workers, employers and government became ever stronger, because this was considered to serve the interests of all parties best. The increasing state involvement through labour and social laws, the first attempts to create joint consultative boards at industry level, and the contacts between employers' associations and trade union federations at the national level, were all manifestations of this Dutch system of labour relations in the process of formation. After World War II, this course would be continued (Windmuller, pp. 81-84). To round off this section on pre-war unionization in the Netherlands, the union membership developments will be summarized in the demand and supply framework. Just as in the 19th century, the supply of unions was not obstructed by illdisposed laws or political parties, which was rather different in most other western countries (cf. Eickhof; Owen Smith). On the contrary, owing to the deep-rooted pillarization most federations maintained close relations with political parties of the same signature as theirs, through which they improved the social and legal position of the working class. The rise of the modern trade unionism, embodied in the social-democratic ANDB and the NVV, gave the impulse to the emergence of employers' associations on the one hand, and to the foundation of Christian union federations on the other. At first the employers offered much resistance to union demands, especially in periods of an economic downturn; this would lead to lost strikes and great losses of union members, who primarily reacted to the ups and downs of the business cycle during the first decades of the century. Gradually however, employers came to recognize the unions as bargaining partners, which was facilitated by their joint session in the High Council of Labour, and by the hardships of the Great Depression. The actual threat of Christian workers joining a socialistic union, made that the Protestant as well as the Roman Catholic Church instigated the formation of their own trade union federations. According to Windmuller, the existence of these religious union centres added to the size of the total union movement. This opinion might be questioned, because the fractionalization forestalled unity in the labour
46
movement, which weakened its position. After all, the Christian unions were much more complaisant with the employers, less willing to strike, and more easily gave in when they did go on strike. This attitude probably did not contribute to their recruitment power. On the other hand, all unions clearly provided for the wants of the workers, by offering all kinds of services. Before World War II, many social insurances could only be obtained through the membership of a trade union. Especially the Ghent system of unemployment funds, and the formation of special unemployment unions during the Great Depression, prevented that workers en masse renounced their membership. The unions' dedication to more and better labour and social legislation, and their supply of many facilities greatly contributed to the emancipation of the labour force. In a way, it can be concluded that in the period 1900-1940 the supply of union services linked up with the demand. Yet, in spite of this good match the union density rate almost never surpassed the 30%, but this was also the case in most other OECD-countries (cf. Neumann et al.): the demand for unionism in the pre-war period was everywhere more or less hindered by the downs of the business cycle, resistance from employers and governments, and, at least in the Netherlands, by the pillarization and the consequent mental attitude.
47
Table 3.1 Members of trade union federations 1899-1939 (end-of-year data, in thousands) NAS
NVV
RKWV*
CNV**
ANV***
VVH***
NSV
CMHA
CRP
Other
Total
1899
12,4
1900
8,9
0,7
1901
10,5
0,9
1902
7,9
1,4
1903
6,0
3,0
1904
5,0
1,8
1905
3,3
19,0
2,3
1906
3,7
26,2
2,5
96,6
129,0
1907
3,4
32,3
1,1
103,2
140,0
1908
3,7
36,7
0,8
103,8
145,0
1909
3,5
40,7
11,7
6,6
122,6
185,1
1910
5,2
44,4
15,5
7,5
123,9
196,5
1911
6,2
52,2
16,4
7,8
130,2
212,8
1912
8,1
61,4
21,1
7,9
2,8
132,5
233,8
1913
9,7
86,1
29,4
11,1
3,9
128,1
268,3
1914
9,2
89,3
35,3
12,4
4,7
124,4
275,3
1915
10,5
103,2
42,1
15,2
5,0
128,8
304,8
1916
14,4
133,8
58,8
20,9
5,6
4,0
124,7
362,2
1917
23,7
166,3
73,7
28,8
7,8
4,7
129,2
434,2
1918
34,0
200,0
98,2
47,9
10,6
6,8
135,8
533,3
1919
52,6
259,7
151,0
70,4
40,3
9,9
124,3
710,7
2,5
48
1920
37,4
224,7
158,6
76,3
52,7
2,7
11,6
111,9
674,9
1921
31,6
223,8
150,9
73,4
50,1
3,7
20,8
125,1
658,6
1922
23,7
201,1
122,8
62,6
47,8
3,9
8,6
123,8
585,7
1923
13,9
183,3
101,2
54,0
32,5
7,8
3,7
7,9
129,0
525,4
1924
14,0
187,0
94,4
50,7
24,0
16,6
7,4
3,5
6,4
99,9
503,9
1925
13,8
193,2
93,0
49,0
13,3
16,3
6,2
3,2
6,9
105,3
500,2
1926
14,0
199,1
99,6
52,0
11,7
15,7
5,0
3,2
7,2
104,0
511,5
1927
14,5
205,9
104,8
53,6
12,4
15,7
4,8
3,4
7,2
105,7
528,0
1928
16,3
220,8
114,9
58,6
12,7
15,9
2,1
3,5
7,6
110,9
563,3
1929
17,5
255,6
136,6
72,9
36,7
2,8
3,7
7,7
104,1
637,6
1930
17,5
276,2
152,9
82,1
40,8
2,3
3,8
8,0
107,7
691,3
1931
20,7
320,3
183,1
103,3
46,1
2,7
3,9
8,4
106,6
795,1
1932
23,0
340,1
198,1
116,9
49,3
2,8
3,8
8,2
99,6
841,8
1933
19,6
324,9
194,6
116,8
47,2
2,4
3,8
7,7
88,6
805,6
1934
13,0
300,3
182,7
114,0
46,7
2,0
3,7
7,7
88,9
759,0
1935
12,0
287,4
175,9
109,3
44,7
2,0
3,7
8,0
90,6
733,6
1936
11,4
285,3
171,0
109,1
44,7
1,9
3,6
8,4
94,2
729,6
1937
11,2
296,0
173,3
111,5
45,6
2,0
3,6
9,0
97,9
750,1
1938
10,8
309,0
182,0
115,1
46,4
1,6
3,6
9,7
100,4
778,6
1939
10,5
322,3
191,7
120,3
50,9
1,6
3,6
10,6
97,5
809,0
49
Elucidation to table 3.1 *including its predecessor, the Bureau for Roman Catholic Trade Unions. ** including its predecessor, the Christian Labour Secretariat (CAS). *** after a merger in 1929 the name changed into Netherlands Trade Union Centre (NVC)
Source:a) 1899-1912: Harmsen & Reinalda (1975), CNV 1900-1908 in Kruithof (1981), and Total 1906-1912 in CBS, Omvang der vakbeweging. b) 1913-1939: Visser (1989b).
List of abbreviations: NAS: NVV: RKWV: CNV: ANV: VVH: NSV: CMHA: CRP:
Nationaal Arbeids-Secretariaat Nederlands Verbond van Vakverenigingen Rooms-Katholiek Werkliedenverbond in Nederland Christelijk Nationaal Vakverbond Algemeen Nederlands Vakverbond Verbond van Vakorganisaties van Hoofdarbeiders Nederlands Syndicalistisch Vakverbond Centrale van Middelbare en Hogere Ambtenaren Centrale van Rijkspersoneel
Table 3.2 Union density rate* 1913-1940 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926
16.9 17.0 18.5 21.4 25.1 30.0 38.6 35.8 33.9 29.6 26.9 25.4 24.8 24.8
1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
25.1 26.0 28.5 30.1 34.3 37.1 35.0 32.4 31.0 29.9 30.4 31.1 32.5 29.4
The union density rate is defined as: total union membership divided by the number of full-time workers (> 20 hours) in the dependent labour force and those looking for such a job (x 100) *
Source: J. Visser (1989a), European trade unions in figures
50
Table 3.3 Number of collective labour agreements & workers involved agreements
workers
1904
1
-
1905
1
-
1906
3
-
1907
13
-
1908
19
-
1909
23
-
1910
46
-
1911
87
23002
1912
119
-
1913
178
-
1914
474
-
1915
614
-
1916
586
-
1917
827
83119
1918
932
118371
1919
1031
-
1920
974
273598
1921
959
-
1922
728
257592
1923
677
238145
1924
830
285607
1925
764
267016
1926
810
264095
1927
894
267791
1928
1016
279595
1929
1254
291831
1930
1554
385900
1931
1509
359300
1932
1351
252900
1933
1237
244200
1934
1165
259500
1935
1259
211650
1936
1239
253450
1937
1245
283260
1938
1742
293300
1939
1667
323350
1940 1552 351800 Source: CBS (several editions), Maandschrift; Statistiek der lonen; Regelingen van lonen en andere arbeidsvoorwaarden in Nederland.
51
3.4 The interlude of the second World War The growth of the trade union movement was interrupted by the German occupation of the Netherlands from May 1940. At first, the executive committees of the confessional federations were allowed to remain in function, while the NVV was seized and its executive board replaced by a Dutch national socialist. The leaders of the NVV unions were replaced as well. Despite this take-over at the top, the NVV decided to continue operating, for which it was later heavily criticized. The two revolutionary federations, NAS and NSV, were dissolved, while the liberal NCV associated with the NVV (Lammers, pp. 61-63). The NVV had several reasons to remain in being. Some leaders thought that the war would not last very long, and that it was better to keep the union apparatus up, because a renewed establishment after the liberation would be difficult and time consuming. Also, the NVV had 800 paid functionaries, who would otherwise become unemployed. The main motivation however, was the protection of the workers' interests in a time of rising prices and unemployment; discontinuation would have meant that their members would become unorganized, or become members of a confessional or national-socialistic governed union. Indeed, thanks to union efforts the terms of employment did improve, especially in agriculture, with the result that most members remained loyal to their unions25) (Windmuller, pp. 89-90; H & R, pp. 205-207, 217-218; Hueting et al., pp. 125-126). From July 1941, the Germans started with the nazification of Dutch institutions. This time the RKWV and the CNV were also placed under direct supervision of the national socialists, but unlike the NVV, the confessional union leaders and their staff resigned, followed by about 90% of the Catholic and 75% of the Protestant members. The Roman Catholic Church forbade the membership of any national socialist organization, which in her view included the NVV, but some ten thousands CNV members joined the NVV out of free will or were simply transferred by NVV agents, put in charge by the Nazis. This would only widen the gap between the NVV and the confessional unions. Thereupon, the employers associations dismantled the newly founded Governing Board for Labour Affairs (Raad van Bestuur in Arbeidszaken), in which they used to negotiate with the three union federations, within the limits set by the occupants. As the captive NVV was not regarded sufficiently representative of the labour force, the employers withdrew, showing their solidarity with the suppressed unions (Windmuller, pp. 91-93, 435 (note 15); H & R, pp. 219-220). The NVV finally liquidated on May 1, 1942, when the Germans wanted to transform the remaining union movement into a Dutch Labour Front (NAF), to become the only functioning federation, according to national-socialistic principles. Although most NVV members resigned, approximately 100,000 members joined, notably rural workers. However, the NAF did not succeed in exerting a real influence, as it could not acquire sufficient support (despite intimidation tactics) nor authority among the employers Windmuller, pp. 94-95; Lammers, pp. 64-66). Another institution created by order of the Germans, was the Board of Government Mediators (College van Rijksbemiddelaars), consisting of four Dutchmen who had In 1940, the NVV even increased by 3.5% to 333,700 members, thanks to the affiliation of the NCV. The RKWV and the CNV declined only by 7.4% and 5.3%, respectively (Visser (1989a), p. 147).
25)
52
already been mediators in labour disputes before the war. Their new task was to extend collective labour agreements - which had hardly occurred since the Act of 1937 -, and to issue binding wage orders when both negotiations and mediation had failed. This Board existed between May 1940 and October 1942, and developed several principles which after the liberation would be incorporated in the new wage policy; among those were the institution of a minimum wage and the imposition of compulsory wage orders in a variety of circumstances (Windmuller, pp. 87-88; Hueting et al., p. 126). When it became clear that Germany was not invincible, the unions and employers were encouraged to consult each other secretly on the subject of post-war labour relations. The NVV approached the two confessional federations several times on the matter of joining forces in one unified movement. To that end, the political ties with the socialistic SDAP were severed. The three did not reach an agreement, partly because of the alleged collaboration of the NVV with the Nazis, but most of all because the CNV and the RKWV held on to their religious conviction. Still, in May 1943 it was jointly decided to found a Council of Trade Union Federations (Raad van Vakcentralen), in which the participating federations would remain autonomous. The purpose was interorganizational co-operation at all levels of decision making. To that aim, it was essential that the organizational structure of the federations would become identical. So, it was resolved that after the war workers would be organized according to industrial sector and company, irrespective of their profession26). To eliminate unfair competition between similar unions of different ideology, it was decided that the rights and duties of the members would be the same in all three pillars.27) In this way, the choice which union to join could be taken purely on principle (Lammers, pp. 71-72, 74-75; Windmuller, pp. 96-97; H & R, pp. 228230). When the South of the country was liberated in the autumn of 1944, in this area many union leaders and active members made a new attempt to form labour unity, disappointed with the role of the NVV during the war. They grasped the opportunity while the old leaders were still in occupied territory, and appealed to the desire for an undivided union movement, which would be much more powerful than a segmented one. They called themselves the Trade Union Unity Movement (EVB) and at first attracted all kinds of followers, from left-wing to Catholics. Driving force behind the EVB was the Dutch Communist Party (CPN), which nevertheless refrained from imposing its will on the EVB, so as to keep the movement open to all ideologies (H & R, pp. 221-227; Windmuller, pp. 99-101). In clandestine meetings, representatives of the three major federations and of the employers' associations also discussed on future co-operation. In 1943, this led to the creation of the Foundation of Labour (Stichting van de Arbeid), in which all interest groups - industrial and agricultural employers, small businessmen and trade unions Originally, unions protected the interests of one occupational group. With the emergence of many large firms in which a lot of different professions were practised, it would be more efficient to organize workers according to industrial sector (Lammers, p. 74).
26)
The associated unions would all charge the same subscription fee, an average of 2% of the member's income; only a few unions deviated from this scheme. Furthermore, all unions would set up uniform funds for funeral benefits, and marriage allowances for female members (Lammers, p. 84).
27)
53
would consult each other. The readiness to co-operate already stemmed from the depression years, but was greatly encouraged during the war by the joint resistance of employers and workers, and by the complete devastation of key sectors of the economy. In exchange for this enhanced influence on the national level, the unions had to pay a high price: for the time being, they renounced the claim of worker participation on the plant level. The developments of 1943-1944 bring Harmsen & Reinalda to the following conclusion: while the EVB was striving after unity at the basis, continuing the class struggle, the old federations were working at the top on maintaining the segregation, meanwhile aiming at co-operation between capital and labour (H & R, pp. 230-231; Windmuller, pp. 87, 97-99). 3.5 Trade unions in the welfare state, 1945-1990 With respect to developments within the trade union movement, the postwar era can be divided into three subperiods, even though the borderlines are not always very clear. Moreover, depending on which approach is taken, scholars choose slightly different demarcations, or prefer to distinguish more subperiods. In this section the following tripartition is made: a) 1945-1959, a period characterized by very centralized decision-making, in which the power was shared between government, employers associations and the union federations; union membership rose almost continuously, with a temporary stagnation towards the end of the 1950's. b) 1959-1975, a period in which the decision process gradually shifted to industry level, while the unions took the greater part of the power, thanks to favourable economic conditions; apart from some minor backslidings, there was a steady rise in union membership. c) 1975-1990, a period in which the unions lost their superior power again, as result of the longest lasting recession since the war; for the first time, the size of the union movement decreased substantially until 1986, after which it regained part of its strength.
3.5.1 1945-1959: centralization and co-operation In order to rebuild the badly damaged Dutch economy, the government, labour movement and employers associations were prepared to work closely together. From the point of view of the unions, this readiness was greatly facilitated by the fact that for the very first time the Labour Party PVDA (successor of the SDAP 28)) took part in the government. The most effective way to realize this reconstruction, was to centralize the decision-making in social-economic policy, which was expressed in the 'guided wage policy', formulated in the Extraordinary Decree on Labour Relations (BBA) of October 1945.29)
The PVDA was still a leftist party, but had shaken off her socialistic image. By turning into a people's party, she hoped to gain a greater part of the electorate, notably tradespeople, farmers and intellectuals.
28)
The rather complex BBA will only briefly be dealt with here; more information can be found in Windmuller, pp. 121-131.
29)
54
Apart from regulating such matters as hours of work, dismissals and voluntary severance, the BBA had its foremost impact on the wage policy. It stipulated that the reinstalled Board of Government Mediators, consisting of independent, authoritative persons, had the final say in the determination of most wages and other terms of employment in the private sector. After obligatory consultation with the Foundation of Labour, the Board would set the wage guidelines in accordance with the governments wishes. Next, the contracting parties could negotiate within these strict limits, after which the bargaining result had to be approved by the Board again. If no agreement was reached, the Board would lay down a wage order. By recognizing the Foundation of Labour, the government involved the top representatives of workers and employers in the central decision process. The three union federations and the employers' associations agreed to participate, because they pursued a common interest that was not served by mutual conflicts. In order to make the economy competitive again, price and wage increases had to be limited as much as possible; the unions voluntarily accepted these restraints until approximately 1958. Although the Foundation of Labour would remain influential in the area of wage regulations, in 1950 all its other advisory functions were transferred to the SocialEconomic Council (SER). This was a consequence of the Industrial Organization Act (PBO), the second legal measure that stressed the importance of centralization. The purpose of the PBO was to form supervisory boards of employers and workers in each industrial sector, and on top of this structure the SER would function as the official advisory council to the government. This council consisted of the three recognized federations, the employers' associations and a group of independent crown members. The SER would become an influential organ, but on the industry level the PBO boards became a failure, because most employers were opposed to formal arrangements and restrictions and prefered to make decisions on their own (Windmuller, pp. 136-142). At the plant level, the unions made no headway for some time; the Works Council Act (Wet op de Ondernemingsraden) of 1950 was meant to give the workers some participation in the management of individual firms, but it left the employers with all the power; in case of non-compliance with the law no sanctions were imposed (H & R, pp. 317-321). The result of this new legal framework was the complete dominance of the central organizations of employers and unions. Since only the recognized federations had a say in social-economic policies, it isolated all the other unions. According to Windmuller (p. 105), this hastened the decline of the fourth federation, the Unity Centre (EVC, successor of the EVB). By organizing economic as well as political oriented strikes, the EVC attracted a lot of disappointed workers, notably among urban and dock workers. Especially the NVV lost many members to the Unity Centre in the first post-war years. It was the only federation which protested against the wage and commodity controls. When the EVC did not achieve its goals, and became ever more dominated by the Communist Party, it suffered from a split off by syndicalists in 1948 and started to lose strength quickly. Another cause of this decline was the fact that both the Protestant and the Catholic Church forbade their congregation to be a member of the EVC (H & R, p. 255). As the EVC was suspected of publishing flattering membership figures for some time, it was excluded from the statistics since 1953. Its membership is believed to have been dwindling ever since, until the EVC ceased to exist in 1964 (Windmuller, pp. 103-106).
55
Hence, the hegemony of the existing three pillars returned soon after the war. Although now working together in the Council of Trade Union Federations, the Foundation of Labour and in the SER, pluralism remained the characteristic of the Dutch union movement. The Protestant CNV and the Roman Catholic federation, now called KAB (Catholic Labour Movement) already in 1946 reached their prewar levels again; moreover, for the first time the confessional federations together surpassed the size of the NVV, and would keep that position until 1972. This clearly shows that the confessional union movement remained deeply rooted. The NVV first had to shake off its blemished image from the war, and had to contend with competition from the EVC: in the summer of 1945, the latter had even more members than the former, 170,000 against 150,000 (H & R, p. 282). Between 1945 and 1947 the two leftist federations attempted to realize a merger, which failed (H & R, pp. 282-290). By subscribing to the principles of the Foundation of Labour, the NVV had committed itself to the acceptance of the capitalistic structure of the economy, and to maintenance of industrial peace. Moreover, by exchanging its socialistic principles for social-democratic principles, the NVV hoped to attract more workers, irrespective of religious or political persuasion. (H & R, pp. 248-250; Lammers, p. 97 f). The reconstruction of the economy passed off very well, thanks to the strong international recovery and the Marshall Aid, and internally due to a rigorous income policy, the guided wage policy, and a strict system of price controls. Already at the end of the 1940's the Dutch economic indicators had surpassed most of their prewar levels; only labour productivity did not reach this level until 1952. Thereafter, the Dutch economy experienced an almost uninterrupted period of growth until 1973, with the exception of the year 1958. (Z & G, pp. 184-204, 210; Windmuller, pp. 163164) Alongside this economic expansion, social security increased tremendously. Prompted by the miseries during the Great Depression, and incited by the efforts of especially the NVV, the Roman-leftist government enacted a series of important social insurances laws. During the 1950's, thanks to the solid economic foundations that were laid, a safety net was formed not only for the workers (Unemployment Act of 1952), but also for the non-working population (General Pensions Act of 1957, General Widow's and Orphans' Act of 1959). As the welfare increased, so did the social security in the 1960's; among others, this involved the General Child Allowance Act, the Social Benefits Act, several acts concerning health and disability benefits, and the Minimum Wage Act. (Hueting et al., pp. 279-286). The three federations supported the policy of wage restrictions, as this not only served the common interest, but also their own: they were now recognized as official advisors to the government, and the extension of collective agreements expanded enormously 30) through the BBA, therefore their position was greatly strengthened. Until 1953, the wages were allowed to rise just as much as the consumer prices in annual national pay rounds. Only in 1951, when the 'social partners' could not reach an agreement, the government unilaterally imposed a restriction on spending because the balance of payments worsened considerably in the second half of that year. The federations did not protest against the resulting loss in purchasing power. For one The percentage share of the dependent labour force covered by a collective agreement was 32% before World War II, while this increased to over 80% in the 1950's (Z & G, p. 95).
30)
56
thing, they prefered to remain loyal to the 'Roman-red coalition' (Z & G, pp. 95-97; Windmuller, pp. 172-180; Lammers, pp. 76-79). Between 1954 and 1959 the pay rounds became known as welfare rounds: the wage increases exceeded the risen costs of living, since the average increase in labour productivity was also taken into account. For the first time, the Dutch workers could share in the expanded welfare, except in the period 1957-58, when the government once again applied a retrenchment policy in order to stimulate exports and employment. Just like 1951, the union federations supported this measure (Windmuller, p. 183). Hence, between 1945 and 1958 the three major pillars went along with the strictly guided wage policy. Especially the NVV was very much in favour, because of its solidarity principle: equal proportional pay rises for everyone (Reynaerts & Nagelkerke, pp. 180-181). At first, the rank and file could accept the lagging behind of their real wages; after all, their sacrifice led to the expansion of production and exports, which in turn promoted employment growth. Their approval was reflected in the rapid increase of the union movement as a whole31), but most of all the three recognized federations gained new members (see table 3.4). However, when the Dutch economy began to boom, many union members refused to accept this retrenchment policy any longer and started putting pressure on their leaders. This contributed to the welfare rounds mentioned above. Union membership figures kept rising until the restriction on spending, imposed in 1957. The workers not only had to experience another cutback in their real income, but on top of that unemployment more than doubled. Their opposition to these developments found expression in a considerable loss of the PVDA during the provincial elections of 1958, while the NVV and the KAB lost a lot of members for the first time since the liberation. In two years time, their size decreased by 23 and 17 thousand members, respectively.32) The CNV did not decline, but its growth came to a complete standstill. (Windmuller, pp. 181-184; H & R, pp. 355-356)
The union density rate, which had been fluctuating around 30% in the second half of the 1930's, surpassed the 40% immediately after the war (if the EVC is included); see also table 3.5. It should be mentioned that the decline of the KAB was not caused by a resignation of individual members, but by the withdrawal of the affiliated Catholic union of civil servants (counting 20,000 members) in october 1957 (H & R, p. 357).
31)
32)
57
Table 3.4 Members of trade union federations 1945-1989 (end-of-year data, in thousands) NVV
NKV*
CNV
MHP**
AC***
Other****
Total
1945
237.7
182.8
94.0
4.1
6.8
270.4
795.8
1946
301.3
224.9
119.1
4.3
17.4
293.8
960.7
1947
331.5
251.5
131.6
5.1
18.4
308.6
1046.7
1948
368.2
269.6
147.6
5.4
19.9
296.5
1107.1
1949
381.6
296.4
156.0
5.6
20.9
298.5
1159.1
1950
405.9
311.4
166.2
5.6
22.1
303.3
1214.6
1951
418.8
321.5
174.8
5.9
22.2
150.3
1093.4
1952
433.7
334.7
182.0
6.2
23.6
150.6
1130.9
1953
454.1
347.3
191.1
6.7
24.6
156.6
1180.4
1954
463.5
361.0
199.7
7.3
26.2
164.3
1221.9
1955
467.3
381.7
206.3
9.1
27.5
168.0
1259.9
1956
500.3
412.0
216.0
9.2
27.9
171.1
1336.5
1957
486.2
395.0
218.5
11.0
27.0
195.0
1332.8
1958
476.9
395.9
218.4
11.6
29.8
202.6
1335.1
1959
486.7
400.4
219.0
12.0
29.6
207.0
1354.9
1960
507.0
411.8
223.8
12.5
30.1
214.5
1399.6
1961
507.2
417.8
224.9
12.9
30.4
226.0
1419.1
1962
512.2
418.5
227.5
12.9
30.7
234.2
1436.0
1963
528.6
418.8
229.8
13.0
33.3
246.8
1470.3
1964
527.2
407.7
229.3
13.1
33.1
248.1
1458.5
1965
535.7
412.0
234.5
13.7
34.6
251.1
1481.7
1966
556.3
425.3
240.7
19.5
35.2
254.9
1531.9
1967
563.2
428.4
240.7
19.7
35.5
252.0
1539.5
1968
560.6
409.4
240.0
26.2
36.1
277.4
1549.7
1969
564.8
400.2
238.2
27.3
36.7
278.7
1546.0
58
Table 3.4 continued NVV
NKV*
CNV
MHP**
AC***
Other
Total
1970
611.4
401.8
238.2
32.4
39.5
262.1
1585.4
1971
623.8
399.7
238.9
45.3
43.6
254.5
1605.8
1972
633.1
395.4
235.8
51.1
46.7
257.8
1619.8
1973
664.1
397.9
230.0
64.0
61.6
249.6
1667.3
1974
677.1
355.2
224.5
65.1
85.8
276.2
1683.9
1975
702.1
356.5
226.0
110.6
89.7
238.5
1723.3
1976
706.6
345.3
258.3
114.4
101.0
194.1
1719.6
1977
730.4
341.9
295.3
114.6
102.9
162.5
1747.7
1978
745.3
335.5
301.0
117.9
105.4
153.7
1758.8
1979
749.9
326.9
303.8
119.1
112.2
156.3
1768.2
1980
744.1
310.0
302.3
114.1
110.2
160.1
1740.8
1981
1026.7
346.0
114.1
112.5
103.5
1702.9
1982
998.8
335.2
113.3
108.1
108.0
1663.5
1983
946.2
316.9
111.5
108.6
112.0
1595.2
1984
918.5
302.8
109.8
102.0
117.9
1551.1
1985
903.3
298.3
107.3
100.7
121.8
1531.5
1986
900.8
293.8
108.3
100.5
123.6
1527.0
1987
911.7
291.6
110.9
101.0
129.7
1544.8
1988
931.4
291.1
121.3
101.6
134.6
1580.0
1989
967.1
298.8
123.6
107.0
139.4
1635.9
*Including its predecessor, the KAB. ** Including all its predecessors, among which the CMHA. *** AC stands for Ambtenaren-Centrale, and is the successor of the CRP. ****This includes the EVC, which had the following membership figures: 1945: 162,300; 1946: 169,400; 1947: 176,900; 1948: 163,400; 1949: 163,200 and 1950: 163,800.
59
Table 3.5 Union density rates according to four different definitions (1) 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989
41.1 40.2 40.6 41.7 43.0 38.2 38.9 39.9 40.2 40.6 42.2 41.4 41.2 41.4 41.8 41.7 41.0 41.2 40.0 39.9 40.7 40.4 40.2 39.5 40.8 41.1 41.5 42.6 42.8 43.3 42.6 43.0 42.7 42.3 40.9 36.9 35.3 33.6 32.4 31.6 31.1 31.1 31.3 31.9
(2) 41.1 40.1 40.6 41.7 43.0 38.2 38.9 39.9 40.2 40.6 42.2 41.4 41.2 41.4 41.8 41.7 41.0 41.2 40.0 39.9 40.7 40.4 40.2 39.5 38.1 38.0 38.0 38.6 38.4 38.8 38.1 38.2 37.9 37.0 35.4 33.6 32.0 30.4 29.2 28.4 27.8 26.2 26.1 26.7
(3) 40.4 39.4 39.8 40.8 42.0 37.1 37.7 38.5 38.7 38.9 40.4 39.4 39.0 39.1 39.4 39.1 38.4 38.5 37.3 37.0 37.6 37.4 37.1 36.2 37.3 37.3 37.6 38.6 38.6 38.9 38.0 38.3 38.0 37.5 36.2 32.2 30.6 28.8 27.5 26.6 30.7 30.4 30.6 31.3
(4) 40.4 39.4 39.8 40.8 42.0 37.1 37.7 38.5 38.7 38.9 40.4 39.4 39.0 39.1 39.4 39.1 38.4 38.5 37.3 37.0 37.6 37.4 37.1 36.2 34.9 34.6 34.5 35.0 34.6 34.9 33.9 34.1 33.7 32.8 31.3 29.3 27.7 26.0 24.7 23.9 27.5 25.6 25.5 26.2
1 [total membership/full-time dependent labour force* (> 20 hours)] x 100 2 [total membership/full-time and part-time dependent labour force] x 100 3 [active** membership/full-time dependent labour force (> 20 hours)] x 100 4 [active membership/full-time and part-time dependent labour force] x 100 * **
dependent labour force: wage earners and those looking for a job active membership: members who are not pensioners, conscripts or disabled.
60
Source: A) total and active membership: Ebbinghaus & Visser (1993), Trade union systems in Western Europe; data handbook vol. 1 (1945-1989). B) dependent labour force: Visser (1989a), European trade unions in figures (regarding 1946-1969) and Department of Social Affairs and Employment (1985 up to 1991), Rapportage Arbeidsmarkt (regarding 1970-1989).
Still, for the whole period 1945-1959, the union movement had grown very rapidly. The first post-war years witnessed an explosive growth with an annual average of 11.6% between 1945 and 1948. In the decade 1949-1959 this overall growth slowed down to about 1.6% annually, but pro rata the three pillars increased much more in these ten years: the NVV by on average 2.5% per year, the KAB by 3.1%, and the CNV even by 3.5%. In most other OECD-countries the same rising membership trend could be detected. The exceptions were France and Italy, where pluralism and internal conflicts obstructed growth, and the United States, where a hostile public opinion and anti-union legislation (since 1947) formed major obstacles (cf. Visser (1989a); Owen Smith). The fractionalization remained firmly founded despite close interorganizational cooperation. But the increasing contacts with non-Catholics were a thorn in the flesh of Dutch Roman-Catholic Church leaders, who issued an edict in 1954 stating that membership of any socialistic organization, including the NVV, was strictly forbidden for Catholics. Thereupon, the KAB immediately tried to remain on good terms with the NVV, but at the same time the leader of the CNV openly welcomed this edict. Next, the NVV reached the conclusion that co-operation on this basis was not possible anymore and withdrew from the Council of Trade Union Federations, which then ceased to exist. In practice, however, the three pillars kept working together informally, as well as in the Foundation of Labour and the SER. In 1958, the three resumed top-level consultations in the newly formed Consultative Board (Raad van Overleg) (H & R, pp. 328-333; Windmuller, pp. 112-115).
3.5.2 1959-1975: towards more decentralization and conflicts Windmuller regards 1959 as a turning point in the postwar industrial relations (pp. 120, 142-143). For the sake of solidarity with the weak sectors, especially the PVDA and the NVV had defended the guided wage policy for a long time. In the Spring of 1959, when a new government was formed without the PVDA taking part in it, the time was ripe to switch to a more free wage formation. Differentiation in the terms of employment had to be reached by linking the wage development in a certain industrial sector to the rise in productivity in that sector. By forbidding the wage increases to be passed on to the customer, the government kept pursuing an active price policy. While the negotiations shifted more towards the industrial level, the final review still remained centralized, at the Board of Government Mediators (Windmuller, pp. 184186; H & R, p. 359). The new system had a large problem with the assessment of productivity rises. But as long as the business cycle was very favourable the Board gave its consent to wage increases. Meanwhile, the labour market became extremely tight. Foreign labour was brought in, but at the same time many Dutchmen went working in Germany, where the wages were much more attractive (Windmuller, p. 445 (note 30); H & R, p. 367). The shortage of labour gave the unions a strong bargaining position, leading to a reduction of the working week from 48 to 45 hours in 1961. This immediately led to a rise in the labour costs, and a temporary decline in industrial production and ex-
61
ports, which incited the government to intervene once again in the wage formation for 1962 (Windmuller, pp. 186-192). So, the new wage policy was still dominated by central dirigisme. There was some differentiation, but the big contrast to the preceding period was that, one of the bargaining parties did not abide by the rules. Although the affiliated unions felt passed over, they accepted the decisions of their federations instead of listening to their members' demands. But individual employers complied ever less with the central guidelines. To secure themselves enough personnel, they were prepared to pay black wages far above the official level. As it was recognized that this new system fell short of the expectations, it was adapted in 1963. The responsibility for the wage formation was shifted to the Foundation of Labour, a first move towards decentralization. (Windmuller, pp. 142-150, 192-193) In 1963, the actual wage increase was far above the pay rise that was agreed upon in collective contracts. Apart from black wages, the labour contractors formed the largest problem because they delivered temporary personnel, who received much more pay then the workers with tenure. This generated bad blood, leading to ever more short-lived wildcat strikes. The trade union federations condemned this, but the affiliated unions realized that support of these dissatisfied workers was the only way to get hold of the situation and to retain their members. Forced by the circumstances, the negotiating parties agreed on a wage increase of 10% for 1964, to get the pressure off. This resulted in the well-known 'wage-explosion', which turned out to become 15% in reality. (H & R, pp. 367-369; Windmuller, pp. 193-195) Between 1963 and 1965 the Dutch wages were raised to the European level (H & R, p. 369). Especially in the period 1966-1967 the government again gave wage instructions, because of a temporary stagnating economy, along with a rising unemployment, foremost in the textiles industry. The economy quickly recovered in the second half of 1967, while the unemployment stabilized, but the times of full employment were definitively over. (Van Doorn et al., pp. 247-250; Windmuller, pp. 196203) At the end of 1967 the parties in the Foundation of Labour agreed that the dirigisme should be replaced by autonomous bargaining in each industrial sector. For one, the more relaxed labour market was giving the employers a bit more bargaining scope. Also, employers and unions both agreed that the gap between the strong sectors (metal, chemical industry) and the weaker ones (textiles, clothing and shoe industry) had become too large to justify the same wage increases. Still, the Wage Determination Act only passed in Parliament in April 1970 after long discussions and protests of the union movement, because the Minister of Social Affairs not only retained the power to issue a general wage control, but also the right to interfere with separate collective labour agreements. Despite protests against this last clause, the act was passed intact. Thereupon, the NVV and the NKV (Netherlands Catholic Trade Union Federation, successor of the KAB in 196433)) both decided to withdraw temporarily - from the SER and from the Foundation to express their anger. This was a clear indication of the growing polarization between a large part of the union movement on the one hand, and the government and employers on the other. (Van Doorn et al., pp. 251-258; Windmuller, pp. 224-226) With the foundation of the NKV, the Catholic labour movement finally got rid of the dual organizational structure, of unions and estate organizations. For the sake of effectiveness, the NKV now became a pure federation of trade unions, with the same organizational structure as the NVV (Windmuller, p. 116; H & R, p. 375).
33)
62
All in all, the 1960's had not been easy for the trade unions. As a whole, the movement grew in this decade by 10.5%, but the dependent labour force rose by 17%, implying that the union density rate declined a little (see table 3.5). Most union movements abroad performed better in this period, but in Australia, Switzerland and the U.S.A the union density rate went down more than in the Netherlands, for divergent reasons (Neumann et al.; Owen Smith). Of the three Dutch federations, the NVV achieved best with a rise of 11.4%, but the CNV only increased by 6.4%, while the NKV did not grow at all (see table 3.4). Already since the middle of the 1950's the NVV started worrying about the high membership turnover, which became a persistent phenomenon in the years to come (Van Doorn et al., p. 211; H & R, p. 359, 385). The drop in membership figures in 1957 had been another sign, that the (recognized) union movement was losing its attraction to a part of the workforce. Since that year, the so-called categorial unions were clearly making headway.34) Although much smaller in absolute numbers, the categorial unions experienced a strong growth. Between 1957 and 1964, the share of the categorial unions rose to 20% of the total membership; this is shown in table 3.6 below. Workers choosing for such a union could primarily be found among civil servants (teachers, police), the nursing profession, and the higher personnel. This development caused the recognized movement a lot of concern, judging from the number of reports and publications on this matter (see, among others, Top; Lulofs). It is not very difficult to explain the rise of these categorial unions. Ever since the end of the war, the recognized federations were emphasizing the promotion of collective interests; with the booming economy and the expansion of the welfare state, these common interests were already fulfilled in the course of the 1950's. Instead, the need for a more individual, or specific group approach, was increasing. The categorial unions, organized by profession or sometimes by income bracket (such as higher officials), took advantage of these specific needs. Standing closer to their members, they knew much better the mind of the workers. This in contrast with the recognized unions, which had more means and could therefore provide more services (legal, educational) to the members, but which were at the same time invisible on the work floor. On top of that, the categorial unions charged a much lower membership fee35). (Top (1965), p. 453; H & R, pp. 358-359; Windmuller, pp. 118, 290-292)
Scholars most often use this term to denote all non-recognized unions, which implies not affiliated to the main three federations. The most important characteristics of this large group of - mostly small - labour organizations, were ideological independence, emphasis on economic goals, and a strong orientation towards a specific occupation (Windmuller, pp. 117, 290-292; Top (1970), pp. 1-12).
34)
In 1963, the average subscription fee of the categorial unions came to ƒ 30 per year, while the recognized unions charged about ƒ 50 (Hueting et al., p. 307). This latter amount did no longer come up to the 2%-norm that was agreed upon after World War II, since the members refused an automatic increase of their fee as the result of pay rises (Windmuller, p. 304).
35)
63
Table 3.6 Percentage membership share of the recognized federations versus the categorial unions, 1945-1989 NVV
NKV
CNV
MHP*
1945
29.9
23.0
11.8
35.3
1947
31,7
24.0
12.6
31.7
1949
32.9
25.6
13.5
28.0
1951
38.3
29.4
16.0
16.3
1953
38.5
29.4
16.2
15.9
1955
37.1
30.3
16.4
16.2
1957
36.5
29.6
16.4
17.5
1959
34.6
29.6
16.2
19.6
1961
35.7
29.4
15.8
19.1
1963
36.0
28.5
15.6
19.9
1965
36.2
27.8
15.8
20.2
1967
36.6
27.8
15.6
20.0
1969
36.5
25.9
15.4
22.2
1971
38.8
24.9
14.9
22.4
1973
39.8
23.9
13.8
22.5
1975
40.7
20.7
13.1
25.5
1977
41.8
19.6
16.9
6.6
15.1
1979
42.4
18.5
17.2
6.7
15.2
Other**
1981
60.3
20.3
6.7
12.7
1983
59.3
19.9
7.0
13.8
1985
59.0
19.5
7.0
14.5
1987
59.0
18.9
7.2
14.9
1989
59.1
18.3
7.6
15.0
The MHP became recognized in April, 1976. Between 1945 and 1950, this figure includes the EVC, the share of which amounted 20.4% in 1945, 16.9% in 1947, and 14.1% in 1949.
*
**
64
Since 1957 the union density rate had stagnated and declined to an average of 40.5% during the sixties. The recognized unions realized they had to reflect on the high turnover problem and the growing attraction of the categorial unions. Already in 1955 the NVV assigned the sociologist Van de Vall to examine the motives of resigning members. From his research, conducted between 1956 and 195836), it became clear that the two main motives to resign were the notion that membership had become useless (due to the merits of the welfare state), and that the benefits did not compensate for the high union dues (Van de Vall, p. 239). To make the membership more attractive again, it was most important to reduce the distance between union (leaders) and members. In the course of the 1960's, this aim was pursued in several ways. For one, the unions realized they had to go along more with the existing needs among their members. Especially the unions in the booming sectors would profit from a more decentralized policy, as they could press for higher demands than average. Gradually, there was a shift of power towards the affiliated unions, but there would remain occasions, in which the unions were prepared to let the national interest prevail. (Windmuller, pp. 119, 231, 271-272) Facilitated by the gradually more differentiated wage policy as well as by the favourable economic circumstances, the annual wage consultations changed into wage negotiations. Little by little, unions started to engage in strikes again. In the first half of the decade, the workers fought for higher rewards, but in the second half of the 1960's, the unions gradually had to call strikes - though often reluctantly - against a new upcoming phenomenon: massive redundancies caused by shut-downs and reorganizations. (H & R, pp. 369-371, 382-385, 394-404) In the mid-1960's two more issues were raised concerning the improvement of the bond between members and union, namely the institution of plant representatives and the question of preferential treatment to members. In 1964, the largest NVV union, of the metal workers (ANMB), first came with the idea to strengthen the position of the union on plant level, by installing a contact man (group) in all companies. Up to then, members on the local level were organized according to place of residence. Goals of the plant representation were: achieving more involvement and activities at the grassroots level, and more influence on the firm's development. However, this new institution failed to break through on a large scale for three reasons: first, it created tensions within the unions themselves, as the local and district leaders felt bypassed. Second, conflicts emerged between the more radical plant representatives and the moderate members of the works councils. Third and foremost, the plant representation met with great opposition from the employers, who wanted to keep the upper hand in their own firm. (H & R, pp. 385-389; Windmuller, pp. 385-387) The free rider problem was tackled in 1965, after years of discussion. The printing industry was the only sector with a closed shop, but in all other sectors it was decided that coercion was fundamentally wrong and would only weaken the organization. However, given the rather low union density rate while all workers benefitted from the negotiation results, the union movement sought for a way to discourage this freeriding conduct. Originally, the plan was to raise a fund to which the employers would allocate a small percentage of the total wage pay, which would then be used primarily 36)The results of this study were first published in his PhD thesis in 1963. Although the main findings are not contested, it must be kept in mind that the research period was an exceptional time, in which for the first time the union members actively protested against the compliant attitude of the NVV towards the retrenchment policy.
65
to refund the subscription fees to the union members. This idea was opposed by the central employers' associations who declared themselves against any discrimination of nonmembers, but also some unions were not in favour of it, because it would make them dependent on the employers. Finally, it was agreed that employers would annually add ten guilders per member to a special fund, from which certain union activities would be financed such as schooling.37) (Van Doorn et al., pp. 211-224; H & R, pp. 371-373; Windmuller, pp. 315-319) Another way to increase the alertness of the union movement was by means of merger. As a consequence of industrial concentration at the end of the 1960's, the unions had to follow suit. Not everyone favoured such a development; some leaders did not want to lose their power, and many members were reluctant to let go of their familiar union. Still, several unions joined forces in the beginning of the 1970's, among which the Industriebond NVV and the Industriebond NKV (Socialist and Catholic Industrial Union, formed by the unions of miners, metal, textile, and factory workers), both founded on January 1, 1972. (H & R, pp. 390-392) Between 1972 and 1974 these two unions, especially the former, took the initiative in many strikes. A certain radicalization had set in, aimed at a "redistribution of knowledge, power and income". In the Spring of 1973, a long lasting conflict occurred with industrial actions all over the country, supported by the federations NVV and NKV; the central issue was the desired levelling of income, against which the employers' associations took a firm stand. In reaction to this, the NVV and NKV temporarily withdrew from the SER. (H & R, pp. 393-399; Windmuller, pp. 228-234; Reynaerts & Nagelkerke, hereafter referred to as R & N, pp. 199-201) Owing to the more active attitude of the NVV in particular, membership rates showed quite an improvement in the first half of the 1970's. The union movement as a whole expanded, and the union density rate reached the level of the 1950's again; at its height in 1975, the density rate became 43.3%. Yet, the membership rose only at the NVV (+14.8% between 1970 and the end of 197538)) and at the categorial unions (+ 31.4%), while the CNV and the NKV fell back by about 5% and 11%, respectively. Since the end of 1972, the confessional federations together would remain smaller than the NVV, also indicating a lessening of the influence of confessionalism. (See tables 3.4 and 3.5; H & R, p. 394) Within the NKV, the majority could accept the waning importance of religion; already in 1968, two thirds of the NKV members were in favour of a unified movement. As always, the CNV was much more reluctant on this matter, because her christian identity was at stake. Between 1970 and 1974 the three federations made serious attempts to seek for more unity, but finally the CNV decided to retreat, also due to the radicalization of the NVV. (H & R, p. 392-393) Another effect of the more radical policy of the NVV was the growth of categorial unions for higher personnel, whose interests were not served by any levelling of income. When the NKV chose to form a federation with the NVV, this was the direct cause for two unions, among which the Unie BHLP (about 32,000 members) to leave the NKV in 1975 and to join the new categorial Federation MHP (Middle 37)These employers' contributions to union activities became known as the "union ten- guilder piece" (vakbondstientje) and have continued until today. Over the years, the money of these build-up funds has been used to finance branch specific schooling. In the 1980's, employment measures have been financed as well out of these funds. 38)Part of this increase was owing to the entry of the categorial Union of Educational Personnel (ABOP) per January 1, 1971.
66
and Higher Personnel), founded in April 1974 (Conen, p. 55).39) Gradually the higher echelons began to realize that their group too, was ever more threatened by unemployment. This, together with the equalizing policy of the other federations, fastened the growth and co-operation of the categorial unions who organized this higher personnel. Also, the establishment of a federation would greatly increase the possibility that these unions would finally be recognized at the (central) bargaining table; this indeed happened in 1976, when the MHP was admitted to the SER. (Windmuller, pp. 296-298; R & N, pp. 69-72)
3.5.3 1975-1990: shift of power, deterioration and some recovery Between 1974 and 1976 the circumstances for the trade union movement changed to such an extent, that it is justified to distinguish the beginning of a new (sub)period in the post-war labour relations. The main cause of this alteration could be found in the business cycle. After the oil crisis at the end of 1973, a worldwide depression set in. Long before 1973, several unfavourable processes and external shocks had already weakened the Dutch economy: ever since 1960 the average wage had been increasing faster than labour productivity, causing a declining rentability. This resulted in growing inflation since 1968, while unemployment started to rise, in the weak branches first. When the oil price increased by 130%, the great dependency on oil became unravelled. It came as a huge shock, and made that faith in the economy decline, causing an enormous drop in the stock markets. For the Dutch labour force, the main impact was the rise of structural unemployment, which required a change in trade union policies. (Z & G, pp. 102-105, 255-256) The beginning of the depression does not coincide exactly with the chosen caesura of 1975/1976, for the union movement first declined after 1975. Through 1975, total membership was increasing steadily; the union density rate reached its top in that year as well. The explanation can largely be found in the policy of the new leftist government, which at first benefitted the unions. At the end of 1973, when the bargaining parties could not agree on degressive indexation40), the government issued a temporary law for 1974, to regulate all incomes; everyone would receive the same absolute pay rise. As this implied a levelling of income, while the Cabinet was also programming other measures in favour of the work force, the unions did not protest. In 1976 the government limited the wage increase again, but this time the regulation met much more the wishes of the employers, and as the announced pro-worker reforms hardly came about, the initial good relationship between unions and government cooled off. (Z & G, pp. 256-261; Windmuller, pp. 233-237; Van Waarden, pp. 92-95) During 1975-1976 there was also a certain shift of power within the union movement, justifying a demarcation with the preceding period. In 1975, the newly founded MHP was greatly reinforced owing to the affiliation of the Unie BHLP. This led one year later to admission in the SER. Since the federation between the NVV and NKV also commenced from 1976, the arisen FNV (Federation of Dutch Trade Unions) was by far the strongest centre (FNV: 1,058,000 members, CNV: 226,000 members, and 39)The three largest unions within the ideologically neutral - but rather conservative -MHP were (and still are) Unie BHLP and NCHP (both higher personnel and managers in the market sector), and CMHA (intermediate and higher civil servants). 40)This implied a wage compensation for the risen consumer prices, with a certain minimum, and declining according as the wage level was higher.
67
MHP: 110,000 members). While the NVV and NKV were heading for a final merger (in 1982), the distance between FNV and CNV only became greater. The Consultative Board was dissolved, and mutual co-operation decreased. The main reason for this alienation was, that the CNV remained moderate-minded in its views regarding co-operation with the employers. At the same time, the FNV underwent an ideological reorientation towards more socialistic reforms, for which a more regular use of strikes was permitted. (Windmuller, pp. 268-269, 301-303; Peper, p. 1364) From 1976 onwards, the union movement was forced back on the defence. As business profits deteriorated ever faster, so did the bargaining power of unions. The key problem in the Dutch economy was the disproportionate increase in the costs of labour, which already since the 1950's rose much quicker than the costs of capital; the combination of over investments and a very tight labour market drove up the wages even faster in the 1960's and early 1970's. The mechanism of wage indexation, the introduction of a minimum wage in 1968, and the rising tax and premium burden were also forcing up the gross wages. This unbalanced wage growth took a heavy toll of the work force. It was leading up to the substitution of labour for more capital, and it injured Dutch competitiveness to such an extent that it resulted in mass lay-offs. (Z & G, pp. 262-265) Where the first half of the 1970's were characterized by a radicalization of a large part of the union movement - by demanding more equality of income, and more influence on the work floor -, the second half of the decade showed yet another change of course. Forced by the worsened conditions, unions started to moderate their wage claims. These developments benefitted the employers, who had already reinforced themselves since 1968 when a merger led up to the foundation of the VNO (Dutch Federation of Enterprises). Both the VNO and the Christian Federation of Employers, NCW, realized that centralization would be the effective weapon against the workers' demands and actions. Full-time professional leaders were installed, much more effort was undertaken to influence the public opinion as well as the government's policies, and a 'counterstrike-fund' was set up so that individual employers would not have to give in to the unions. (R & N, pp. 201, 210-211, 223-224; Peper, p. 1365). During the negotiations for 1977 the employers tried to exploit their increased power so as to get rid of the wage indexation. The union movement reacted with carefully planned strikes in remunerative companies, creating a hole in the employers' front. The employers had to agree on maintenance of the automatic price compensation, at least for the time being. As the average pay rise also became more than the employers had wanted, the unions gained an almost complete victory, which was once more translated into a reasonable growth of the total membership rate by 1.6%. (Albeda & Dercksen, pp. 178-179; Windmuller, pp. 237-239). Yet, the union movement began to realize that moderation of its wage demands became a necessity for two reasons. To improve the firms' profits, which should slow down further dismissals, and out of solidarity with the non-active population, whose social benefits were linked to the minimum wage; if the wage increases became too high, this linkage would be endangered. The unions aimed at a trade-off: for 1978 they were willing to accept a wage sacrifice of 1.5% if they could reach agreements on the preservation of jobs (arbeidsplaatsenovereenkomsten), to be accomplished by early retirement, and by a transition to five shifts in the continuous industry. For 1979, again such a trade-off was offered, and now the industrial unions aimed at a reduction of the working week to 35 hours, to create more employment. However, in
68
both years the unions could not retain a united front 41), while employers would rather concede to a pay rise, than to give the workers any say in their (employment) policies. (R & N, pp. 211-215; Windmuller, pp. 240-245; Peper, pp. 1361, 1364-1365) Meanwhile, the government had unilaterally decided to curtail the social benefits payments, and the salary growth of all its personnel from 1978 onwards. Hereupon, the civil servants unions organized large demonstrations, which was a novelty, but failed to alter the government’s plans. Whereas strikes were reduced to a minimum in the market sector, labour disputes increased in the public sector. (Windmuller, pp. 241242; Peper, p. 1362; Albeda & Dercksen, hereafter referred to as A & D, p. 181) Between 1980 and 1982, triggered by the second oil crisis of 1979, a severe depression hit the Dutch economy. The budget deficit rose quickly, and so did the interest rate, which forced the government to cut down expenses by an extra amount. The private consumption also decreased heavily, causing a large fall in economic growth. The risen interest brought firms in great difficulties, as many had become very dependent on borrowed capital. Many bankruptcies caused unemployment to rise dramatically, aggravated by the fact that the labour supply, especially of women, was also increasing since the seventies. Between 1975 and 1979 about 5% of the labour force was unemployed, rising to 17% in 1983/442). (Z & G, pp. 256-262) Each year, the government tried to reach a central agreement with the social partners. Yet the interests and objectives were too wide apart, not only between unions and employers, but also among unions themselves (serving primarily the interests of their own members versus solidarity with the rest of the society). From 1979 to 1982, the government took measures to slow down the wage growth in the market sector. The salaries in the collective sector were reduced most, while the social benefits payments also decreased. In 1980 and 1981, real wages of all workers declined. (Windmuller, pp. 241-252 R & N, pp. 235-236; A & D, pp. 64-65) The union movement stood with its back against the wall, which showed itself in a loss of members for most unions, except at first for the CNV. During 1981, this federation had gained a lot of former members of the NKV (about 40,000), who were against the upcoming merger with the NVV. Between 1979 and its low in 1986, the FNV lost most members, about 175,000 (16%). The union movement as a whole declined by about 13.5% in that period (table 4.4; R & N, p. 235). Graph 3.2 shows the post-war developments of the recognized union federations.
41)As the economic situation could differ considerably between branches, coordinated actions of unions became more difficult. Some unions still strove after wage improvements, whereas others were willing to renounce a pay rise altogether if agreements regarding employment could be made (Peper, p. 1364; Windmuller, p. 243). 42)This percentage differs considerably - from 14% to 17% -, depending on which definitions of the labour force and of the unemployment are used. But it is agreed upon that the percentages were extremely high in the first half of the 1980's.
69
Graph 3.2 Union membership of the FNV, NVV, NKV*, CNV, and MHP 1945-1989 (end-of-year data, in thousands)
*
Including its predecessor, the KAB.
Until 1982, the unions still aimed at preserving the spending power, by holding on to their acquired right of automatic wage indexation. However, in that year the unemployment increased even faster, so that unions let go of their last demands. The sole strategy now became the preservation of employment, to be accomplished by work-sharing through shorter working hours. There were two reasons to give absolute priority to employment policies. In the first place, due to the huge unemployment the union movement had lost most of its bargaining power over wages. Secondly, it was realized that the social security system was very much endangered, as the ratio between the employed and people receiving some sort of social benefit had declined from 7:1 in 1970 to 3:1 in 1981. (Windmuller, pp. 419-420; Visser (1989b), pp. 238239) The employers were not really against a reduction of working hours, but they did oppose any central agreement that would apply to all categories of the work force, as they wanted to become more flexible. For the same reason, they rejected the government's wage restraints, which obstructed their personnel policies. Facing the threat of another wage intervention for 1983, the employers were finally prepared to come to an agreement with the union movement. (Visser (1989b), p. 239; A & D, p. 68) Hence, in November 1982 the Central Agreement came about, with the following recommendations. In order to restore economic growth, competitiveness and business profits, workers would renounce wage claims. In return, the available employment would be distributed among more people through several measures, to be taken at a decentralized level in order to be effective. It was agreed upon that current contracts could be reopened, after which the wage settlements would be used, among other things, to finance the costs of shortening the working-time. The government was urgently asked not to interfere anymore; from that time the social partners
70
wanted to take responsibility themselves for improving the employment situation. (Windmuller, pp. 418-419; R & N, pp. 242-243) Thereupon, the government gave its consent and soon afterwards many initiatives were taken. Within one year, in two-thirds of all 481 collective agreements, covering about 85% of the 2.3 million private sector workers, wage claims had been postponed. However, the reduction of the working time was a very gradual process. In the public sector the 38-hours working week became the norm as from 1986, but this would not become the general outcome in the market sector. Other measures to create more employment were voluntary early retirement schemes and jobs to a maximum of 32 hours for workers below the age of 25. (Visser (1989b), pp. 240-241) Owing to the reduction or withholding of wage increases, the profitability and competitiveness of Dutch firms greatly improved. The wage cuts combined with the reduction of the least productive hours, led to a substantial decline of the labour costs: the labour income ratio to national income fell from 90.5% in 1980 to 83.5% in 1985. However, the refilling of vacant hours and additional recruitment hardly came about. Still, it did slow down further unemployment growth and contributed to job maintenance. The unions were in no position to enforce the refilling of vacant time. (Visser (1989b), pp. 241-242) Up to and including 1986, the union movement lost more members than it enrolled new ones (see table 3.7 below), though the absolute loss was becoming smaller as the economy slowly recovered. With the upswing of the business cycle and the rising employment in the second half of the decade, total membership finally started to rise again. At the beginning of the 1990's, 1,636,000 people were union members (see table 3.4). If the union density rate is defined as the ratio of total membership and the dependent labour force with (or looking for) a job over 20 hours per week, it declined from 42.7% in 1978 to 31.1% at the end of 1986, after which it stabilized. If however total membership is divided by the total dependent labour force, the union density rate declined from 37.9% in 1978 to 26.2% in 1987, at which level it remained thereafter. (see table 3.5)
71
Table 3.7 Inflow to and outflow from unions (per year, in % of total union membership* at the beginning of that year) Inflow
Outflow
Net growth
1972/73
12.4
11.0
1.4
1974/75
11.9
10.7
1.2
1976/77
11.4
10.0
1.4
1978/79
10.2
9.7
0.6
1980/81
8.3
11.1
-2.8
1982/83
7.3
10.0
-2.7
1984/85
6.1
8.8
-2.8
1986/87
8.3
7.7
0.5
1988/89
10.9
7.2
3.7
1990/91
10.9
7.6
3.3
1992/93
10.3
7.8
2.5
Source: CBS, Leden van de vakbeweging (two-yearly union survey); as from 1989 published as an article in Sociaal-economische Maandstatistiek. *
These figures are based on a response of about 92% of all members.
Between 1974 and 1986 most OECD-countries were confronted with fast rising unemployment, leading to a strong decline of most union movements, except in Sweden, Denmark, Finland and Belgium. There are strong indications that the union success in these four countries can be attributed to favourable institutional conditions, namely a very centralized bargaining system combined with a union-organized unemployment insurance (Ghent) system, making membership very attractive if not indispensable (cf. Neumann et al.; Western). Finally, if the post-war developments are summarized in terms of supply and demand, the following image emerges. The supply of unions directly after the war had the same features as before: the tripartition in the union movement remained the dominant characteristic, despite efforts of the NVV, and the EVC, to create unity. The three major federations had agreed to eliminate competition between them, by charging approximately the same subscription fee, and by supplying the same kind of services. Since these pillars (NVV, KAB and CNV) were recognized as official advisors to the government and in the central bargaining process, they became much more influential than all the other unions and could therefore attract the majority of the union members. After the events of the Great Depression and World War II, there was a great consensus about the reconstruction of the Dutch society among government, employers and the recognized federations. In just a short period of time the parties succeeded in rebuilding the economy, accompanied by full employment and an enormous increase of social security.
72
At first, workers could appreciate the union policies, which expressed itself in a fast growing membership and a rising union density rate. But already early in the 1950's this density rate was over its top and stagnated completely, and only shortly increased again in the mid-1970's. Until 1979 the number of union members kept rising, apart from a few minor setbacks, but the total labour force expanded at least as much. As from 1980, the recession caused a sharp drop in the membership and in the union density rate. Recently, the beginning recovery of the union movement has not been accompanied by a noteworthy improvement of the density rate. Explanations for the unions' difficulties can be found both on the demand and on the supply side. One of the most striking explications lies in the bad match between demand and supply of services. While the (recognized) unions kept on emphasizing the interests of the whole society, the increasing welfare incited workers to claim a higher share, to which unions only reluctantly reacted. Their influential role on the central level, for a long time facilitated by favourable economic developments, made them powerful, irrespectively of the number of their members. Dissatisfied workers reacted in two ways: either they joined one of the categorial unions, whose share in the total membership rose steadily from the mid-1950's; or they renounced their membership altogether, which could not really harm them. Owing to the merits of the welfare state many people had come to consider union membership as useless, and as a consequence of the act on the extension of collective labour agreements workers benefitted from the unions' bargaining results anyway, without the necessity of being a member. The obligation to join a union only existed in the printing industry. From the second half of the 1960's especially the NVV tried to make the membership attractive again by installing plant representatives, engaging in strikes and demanding more worker influence at company level. This more active attitude bore fruit for the NVV, but the CNV and NKV dropped behind. The role of confessionalism started to wane. The NKV and NVV were heading for a merger, but the CNV remained independent. This did not lead to less fractionalization though, because of the rise of the MHP, which protected the interests of the higher personnel who felt endangered by the equalizing policies of other unions. The Dutch union movement remains fragmented and therefore less powerful. In the 1970's the economic circumstances deteriorated, but the unions still fought for more worker participation, better terms of employment and conservation of jobs. This short-lived radicalization was accompanied by an increase of the union density rate. Apparently, unions at that time appealed to the workers' needs. However, with the huge rise in unemployment from 1980 onwards, the union movement lost its power and with that its force of attraction. Without the existence of the Ghent system, the very high level of unemployment only had a negative effect on Dutch unionization. Again unions tended to rank collective interests prior to those of their working members, which created tensions. With the upswing of the business cycle the union density rate did not follow suit, which can be attributed to two different structural developments. Firstly, accelerated by the Central Agreement of 1982, the bargaining over labour contracts is gradually becoming more decentralized (R & N, pp. 237-240; Visser (1992), pp. 33-36; Visser (1989b), pp. 246-247). This is favoured by the employers, who prefer to reach deals that are geared to the conditions in the individual firm. They still prefer to bargain over wages with the unions because of their professionalism and reliability. Concerning other terms of employment, employers would rather negotiate with worker representatives within their own firms, for instance the works councils. Until now, unions have not co-operated much with the works councils, as they are still often
73
looked upon as rivals, even though about 70% of all works councillors are union member. Due to the tendency of decentralization, the weak performance of the unions at plant level becomes more noticeable. Especially in the first half of the 1980's, when the unions had not much to offer, this absence on the work floor had a negative impact on the bond between members and union. For several years a strong appeal was made to the working members, to accept a deterioration of their terms of employment, out of solidarity with the unemployed. Such a solidarity is difficult to preserve for a longer time, especially when the employment policies do not come off very well. (Visser (1989b), pp. 244-245; Windmuller, pp. 422-423) Hence, the union movement is faced with conflicting interests of the unemployed and its own members. It can not afford to neglect the latter group any longer, as many reports (cf. Van Ham et al.; Visser (1986)) point exactly at the danger of losing more and more members, who expect more and better service from their union. The second development, obstructive to union growth, can be found on the demand side and concerns structural changes in the working population. Traditionally, the demand for unions has been low in agriculture and in private services, and high in manufacturing, transport and public services. The early and vast decline in agricultural employment therefore had no negative effect on Dutch unionism, contrary to the later employment shifts in other sectors. Since the 1970's, the structure of the work force gradually underwent several changes. Industrial employment collapsed, but a lot of new jobs were created in dynamic branches of the service sector. This shift from manufacturing towards services has had a negative impact upon the union density rate, as the industrial sector used to be the unions' stronghold, while unions have not yet gained a foothold in those new employment sectors. Thanks to the strong job growth in the public sector in the 1960's and 1970's, the union movement was able to compensate its losses until about 1980. (see Visser (1990), pp. 41-51) Another unfavourable process is the rise of categories of personnel, which are more difficult to recruit. Examples are: the increase of female employment, the prevalence of white-collar, over blue-collar workers, in the private sector, and, partly overlapping, the rise of jobs for higher educated personnel. These sectoral and occupational shifts affect the union density rate negatively, not only in the Netherlands, but elsewhere as well (Visser (1990), pp. 61-65). Finally, the pattern of labour has been changing substantially, in the sense that the standard 38-hours week, combined with tenure, is losing its preponderance fast. Ever more people are working temporarily, part-time, or on the basis of leasing contracts. Those kind of jobs lead to less commitment to the labour market, and thus to the union movement, than the 'traditional' jobs. (see Visser (1986), pp. 134-136; A & D, pp. 47-48; 239-241; Van Ham et al., p. 470) Unions have great difficulties in adjusting their strategies to all these changes. One of their major problems might be their division; not only between the different federations but also within. The largest centre, the FNV, has 19 affiliated unions which are not co-operating enough. Some have many members and large resources, which they refuse to share with the smaller unions, which are therefore not capable enough to attract more members. This segregation should be broken through.
74
3.6 Conclusions The Dutch trade union movement has gone through a lot of changes between 1860 and the present. Growth and decline of the union membership have succeeded each other, and these fluctuations stem from a variety of causes, which have been described in this chapter. In this last section the most important determinants of unionization will be lined up once more, and will be placed in a theoretical framework. Since closed shops have never become an important feature in the Netherlands, the demand for unions and their services has always been a matter of personal weighting, influenced by a lot of environmental factors, in particular the economic and social situation, the mental climate and the institutional setting. These same determinants also influence the supply of trade unions, and the degree to which unions can fulfill their goals. In the 19th century, the circumstances did not create a fertile soil for large scale unionism. The relative late industrialization delayed a concentration of the workforce, whose backwardness and prevailing opinion about obedience to the employer obstructed the development of a class-consciousness; religious beliefs played an important role in this. When their working conditions deteriorated, educated craftsmen became accessible to new ideas about defending their own interests, and founded the first trade unions. However, the unions that really aimed at improving the terms of employment, if necessary by striking, did initially not increase very much. This left wing of the union movement was not only fought against by employers, but also by several confessional and conservative groupings, who reacted by forming their own organizations, which could hardly be regarded as real trade unions. Moreover, also within the progressive camp differences of opinion about goals, methods and organizational structure impeded their further development. During an economic downturn the unions were even more weakened, so that many members would resign. With the foundation of the first efficiently organized union federation, the socialist NVV, in 1906, it instigated the Protestants and the Roman Catholics to establish their own centres in 1909, which was the start of a long-lasting tripartite division within the Dutch trade union movement. In the first decades of the 20th century, the unions not only strove after improving the material terms of employment, but also dedicated themselves to emancipate the working class. They aspired to more extensive education and universal suffrage, so that the voice of the workers would be heard in Parliament. By constituting more or less formal alliances with political parties, the three main federations pursued these emancipatory goals. Through these channels, they also attempted to improve social and labour legislation, because the workers were hardly protected against the risks of accidents, illness, unemployment and so forth. Since the state scarcely provided social services as yet, the unions set up social benefit funds themselves, which also served as a means to hold on to their members. The main difference between the three pillars was the NVV's adherence to the class struggle and its consequent willingness to strike to reach its goals, while the confessional unions adopted a much more moderate attitude. The rigorous pacification forestalled a unified labour movement, and during slumps this lack of concord weakened the unions' bargaining position and thereby their power of recruitment. As spokesman of a growing part of the labour force, the union movement gradually became an accepted institution by the workers, but also by employers and
75
government. Recognition by employers' associations revealed itself on the bargaining levels, in an increasing number of firms subscribing to collective labour agreements; and on the national level, in the joint consultation in the High Council of Labour. Recognition by the government first became apparent after the outbreak of World War I. It was reasoned that involving the unions in the combat against the arisen chaos and unemployment as a result of the war, would prevent social upheaval. The unions' unemployment funds became heavily subsidized by the state under the Ghent system, which would last during the whole interwar period. Furthermore, at the instigation of the union movement the government introduced a number of laws beneficial to the working class. Special mention should be made of the 1927 Law on Collective Bargaining Agreements, and the Act on Extension or Nullification of Provisions of Collective Labour Agreements of 1937. Although the unions had pleaded for the introduction of these laws, their effects on the size of the union movement could be ambiguous. On the one hand, they protected the union members in the firms, as employers became obliged to provide the same terms of employment to members and nonmembers alike. On the other hand, extension of the agreements to all workers in the firm and in the rest of the sector could lead to free-riding conduct among workers, as the unions' bargaining results would accrue to them anyway. At an early stage, unions learned to take advantage of the business cycle. In periods of boom, they successfully demanded higher wages, and the number of collective labour agreements increased. These bargaining results, and the growing recognition by the employers, strengthened the position of the unions, thereby attracting new members. Hence, when the economy was booming, membership figures rose. Slumps however, were not all the time accompanied by declining membership rates in the period before 1940. During the first World War for instance, the unequal distribution of (un)employment and welfare caused much social unrest, which prompted many people to join the union in hopes of changing the society. Conversely, the recession between 1920 and 1924 did lead to a declining membership rate. However, the Great Depression of the 1930's did not result in a mass resignation of union members; at first, their number even increased. This phenomenon could partly be ascribed to solidarity, and partly to the fact that unions remained useful for unemployed members, as they provided social benefits payments and the possibility to join special unemployment unions. Graph 3.3 below shows the development of total unemployment and total union membership in one picture. Towards the end of the 1930's, the union movement had gained considerable results. The unions had contributed much to the extended welfare of the working class. Not only had the terms of employment been improved (higher wage levels, shorter working weeks), but the spiritual well-being of the workers as well. The introduction of universal suffrage, and the improvement of social and labour legislation could largely be ascribed to the efforts of trade unions. During and immediately after the second World War, attempts to create unity in the trade union movement failed, but the three main union federations (NVV, KAB and CNV) did come to much more interorganizational co-operation. In order to rebuild the heavily damaged economy, collaboration between union federations, employers associations and government also greatly increased. At least until the 1960's the Dutch corporative economy ("overlegeconomie") remained characterized by very centralized decision-making, which only gradually started to subside.
76
Graph 3.3 Connection between unemployment (U) and membership (M)
Initially with approval of their members, the recognized unions gave priority to serving the common interests, which implied that for several years real wages lagged behind productivity growth. The guided wage policy and other measures quickly yielded rewards. Already in the early 1950's the Dutch economy had recovered, and subsequently experienced an almost uninterrupted period of growth until 1973. Gradually at first, and in the 1960's at increasing speed, the labour force shared in this wealth as wage levels rose. Upon the solid economic foundations the welfare state developed; many social insurance laws were introduced or extended, as well as a range of labour laws concerning the security and safety of work. Unemployment remained very low until 1971. In the course of the 1970's, triggered by the 1973 oil crisis, the economy deteriorated. Dutch competitiveness declined, leading to ever more lay-offs. Between 1980 and 1986 unemployment rose to unprecedented levels, and the upswing of the business cycle in the second half of the 1980's only slowly reversed this trend, because a quick growth in employment was partly counterbalanced by persistent long-term unemployment. The fluctuations in the union membership can be placed against the background of these post-war economic and institutional developments. At first sight, a strong connection between (un)employment growth and membership can be detected, as Graph 3.3 suggests. Apart from a few minor exceptions, the union movement gained members during the upswing of the business cycle; the downfall of members in 1951 was caused by the exclusion of the EVC from the statistics. Only when unemployment really shot up from 1980, did total union membership started to fall significantly. With the recovery of the economy in the second half of the 1980's, union membership also increased again.
77
However, a few remarks should be made on this apparent relationship between membership and business cycle. Firstly, rising unemployment does not always lead immediately to a declining membership rate, which can be seen in the years 19301932 as well as in the second half of the 1970's. The threat of becoming unemployed may induce people to join a union in order to be protected, and when union members indeed lose their jobs, they often - at first - stay in the union either because the union provides them with benefits payments (in the pre-war period) or because the contact with the labour market is maintained through the union. Moreover, feelings of loyalty may play a role. Only persistent unemployment reduces faith in a union, leading to resignation. Secondly, a fall in the union membership rate is not always caused by increased unemployment. Union members can have many other reasons to resign. For instance, in 1957/58 the NVV lost a considerable number of members who disagreed with the federation's support of the retrenchment policy. The rise of the categorial unions at the cost of the recognized unions since the second half of the 1950's, was another sign that ever more union members rejected the establishment's policy of promoting the interests of the whole society, instead of just theirs. The centralized structure of union federations, and the invisibility of unions at the workfloor contributed to this dissatisfaction, which also showed itself in the high turnover of members, a problem that still exists. And thirdly, the development of the post-war union density rate hardly shows any causal relation to the business cycle. Comparing Graph 3.3 with Graph 3.4, it becomes clear that the fluctuations in the density rate followed the same course as the variations in the union membership level in the pre-war period. After the second World War however, these two developments separated. Since the mid-1950's, the growth of the union density rate stagnated, whereas the number of union members kept rising almost continuously. As from 1980, the density rate declined much stronger than the union membership, and the subsequent recovery was much more modest. This implies that the growth in the number of union members has not kept up with the growth of the dependent labour force. Due to the international recession most OECD-countries have experienced a downward membership trend in the 1980's, which was hardly or not reversed by the following upturn of the business cycle. Several structural developments account for this; according to Flanagan et al. (pp. 428-430), changes in the employment structure and in the occupational structure have been unfavourable to union growth, while the legal extension of labour contracts has been leading to free rider conduct in ever more countries. In addition, unions in the United States and in Great Britain have been subject to increased employer resistance, and in the latter country to obstructive public policy as well. The causes of this waning attraction of unions to Dutch workers are also very diverse, but can be resolved into a few categories. The explanations must be sought in the changing composition of the workforce, consisting of ever more people who are difficult to organize; and in the changing bargaining structure, in which ever more stress is laid on negotiations on the company level, where unions are hardly if at all represented. Until now, unions have not reacted adequately enough to all these changes. Despite the secularization and the waning importance of ideologies, the union movement has remained divisioned, and also within the pillars unions are often divided with respect to joint strategies.
78
Graph 3.4 Union density rate* (D, in %) and unemployment (U, in 1000's)
* = total membership/full-time and part-time workers _ _= total membership/full-time workers (> 20 hours)
However, since the union density rate already stopped growing after 1956, only to increase a little more between 1973 and 1979, the original cause(s) of the unions' organizing difficulties go back to the 1950's and 1960's. It can't be accidental that the expansion of the welfare state coincided with the stagnation of the union density rate. All the improvements in social insurances and working conditions, for which the union movement had fought before the war, had to a large extent been realized. Together with the fact, that the bargaining results of the unions apply to members and nonmembers alike, a number of benefits from union membership have disappeared, and are in fact substituted by legal arrangements. In a way, the unions have taken the ground from under their own feet. Finally, the description of the history of the Dutch trade union movement can be related to the demand and supply functions of the theoretical framework. In chapter 1, these functions have been formulated as Ud = f(p, y, diff, s, z, t) and Us = g(p, CO, CS, G). Given the specific developments of Dutch unionism, the following remarks can be made with respect to each of the variables involved. Regarding union dues (p), the historical literature does not give much information. From the archives it becomes clear that membership fees before the second World War differed very much between trade unions as well as within unions, dependent on age, experience and income of workers. This makes it difficult to draw any conclusions on the relation between dues and membership changes. Especially in the pre-war period members could voluntarily pay additional dues to insure themselves through the union against the loss of income due to illness, unemployment, and the
79
like. Until then, the state hardly provided such social insurances, so that it formed an important motive to join and remain in the union. After World War II, the three major federations charged approximately the same subscription fee, which was much higher than at the categorial unions. From the late 1950's, this disparity would become one of the reasons why the membership share of the categorial unions rose at the expense of the recognized unions. Gradually, unions also had to deal with another kind of competition, namely from the state. With the introduction and extension of ever more publicly provided social benefits, the need for unions to provide these insurances disappeared. This development is a typical example of declining costs of substitute services (s), which forms a threat to union growth. The next variable in the demand function is income (y). One of the primary goals of unions has always been the improvement of the wages, if necessary by instigating strikes. When successful, this would lead to membership growth. In the early phase, unions bargained for their members exclusively, but owing to the legislation on extension of the labour contracts all workers in an industry could profit from the negotiated results. After the second World War, this 1937 act was implemented on a large scale and would lead to an increasing risk of free riding. The so-called relative union-nonunion compensation differential (diff) measures the wage difference between unionized and nonunionized workers with otherwise identical characteristics, and is as such not applicable to the Dutch situation, where union members and nonmembers with the same job in the same sector receive the same wage. But even in American research the variable diff usually cannot be measured directly, so that proxy variables are used instead. Workers with certain personal and industry characteristics are believed to gain more from union representation than others, which increases the possibility that they will become a member. Therefore these characteristics serve as proxies for the expected benefits of union membership. In terms of these proxies it is possible to make a few remarks with respect to the Dutch situation. In the Netherlands, the composition of the workforce has gradually undergone several changes. The employment shift from industry towards services affected unionism negatively, while the job growth in the public sector stimulated union membership. However, the rise of female - and part-time employment obstructs union growth, to recall just a few adverse developments. In this context the role of unemployment can be mentioned, which most often has a decreasing effect on trade union membership, at least eventually. In the first years of rising unemployment in the 1930's and 1970's many people held on to their union, but persistent unemployment at a high level reduces the power of unions and with that their force of attraction to workers, with or without a job. With respect to nonpecuniary benefits from a unionized work environment (z), the historical analysis has shown that unions certainly have not only exerted themselves to raise the wages. From the very beginning they fought over numerous nonmaterial issues as well. This ranged from more worker influence on politics (universal suffrage) and on industrial policies (at all levels of decision making), to improvement of the working conditions on the work floor. In the course of the century, the union movement has scored a lot of success in raising the participation and security of workers in all sorts of ways. The last variable in the demand function concerns taste for unionism (t). Attitudes towards union membership have changed through time. Up to and including the first decades of the 20th century unions were not quite common yet, due to the prevailing climate of opinion, which in turn was influenced by religious beliefs and resistance
80
from employers. Gradually however, unions became an accepted phenomenon, which was also ex-pressed by their increasing participation in advisory councils jointly with employers' associations, notably after World War II. The rise of the welfare state partly reversed this positive attitude towards unionism, as many workers came to consider the membership as superfluous. Turning to the supply function, the variables CO and CS refer to organizing and servicing costs of unions, respectively. Both types of supply costs become relatively lower if they are spread among more members. In other words, if a lot of workers are concentrated in one area, industry or factory, unions can profit from this: the same organizing efforts will yield more results (read: more members). In the same line of thought, more members will reduce the average servicing costs. From this perspective, the employment shift from the industrial - towards the service sector has been an adverse development, because the concentration of workers is on average much lower in services. Ultimately, the willingness to spend extra resources on recruitment efforts and on servicing depends on which goals (G) unions pursue. Within the Dutch trade union movement, unions and federations strove, and strive, after divergent goals. Depending on ideology or religion, unions have been more or less radical, more or less willing to organize strikes, and so forth. Their attitude, and their degree of success in the bargaining rounds, have attracted various categories of workers to a more or lesser extent. Generalizing, Christian workers feel less at ease with the fighting spirit of the FNV, while higher salaried employees find their interests best served by the MHP. Directly after the second World War the three major pillars did not really compete with each other, nor did they put much effort in bringing in new members. At that time, their power was not so much determined by the size of their movement, but more by their recognition and participation in national social economic policies. For the sake of rebuilding the Dutch economy, the three federations deliberately subordinated their members' interests to the common good. At first the workers approved this course of action, and membership rates grew. But eventually criticism gained strength and many workers temporarily quit their membership: they demanded a greater share in the increased welfare. Due to the gradual decentralization of the bargaining process and the growing competition from categorial unions in the 1960's, the recognized unions began to realize that they had to make the membership more attractive, to fight the high turnover rate. However, the installation of plant representatives became a failure mainly due to opposition from employers, while the attempt to give preferential treatment to members also foundered on rejections of the employers' organizations. From the end of the 1960's unemployment and inflation both started to rise. In order to withstand the growing resistance from employers, many unions reinforced themselves by means of mergers. A detrimental side effect of this was the increasing distance between union leaders and their members. Meanwhile the decentralization developed further, notably after 1982. Negotiations mainly take place at industry level, but ever more arrangements are left to be elaborated at the company level, where unions have little influence. At the same time, many unions still feel a lot of responsibility for the whole labour force; after the second oil crisis of 1979 they ranked the preservation of employment above improvement of the wages. Hence, it is not so surprising that the union density rate has declined and remained low at the outset of the 1990's. During a downturn of the business cycle unions are not able to serve the interests of their members very well; moreover, they ask
81
sacrifices from their members to defend the position of nonmembers as well. And when the economy is recovering, unions may be able to increase the wages, but nonmembers also profit automatically. With respect to other terms of employment, these are ever more determined at the company level, beyond the reach of unions. In the end, this inevitably leads to growing dissatisfaction among members, who expect more worth for their money. The aim of the next chapters is to establish to what degree all these described developments and causal relationships can be tested empirically, with the aid of formalized time-series and cross-sectional analyses, respectively.
82
APPENDIX List of abbreviations ANDB ANTB ANWV BBA CAS CMHA CNV CNWB EVB EVC FNV KAB MHP NAF NAS NCHP NCW NKV NVV PVDA RKWV SDAP SDB SER Unie BHLP VNO VNW VPCW
Algemene Nederlandsche Diamantbewerkersbond (General Netherlands Diamond Workers Union) Algemene Nederlandse Typografenbond (General Netherlands Typographers Union) Algemeen Nederland Werklieden Verbond (General Netherlands Workingmen's Association) Buitengewoon Besluit Arbeidsverhoudingen (Extraordinary Decree on Labour Relations) Christelijk Arbeidssecretariaat (Christian Labour Secretariat) Centrale van Middelbare en Hogere Functionarissen bij Overheid en Onderwijs (Centre for Higher Civil Servants) Christelijk Nationaal Vakverbond (Christian National Trade Union Federation) Christelijk Nationale Werkmansbond (Christian National Workmen's Union) Eenheidsvakbeweging (Trade Union Unity Movement) Eenheidsvakcentrale (Unity Centre) Federatie Nederlandse Vakbeweging (Federation of Dutch Trade Unions) Katholieke Arbeidersbeweging (Catholic Labour Movement) Vakcentrale voor Middelbaar en Hoger Personeel (Federation of Middle and Higher Personnel) Nederlands Arbeidsfront (Dutch Labour Front) Nationaal Arbeids-Secretariaat (National Labour Secretariat) Nederlandse Centrale voor Hoger Personeel (Dutch Federation of Staff Associations) Nederlands Christelijk Werkgevers Verbond (Christian Federation of Employers) Nederlands Katholiek Verbond van Vakverenigingen (Netherlands Catholic Trade Union Federation Nederlands Verbond van Vakverenigingen (Netherlands Federation of Trade Unions) Partij van de Arbeid (Labour Party) Rooms-Katholiek Werkliedenverbond in Nederland (Catholic Workers Federation of the Netherlands) Sociaal-Democratische Arbeiders Partij (Social Democratic Labour Party) Sociaal Docratische Bond (Social-Democratic League) Sociaal-Economische Raad (Social-Economic Council) Unie van Beambten, Leidinggevend en Hoger Personeel (Union of Functionaries, Executives and Higher Personnel) Verbond van Nederlandse Ondernemingen (Dutch Federation of Enterprises) Vereniging van Nederlandse Werkgevers (Association of Dutch Employers) Verbond van Protestant-Christelijke Werkgevers in Nederland (Federation of Protestant Employers)
83
84
4. TIME-SERIES ANALYSIS ON TRADE UNION GROWTH 4.1 Introduction In international literature, two econometric models of union growth have been very influential. The first one comes from Ashenfelter and Pencavel (A & P 1969), the second from Bain and Elsheikh (B & E 1976). Both models have not gone completely uncontested, but nevertheless they stimulated a lot of further research into this area. This has been discussed in chapter 2. Both researches contend that one model is capable of explaining union growth in the whole period. Phrased differently, both couples claim that there was not any structural break in the period under consideration. Apart from other criticisms on both models, the absence of a structural break has been contradicted by several researchers in both cases. It would seem that economists strive to find one relationship that can explain the whole period, as if it is the best possible result one can get. However, from the point of view of an economic historian, it would be very surprising indeed to find that only one model will suffice, if we speak of a period that lasts more than a few decades. In a 60 to 70 year period so many things change in a society, which may have an impact on the unionization decision. One important category of changes that will often lead to a structural break, concerns changes in institutional arrangements and laws. Most countries have experienced such changes, and in many cases researchers have indeed shown that these events have been responsible for a structural break in a model on union growth. Among others, this has been claimed for the U.S.A, the U.K., Sweden and Denmark. Until now, not much has been published on trade union membership developments in the Netherlands. As far as time-series analysis is concerned, there is an article by Van Ours (1992) that only covers a 28-year period and employs just a few variables, and the research by Visser (1987), who analyzed a longer period (1913-1985), but he presents the main findings only very roughly. Finally, there is an M.A. thesis by Hendriks (1986) on the period 1910-1985, which is very detailed but rather explorative. In this chapter, a much more extensive time series analysis on Dutch union membership will be presented. A short overview of the relevant variables will be given in the next section, followed by a discussion of the availability and peculiarities of the Dutch data in section 4.3. Thereafter section 4.4 presents a replication of the A & P and B & E models, while in section 4.5 models are constructed that are tailored to the Dutch circumstances. After the conclusions in section 4.6, some more statistical information can be found in the appendix. 4.2 Short review of the dependent and independent variables In this section the theoretical assumptions mentioned in chapter 2.2.2 will be briefly repeated, supplemented with the additional insights gathered from chapter 3. Concerning the specification of the dependent variable, the annual rate of change in union membership is chosen: ΔMt = 100*(Mt - Mt-1)/Mt-1. Several different specifications occur in the literature, but this form is most common. The most logical alternative form would be the union density rate, but following the argument by Bain & Elsheikh (1976, p. 59) this has several drawbacks (see also chapter 2, page 18). For one, the density rate shows only small annual variations, which is a disadvantage in time-series analysis. Furthermore, the density rate can be an explanatory variable as well, so that it makes
85
more sense to use the annual fluctuations in the membership rate on the left side of the equation. Table 4.1 presents an overview of union membership fluctuations in the Netherlands. Next, all the relevant and available independent variables will be mentioned, starting with the basic ones. Inflation. Price increases are supposed to radiate a so-called "threat effect" on employees, inducing them to join or remain in a union, which is expected to defend the real wages from inflation. In the Dutch case, it is questionable whether this effect still played a role in the post-war era: at first increasing prices were balanced by pay-rises for everyone on account of the guided wage policy, and from the 1960's until 1982 the system of automatic price compensation was in force in all branches of industry. Wages. A rise in the wages is expected to stimulate union growth, as the workers credit such an income improvement to the trade unions. The Dutch legislation on extension of collective labour contracts to a whole sector might have mitigated this credit effect after World War II. Unemployment. This variable can affect union membership in different ways, and can be specified either as a rate of change or by using the absolute level. In empirical work, both measurements are used. When unemployment first starts to rise, workers might decide to become or remain a union member in order to be protected against discharge. Furthermore, unemployed members often pay a reduced membership fee. However, as unemployment persists, the advantages of being a member might not weigh up against the costs anymore, and then membership rates start to drop. In addition, during a period of high or rising unemployment unions are not able to obtain good results out of the wage bargaining, which also does not attract new members. As has become clear from the historical analysis, rising unemployment in the Netherlands often did not lead to resignation of members right away, but first after some time. This suggests that the unemployment variable probably performs better in the model with a lag. Employment. When employment rises the labour market is tightening, which increases the bargaining power of unions; this is supposed to exert a positive impact on union growth. This variable measures just the opposite as the effect of unemployment on membership, and is only mentioned here because it is used in the A & P model, which will be replicated in section 4.4. Trade union density rate. If union density begins to increase from a low level, at first non-members are incited to join as well (enforcement effect). However, after a certain level a further extension of the labour movement becomes ever more difficult (saturation effect). Therefore, it is best to incorporate this variable in two forms: linear and quadratic. In the Netherlands, the union density rate has never been very high; before World War II the average percentage lay at 33% in the 1930's, which was also the case in most other countries. After the war the density rate rose quickly to 43% in 1950, but subsequently stagnated completely; after 1980, it fell back to the pre-war level. The initial rapid increase may be associated with the enforcement effect, while the long-lasting stagnation might be related to the saturation effect, but given the relatively low level of the density rate this assumption can only be made with much caution. Institutional variables. The percentage of leftist parties in parliament can serve as a proxy for the extent of pro-labour sentiment. Alternatively, it is also possible to examine the impact of a certain law on union membership, by using a dummy variable that takes the value 1 during the years that such a law is in force.
86
The political climate has played a very important role in the determination of union growth and decline in several countries. However, Dutch union goals and activities were more hindered by employers than by laws or certain political parties. Owing to the pillarization, all major parties maintained close relations with their allied union federations. The Labour Party (SDAP, later PVDA) made the most efforts to emancipate and benefit the working class, which might have had a stimulating effect on union membership rates. As far as legislation is concerned, the already mentioned 1937 Act on extension of labour agreements created the possibility of free riding among workers. Ever since the end of the war the implementation of this law made that the majority of the dependent labour force could profit from the unions' bargaining efforts without the need to be a member. Strikes. (Successful) strikes often induce workers to join a union, not least in order to receive strike pay; this membership accretion is called "war profits", and it is the difficult task of the union to hold on to these members. In the Netherlands, the number of strikes and of workers involved was much higher before than after the second World War. Directly after the liberation all emphasis was voluntarily put on rebuilding the economy, in which there was no place for counterproductive labour disputes. Only the EVC did not go along with this co-operative attitude and organized a lot of strikes in the first post-war years. Thereafter the EVC lost its power, and the Dutch labour relations became characterized by a large degree of industrial peace. Social benefits payments. The increasing government expenditures on social welfare can be regarded as a substitute for the social security insurances that unions used to offer, thereby reducing the attractiveness of union membership. The extension of the Dutch welfare state resulted in a huge growth of the social benefits payments by the government. As a percentage of the net national income, this rose from 4% in 1946 to 33% in 1990.
87
Table 4.1 Union membership developments in the Netherlands 1910-1989 (ΔMt = annual % change) (based on end-of-year data) year
ΔMt
year
ΔMt
year
ΔMt
year
ΔMt
1910
27.6
1930
8.4
1950
4.8
1970
2.5
1911
6.2
1931
15.0
1951
-10.0
1971
1.3
1912
8.3
1932
5.9
1952
3.4
1972
0.9
1913
26.1
1933
-4.3
1953
4.4
1973
2.9
1914
2.6
1934
-5.8
1954
3.5
1974
1.0
1915
10.7
1935
-3.3
1955
3.1
1975
2.3
1916
18.8
1936
-0.5
1956
6.1
1976
-0.2
1917
19.9
1937
2.8
1957
-0.3
1977
1.6
1918
22.8
1938
3.8
1958
0.2
1978
0.6
1919
33.3
1939
3.9
1959
1.4
1979
0.5
1920
-4.9
1940
-9.6
1960
3.3
1980
-1.5
1921
-2.6
1941
-1.1
1961
1.4
1981
-2.2
1922
-11.1
1942
.
1962
1.2
1982
-2.3
1923
-10.3
1943
.
1963
2.4
1983
-4.1
1924
-4.1
1944
.
1964
-0.8
1984
-2.8
1925
-0.7
1945
.
1965
1.6
1985
-1.3
1926
2.3
1946
20.7
1966
3.4
1986
-0.3
1927
3.2
1947
9.0
1967
0.5
1987
1.2
1928
6.7
1948
5.8
1968
0.7
1988
2.3
1929
13.2
1949
4.7
1969
-0.2
1989
3.5
Source: CBS (1959) 60 jaren statististiek in tijdreeksen (concerning 1910-1912); J. Visser (1989), European trade unions in figures (concerning 1913-1941); B.O. Ebbinghaus & J. Visser (1993), Trade union systems in Western Europe; data handbook volume 1 (concerning 1946-1989).
88
4.3 The Dutch data and their peculiarities Given the expectations that follow from the theoretical and historical literature, as many relevant explanatory variables as possible have been collected, for a period as long as possible. Unfortunately, the data are far from perfect for three reasons. Firstly, about a few variables no statistical information exists at all, a question which cannot be redressed. Secondly, some variables are only recorded from a certain year onwards, or are stopped from being recorded after a certain year; this problem can only partly be obviated by approximating the missing values. Although reliable membership figures are available from 1907 onwards, some of the important explanatory variables start off from a later date. Furthermore, hardly any statistics are available concerning the war years (19401945). Thirdly, some important variables suffer from changes in definitions, which obstructs creating consistent time-series. It is tried to solve this problem as good as possible. In effect, this analysis employs the following measurements of the variables mentioned in section 4.2. In the appendix of this chapter (table 4.10) all series are shown in full, with mention of definition and source. The dependent variable, the rate of change in union membership (ΔMt), is largely based on the series compiled by Visser and Ebbinghaus, regarding the period 1913-1989; the years 1907-1912 are taken from the Central Bureau of Statistics (CBS). Visser made a few meaningful adjustments to the existing CBS statistics, making the series more complete and consistent. The inflation (rate of change in consumer prices, ΔPt) is derived from the cost-of-living indices of the CBS, and covers the longest period of time, from 1901 continuously up to and including 1989. This series is not completely consistent though, as the base year of this index has changed several times through the century, just like all the other indices that are employed in this chapter. The longest more or less consistent series on wages concerns the gross nominal hourly wages of male industrial workers, which first starts in 1926. On the basis of given wages in the metal- and mining industry from 1910 onwards, the series has been extrapolated backwards to 1910. As it is assumed that workers react to wage changes, this variable is transformed to the rate of change in wages, ΔWt. As it is conceivable that, certainly in the post-war period, the real wage development (ΔRWt) might have been influential, this variable was also constructed on the basis of ΔWt and ΔPt. The measurement of the unemployment, in absolute numbers (Ut) as well as in a percentage of the dependent labour force (U%t) causes the most problems. The pre-war figures of the CBS are unreliable, while the post-war series suffers from several changes in the definition of unemployment. A thorough investigation of the pre-war statistics by Den Bakker & Van Sorge (1991) has led to a better unemployment series for the years 1920-1939, while the numbers for 1911 to 1919 have been calculated on the basis of Mitchell (1975). A related problem regards the data on the dependent labour force, which also underwent several definition changes. The trade union density rate (taken up in a lagged form: Dt-1) is based on that figure, because it is defined as the ratio of total union membership and the dependent labour force. The longest consistent workforce figure is calculated by Visser, and runs from 1913 to 1986, the war period excepted. It has been impossible to supplement this series with the aid of recent CBS-statistics, as the numbers differ too much. This implies that a model which includes Dt-1 can only cover the period until 1986.
89
To replicate the A & P model, data had to be collected with respect to the level of employment in the "highly unionized sectors". Given the scarce availability of sectoral employment over a long time-span, it was best to construct a series of employment in the industrial sector (transformed to the rate of change: ΔE), which unfortunately only starts in 1922. Turning to the institutional determinants, a variable is compiled which measures the proportional share of leftist parties in Parliament (Leftt). This information is available for the whole period, except for the years 1940-1945. Next, it has been considered to incorporate the effect of the 1937 Law on extension of labour agreements into the modelling, but that idea has been abandoned. It has no use to create a dummy variable with value one for all years since 1937, because this law was first really administered during the German occupation, and from 1945 by the new Dutch government. As will become clear later on in this chapter, it is necessary to estimate a pre- and post-war model separately; as the 1937 law was in force during the whole post-war era, its effect on union membership rates can therefore not be measured directly. The strike variable (St) is taken from the CBS statistics and measures the number of strikes per year. The series is consistent and covers the whole period, other than the war years. Finally, information has been gathered about the total sum of social benefits payments by the government, which is set off against the level of the net national income. The resulting variable (SOCt) also runs from 1900 to 1989, except for the period 1940-1945. The most important variable of which a series can not be computed, concerns the changes in the union dues. As has been mentioned earlier, the subscription fees widely differed between and also within unions. Even if all the information would have been available, it still would make not much sense to construct one series of aggregate figures, given the large differentiation. To round off this overview, some simple statistics of all employed variables are shown in table 4.2. Table 4.2 Simple statistics of the variables over the period 1900-1989 Variable ΔM ΔP ΔW ΔRW U U% D E Left S SOC t
t
t
t
t
t
t-1
t
t
t
N
Mean Std Dev Min.
78 3.48 8.16 -11.07 90 3.33 5.44 -13.29 78 6.32 8.41 -13.64 78 2.79 6.38 -8.41 73 217.04 234.07 20.70 71 6.55 6.19 0.60 68 34.67 6.78 16.87 62 1.04 4.21 -13.08 84 33.01 8.65 15.00 86 136.94 122.18 8.00 85 11.61 11.99 0.43
Max. 33.26 19.01 50.00 37.58 822.40 22.50 42.96 15.40 46.70 649.00 35.56
Label % change in membership % change in consumer prices % change in nominal wages % change in real wages level of unemployment (in 1,000's) % unempl. of the dep. labour force lagged density rate % change industrial employyment % share left in Parliament number of strikes social benefits in % of nni
Two aspects of the Dutch data must be taken into consideration carefully, before the actual estimations can take place. Firstly, there are hardly any observations during the second World War; in other words, we have to deal with "missings" in the time-series. Secondly, as from the year 1951 the CBS omitted the EVC membership from the statistics; this led to a sudden and large decline of the annual rate of change in union
90
membership in 1951 (see table 4.1, and for more details chapter 3). This implies that the observation in 1951 obtains an extreme value, namely -10%. Both peculiarities in the data should be dealt with, in order to derive reliable estimates in the modelling. Correcting for the extreme value in 1951 is relatively easy. Because the dependent variable is measured in first differences, the most obvious solution is the introduction of a dummy that obtains the value one for the year 1951 only. Alternatively, we can either start our analysis in the year 1952, or we can subtract the EVC membership figures of 1945-1950 from the total number. At first sight, choosing for the option of including a dummy variable for the year 1951 seems the best solution. In this way, the maximum number of observations can be taken into account, contrary to the second option, which takes 1952 as starting year. In the third alternative, also the maximum number of observations is possible; but excluding a group of about 163,000 members from the analysis seems quite a radical step, because this made up approximately 13% of the total number of union members. The occurrence of the missing observations constitutes a larger problem. It means that the time-series is not continuous, which in turn implies that it is difficult to fully exploit the time-series nature of the data. When using for instance autoregressive models, it is assumed that variable values in subsequent years are related. Therefore, each interruption in the series causes an additional loss of observations that can be used to estimate the model. In particular the first few observations (depending on the order of the autoregressive model) after the interruption in the time-series will not be available for analysis. Furthermore, if observations are missing over a fairly large coherent subperiod this may also affect the efficiency of the model estimates, which makes it very hard to determine a correct model specification. The Ordinary Least Squares (OLS) method, on the contrary, starts from the assumption that the observations in all periods are independent drawings from an identical distribution. On the one hand this means that the specific time-series structure of the data cannot be exploited when using OLS, but on the other hand it also means that breaks in the time-series will not affect the efficiency of the model estimates, presuming that all OLSassumptions are met. After the estimation of any model it is necessary to analyze the residuals (i.e., the difference between the actual and the estimated value of the dependent variable), to verify whether these OLS-assumptions hold true, so that the model specification is correct. The three most relevant OLS-assumptions are the following (see D.L. Harnett (1982), pp. 487-494, for a detailed discussion of the assumptions with regard to the application of OLS in the linear statistical model). Firstly, the residuals are assumed to be statistically independent from each other, otherwise the model outcome points at autocorrelation. Secondly, the residuals are assumed to have a constant variance, otherwise the model outcome points at heteroscedasticity. And thirdly, the residuals are assumed to be normally distributed. All three assumptions have to be met in order to pronounce upon the significance of the parameter estimates. As will be shown in the next sections, all models that estimate the whole period produce residuals that are not identically distributed. More specifically, residuals have a much larger variance before World War II than thereafter. This points at a specific form of heteroscedasticity, namely that the sample can be partitioned into two subsets of observations, each subset corresponding to a different error variance. The error variance is assumed to be constant within each subset, so that the complete model has only two
91
unknown variances (see for the several forms of heteroscedasticity: Judge et al. (1988), p. 356 f). To determine these variances and to investigate if they really do differ significantly, the same model must be estimated for the pre- and post-war period separately; after that the resulting residuals of these separate models should give the decision. If it turns out that within each subperiod the residuals are independent with an identical variance, then the problem with regard to the estimation of the whole period is just heteroscedasticity as specified above, which can be corrected for.43) A formal test on heteroscedasticity can be found in Judge et al. (1988, pp. 362-363), who use the following test statistic: MSE1/MSE2 ~ F(n1-k,n2-k), where n1 and n2 are the number of observations in the two subperiods, and k stands for the degrees of freedom, while the Mean Squared Error (MSE) stands for the variance of the residuals. The problem of heteroscedasticity can be solved by means of transformation: the observations before the war are divided by the root Mean Squared Error (root MSE: the standard deviation of the residuals) that was found in the pre-war model, while the observations after the war are divided by the root MSE that was found in the post-war model. The result of this operation will be that we have created homoscedasticity, so that this OLS-assumption is met again.44) The final question that can be raised after this rather technical exposition is whether it is wise to try estimating one model which covers the whole period. If it serves the purpose of replicating the A & P and B & E models, the answer can be affirmative. If we restrict ourselves to the Dutch case, and judge on the basis of the large differences before and after the war not only in the outcome of the modelling, but also consequent on the historical description, there is much to be said for the construction of two separate models. The fact that the structure of the residuals before the war is so different from the one after the war, is a first indication that the estimations should be based on two separate subperiods. A formal test on the occurrence of a structural break can not be performed on the Dutch data. According to Godfrey (1990, p. 141) "Chow's tests will not be robust to violations of the assumptions that the disturbances are NID (0,σ2) variates45). Autocorrelation, heteroscedasticity and non-normality will all affect the validity of the tests to some degree". Instead, the coefficients of the two submodels can be compared with each other, to check whether these are stable or not. If the latter case holds, this will further support the argument in favour of separate modelling. One last tentative indication of separate modelling is shown in table 4.3, in which the same simple statistics are given as in table 4.2, but now with respect to the two separate periods. When comparing these two periods, the relative differences between the means of the variables within each subperiod are very distinct.
43)The absence of heteroscedasticity in each subperiod does not yet imply that autocorrelation is absent as well; this still has to be checked. 44)A full description of this estimation method, also known by the term EGLS (Estimated Generalized Least Squares) can be found in Judge et al. (1988), pp. 361-365. 45)This concerns the standard OLS-assumptions, namely Normally, Indepently Distributed, with expectation 0 and an identical variance of σ . 2
92
Table 4.3 Simple statistics of the variables over two subperiods
93
4.4 The models of Ashenfelter & Pencavel and Bain & Elsheikh It can be useful and interesting to check whether Dutch trade union growth and decline can be explained by the same determinants that have influenced American and British union membership developments. To that end, the A & P model and the B & E model are replicated on the Dutch data. Because of the missings during the period 1940-1945, the estimations will be performed by means of OLS, which A & P and B & E also used. The A & P model seeks to explain the annual percentage change in American trade union membership by price changes (ΔPt), employment changes in the highly unionized sectors (ΔE, ΔE-1, and ΔE-2), the unemployment rate (U%t), the union density rate (Dt-1) and finally the percentage of Democrats in the House of Representatives (Leftt). The B & E model (on the U.K.) uses the same dependent variable, but differs somewhat in the choice of explanatory variables; next to price changes and the union density rate, they use the unemployment variable in a different way (Ut-1 combined with Ut-2), and finally they include a wage variable (ΔWt). When these two models are applied to the Dutch data, it is always necessary to correct for the extreme value in 1951; therefore a dummy is incorporated. Trying to replicate the A & P model, the first step consists of estimating a model which includes all the variables that A & P employ. The plot of the estimated residuals for the A & P replication is shown in graph 4.1. Graph 4.1 Residuals of the replicated A & P model
This plot clearly demonstrates the occurrence of a widely different error structure before the war, against the one after the war. In particular: the errors tend to be larger in the prewar period. The next step then is to apply the procedure described in section 4.3: the same parameters are estimated in separate pre- and post-war models, and after that the residuals are analyzed again. In the first two columns of table 4.4 the results of the 'submodels' with respect to the A & P specification are shown.
94
Table 4.4 Replication of the A & P model and the B & E model on Dutch data, two subperiods (t-values in brackets) A & P model 1924-1939 2
adj. R D.W. Root MSE
B & E model
1949-1986
0.522 1.745 4.184
1914-1939
0.753 1.894 1.433
0.537 0.995 7.245
16
Constant
-56.642** (2.021)
17.272 (1.413)
24.069** (2.700)
-0.645 (0.043)
0.718 (1.091)
-0.073 (0.773)
-0.117 (0.289)
-0.190 (1.033)
DummyΔPt
0.036 (0.005)
0.288 (0.267)
ΔWt
0.611*** (3.187)
0.022 (0.255)
ΔUt-1
0.019 (0.617)
-0.014* (1.824)
ΔUt-2
-0.016 (0.564)
0.010 (1.472)
-0.798** (2.253)
-0.083 (0.228)
ΔE
-0.736 (1.781)
0.198 (1.581)
ΔE-1
0.023 (0.082)
-0.012 (0.081)
ΔE-2
0.393 (1.470)
0.202* (1.997)
U%t
-0.537 (1.722)
-0.576*** (2.930)
Dt-1
-0.112 (0.153)
-0.463 (1.660)
Leftt
2.193** (2.511)
0.116 (1.311)
Dummy51
-10.676*** (5.795)
26
0.597 1.369 1.830
# obs.
ΔPt
38
1949-1986
38
-11.180*** (5.123)
Explanation of the symbols: ΔPt: % change of consumer prices DummyΔPt: dummy with value 1 if inflation exceeds 4% ΔWt : % change of gross nominal wages ΔE-1 : % change of employment in industrial sector Ut-i : level of total unemployment (in 1,000's) Explanation of the asterisks: U%t : unemployment in % of the dep. labour force Dt-1 : union density, lagged with 1 year * significant at 10% level Leftt: % share of left parties in Parliament ** significant at 5% level Dummy51: dummy with value 1 for the year 1951 *** significant at 1% level
95
Regarding the first two columns of this table, several conclusions can be drawn. Firstly, the formal test statistic on heteroscedasticity produces an F-value of 8.52, while F(8,29) in the F-table yields a value of 3.20; this implies that the null hypothesis (MSE1 = MSE2) is rejected, meaning that heteroscedasticity is present. Secondly, comparison of the results of the two subperiods teaches that the coefficients are not stable over time; except for the unemployment variable, all other coefficients yield widely different values; this points at a structural break. Moreover, neither in the pre-war model nor in the post-war model are many significant coefficients found, suggesting the wrong model specification in both subperiods. Finally, on the basis of a histogram of the estimated residuals there is no reason to call a normal distribution of the error terms in the two subperiods in question; in addition, these errors are independent as well: the autocorrelation function suggests that (higher order) autocorrelation is absent. This implies that the A & P model can be estimated over the whole period using OLS, after a correction for the heteroscedasticity, by means of transformation of the variables (the description of this procedure was given in the preceding section). The resulting transformed model now meets the requirements with respect to the residuals, so OLS in the transformed model yields unbiased and efficient estimators. This time the outcome will show reliable coefficients and t-values (indicating the degree of significance). Only the Durbin-Watson (DW) statistic, indicating the extent of autocorrelation in the residuals, is still not reliable: due to the missings the computed DW is incorrect. However, since it is established that the A & P specification does not hold for the two subperiods and that the model is characterized by a structural break, it does not make much sense to discuss the results of the replication with respect to the whole period, not even if the heteroscedasticity is corrected for. Therefore these results are only taken up in the appendix, from which it becomes clear that the specific combination of variables that could explain American union growth, is not able to explain Dutch union membership developments very well. Turning to the B & E model, the following results are obtained. After the first estimation of the whole period with the variables that draw most closely to the ones used by B & E, the residuals were analyzed again. The outcome of this procedure is shown in graph 4.2. Graph 4.2 Residuals of the replicated B & E model
Again it can be detected that the residuals are not identically distributed, because the variance is much greater before, than after the war. Subsequent estimation of the two subperiods (shown in the third and fourth column of table 4.4) again confirms that the deviant error structure is caused by heteroscedasticity: using the Mean Squared Errors of
96
the two submodels, the test statistic gives an F-value of 14.79, while F(19,30) yields a value of 2.14. Comparing the coefficients of the pre-war model with those of the post-war model, shows that none of the coefficients is stable, which clearly indicates the occurence of a structural break. The small number of significant estimates, be-fore as well as after the war, leads to the conclusion that the model specification of B & E, just like A & P's, is not suitable for the Dutch data. This is remarkable so far as that B & E have concluded that their model containing four basic variables (ΔP, ΔW, U, D), with or without lags, sometimes supplemented with a political variable, can be applied to no fewer than four countries, namely Great Britain, the United States, Australia and Sweden (B & E, pp. 105-114). Our result implies that a correction for heteroscedasticity, followed by a replication of the B & E model to estimate Dutch union membership fluctuations during the whole research period, will not lead to a meaningful outcome. Nevertheless, for the sake of comparison the results of the uncorrected and corrected model covering the whole period are taken up in the appendix (table 4.11). 4.5 New modelling A new Dutch model should be based on the theoretical assumptions that spring from the international literature, taking the specific developments of the Dutch trade union movement into consideration as well. In sections 4.2 and 4.3 all the relevant and available variables have been discussed in detail, so here it will be tested to what degree these assumptions are met by the estimation results.
4.5.1
Building a model covering the whole period
To begin with, an attempt will be made to estimate a model which explains the membership fluctuations through the whole period; the main reason to do this is to examine the correctness of the model specification. On the basis of the replication results in section 4.4, it can be suspected that the error structure does not meet the assumptions with regard to the application of OLS. Including all the relevant variables yields estimates which are shown in table 4.5, in the first column. Only the coefficient of ΔPt is insignificant, that is why the second equation includes ΔPt-1 instead (there are no strong theoretical indications whether to use the price variable with or without a lag), which does yield a significant coefficient. In the third column the least significant variable, Ut, is left out but this does not improve the model. Therefore, the second model is selected as the best alternative because it obtains all theoretically important variables, all of their coefficients with the expected signs and mostly high t-values.
97
Table 4.5 Determinants of union growth 1914-1986 (t-values in brackets) (1) 2
(2)
(3)
adj. R D.W.
0.675 1.077
0.696 1.052
0.689 0.993
Root MSE
4.154
4.014
4.062
# obs.
65
65
Constant
14.443 (1.826)
ΔPt
0.107 (0.776)
ΔPt-1
65 16.086 (2.171)
14.777 (1.984) % change of consumer prices
0.299 (2.111)
0.308 (2.145)
% change of consumer prices, lagged with 1 year
0.306 (3.180)
% change of gross nominal wages
ΔWt
0.432 (4.229)
0.354 (3.535)
Ut
0.017 (1.647)
0.015 (1.525)
Ut-1
-0.023 (2.306)
-0.021 (2.164)
-0.007 (2.173)
level of unemployment, lagged with 1 year (in 1,000's)
Dt-1
-0.555 (4.551)
-0.552 (4.869)
-0.548 (4.781)
union density, lagged with 1 year
Leftt
0.359 (1.581)
0.326 (1.485)
0.347 (1.565)
% share of Left in Parliament
St
0.019 (1.689)
0.017 (1.657)
0.020 (1.940)
number of strikes
St-1
-0.034 (3.967)
-0.035 (4.192)
-0.034 (4.031)
number of strikes, lagged with 1 year
SOCt
-0.390 (3.612)
-0.422 (4.016)
-0.376 (3.691)
social benefit payments as a % of nni
Dummy51
-11.859 (2.690)
-13.395 (3.126)
-13.313 (3.070)
dummy with value 1 for the year 1951
level of total unemployment (in 1000's)
The residuals of this second model are plotted in graph 4.3 and from this plot it becomes clear that the estimated residuals are not identically distributed. Graph 4.3 Residuals of the Dutch model, 1914-1986
98
The next procedure is to estimate the same parameters (of the second equation) in two separate models, covering the pre- and post-war period respectively. Analysis of these results can give decisive answers about the presence of a structural break, heteroscedasticity, and/or autocorrelation. The results are shown in table 4.6. Table 4.6 Determinants of union growth, before and after the war (t-values in brackets) 1914-1939 2
1948-1986
adj. R D.W. statistic
0.867 1.986
0.803 2.001
Root MSE
3.883
1.310
# observations
26
39
Constant
***
90.800 (3.643)
21.636* (1.854)
ΔPt-1
0.269 (1.341)
0.048 (0.532)
ΔWt
1.074*** (4.388)
0.036 (0.634)
Ut
0.045* (1.948)
-0.011* (2.188)
Ut-1
-0.099*** (4.852)
-0.001 (0.282)
Dt-1
-0.044 (0.918)
-0.578* (2.124)
Leftt
-2.026** (2.522)
0.153 (1.604)
St
-0.066* (2.080)
0.015* (2.038)
-0.075*** (3.798)
0.003 (0.605)
3.980 (0.769)
-0.180** (2.672)
St-1 SOCt Dummy51
-12.058*** (7.242)
Calculating the test statistic on heteroscedasticity (MSE1/MSE2), a value of 8.79 is found, while F(17,29) yields a value of 2.67. This confirms the presence of heteroscedasticity. On the contrary, further analysis of the autocorrelation function does not give rise to assume the presence of autocorrelation in the errors within the two subperiods. There are however strong signs of a structural break. Firstly, the root MSEs of the two submodels are very different. And secondly, the estimated parameters are clearly not stable over time; half the coefficients obtain the opposite sign, while the other half show a very different value.
99
It is remarkable that in two other studies on Dutch membership fluctuations, by Visser (1987) and Hendriks (1986), also the whole period is analyzed by means of just one model, without any mentioning of a heteroscedasticity problem. Working with, by and large, the same data, they presumably would have found about the same error structure as in this research; but they probably did not check this out. Using OLS, both researches yield low DW statistics. Hendriks does comment on this finding, but he does not draw the conclusion that it means that the significance of the estimates cannot really be interpreted anymore. Although both authors do mention the possibility of a structural break in the Dutch data, their emphasis lies on the interpretation of an (uncorrected) model which covers the whole period. The following question then becomes whether we have to proceed with one model, corrected for heteroscedasticity; or whether it is better to start directly with estimating the pre-war and post-war period separately. The latter has been chosen, because there are enough indications which justify this decision. From the historical analysis in chapter 3 it has become clear that the working - and living conditions of Dutch workers before World War II differed much from thereafter; and in the following discussion of the relevant variables (in sections 4.2 and 4.3) it was already noted that some variables probably had more impact on union membership developments before the war, while others were assumed to be more influential after the war. This was not the result of one specific event such as a new law or institution. Much more, the causes must be sought in the combined effect of the economic depression of the 1930's and of the destructions during the war. When the Netherlands were liberated, the ideas already existing of rebuilding Dutch society (not just economically, but also institutionally) could be put into practice. When comparing pre- and post-war Holland with respect to our research object, it can be stated that after the second World War, gradually, security for the workforce increased tremendously. Price increases were compensated through wage increases, collective labour agreements were usually extended to a whole branch of industry, protection against discharge was greatly increased by law, and a variety of social benefits prevented the unemployed, disabled and retired workers from being thrown into poverty. All this, and more, has changed the position of the workers. Most probably, these changes must have had their impact on the decision making process whether or not to join, to leave, or to stay in the union. It is conceivable that the considerations to become a member before the war, were different from those after the war. In any case, afterwards the union movement experienced a much more stable growth pattern than in the pre-war era. Therefore, the remainder of this chapter will be dedicated to the construction of two separate models, specified according to the theoretical expectations with regard to the determinants of pre- and post-war unionization, respectively.
100
4.5.2 The pre-war model Section 4.2 dealt at length with the question which variables were expected to be of influence on union growth and decline in the Netherlands, and whether these determinants would be of importance during the whole century, or just in one of the two subperiods. In short, it is assumed that the rate of inflation, wage changes, the level of unemployment, the number of strikes and the proportional share of leftist parties in Parliament might have played a role before the second World War. The remaining candidate variables are supposed to have been more influential in the post-war era, and will not be included in the pre-war model. These regard the union density rate, because a saturation effect is not likely to have occurred in the pre-war Netherlands, since the density rate was rather low at that time; and the social benefits payments by the government, because the radiating substitution effect could only unfold after the war, when the state gradually provided ever more social benefit payments. Both suppositions are confirmed by the results shown in table 4.6 in the preceding subsection, where Dt-1 and SOCt remained insignificant in the pre-war model containing all variables. Beforehand, it was assumed that inflation will exert a stimulating (threat) effect on union growth, just as pay rises will lead to a stimulating (credit) effect. Visser and Hendriks both found that ΔP and ΔW could not be combined in one and the same model due to multicollinearity. In general it is said that if the correlation coefficient between two variables is greater than 0.9, multicollinearity may be severe. In this research, between ΔPt and ΔWt a correlation coefficient of 0.63 is found in the pre-war period (and between ΔPt-1 and ΔWt a coefficient of 0.84); this points to little problems with regard to the inclusion of both ΔPt and ΔWt. All pre-war correlation coefficients are shown in the appendix (table 4.12). The unemployment variable will be taken up in the model with and without a lag, to test whether unemployment will in first instance (Ut) induce workers to join or to stay in the union in order to be better protected against discharge; if unemployment persists (Ut-1) it is assumed that it will discourage people and prompt them to quit. The strike variable will also be incorporated with and without a lag: there are indications that strikes might have incited workers to join a union at first (St), but that they changed their mind later on (St-1) if the immediate cause of the labour dispute was solved, or before World War II because there was more danger of employer retaliation. Finally, the variable Leftt is included to estimate if, and to what degree, the influence of leftist parties in the Second Chamber had a positive impact on unionization. In table 4.7 the results of the pre-war modelling are shown. Each estimation has used 28 observations, which means that the research period covers the years 1912 up to and including 1939.
101
Table 4.7 Determinants of union growth before the war (t-values in brackets) (1) (2) (3)
(4)
adj. R2 D.W. root MSE # obs.
0.738 1.894 5.636 28
0.712 1.798 5.843 28
0.749 1.809 5.518 28
0.760 1.810 5.392 28
Constant
30.700** (2.302) 0.356 (1.602) 0.629*** (3.296) 0.049** (2.308) -0.079*** (3.592) -0.002 (0.072) -0.067*** (4.967) -0.131 (0.362)
38.157*** (2.945)
26.485 (4.178) 0.384* (1.879) 0.595*** (3.672) 0.047** (2.329) -0.079*** (3.658) 0.002 (0.077) -0.067*** (5.113)
26.834*** (6.169) 0.382* (1.923) 0.603*** (5.157) 0.047** (2.417) -0.080*** (4.017)
ΔPt ΔWt Ut Ut-1 St St-1 Leftt
0.798*** (4.833) 0.040* (1.893) -0.070*** (3.173) -0.009 (0.404) -0.067*** (4.790) -0.330 (0.940)
-0.067*** (5.484)
*significant at 10% level **significant at 5% level ***significant at 1% level
The first column gives the estimates of a model which includes all variables that are supposed to be of importance. As can be seen, not all coefficients are significant. The inflation nearly reaches the 10% significance level, while the parameters of St and Leftt are very insignificant. Dropping one or more of these three in subsequent estimations shows that the other four parameters (of ΔWt, Ut, Ut-1 and St-1) always stay very significant with the expected signs; phrased differently: these four are robust estimates. If St and/or Leftt are removed from the model, the price coefficient becomes significant as well (see equation no. 4). All model specifications come up with a DW value that lies in the so-called inconclusive area: this means the DW statistic does not give a decisive answer about the presence of autocorrelation. With respect to the fourth equation, additional analysis of the autocorrelation function reveals that there is no autocorrelation. A rise in the consumer prices as well as wage increases lead to membership accretion. Especially the latter has a strong impact: a pay rise of 1% point results in an extra growth in union membership of 0.6% point. The outcome regarding the unemployment variables is very interesting: Ut has a significantly positive effect, while Ut-1 has a significantly negative effect. This confirms the
102
theoretical notions. The corresponding coefficients of +0.047% and -0.080%, respectively, imply that eventually the negative impact dominates: in the end persistent unemployment reduces the size of the union movement. A remarkable result is found with respect to the strike variables. The assumption that strikes induce people to join a union does not hold true in the case of the pre-war Netherlands: the parameter of St remains insignificant in all different model specifications. On the contrary, St-1 has a very significantly negative effect on unionization, and it is not possible to get round the robustness of this finding. It means that workers leave the union as a result of increased strike activities in the past. This outcome must be placed against the background of the Dutch society in the first decades of this century. On the one hand, workers enjoyed much less security in those days: strikers could easily be dismissed. On the other hand, partly due to the strong influence of religious beliefs, many people were unfavourably disposed towards strikes as this contravened their principles of co-operation. Lastly, the supposedly stimulating effect of leftist parties on union membership rates, is rejected in all estimations. This might be explained by the fact that owing to the pillarization, all political parties, from left to right, maintained relations with their allied union organizations. In addition, the variable Leftt consists before the second World War almost exclusively of the percentage number of seats of the SDAP, which never became a government party. To round off this subsection, graph 4.4 shows the actual and predicted values of the dependent variable based on the fourth pre-war model. It can be seen that the predicted values follow the actual ones quite reasonably, except for the period 1925-1927. Looking at the data in the appendix, this wrong prediction is probably due to the strong wage growth in the years 1925 and 1926, after a period of rapidly declining wages. It would seem that the model gives too much weight to this recovery. Graph 4.4 Actual and predicted values of the rate of change in Dutch union membership before the second World War
103
4.5.3 The post-war model With respect to the theoretical assumptions that underlie the post-war model, the following considerations have been made. Given the description of historical developments in the Netherlands, it has already been argued that the influence of inflation and wage increases on unionization will probably be much smaller after World War II. This is based on the fact that during the major part of this period, price rises were automatically compensated for by pay rises, while (union negotiated) wage increases applied to all workers, union members and nonmembers alike, owing to the Act on extension of collective labour agreements. As can be seen from the second column of table 4.6 in section 4.5.1, both assumptions hold true: the coefficients of the price variable and the nominal wage variable become very insignificant and obtain much smaller values than before the war. It might still be possible though, that people credit unions for an improvement of the real wages (ΔRWt), and think that joining a union increases its power and with that the chance that more wage growth can be realized. Concerning the impact of unemployment, there is no cause to assume another relationship with unionization than before the war. The same goes for the variables Leftt and strikes. The trade union density rate (Dt-1) will be included in linear and quadratic form, to test to what extent the Dutch union movement has experienced an enforcement and/or a saturation effect. Next, the variable SOCt will enter the modelling to find out if the growth in government expenditures on social benefits has really exerted a substitution effect on union membership. Finally, we must consider the extreme value of the dependent variable in 1951. Since that year, the membership figures of the radical communist federation EVC, are not included anymore. As has been expounded in section 4.3, correction for this one-off outlier can be accomplished by either including a dummy for the year 1951, or starting the estimations as from 1952, or excluding the EVC-members from the analysis. The tables 4.8 and 4.9 show the estimation results of all three alternatives. The corresponding postwar correlation coefficients can be found in the appendix (table 4.13). As becomes clear from table 4.8 below, the first equation obtaining all the theoretically relevant variables does not come up with significant coefficients only. Further variations in the model specifications, which are not shown here, indicate that the real wage variable as well as the lagged strike variable never yield significant parameters. Therefore, these two are left out in the second equation. In this second model the lagged unemployment variable remains insignificant, so it is decided to drop this one as well. Although the first three equations all show the expected signs with respect to the union density variables, their combined effect is not significant. Therefore the final equation only includes the density rate in linear form.
104
Table 4.8 Determinants of union growth after the war,including the EVC (t-values in brackets) (1)
(2)
(3)
(4)
adj. R2 D.W. root MSE # obs.
0.819 1.954+ 1.255 39
0.825 1.888+ 1.233 39
0.852 1.858+ 1.222 40
0.850 1.847++ 1.230 40
Constant
-46.955 (1.149) 0.040 (0.634) -0.012** (2.492) 0.008 (1.226) 3.073 (1.449) -0.048* (1.729) 0.129 (1.384) 0.014* (1.898) 0.005 (0.842) -0.196*** (3.071) -10.979*** (7.039)
-38.531 (0.984)
-8.112 (0.304)
21.674** (2.359)
-0.012*** (2.781) 0.007 (1.117) 2.690 (1.315) -0.043 (1.618) 0.134 (1.499) 0.016** (2.536)
-0.009** (2.633)
-0.010*** (3.031)
1.139 (0.773) -0.024 (1.186) 0.163* (1.954) 0.019*** (3.748)
-0.590*** (2.740)
-0.203*** (3.377) -11.275*** (7.562)
-0.201*** (3.442) -11.389*** (7.816)
-0.179*** (3.212) -11.885*** (8.463)
ΔRWt Ut Ut-1 Dt-1 (Dt-1)2 Leftt St St-1 SOCt Dummy51
0.180** (2.187) 0.019*** (3.772)
*significant at 10% level **significant at 5% level ***significant at 1% level +the null-hypothesis of serial independence is accepted, implying that autocorrelation is absent ++the D.W. value is just in the inconclusive area, but additional analysis of the autocorrelation function indicates that autocorrelation is absent
105
Interpreting the results of this fourth model, which explains the rate of change in total union membership (including the EVC), the following remarks can be made. The dummy, which always obtains the expected significantly negative sign, and the five other variables appear to be quite robust. Contrary to the pre-war model, Ut is only significantly negative without a lag. This would imply that union members in an early stage decide to quit the union in times of unemployment. The fact that before the war members who lost their job received a (higher) unemployment benefit through their union, might explain why people renounced their membership at a later moment than after the war. The rather small coefficient value of -0.01 implies that a rise in unemployment with 10,000 persons decreases the union growth by 0.1% point; so, only if unemployment really shoots up, this has a serious downwards effect on union membership. This was the case in the 1980's. The outcome of the first three models with respect to the union density rate tentatively suggests that there has been an enforcement effect as well as a saturation effect, but the rather low t-values make this interpretation unreliable. Including only Dt-1 in the model, yields a robust significant estimate with a negative sign, supporting the saturationist assumption: further growth of the union movement is hampered by its own size. This is quite remarkable, given the relatively low union density rate that the Netherlands have always had. According to Visser (1990, p. 73) this finding must be ascribed to the rather small recruitment efforts of Dutch unions as a result of the high degree of institutional security: they were accepted as negotiating and consultation partners at the macro and meso level anyway, irrespective of their size. Also contrary to the pre-war model, the variable Leftt now does seem to have been a stimulating determinant of union growth. The coefficient value of +0.18 implies that if the proportional share of leftist parties in Parliament rises by 1% point, this leads to an additional growth in union membership by 0.18% point. This finding could be related to the increase of a general leftist sentiment in the Dutch society, which was also reflected in the governmental power of the Labour Party in about half of all post-war years. Ever more people sympathized with the ideas of the trade union movement, which had also become less radical (at least as far as the largest federation, the NVV, was concerned). Where the pre-war model came up with a significantly negative sign of St-1, the post-war model shows a robust outcome with respect to St, but now with a significantly positive effect on union membership. The lagged strike variable obtains insignificant parameters in all estimations. Hence, in the period after World War II strikes had quite another impact on workers; according to the coefficient value of +0.019, a year in which there are 100 more labour disputes than in the preceding year, leads to an additional membership rise of almost 2% point. In the discussion of the next table we will come back to the interpretation of this result. The last robust finding concerns the social benefits payments. The negative value of SOCt shows that the extension of the social security system indeed caused a substitution effect. Between the nineteen fifties and eighties, the social benefits payments increased from about 10% of the net national income to over 30%, a rise of 20%. The value of SOCt is -0.18; this suggest that the growth of the union membership now is 0.18 x 20% = 3.6% point lower than would otherwise have been the case. For the sake of completeness it can be noted that the addition to this fourth equation in table 4.8 of ΔP (with or without a lag), ΔW or ΔRW successively, does not lead to any improvement of the model. All added variables do not exert a significant effect, confirming the expectations with regard to these variables. The models in table 4.9 explain union membership fluctuations without taking the EVC into account. The first two model specifications show estimates over the period 19521986, while the third and fourth equations have estimated the period 1947-1986, after
106
having subtracted the EVC members from the total number of members. In essence the results do not differ very much from those of table 4.8, which is why only two model specifications of both estimation periods are shown. Table 4.9 Determinants of union growth after the war, excluding the EVC (t-values in brackets) 1952-1986
1947-1986
(1)
(2)
(3)
(4)
adj. R2 D.W. root MSE # obs.
0.688 2.102+ 1.181 35
0.694 1.989+ 1.199 35
0.857 2.146+ 1.095 40
0.844 2.077+ 1.144 40
Constant
25.456** (2.607) 0.095 (1.360) -0.011*** (3.156) -0.688*** (3.111) 0.177** (2.093) 0.009 (1.082) -0.158*** (2.838)
28.559*** (2.965)
24.478*** (4.264) 0.109* (2.033) -0.011*** (4.809) -0.663*** (5.141) 0.173** (2.320) 0.008 (1.342) -0.155*** (3.111)
21.771*** (3.731)
ΔRWt Ut Dt-1 Leftt St SOCt
-0.013*** (3.691) -0.741*** (3.352) 0.159* (1.857) 0.012 (1.516) -0.150** (2.670)
-0.011*** (4.488) -0.570*** (4.523) 0.146* (1.908) 0.011* (1.813) -0.140*** (2.712)
*significant at 10% level **significant at 5% level ***significant at 1% level +the null-hypothesis of serial independence is accepted
107
Again, the variables Ut, Dt-1, and Leftt are robust and obtain the expected signs. However, the variable St is not so robust anymore. Inclusion of the variable ΔRWt reduces the significance of the strike parameter; in equation no. 3 it is even the latter variable that obtains a small significant coefficient. This result can not be attributed to multicollinearity between these two variables. Comparing this with the findings in table 4.8, it can be concluded that strikes have a smaller effect on union membership growth after 1951 or after 1946, when the EVC members are excluded from the analysis. This suggests that strikes have primarily been an important determinant in the membership decision for workers choosing for the EVC. This is not inconceivable, because the EVC was the only trade union federation that was actively involved in organizing strikes during the first after-war years. Otherwise the post-war Netherlands are characterized by a relative low number of labour disputes, which makes the small effect of the strike variable understandable. There exists only one other time-series analysis with respect to post-war Dutch unionization, namely the research by Van Ours (1992), who explains the fluctuations in the FNV-membership between 1961 and 1989. In this analysis only the variables unemployment, labour income ratio (to measure the share of real wages in the national product), union density rate and strikes are employed. A good comparison between his and this research is difficult to make, given the fact that different dependent variables are used and that in his research fewer independent variables are included. Van Ours' findings confirm the significantly negative effect of unemployment and the density rate. On the other hand, he finds a significantly positive effect of the labour income ratio, while the strike variable turns out to be insignificant. This outcome is partly supported by the results in table 4.8, which show that inclusion of both St and ΔRWt does not lead to significant coefficients. But ΔRWt and the labour income ratio are not identical, and it is doubtful whether this latter variable should be incorporated: individuals may credit a union for improving their real wages anyhow, even if at the same time the wage share of the total labour force in the national product remains constant or declines. This subsection is also brought to a close by presenting a graph which shows the predictive power of the most adequate model, now with respect to the explanation of post-war unionization. Although the results of the tables 4.8 and 4.9 are not identical it is hard to say which model is really the best; the differences are not that large. Here the results of the fourth model of table 4.8 are chosen to be used, as this contains information about all Dutch union members: even though the membership figures of the EVC are not fully reliable, its size was certainly appreciable. A study of graph 4.5 teaches that the chosen model predicts very well; only in three separate years the actual and predicted values differ somewhat, namely in 1957/58 (when dissatisfaction about union policies led to a much smaller actual growth), in 1964 (probably for the same reason, but less clear), and in 1975 (when the NVV as well as the government actively supported the interests of the lower paid workers, leading to membership growth). The model is not able to explain the distinct impact of specific union or government policies on union membership fluctuations.
108
Graph 4.5 Actual and predicted values of the rate of change in Dutch union membership after the second World War
actual values _ _ _ predicted values (based upon table 4.8, equation 4)
4.6 Conclusions The objective to build a time-series regression model that can adequately explain Dutch union membership developments, is not realized by replicating the exemplary models of Ashenfelter & Pencavel or Bain & Elsheikh. This attempt fails for two reasons. Firstly, the model specification of A & P and B & E does not fully meet the assumptions that can be made with respect to the explanation of union growth in the Netherlands. And secondly, because the Dutch data clearly point at a structural break around the second World War, it is not possible to find a single relationship explaining the whole period. Therefore, two subperiods are analyzed instead of one. The resulting models of the preand post-war period respectively, are quite satisfactory. Before the war, we have less observations and much more fluctuation in the union membership figures, which both make a good fit more difficult. The root MSE, indicating the degree in which the predicted values of the dependent variable differ from the actual values, is very high, but this is due to just a few large misfits in the prediction. As can be judged from graph 4.4, on the whole the pre-war model predicts the actual peaks quite well. The post-war model, consisting of five or six (if we count the dummy in) significant variables, performs very well: graph 4.5 shows that the predictive power of the model is very high. The adjusted R2, indicating the degree in which the variation in the dependent variable is explained by the variations in the explanatory variables, is high as well, compared to those of the A & P and B & E models. The very different outcome of the pre-war model as compared to that of the post-war model lends strong support to the decision to estimate the two subperiods separately. Before the second World War, Dutch unionization was predominantly stimulated by price and wage increases, and obstructed by the unemployment (although this increased
109
the membership in first instance) and by strikes. After the war, unionization was stimulated by strikes (though not in such a strong degree) and by the impact of leftist political parties, while the unemployment, the density rate and the rise in social benefits payments exerted a negative impact on union growth. This conclusion will be rounded off with discussing the question as to what extent this time-series analysis contributes to the explanation of union growth and decline; the findings will be interpreted in terms of demand and supply according to the theoretical framework of this thesis. Most of the theoretical expectations formulated in section 4.2 are confirmed by the outcome of the two models. Deviating results concern the variables Leftt and St in the pre-war period, which were both insignificant. Regarding the post-war period, the lagged unemployment variable had no significance anymore, while the expected enforcement effect of the union density rate was not found either. In terms of the demand and supply framework the results of these time-series analyses give the following picture. The demand for unions before World War II was influenced by the threat effect of inflation and by the credit effect of wage growth; as pay rises are often the result of the unions' bargaining efforts, this makes it a supply-side feature as well. The same goes for the unemployment development: on the demand-side, unemployed workers eventually became discouraged and quit the union; on the supply-side, high unemployment forced the union movement onto the defence and reduced its power: this made the membership less attractive. The last significant and negative determinant in the pre-war pe-riod concerns the lagged strike variable, which can also be placed on the supply and on the demand-side. Apparently, many workers did not approve of union instigated strikes or were afraid of being dismissed if they would participate. Regarding the results of the post-war model, now the unlagged strike variable had a positive impact on the demand for unionism. As far as strikes were organized by unions, those can be placed on the supply-side as well. The same goes for the variable Leftt: the positive impact of the share of leftist parties in Parliament could reflect a pro-union attitude among workers, but it could also facilitate the unions' power and possibilities to recruit new members. The above argument with respect to unemployment also holds for the post-war period. The union density rate is a demand-side feature if it concerns an enforcement effect. In the Netherlands, primarily a saturation effect was found, which links up better with the supply-side: it indicates that the recruitment possibilities of unions become smaller after a certain rate of organization level. It should be kept in mind however, that density rates can differ substantially between sectors, companies and regions, so that on a less aggregate level the presence of a possible enforcement effect may not be ignored. Finally, the significant impact of the growth of social benefits payments by the state on union membership rates reflects the declining costs of substitute services, which affected the demand for unions negatively. The fact that most theoretical expectations are confirmed by the results of these timeseries analyses does not mean that all determinants of Dutch unionization are now mapped out. As has been expounded in the first three chapters of this thesis, the union membership is influenced by a lot of factors, only a few of which have been taken into consideration in this chapter. This time-series analysis only employs so-called 'aggregate data', which means it attempts to explain the fluctuations in the total Dutch union membership with the aid of variables such as the development of total unemployment, wage growth of industrial workers, the total number of strikes and the overall trade union density rate. As the union membership and density rate differ considerably between the branches of industry, it
110
would be better to perform a time-series analysis on sectoral data as well, but these are too incomplete to be used. Whether we have the disposal of nation-wide or sectoral data, in the end union membership is an individual choice. In the Netherlands, there are no closed shops (except for the printing industry), so workers can freely decide to join a union or not. Hence, individual circumstances, which are in turn embedded in socio-economic developments on the macro and the meso level, play an important role in the unionization decision. These individual circumstances concern personal characteristics, job and industry related characteristics, wage related characteristics and regional characteristics. All of these aspects influence the disposition of workers towards unionism, as has been theoretically expounded in chapter 2. Moreover, since individuals usually don't work or live isolated, their attitudes are also affected by their social environment, at home as well as at work. Finally, the choice whether or not to become a union member is also influenced by the unions themselves; by their performance in the bargaining over wages and employment, their attitude towards all kinds of socio-economic issues, and last but not least by their services for the members. Even though the time-series analysis did contain several variables that could be placed on the supply-side of the theoretical framework, these variables were rather onesided because they were all more or less linked to the unions' performance in the bargaining (striking) over wages and employment. No variables could be employed which measure the (dis)satisfaction of workers (members) regarding the direct costs and benefits of individual membership, in the form of union dues and service facilities. Overlooking all these determinants of trade union membership, it becomes clear that the time-series analyses of this chapter only give a part of the explanation why unions grow and decline. Very substantial, additional explanations can be derived from the crosssectional analysis. Since this research method is based on (extensive) enquiries, far more variables can be taken into account, which give information about the 'individual circumstances' mentioned above, and possibly also about several environmental determinants, and attitudes or opinions regarding trade union policies. Therefore, the next chapter will give a detailed discussion of the results obtained from two cross-sectional analyses, after which again the question will be posed, to what extent Dutch union membership is explained by the determinants brought to the fore by employing cross-sectional data.
111
Table 4.10 Values of all variables used in the time-series analyses Year 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945
ΔMt
. . . . . . . . 8.5271 3.5714 27.6379 6.1721 8.2952 26.0820 2.6090 10.7156 18.8320 19.8785 22.8236 33.2646 -4.8966 -2.5596 -11.0689 -10.2954 -4.0921 -0.7343 2.2591 3.2258 6.6856 13.1901 8.4222 15.0152 5.8735 -4.3003 -5.7845 -3.3465 -0.5453 2.8098 3.7995 3.9044 -9.5550 -1.0872 . . . .
ΔPt
. 5.1948 -2.4691 0.0000 2.5316 0.0000 0.0000 1.2346 3.6585 -1.1765 2.3810 1.1628 1.1494 1.1364 0.0000 14.6067 11.7647 6.1404 19.0083 9.0278 10.1911 -13.2948 -11.3333 -3.7594 0.7813 -0.7752 -3.9063 0.0000 0.8130 -0.8065 -4.0650 -5.9322 -7.2072 -0.9709 0.0000 -2.9412 -5.0505 6.3830 1.0000 0.9901 14.7059 14.5299 7.4627 3.4722 2.6846 15.0327
ΔWt
. . . . . . . . . . . 5.5556 0.0000 5.2632 0.0000 5.0000 14.2857 20.8333 17.2414 50.0000 25.4902 3.1250 -13.6364 -8.7719 -7.6923 6.2500 5.8824 1.8519 1.8182 3.5714 0.0000 -1.7241 -8.7719 -1.9231 -3.9216 -4.0816 -2.1277 4.3478 2.0833 2.0408 6.0000 9.4340 6.8966 6.4516 . .
ΔRWt
. . . . . . . . . . . 4.3423 -1.1364 4.0804 0.0000 -8.3824 2.2556 13.8430 -1.4847 37.5796 13.8842 18.9375 -2.5974 -5.2083 -8.4079 7.0801 10.1865 1.8519 0.9971 4.4135 4.2373 4.4734 -1.6863 -0.9615 -3.9216 -1.1750 3.0783 -1.9130 1.0726 1.0404 -7.5897 -4.4495 -0.5268 2.8794 . .
Ut . . . . . . . . . . . 23.0 37.0 48.0 134.0 118.0 51.0 66.0 78.0 81.0 71.8 125.6 161.0 171.0 143.1 138.4 121.0 127.0 108.5 133.5 133.1 267.0 475.4 530.4 517.1 587.4 601.4 525.1 488.2 383.7 . . . . . .
U%t . . . . . . . . . . . 1.8 2.8 3.5 9.7 8.5 3.6 4.6 5.3 5.4 3.3 5.7 7.2 7.5 6.2 5.9 5.1 5.3 4.5 5.4 5.4 10.6 18.6 20.5 19.7 22.1 22.5 19.4 17.9 13.8 . . . . . .
Lt . . . . . . . . . . . . . 1590000 1620000 1650000 1690000 1730000 1780000 1840000 1890000 1940000 1980000 1950000 1980000 2020000 2060000 2100000 2170000 2240000 2300000 2320000 2270000 2300000 2340000 2370000 2440000 2470000 2500000 2490000 2490000 . . . . .
Dt-t . . . . . . . . . . . . . . 16.8742 16.9938 18.4727 21.4320 25.0983 29.9607 38.6250 35.7619 33.9485 29.5808 26.9436 25.4495 24.7624 24.8301 25.1429 25.9585 28.4643 30.0565 34.2716 37.0837 35.0261 32.4359 30.9536 29.9016 30.3684 31.1440 32.4900 29.3855 . . . .
112
ΔE . . . . . . . . . . . . . . . . . . . . . . 2.2728 -1.1111 4.4943 4.3010 3.0928 3.0000 6.7961 4.5455 1.7390 -8.5470 -13.0841 0.0000 2.1506 -3.1579 1.0869 9.6775 3.9215 6.6038 . . . . . .
Leftt . 15 15 15 15 19 19 19 19 15 15 15 15 25 25 25 25 25 30 30 30 30 27 27 27 32 32 32 32 33 33 33 33 33 33 33 33 34 34 34 . . . . . .
St . 122 142 163 102 132 181 154 135 189 146 217 283 427 271 269 377 344 325 649 481 299 325 289 239 262 212 230 205 226 212 215 216 184 152 152 96 95 141 90 23 . . . . 118
SOCt 0.4454 0.4294 0.4734 0.4611 0.4516 0.5345 0.5566 0.5467 0.6208 0.6891 0.6645 0.7168 0.6696 0.7125 0.7326 0.7437 0.7584 0.8729 1.0000 0.9042 0.8910 1.0727 1.2639 1.4329 1.5157 1.4984 1.6055 1.6262 1.5728 1.6318 2.0119 2.6047 3.1250 3.2434 3.2604 3.2395 3.2709 3.0998 3.2413 3.0735 . . . . . .
Year 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
ΔMt 20.72130 8.95180 5.77050 4.69700 4.78820 -9.97860 3.42970 4.37700 3.51580 3.10990 6.07985 -0.27684 0.22509 1.42985 3.29914 1.39326 1.19090 2.38858 -0.80256 1.59068 3.38800 0.49612 0.66255 -0.22585 2.53525 1.28674 0.87184 2.93246 0.99562 2.33981 -0.21470 1.63410 0.63512 0.53446 -1.54960 -2.17716 -2.31370 -4.10580 -2.76454 -1.26362 -0.29383 1.16568 2.27861 3.53797 .
ΔPt 9.0909 3.6458 3.5176 6.3107 9.5890 9.1667 0.0000 0.3817 4.5627 2.1818 0.3559 10.6383 2.4359 0.8448 2.4201 1.0906 3.2664 4.1788 5.9889 4.7306 5.3952 3.4286 4.1206 7.1634 3.1360 7.5215 7.0326 9.2126 8.4355 11.1111 8.9960 6.4235 3.4165 4.4488 6.3785 6.9970 5.6132 2.6487 4.2550 2.0325 0.0000 -0.9960 2.0201 0.9861 1.9608
ΔWt 13.9241 5.5556 5.2632 1.0000 10.8911 6.2500 4.2017 1.6129 16.6667 2.7211 9.2715 11.5152 0.5435 3.7838 10.9375 12.6761 9.1667 6.8702 14.6429 8.7227 10.0287 7.0313 8.2725 9.8876 13.7014 15.6475 11.0420 14.0056 16.3391 13.6220 7.6208 10.5354 6.5625 6.5982 5.0894 5.0393 6.1059 0.9982 0.8140 2.1915 1.2980 1.7827 8.4839 1.8668 1.3868
ΔRWt 4.4304 1.8425 1.6863 -4.9954 1.1881 -2.6718 4.2017 1.2265 11.5758 0.5278 8.8840 0.7925 -1.8474 2.9144 8.3161 11.4605 5.7136 2.5835 8.1650 3.8118 4.3962 3.4833 3.9876 2.5421 10.2442 7.5575 3.7460 4.3887 7.2888 2.2598 -1.2617 3.8638 3.0421 2.0578 -1.2118 -1.8297 0.4665 -1.6078 -3.3006 0.1558 1.2980 2.8067 6.3357 0.8721 -0.5629
Ut
U%t
. 31.8 29.9 42.9 58.8 68.4 105.7 76.7 51.8 32.9 23.5 33.1 69.0 48.5 28.8 21.2 21.2 23.5 20.7 25.2 35.6 74.8 68.2 49.4 44.5 62.0 108.0 109.9 134.9 195.3 210.8 203.6 205.6 210.0 247.9 385.3 541.7 800.6 822.4 761.0 710.7 685.5 682.2 639.2 595.2
. 2.8 1.9 1.7 2.1 2.5 3.8 2.7 1.7 1.1 0.8 1.0 2.2 1.5 0.9 0.6 0.6 0.7 0.6 0.7 1.0 2.0 1.8 1.3 1.1 1.6 2.8 2.8 3.5 5.0 5.3 5.1 5.1 5.1 5.9 9.1 12.5 17.0 17.3 15.9 14.7 13.9 13.7 . .
Lt 2339000 2609000 2729000 2782000 2827000 2864000 2906000 2956000 3040000 3103000 3165000 3219000 3243000 3276000 3345000 3404000 3500000 3571000 3642000 3715000 3768000 3808000 3858000 3914000 3971000 4041000 4112000 4266000 4334000 4396000 4469000 4520000 4581000 4728000 4879000 5016000 5151000 5195000 5294000 5387000 . . . . .
Dt-t
ΔE
. 41.0731 40.1188 40.5680 41.6643 42.9643 38.1774 38.9160 39.9323 40.1941 40.6026 42.2275 41.4042 41.1903 41.3584 41.8416 41.6892 41.0286 41.1733 40.0467 39.8843 40.6555 40.4280 40.1685 39.5043 39.9245 39.7377 39.3920 39.0835 38.8533 39.2015 38.4784 38.6659 38.3934 37.3985 35.6794 33.9494 32.2947 30.7064 29.2992 28.4296 . . . .
. 15.4026 9.3117 4.3210 5.6686 2.8671 -1.3936 3.8755 5.5804 2.8894 1.64384 1.10705 -1.87565 2.05705 3.47024 1.75503 3.43146 1.32705 1.84387 0.57530 -0.69818 -3.91360 -0.78462 0.76417 -0.87302 -1.53011 -1.84298 -1.21491 -1.34166 -4.54245 -2.07756 -0.42433 -2.66843 -0.69843 -1.20722 -3.18776 -4.35737 -3.80996 -0.58459 2.12408 1.53937 1.04155 -0.49250 1.59991 1.88059
113
Leftt 39 39 35 35 35 35 36 36 36 36 38.0 38.0 38.3 35.3 35.3 35.3 35.3 34.0 34.0 34.0 34.0 35.3 35.3 35.3 35.3 40.0 43.3 43.3 43.3 43.3 43.3 44.7 44.7 44.7 44.7 46.7 40.7 40.7 40.7 40.7 42.7 42.7 42.7 44.7 44.7
St 270 272 183 116 79 85 40 58 91 63 80 37 73 48 121 43 24 104 53 60 20 8 11 28 124 25 58 29 41 15 28 44 32 57 22 11 12 9 11 45 35 28 38 27 29
SOCt 4.2002 4.2268 10.3036 9.8742 9.8580 9.3418 9.7452 9.9727 10.0865 10.1595 10.1836 11.9566 13.3458 12.9249 12.8998 13.1648 13.6937 15.6857 15.7419 16.9297 18.2811 19.1832 19.7060 19.5865 20.3355 21.8583 23.0174 23.5318 25.0871 28.2919 28.4530 29.4052 30.4101 31.4790 32.0525 33.3734 35.2030 35.5627 34.2450 32.5567 32.3586 33.5342 32.5180 31.9117 33.0104
Definitions and sources ΔMt Annual % change in total union membership. Source: A) CBS (1959) 60 jaren statististiek in tijdreeksen (concerning 19101912); B) J. Visser (1989), European trade unions in figures (concerning 1913-1941); C) B.O. Ebbinghaus & J. Visser (1993), Trade union systems in Western Europe; data handbook volume 1 (concerning 1946-1989). ΔPt Annual % change in consumer prices. Source: CBS (1959, 1969, 1979, 1984, 1989), Statistiek in tijdreeksen; and checked in CBS (1946..1972), Statistical Yearbook. This series is based on indices with several base years (1937, 1951, 1963, 1975, 1980, 1985), which implies that the series is not completely continuous. ΔWt
Annual % change in gross nominal hourly wages of male industrial workers. Source: CBS (1969,1989), Statistiek in tijdreeksen, supplemented with extrapolated figures for the years 1910-1925, on the basis of given wages in the metaland mining industry. From 1970 onwards this figure relates to the wages of men and women. In the earlier period those wages differed considerably, but from the 1960's the wages for women were gradually raised to the level of male workers. ΔRWt Annual % change in real wages. Source: compiled on the basis of ΔPt and ΔWt (see above). Ut Unemployment in absolute numbers (in 1,000's). Source: A) 1911-1919: calculated on the basis of B.R. Mitchell (1975), European historical statistics 1750-1970 (the average ratio of the Mitchell-figures and the Den Bakker-figures of 1920-1939 is 1.4185; subsequently this ratio is employed on the Mitchell-figures of 1911-1919). B) 1920-1939: Den Bakker & Van Sorge (ESHJb 1991), Het onbenut arbeidsvolume in het Interbellum. C) 1947-1990: CBS (1969, 1989), Statistiek in tijdreeksen. This series is not continuous, because of the different sources in the pre-war period and the use of different definitions in the post-war period. Until 1977 registered unemployed looking for a job of 30 or more hours were included; from 1978 this limit was lowered to 25 hours or more, and from 1983 this was lowered again to 20 hours or more. Further, since 1978 unemployed people over the age of 64 are not counted anymore. U%t The number of unemployed as a % of the dependent labour force (Lt). Source: see A), B), and C) right above. Lt
Size of the dependent labour force (until 1969 only full-time workers and people looking for full-time jobs, from 1970 all persons in employment working at least 20 hours, and those willing to do so). Source: compiled by J. Visser (1989), European trade union membership in figures. Dt-1 Union density rate (lagged with one year), defined as the ratio of total union membership and the dependent labour force. Source: constructed on the basis of Visser (1989), European trade unions in figures (see also Lt above). ΔE Annual % change in employment in the industrial sector. Source: A) 1922-1929: CBS (1931), Maandschrift. B) 1930-1939: CBS (1940), Jaarcijfers voor Nederland. C) 1946-1963: CBS, Statistical Yearbook.
D) 1964-1989: CBS, Maandstatistiek van de industrie. This series is not fully consistent, because of the different sources; moreover, before World War II the only information is given in indices, which had to be transformed into absolute numbers on the basis of post-war data, which were at first expressed both in indices and in absolute numbers. Leftt
% share of leftist parties (SDAP/PVDA, CPN, PSP, PPR, D'66) in Parliament. Source: calculated on the basis of CBS, (1989), Statistiek in tijdreeksen. St Number of strikes per year. Source: CBS (1969, 1989), Statistiek in tijdreeksen. SOCt The total sum of social benefits payments by the government, as a percentage of the net national income. Source: calculated on the basis of CBS, (1969, 1989), Statistiek in tijdreeksen. PM All recent figures (1988-1990) are taken from CBS, Sociaal-Economische Maandstatistiek and Statistisch Zakboek.
Table 4.11 Comparison of the replication on results with the original A&P and B&E models, covering the whole period (t-values in brackets) A & P model
°
1 adj. R2 D.W. Root MSE
0.460 1.105 3.033
°°
2
0.651 1.387 1.094
# obs.
54
54
Constant
6.005 (0.905)
1.140** (2.450)
ΔPt
-0.243 (1.499)
-0.152 (1.658)
B & E model
°°°
3
0.700 1.660 57
°
4
0.542 1.243 5.023
°°
5
0.424 1.347 1.270
°°°
6
0.702 1.635 -
65
65
-10.584*** (4.223)
20.045 (5.459)
0.872** (2.213)
6.041*** (4.319)
0.673*** (5.258)
0.001 (0.007)
-0.266 (1.629)
0.583*** (5.860)
DummyΔPt
-2.084 (1.012)
-0.293 (0.356)
-0.386*** (2.686)
ΔWt
0.521 (5.412)
0.286*** (3.156)
0.468*** (4.181)
ΔUt-1
0.001 (0.052)
-0.001 (0.161)
-0.351 (1.325)
ΔUt-2
-0.009 (0.748)
-0.005 (0.609)
0.403 (1.612)
-0.536 (5.423)
-0.016 (0.498)
-0.203*** (5.401)
-7.806 (1.498)
-9.615*** (3.915)
ΔE
-0.180 (1.189)
0.059 (0.506)
0.127 (1.608)
ΔE-1
0.238 (1.487)
0.095 (0.693)
0.110** (2.340)
ΔE-2
0.285 (2.235)
0.201** (2.098)
0.083* (1.886)
U%t
-0.180 (2.007)
-0.316*** (3.772)
0.249*** (2.594)
Dt-1
-0.114 (1.029)
-0.085 (1.271)
-0.063** (2.333)
Leftt
0.041 (0.296)
0.123* (1.703)
0.222*** (3.964)
Dummy51
-11.875 (3.612)
-11.649*** (6.544)
68
Explanation of the symbols: ΔPt: % change of consumer prices DummyΔPt: dummy with value 1 if inflation exceeds 4% ΔWt : % change of gross nominal wages ΔE-1 : % change of employment in industrial sector Ut-i : level of total unemployment (in 1,000's) Explanation of the asterisks: U%t : unemployment in % of the dep. labour force Dt-1 : union density, lagged with 1 year * significant at 10% level Leftt: % share of left parties in Parliament ** significant at 5% level Dummy51: dummy with value 1 for the year 1951 *** significant at 1% level
° uncorrected, °° corrected for heteroscedasticity; °°° original model of A & P; B & E.
Table 4.12 Pre-war Pearson Correlation Coefficients / Prob > |R| under Ho: Rho=0 / 28 observations ΔM
t
ΔM
t
ΔP
t
ΔP
t-1
ΔW
t
ΔRW
t
U
t
U
t-1
D
t-1
(D ) t-1
2
Left
t
S
t
ΔP
t
t-1
SOC
t
t-1
ΔW
ΔRW
t
U
t
U
t
0.53007
0.62008
0.66535
0.44707
0.0
0.0037
0.0004
0.0001
0.0171
0.53007
1.00000
0.50873
0.0037
0.0
0.0057
0.62008
0.50873
1.00000
0.0004
0.0057
0.0
0.0001
0.0001
0.66535
0.63233
0.83702
1.00000
0.79257
0.0001
0.0003
0.0001
0.0
0.0001
0.44707
0.03253
0.69948
0.79257
1.00000
0.0171
0.8695
0.0001
0.0001
0.0
- 0.41000
- 0.28434
-0.36660
-0.39778
-0.30379
0.0302
0.1425
0.0550
0.0361
0.1160
- 0.39989
- 0.13950
- 0.28482
-0.31287
-0.30591
0.0350
0.4790
0.1418
0.1050
0.1134
0.63233 0.0003
D
t-1
1.00000
t-1
Left
2
S
t
S
t
SOC
t-1
t
-0.41000
-0.39989
-0.46299
-0.48047
-0.20041
0.50777
0.00572
-0.37813
0.0302
0.0350
0.0172
0.0130
0.3065
0.0058
0.9770
0.0473
-0.28434
-0.13950
-0.44347
-0.41585
-0.22558
0.36225
0.1425
0.4790
0.0233
0.0346
0.2484
0.0582
-0.36660
-0.28482
-0.21920
-0.19594
-0.15523
0.56900
0.0550
0.1418
0.2820
0.3374
0.4303
0.0016
0.0487
-0.39778
-0.31287
-0.14860
-0.13737
-0.07291
0.72206
0.40346
0.0361
0.1050
0.4688
0.5034
0.7124
0.0001
0.0333
-0.30379
-0.30591
0.14235
0.13690
0.08246
0.63529
0.40631
0.1160
0.1134
0.4879
0.5048
0.6766
0.0003
0.0319
1.00000
0.94864
0.50909
0.49604
0.59499
-0.68514
-0.60377
0.94606
0.0
0.0001
0.0079
0.0100
0.0008
0.0001
0.0007
0.0001
0.94864
1.00000
0.43480
0.41576
0.58216
-0.69528
-0.64518
0.90529
0.0264
0.0346
0.0012
0.0001
0.0002
0.0001
0.55964 0.0030
-0.08842 0.01881 0.53952 0.6676 0.9273 0.0044
0.03253 0.8695
0.83702
(D )
t-1
0.69948
0.0001
0.0
0.16257 0.4085 0.37594
- 0.46299 0.0172
-0.44347 0.0233
-0.21920 0.2820
-0.14860 0.4688
0.14235 0.4879
0.50909 0.0079
0.43480 0.0264
1.00000 0.0
0.99244 0.0001
- 0.48047 0.0130
-0.41585 0.0346
-0.19594 0.3374
-0.13737 0.5034
0.13690 0.5048
0.49604 0.0100
0.41576 0.0346
0.99244 0.0001
1.00000 0.0
0.49966 0.0093
- 0.20041 0.3065
- 0.22558 0.2484
-0.15523 0.4303
-0.07291 0.7124
0.08246 0.6766
0.59499 0.0008
0.58216 0.0012
0.55964 0.0030
0.49966 0.0093
1.00000 0.0
-0.45134 -0.35507 0.72722 0.0159 0.0637 0.0001
0.50777 0.0058
0.36225 0.0582
0.56900 0.0016
0.72206 0.0001
0.63529 -0.68514 0.0003 0.0001
-0.69528 0.0001
-0.08842 0.6676
-0.05537 0.7882
-0.45134 0.0159
1.00000 0.68090 -0.74801 0.0 0.0001 0.0001
0.00572 0.16257 0.9770 0.4085
S
ΔP
-0.37813 0.0473
-0.34887 0.0688
-0.05537 0.08692 0.7882 0.6729
-0.64518 0.01881 0.0002 0.9273
0.08692 -0.35507 0.68090 1.00000 -0.68209 0.6729 0.0637 0.0001 0.0 0.0001
-0.40170 0.0341
0.90529 0.53952 0.0001 0.0044
0.50234 0.72722 -0.74801 -0.68209 1.00000 0.0089 0.0001 0.0001 0.0001 0.0
-0.28582 0.94606 0.1404 0.0001
0.0688 -0.40170 0.0341 -0.42230 0.0252 -0.28582 0.1404
0.50234 0.0089
0.37594 0.40346 0.40631 -0.60377 0.0487 0.0333 0.0319 0.0007 -0.42230 0.0252
-0.34887
117
Table 4.13 Post-war Pearson Correlation Coefficients / Prob > |R| under Ho: Rho=0 / 40 observations ΔM
t
ΔP
t
ΔP
t-1
ΔW
t
ΔRW
t
U
t
U
t-1
D
t-1
(D ) t-1
2
Left
S
t
t
S
t-1
ΔM
1.00000 -0.23998 -0.16563 0.16932 0.36703 -0.48630 -0.42680 0.33187 0.32248 -0.24484 0.55688 0.52445 -0.51631 0.0 0.1358 0.3071 0.2963 0.0198 0.0015 0.0067 0.0364 0.0424 0.1278 0.0002 0.0005 0.0006
ΔP
-0.23998 1.00000 0.35366 0.48673 -0.19881 -0.13302 -0.19875 0.23731 0.22887 0.28530 -0.16618 -0.02743 0.17124 0.1358 0.0 0.0252 0.0014 0.2187 0.4132 0.2251 0.1404 0.1555 0.0743 0.3054 0.8666 0.2907
ΔP
-0.16563 0.35366 1.00000 0.00676 -0.25705 0.02274 -0.03862 0.03521 0.01999 0.34955 0.04782 -0.03743 0.19029 0.3071 0.0252 0.0 0.9670 0.1093 0.8892 0.8154 0.8292 0.9026 0.0270 0.7695 0.8186 0.2395
t
t
t-1
SOC
t
0.16932 0.48673 0.00676 1.00000 0.75907 -0.47745 -0.48204 0.45524 0.44746 -0.03996 -0.03527 -0.01362 -0.12007 0.2963 0.0014 0.9670 0.0 0.0001 0.0018 0.0019 0.0032 0.0038 0.8066 0.829 0.9335 0.4605
ΔW
t
ΔRW 0.36703 -0.19881 -0.25705 0.75907 t
0.0198
0.2187
0.1093
0.0001
1.00000 -0.44187 -0.39725 0.34126 0.33904 -0.26208 0.08718 0.00787 -0.26879 0.0 0.0043 0.0123 0.0312 0.0323 0.1023 0.5927 0.9615 0.0935
U
-0.48630 -0.13302 0.02274 -0.47745 -0.44187 1.00000 0.97066 -0.95496 -0.95087 0.50966 -0.36840 -0.38132 0.76891 0.0015 0.4132 0.8892 0.0018 0.0043 0.0 0.0001 0.0001 0.0001 0.0008 0.0193 0.0152 0.0001
U
-0.42680 -0.19875 -0.03862 -0.48204 -0.39725 0.97066 1.00000 -0.96051 -0.95025 0.47171 -0.34966 -0.37045 0.72079 0.0067 0.2251 0.8154 0.0019 0.0123 0.0001 0.0 0.0001 0.0001 0.0024 0.0291 0.0203 0.0001
t
t-1
0.33187 0.23731 0.03521 0.45524 0.34126 -0.95496 -0.96051 1.00000 0.99860 -0.49345 0.33193 0.35116 -0.74329 0.0364 0.1404 0.8292 0.0032 0.0312 0.0001 0.0001 0.0 0.0001 0.0012 0.0364 0.0263 0.0001
D
t-1
(D ) t-1
0.32248 0.22887 0.01999 0.44746 0.33904 -0.95087 -0.95025 0.99860 1.00000 -0.51444 0.34172 0.35970 -0.76005 0.0424 0.1555 0.9026 0.0038 0.0323 0.0001 0.0001 0.0001 0.0 0.0007 0.0309 0.0226 0.0001
2
-0.24484 0.28530 0.34955 -0.03996 -0.26208 0.50966 0.47171 -0.49345 -0.51444 1.00000 -0.31446 -0.30094 0.76415 0.1278 0.0743 0.0270 0.8066 0.1023 0.0008 0.0024 0.0012 0.0007 0.0 0.0481 0.0592 0.0001
Left
t
S
0.55688 -0.16618 0.04782 -0.03527 0.08718 -0.36840 -0.34966 0.33193 0.34172 -0.31446 0.0002 0.3054 0.7695 0.8290 0.5927 0.0193 0.0291 0.0364 0.0309 0.0481 0.0
S
0.52445 -0.02743 -0.03743 -0.01362 0.00787 -0.38132 -0.37045 0.35116 0.35970 -0.30094 0.73644 0.0005 0.8666 0.8186 0.9335 0.9615 0.0152 0.0203 0.0263 0.0226 0.0592 0.0001 0.0
t
t-1
1.00000 0.73644 -0.59099 0.0001 0.0001 1.00000 -0.58836 0.0001
SOC -0.51631 0.17124 0.19029 -0.12007 -0.26879 0.76891 0.72079 -0.74329 -0.76005 t
0.0006
0.2907
0.2395
0.4605
0.0935
0.0001
0.0001 0.0001
0.0001
0.76415 -0.59099 -0.58836 1.00000 0.0001 0.0001 0.0001 0.0
118
5. CROSS-SECTIONAL ANALYSES 5.1 Introduction46) In this chapter we analyze the determining factors of trade union membership in the Netherlands by means of two separate cross-sectional analyses, based on different data sets. The first data set is derived from an inquiry of 1988, while the second one is derived from an inquiry that was held both in 1979 and in 1987. These surveys will be discussed later on. On the basis of the first data set, we are able not only to analyze the unionization decision, but also the decision which union or confederation the worker will join. A test is presented to determine the nature of the decision process; we ask whether the decision to join a union and the decision which confederation to join are sequential or simultaneous decisions. The answer to this question is not only of consequence for the model specification, but, as will be shown, also affects the conclusions drawn from the estimations. Since the second data set consists of observations with respect to the years 1979 and 1987, we can exploit this "time-series" of cross-sections to analyze changes in individual decision-making with regard to union membership. As has become clear in previous chapters of this book, the Dutch union density rate decreased tremendously between 1979 and 1987. By decomposing the change in the union membership rate into changes in observable characteristics of the employed labour force on the one hand, and in changes in behaviour on the other, it can be determined to what extent the decline in the union density rate is caused by an increase in the share of employees with a low propensity to join a union, and to what extent by a diminishing appeal of unions to workers. The structure of this chapter is as follows. In the next section, we specify the econometric framework which is applied to the first data set, followed by a description of these data in section 5.3. Section 5.4 deals with the empirical findings. The second data set is described in section 5.5. The econometric modelling is expounded in section 5.6, after which the empirical results are analyzed in section 5.7. The overall conclusions are taken up in section 5.8, while the chapter will be rounded off with an appendix.
46)This chapter is based on two articles jointly written with Wim Groot, who constructed the econometric framework. The papers in question are 'Union membership in the Netherlands: a cross-sectional analysis' (1992) and 'Why union density declined' (1994).
119
5.2 Econometric framework One question we want to address in this chapter is whether the decision to join a union and the decision which union to join are sequential or simultaneous decisions. If the individual first decides whether or not to join a union and then, if this decision is positive, decides upon which union to join, union membership is a sequential decision. If both decisions are taken together, membership is a simultaneous decision. To test these alternative models a nested model is estimated in which both simultaneous and sequential decision making processes are special cases. A full description of the nested model can be found in Maddala (1983, pp. 67-70). The structure of the simultaneous decision model is as follows. Each individual in the sample is confronted with four options: to join the FNV, to join the CNV, to join some other union, or not to become unionized. These options are denoted by i, that takes on the values 1, 2, 3, 4, respectively. The individual chooses alternative i if and only if this alternative provides the highest utility. Let Vi be the utility for the individual of choosing option i. We assume that utility is a function of (a vector of) personal and job-related characteristics Z: (5.1) Vi = Zπi + ui where πi is a parameter vector and ui is a random term. Furthermore, the individual chooses the alternative which yields the highest utility: Vi = max {V1, V2, V3, V4}
(5.2)
We assume that the disturbances ui are independently and identically extreme value (type I) distributed. The probability that alternative i will be chosen, Pi, is then defined by: exp(Zπ?i) Pi = 4 ∑exp(Zπ?j)
i=1,2,3,4
(5.3)
j=1
which is the multinomial logit model47). Defining βi = πi - π4 for i = 1, 2, 3, with βi being a vector of parameters determining which union is chosen, the probabilities can be rewritten as: exp(Zβ?i) Pi = 1+exp(I)
i=1,2,3
(5.4)
where I is the so-called 'inclusive value' of union membership: 3
I=log ∑ exp(Zβj) j=1
(5.5)
The probability to join any union, i.e., P1 + P2 + P3, can then be written as: exp(I) PM = 1+exp(I)
(5.6)
47)This is a so-called discrete choice model; since the dependent variable can only take on the values 1, 2, 3, or 4, a linear regression analysis is not an adequate estimation technique in this case.
120
The probability that someone is a member of union i, given that he/she is a member of one of the unions is: exp(Zβ?i) PiM = exp(I) i=1,2,3 (5.7) Alternatively, the sequential model describes the decision to join a specific union as a two stage process. Firstly, an individual decides whether or not to become a member. In the second stage, those who want to join a union decide upon which union to join. Both stages are described by separate logit models. The probability to become a union member is given by: exp(Zα) PM = 1+exp(Zα)
(5.8)
where α is a vector of parameters determining the decision whether or not to join a union. The conditional probability of joining union i, conditional on the choice to become a union member is the same as in the simultaneous model and is given by (5.7). The model that encompasses both the simultaneous and the sequential model must therefore have a specification of PM that contains (5.6) and (5.8) as special cases: exp(σZα+(1-σ)I) PM = 1+exp(σZα+(1-σ)I)
(5.9)
This nested logit model (5.9) can be estimated by first estimating the parameters βi from the conditional-choice equation (5.7), then to calculate the inclusive values I, and finally to estimate α and σ from (5.9). The testing of the alternative models is performed by testing whether σ = 0 or σ = 1. If σ = 0 the model in (5.9) becomes a multinomial logit model supporting simultaneous decision making. If σ = 1 equation (5.9) reduces to a binary choice model of whether or not to join a union. This supports the sequential decision model (cf. Schram & Van Ophem 1989). It is possible that for some workers the decision to unionize and the decision which union to join are sequential, while for others these are taken simultaneously. In chapter 3 it was argued that the federation for higher personnel, MHP, was founded as a reaction to the fact that the FNV and CNV mainly focuss on the interests of the lower income classes. It is therefore not inconceivable that higher educated employees either join the MHP (included in the category 'other unions') or do not unionize at all, and that the possibility of joining the FNV or the CNV is irrelevant to them. In that case higher educated workers take the decision to become a union member and the decision which union to join simultaneously. For lower educated workers, however, the decision structure might be a sequential one. Similarly, the FNV has been most active in recruting female workers, compared to the other federations. Hence, it might be that for female employees the decision is to join the FNV or not to become a union member at all, while for male workers these are sequential decisions. In order to allow for different decision structures between groups of workers we parametrize σ to obtain an even more general framework. We specify: σ = σ0 + ΣσiXi where Xi is a vector of variables and σi is a parameter to be estimated.
(5.10)
121
5.3 The first data set The data are taken from the 1988 wave of the OSA-Labourmarket survey, a national longitudinal survey in the Netherlands48). The sample consists of 4461 randomly selected individuals, from which we have taken a sub-sample of wage earners. After elimination of observations of which essential information on union membership could not be retrieved, 2589 observations could be used in the analysis. In table 5.1 some descriptive statistics of the sample are given. The individuals in the sample have first been asked whether they belonged to a trade union or not. If so, they were asked to indicate to which confederation their union is affiliated: the FNV, the CNV, the MHP or some other trade un-ion. Unfortunately, the share of the MHP was too small to be treated separately. Therefore we shall henceforth distinguish between FNV, CNV and 'others'49), and for convenience we shall refer to them as unions instead of federations. We have tried to include as many variables discussed in chapter 2 as possible, supplemented by variables we thought of interest in explaining union membership. In the last column of table 5.1 the expected signs of all selected variables are shown, mainly based on the theoretical notions and empirical results as discussed in the literature survey. In order not to repeat the contents of chapter 2, the expected effects of the variables will only be commented upon if these have not already been explained there. Of the personal characteristics we incorporated the following variables. Gender (1=male; 0=female) was included to test the assumption that men are more often a member than women. Regarding nationality (1=non-Dutch; 0=Dutch) we expect that foreign workers are less likely to join, because they are less integrated in the society. The predicted effect of age is ambiguous, while the educational level (1=low, 6=high) is assumed to have a negative relation with union membership. Experience is captured by tenure (in months), and by a dummy variable 'general human capital' indicating whether the knowledge and experience gathered in the present job is also applicable in other jobs (1=yes; 0=no)50). Due to divergent empirical results the expected signs of these two variables can not be predicted. For family composition we included marital status (1=married or cohabiting; 0=otherwise) and the number of dependent children. It has been found that people with greater family responsibilities are more likely to be a union member. For religion we incorporated a dummy variable indicating whether the individual is a regular church visitor and three dummy variables for the religious denomination (Roman Catholic,
48)OSA stands for: Organisatie voor Strategisch Arbeidsmarktonderzoek. Even though the OSAsurvey is held every two years, the question whether someone is a union member was, unfortunately, only posed in 1988. 49)The category 'other unions' probably consists of more unions which represent groups of higher personnel, just like the MHP. 50)Unless stated otherwise, all dummy variables are defined by 1=yes; 0=no.
122
Reformed, and Protestant)51). In general, it could be argued that having a religious conviction lowers the probability that one will join a union; the church doctrine preaches placability and co-operation, while trade unions show a readiness to fight. It is not clear however, which sign these religious variables will obtain with respect to the Dutch data. Since workers with a belief have the option to join the Christian CNV, the relationship between membership and religion might be positive. Besides these variables, inclusion of which was motivated by the literature survey, we included a variable indicating whether the father of the respondent had been an employee (i.e., wage earner) as well. Although not very strong, this variable is assumed to capture some of the traditional bonds with the union movement, so that a positive sign is expected. The occupational and industrial characteristics comprehend seven dummy variables for the profession of the individual: 1=agricultural or industrial; 2=commercial; 3=administrative; 4=caring; 5=medical; 6=education; 7=scientific. The classification of professions is taken from Teulings (1988) (cf. Mekkelholt, Praat & Brouwer, 1989). The first category unfortunately comprises professions with opposite expectations regarding the probability of union membership. The second and third category contain professions that are practised in the service sector, which is less unionized. The categories 4 and 6 include professions characterized by a relatively high union density rate, but the expected signs for the categories 5 and 7 are not clear. Unfortunately, this is all the information there was concerning the industrial distribution of the individuals in our sample. Furthermore, we do not know whether the interviewees are blue - or white-collar workers. However, we were able to include a variable indicating whether the individual is working in the public sector, which is highly unionized. A variable whether the individual has a managerial job (measured as the number of people working either directly or indirectly under the respondent 52)) is supposed to obtain a negative sign, because this person might identify more with the employer. Next, the probability of union membership is expected to rise with the company size. Some job characteristics are also taken into account: 1=repetitive work; 2=heavy work; 3=irregular work; 4=dirty work. All these negative aspects of a job are expected to increase the chance of union membership. The dummy variables part-time and temporary job indicate whether the individual has a job of less than 32 hours a week or a job with a labour contract for a specific period of time. Both variables indicate a lower attachment to the labour force, which decreases the probability of joining a union. The dummy variable 'unemployment' denotes whether the individual thinks unemployment in his/her region is high; the dummy variable 'expects unemployment' 51)The denomination, in other countries known as the 'Reformed Church', has been split into two different, though congenial persuasions in the Netherlands, namely the Reformed Church (Gereformeerde Kerk) and the Protestant Church (Nederlands Hervormde Kerk). 52)This variable is divided into six categories, ranging from value 0 to 6, where value 1 stands for 1 to 4 subordinates, value 2 for 5 to 9 subordinates, climbing to value 5 which stands for 50 or more subordinates. The sample mean of 0.628 (table 5.1) therefore underrates the true average number of subordinates.
123
whether the individual is expecting to become unemployed in the coming 12 months. The expected sign of the first unemployment variable is ambiguous, while the second one is positive, as the individual might join or remain in the union in order to be protected against discharge. Job level is classified from 1 (low level) to 7 (high level). The highest job levels are often characterized by executive tasks, and occupied by persons with higher education; therefore a negative relationship with union membership is expected. The dummy variable 'a-d hours' indicates whether the actual hours of work differ from the desired hours of work. The dummy variable 'a-d contract' indicates whether the actual labour contract differs from the desired one. Both variables point at dissatisfaction, which might induce workers to join a union, in order to change their situation. The wage related variables comprehend the net wage rate and the wage rate squared. The latter is added to test whether the probability of joining the union first increases with earnings, to decrease again after a certain income level. We have not accounted for possible endogeneity of the wage rate, as wage packages negotiated by the union in one firm are, by law, also applicable to nonunion members in other firms in the industry: pay differences between union members and nonunion members in the Netherlands are virtually negligible (cf. chapter 3).53) The money value of fringe benefits the individual receives is also included in the equations. The last wage related variables we included are two dummy variables indicating whether the individual is satisfied with his/her job and with his/her wage. When people are more satisfied, unionization becomes less likely. The regional characteristics are represented by a variable for the region in which the individual lives (1=western part of the Netherlands; 0=otherwise) and a variable for the degree of urbanization of the respondent's residence (on a scale from 1 to 13, with 1=small country village, 12=cities with over 100,000 inhabitants, and 13=Amsterdam, Rotterdam and The Hague). Both variables are accompanied by positive expected signs, as a higher concentration of people (in larger cities, and particularly in the West, where the majority of the population and employment is situated) leads to a higher probability of unionization.
53)We have tested this assumption by replicating the Lee (1978) model. The WLS estimates of the wage equation yield very insignificant estimates for the selectivity correction terms, indicating that a union member does not earn more than an average individual in the sample. The same holds for a nonunion member. Inclusion of the expected union-nonunion wage differential in the union membership equation produces bizarre results: the estimated coefficient is negative, suggesting that a positive wage differential decreases the propensity to join.
124
Table 5.1 Means and expected signs° of the explanatory variables total member of union sample FNV CNV OTHER personal characteristics gender (male) nationality (non-Dutch) age tenure (in months) human capital (general) educational level married # children church (regular visitor) Roman Catholic Reformed Protestant father employee
0.653 0.044 36 105 0.881 3.630 0.790 1.085 0.123 0.316 0.122 0.142 0.748
0.816 0.033 39 141 0.867 3.685 0.847 1.271 0.069 0.312 0.130 0.064 0.798
0.820 0.009 40 144 0.820 3.757 0.892 1.432 0.261 0.261 0.234 0.360 0.712
expected sign
0.784 0.049 40 137 0.854 4.249 0.843 1.222 0.141 0.378 0.141 0.124 0.735
+ ─ ± ± ± ─ + + ± ± ± ± +
occupational and industrial characteristics profession 1 (agric./indust.) 0.260 0.342 profession 2 (commercial) 0.046 0.013 profession 3 (administrative) 0.173 0.130 profession 4 (caring) 0.119 0.128 profession 5 (medical) 0.057 0.028 profession 6 (educational) 0.052 0.051 profession 7 (scientific) 0.077 0.077 company size 142 159 manager (# subordinates) 0.628 0.752 civil servant 0.371 0.404 job charact. 1 (repetitive) 0.375 0.348 job charact. 2 (heavy) 0.266 0.338 job charact. 3 (irregular) 0.267 0.376 job charact. 4 (dirty) 0.321 0.389 part-time 0.219 0.118 temporary job 0.079 0.038 unemployment 0.235 0.258 expects unemployment 0.025 0.031 job level 4.149 4.024 a-d hours 0.194 0.642 a-d contract 0.120 0.092
0.288 0.211 0.009 0.016 0.054 0.114 0.144 0.081 0.054 0.070 0.117 0.200 0.072 0.108 132 182 0.730 0.946 0.550 0.697 0.288 0.216 0.297 0.232 0.369 0.249 0.387 0.400 0.090 0.205 0.045 0.032 0.279 0.200 0.036 0.032 4.271 4.874 0.676 0.119 0.072 0.070
± ─ ─ ± ± + ± + ─ + + + + + ─ ─ ± + ─ + +
wage related variables wage rate (per hour) fringe benefits (per month) job satisfaction wage satisfaction
13.26 475 0.884 0.493
13.54 524 0.869 0.482
14.10 507 0.892 0.477
16.18 529 0.897 0.522
± + ─ ─
regional characteristics region (the West) urbanization
0.316 8.660
0.286 8.460
0.288 7.972
0.319 8.681
+ +
number of observations
2589
391
111
185
°)+ and ─ mean respectively: positive, negative effect on unionization decision; ± means: contradictory expectations
125
5.4 Empirical results From the outcome for the nested model (5.9) only the result of σ is of interest here. This will provide guidance in interpreting the results presented in the tables. The nested model yields an estimate for σ of 0.884 with a standard error of 0.333. This result implies that a test on σ = 0 is rejected at the 1% level of significance, while a test on σ = 1 can not be rejected. So, the findings of the nested model clearly show that the simultaneous decision model is decisively rejected against the sequential decision model. Apparently workers first decide whether or not to become a union member and then decide which union to join. The results of the nested logit model demonstrate that the 'Independence of Irrelevant Alternatives' characteristic, which underlies the multinomial logit model, is to be rejected for the union membership decision. The more general model (5.10), in which σ is parametrized, yielded the following coefficients (standard errors in brackets): σ = 2.078 - 0.363 gender - 0.175 educational level (0.552) (0.209) (0.082) The results indicate that there are only significant differences in the decision structure between lower and higher educated workers. We also tested whether the decision structure differs by religion, job level or profession, but none of these variables produced significant results. So, only the higher educated workers are inclined to make a simultaneous decision, instead of a sequential one. Nevertheless, from the overall results of the nested logit model (5.9) we may conclude that we can concentrate our interpretation of the findings on the sequential decision model. If we had not tested for the causal ordering between unionization and the decision which union to join and had wrongly based our interpretation on the simultaneous decision model, some of our conclusions would have been quite different. One example of this is the result on gender. From the multinomial logit model it appears that males are overrepresented in the FNV and other unions. The sequential logit model shows there are no gender differences between unions, but that females have a lower probability of being unionized at all. The full estimation results for both the sequential and the simultaneous decision model are presented in the appendix to this chapter (pp. 150-153). In the former model σ is a priori set to 1, while in the latter σ is a priori set to 0. In a first round of estimates, we included all variables mentioned in section 3; this outcome is shown in table 5.10 of the appendix. Next, we eliminated all variables with an estimated t-value of one or less. This implied that the variables marital status, number of children, the professions 1, 3 and 5 (agricultural/industrial, administrative and medical), manager, a-d contract, fringe benefits and job satisfaction were eliminated from all equations. A likelihood ratio test shows that both for the membership equation and for the 'type of union' equation, the reduced equation can not be rejected against the equations with all variables included. Moreover, elimination of the insignificant variables does not affect the remaining parameters in the model outcome. The results of the reduced sequential model in table 5.2 show quite a large number of significant variables, almost all of which have the expected sign. The first column shows the outcomes for the decision to join a union. The second and third column contain the results of the conditional logit model on the confederation a unionized worker belongs to. The reference category in the conditional model consists of the FNV members.
126
To facilitate reading, all the strongly insignificant coefficients (with a t-value below one) are left out. To predict the effects of changes in one of the independent variables on the probability of becoming a union member, we cannot use the coefficient estimates as shown in table 5.2. For this purpose we have to examine the derivatives with respect to all exogenous variables separately. The interpretation of these derivatives is the same as is the case when interpreting OLS estimates. Table 5.3 presents the mean probability derivatives of the explanatory variables, and is directly derived from the results of table 5.2. Because the probability of becoming a member is a nonlinear function of the exogenous variables, the derivatives with respect to each variable will be different for each value of this variable. In this study we choose to examine the probability derivatives for the average individual in the sample. This is the change in the probability that someone becomes a member as a result of a marginal change in the sample mean for one particular variable. (See Maddala (1983), pp. 23-24). The results show that not all variables affect both the decision to become a member and the choice for a specific union (or federation). Variables that influence the union membership decision but have no effect on which union to join include: gender, age, tenure, a commercial profession, doing heavy or dirty work, having a temporary job, expecting unemployment, the net wage and wage satisfaction, and the regional characteristics. If one has decided to become a member, the choice of a specific union is to a large extent determined by religious conviction, supplemented by the variables civil servant, unemployment, job level, and the difference between actual and desired hours of work. In the subsequent paragraphs, after the presentation of the tables 5.2 and 5.3, we shall discuss all the findings of interest.
127
Table 5.2 Parameter estimates of the union membership sequential model (t-values in brackets) union member intercept
member of union CNV OTHER
-5.469*** -2.139***-3.247*** (11.081) (2.943) (5.254)
personal characteristics gender (male) 0.599*** (4.013) nationality -0.501* -1.503 (non-Dutch) (1.944) (1.364) age 0.037*** (5.622) tenure/100 0.336** (2.432) tenure2/1000 -0.007** (1.966) human capital -0.177 -0.340 (general) (1.189) (1.031) educational level 0.180***-0.099 (3.575) (0.811) church -0.255* 0.563 (1.646) (1.642) Roman Catholic 1.004*** (2.913) Reformed 2.809*** (7.012) Protestant 1.726*** (4.685) father employee 0.285** -0.481* (2.447) (1.769)
0.797 (1.573)
-0.333 (1.170) 0.259** (2.476) 0.541 (1.573) 0.590** (2.489) 0.829** (2.162) 0.411 (1.315) -0.340 (1.454)
occupational and industrial characteristics profession 2 -0.879** (commercial) (2.370) profession 4 0.293* -0.030 -0.645* (caring) (1.753) (0.075) (1.797) profession 6 0.884*** 0.508 0.439 (educational) (4.062) (1.013) (1.151) profession 7 -0.281 -0.526 (scientific) (0.529) (1.352) company size/1000 0.519** -0.417 0.955* (1.982) (0.601) (1.891) civil servant 0.606*** 0.473* 1.009*** (5.551) (1.706) (4.491)
128
Table 5.2 (continued) union member
member of union CNV OTHER
-0.264** -0.313 -0.414* (2.323) (1.134) (1.700) 0.398*** (3.145) 0.400*** 0.109 0.351* (3.781) (0.422) (1.667) 0.495*** (3.893) -0.458***-0.494 0.286 (2.669) (1.164) (0.976) temporary job -0.510** (2.044) unemployment 0.257** 0.232 -0.494** (2.145) (0.830) (1.966) expects 0.749** unemployment (2.433) job level 0.132 0.226** (1.177) (2.566) a-d hours -0.012 -0.033** (0.606) (2.103) wage related variables wage rate (net) 0.187*** (4.228) wage rate squared -0.004*** (4.033) wage satisfaction -0.211** (2.056) regional characteristics region (the West) -0.207* (1.899) urbanization -0.042** (2.513) ______________________________________________________________________
job charact. 1 (repetitive) job charact. 2 (heavy) job charact. 3 (irregular) job charact. 4 (dirty) part-time
# observations 2589 687 Loglikelihood -1264.47 -557.53 pseudo R2 (°) 0.156 0.162 2[logbase - loglikelihood](°°) 466.96 215.84 % union member 0.265 % confederation member 0.162 0.269 * significant at 10% level ** significant at 5% level *** significant at 1% level (°)Definition: 1 – loglikelihood/logbase. Due to the use of discrete variables the real R2 cannot be measured. (°°) This statistic is known as the likelihood ratio test, and indicates the goodness of fit of the model. The values obtained are significant at the 1% level.
129
Table 5.3 Mean probability derivatives° of the explanatory variables union member of union Member CNV OTHER personal characteristics gender 0.097 nationality -0.081 age 0.006 tenure/100 0.054 2 tenure /1000 -0.001 general human capital -0.028 educational level 0.029 church -0.041 Roman Catholic Reformed Protestant father employee 0.046
-0.204
0.194
-0.025 -0.022 0.042 0.090 0.286 0.180 -0.041
-0.041 0.047 0.067 0.057 0.023 -0.002 -0.037
occupational and industrial characteristics profession 2 -0.142 profession 4 0.047 0.023 -0.106 profession 6 0.143 0.040 0.052 profession 7 -0.011 -0.076 company size/1000 0.080 -0.090 0.180 civil servant 0.098 0.013 0.148 job characteristic 1 -0.043 -0.019 -0.056 job characteristic 2 0.064 job characteristic 3 0.064 -0.002 0.054 job characteristic 4 0.080 part-time -0.074 -0.068 0.068 temporary job -0.082 unemployment 0.041 0.047 -0.092 expects unempl. 0.121 job level 0.006 0.032 a-d hours -0.000 -0.005 wage related variables wage rate 0.030 wage rate squared -0.001 wage satisfaction -0.034 regional characteristics region -0.033 urbanization -0.007
° derivatives at the average of the variable values
130
Personal characteristics Females are less organized than males, even after controlling for such variables as part-time or temporary jobs, in which women are overrepresented. This is in accordance with results from other studies and can be explained by the less favourable cost-benefit ratio of membership for women. For many women still counts that they don't intend to stay long in the labour force for family reasons, and that their income is secondary to that of the husband. Moreover, unions on the whole are still a stronghold of men, so that women might have a low assessment of the degree in which unions defend their specific interests. There are, however, no gender differences in the type of union a worker joins, once one has decided to join a union. From table 5.3 it can be concluded that, ceteris paribus, being a male increases the probability of being a union member with 9.7%. Workers who are not born in the Netherlands (non-Dutch for short) are less frequently union member than native workers, as expected. Non-Dutch workers are, however, much more often member of other unions, and much less often member of the CNV than of the FNV. Table 5.3 illustrates this: for a non-Dutch labourer, the chance of joining one of the other unions increases by 19.4%, while the chance of joining the CNV decreases by no fewer than 20.4%. Impressive as these figures may seem, they must be approached with a lot of caution, because the estimates in table 5.2 are only significant at the 20% and 15% level, respectively. Tenure has an inverse U-shaped effect on the membership decision: the probability of becoming unionized first increases with tenure, later it decreases. The tenure effect reaches its peak at 24 years. This is somewhat less than what is found in studies in the United States, where the experience effect peaks between 28 and 35 years for production workers (cf. Hirsch & Addison 1986). This difference probably stems from several causes: firstly, 'tenure' is not exactly the same as 'experience'; secondly, we have controlled for age, which might suppress the tenure effect; and finally, there is the question of a shorter working life in the Netherlands. Tenure does not have a significant effect on the type of union to join. The probability of being a union member rises with age and educational level: older and higher educated workers are more frequently organized than younger and lower educated workers. We more or less expected this result for the age variable, given the theoretical notions, and the fact that Dutch workers below the age of twenty five are heavily underrepresented in the union movement 54), but not for the education variable, since it is assumed that employees with more human capital are less in need of union protection. In two additional tests we left out occupational factors and wage variables respectively, but this did not change the significantly positive relationship between education and membership. An explanation for this remarkable deviation might very well be found in the third column of table 5.2, showing that higher educated workers are more frequently members of other unions than of the FNV or the CNV, probably due to the inclusion of the union for higher staff (MHP) in the category 'others'. As has been mentioned in chapter 3, the MHP emerged in the middle of the 1970's because the higher personnel were beginning to feel the need for a union to protect especially the interests of their own echelon, which deviated ever more from those of the average workers. The figures in table 5.3 support this 54)The overall membership of people below the age of 25 constitutes a meagre 11% of the total union membership (CBS 1989a), whereas their share in the workforce is almost twice as high (about 21%) (CBS 1990).
131
explanation: one step up on the educational scale increases the probability of being a member of another union by 4.7%. Regular church visitors are less often union members. But when they join a union, they more often join the CNV or one of the other unions rather than the FNV. Religious denomination has no effect on the decision to become a member. It has, however, a notable impact on the type of union one joins. Roman Catholics are less frequently members of the FNV and more frequently members of the CNV and of other unions. This is remarkable as the FNV was founded in 1976 by a merger between the socialist and the Catholic confederation. Apparently not all Catholic union members found themselves at home with the non-religious FNV.55) Both the members of the Protestant Church and of the Reformed Churches are more frequently a member of the Christian union CNV than of the general union FNV. Being Reformed or Protestant raises the probability of being a member of the CNV by 28.6% or 18.0%, respectively. If one's father was an employee as well, the probability of being a union member is higher, tentatively suggesting a traditional element in union membership.
Occupational and industrial characteristics
Only a few of the professional dummies have a significant effect. There is no significant difference in profession with respect to the union to be joined once one has decided to join a union at all. Workers with a commercial occupation are less unionized than others, which is presumably related to the low union density rate in the 'commercial part' of the service sector: here employees often work in circumstances that are quite unfavourable to unionization. On the other hand, workers with an educational profession are more frequently union members. Of all the public officers, the educational personnel has been the occupational group with the highest needs for union protection, since the government imposed an extra cutback in their wages in 1983, which persisted throughout the 1980's, whereas their pressure of work has only been increasing. The degree of union membership rises with company size. This might be caused by a smaller attachment to the firm (i.e., the employer) in larger companies. It could also point to the fact that recruitment costs for the unions are lower on average in larger companies. However, from table 5.3 it can be inferred that the magnitude of the effect of company size is very small: a marginal increase of the personnel, with one worker, increases the probability of becoming a member by 0.08/1000 = 0.008%. Individuals working in the public sector are more frequently organized than workers in the private sector, as expected. They are also more frequently member of the CNV and in particular of one of the other unions, than of the FNV. From table 5.3 it can be calculated that being a civil servant has a large impact on the probability of joining a union: ceteris paribus, public sector workers have a 9.8% higher likelihood of being a union member than private sector workers.
55)During the long period of pacification in the Netherlands, there used to be clear links between the unions and political parties in each segment. The merger in the union movement between the socialists and the Catholics, however, was not followed by a same fusion in politics. Instead, a few years later a merger took place between the Catholic and the Reformed political parties into one large Christian party. It would appear that the Catholic workers followed the move of their political party, rather than the move of their trade union.
132
The four job characteristics all have a significant effect on the probability of joining a union, but not really on the choice of a union in particular. Doing repetitive work (job characteristic 1) strikingly enough decreases the likelihood of becoming a member, whereas heavy manual work, dirty work and irregular work increase the overall probability of union membership, as expected. We performed an additional test to check whether this deviating result from monotonous work had been caused by multicollinearity. However, after we eliminated the other three job characteristics, the sign remained negative. We therefore, unfortunately, cannot come up with a satisfactory explanation. Having a part-time job or a temporary job has a significantly negative impact on the likelihood of being a union member, because in those circumstances the costs of membership usually dominate over the gains. For part-time workers the subscription fee usually constitutes a relatively larger share of their income, while the returns are relatively small: work-related matters form a less central part of their life. People with high labour turnover are less inclined to be a member when a change of job to another industrial sector forces them to a change of union (which is common with the FNV and CNV). A temporary job especially has a large negative effect on the probability of becoming a union member, as can be seen from table 5.3. If one has the opinion that unemployment in the region is high, this increases the probability of being a union member, but then it is less likely that one of the other unions is joined. Maybe workers argue that the FNV and CNV have more (judicial) means to prevent discharge. Also, if one expects to become unemployed in the next twelve months, the need for union protection will be higher; in these circumstances the chance of becoming a union member increases by 12.1%. People in higher job levels are not significantly less organized than those in lower job levels. Workers at the higher job levels are, if they are organized, more frequently member of another union, which is probably owing to the share of the MHP in this category. If the desired hours to work differ from the actual hours, this decreases the likelihood of being member of an other union relative to being member of the FNV, but table 5.3 shows that this effect is very small.
Wage related variables
Wages have an inverse U shaped effect on the probability of union membership. This confirms the findings by Bain & Elias (1985), that unionization first increases with wages, but decreases after a certain income level. As the level of earnings rises, the costs of union dues will relatively decline, thereby favouring the benefits/costs ratio of membership; but at the highest income levels this ratio is negatively affected by an increasing chance of employer retaliation, as this usually concerns employees in a managerial position, of whom it is expected that they serve the employer's interests. As has been argued above, these employees often identify more with the employer anyway. In any case, although the wage variables produce significant coefficients in table 5.2, table 5.3 shows that the effect of a pay rise on unionization is quite small. If the net hourly wage rate would increase by 50 cents (which is about 3.75%), this increases the probability of becoming a union member by 1.5%, which is not so much when compared with the derivatives of the majority of the personal and job related characteristics. This links up with the results of the post-war time-series analysis in the preceding chapter, where the wage variable was found to be insignificant. Finally, the wage level has no significant effect on the type of union joined.
133
Workers who are not satisfied with the wage they earn are more often unionized than workers who are satisfied in this respect. Dissatisfied workers may see union membership as a means of improving their conditions.
Regional characteristics
Workers who live in the three western provinces in the Netherlands surprisingly enough have a lower probability of being organized. And finally, contrary to the findings in most empirical work, the degree of urbanization decreases the probability of being organized, although the magnitude of the effect is relatively small. Apparently, the high concentration of the population and of the workforce does not exert the expected positive influence on unionization. The fact that the signs of both regional variables point to the same direction, does not imply that they are necessarily intermingled: if we leave out one of the two, the result for the other variable remains the same. Concerning the negative impact of living in the West, it could be argued that this might be caused, among other things, by the dominance in this area of the service sector, in which, as we have seen, the un-ion density rate is very low. As compared to the rest of the country, the share of the manufacturing industries (higher density rate) is much lower in the West. As for the negative effect of urbanization, this might be explained by the fact that there is less social pressure, but more individualism in the highly urbanized areas of the Netherlands, resulting in a lower degree of unionization. It is possible that on average the inhabitants of Dutch cities have quite a different background and social status than workpeople in American or British cities. Alternatively, one could generalize that the Netherlands are characterized by an urbanization of the entire population, especially if compared to the United States, so that the variable 'degree of urbanization' has a different meaning in this country. 5.5 The second data set As has been announced in the introduction of this chapter, we have the disposal of a second data set, which consists of two repeated cross-sections for the years 1979 and 1987. This enables us to compare the determining factors of unionization in 1979 with those of 1987; these years cover exactly the period of union decline in the Netherlands. This data set concerns the Supplementary Survey of Service Utilization (Aanvullend Voorzieningengebruik Onderzoek, AVO) of the Social and Cultural Planningbureau. The AVO is a random national cross-sectional survey. Both samples consist of over 12,000 individual observations, from which we have taken sub-samples of employees (so, the unemployed, disabled and pensioners were excluded); 5,642 (1979) and 5,350 (1987) observations could be used. The exogenous variables that can be used in the union membership equation, are again subdivided into four categories. Regarding personal characteristics we included gender (1=male; 0=female), marital status (1=married; 0=otherwise), number of children, age, years of education, and political party (1=member; 0= otherwise). We also included four dummy variables for religious denomination: Roman Catholic, Reformed, Protestant, and Other. A dummy indicating whether the individual considered him/herself a 'humanist' was incorporated as well. Inter-industry differences in unionization are captured by five dummy variables for the branch of industry in which the employee works: agricultural, service, transport, finance, and public sector. The reference group consists of workers in the industrial sector. Furthermore, the number of hours worked and a variable indicating the
134
number of sub-ordinates of the individual are included. Unfortunately, the data set does not contain any other industry or plant-specific variables. Wage related variables are covered by net monthly household income56) and net household income squared, and the money value of the fringe benefits. Finally, regional differences in union membership are captured by four variables for the region in which the individual lives (the North, the South, and two separate provinces57)), with the West as reference group, and a dummy variable for the degree of urbanization (1=cities of over 30,000 inhabitants; 0=otherwise). In table 5.4 the mean values of each of the two samples are given. From the data it emerges that the union participation rate has declined from 31.4% in 1979 to 26.4% in 1987; a decline of 5 percentage-points. These figures cannot be fully compared with the actual density rates in the Netherlands as shown in table 3.5 (chapter 3), in which union density is always defined as the number of union members as a percentage of the dependent labour force, including the unemployed. In the AVO-samples we excluded the unemployed.
56)Contrary to our previous analysis on the OSA-data, we do not analyze the effect of individual wages. The reason is, that in the questionnaire individuals were asked after their gross earnings, while households had to indicate their net earnings. It can be assumed that net income exercises a much stronger influence upon union membership than gross income, and this was confirmed by an empirical test: the variable gross individual wage turned out to be insignificant in both years. 57)The North is the provinces Groningen, Friesland and Drente; the South is the provinces Zeeland, Noord-Brabant and Limburg. The Eastern provinces Overijssel and Gelderland are taken separately, because it was empirically not possible to fit them in one of the other two regions in a logical way. When estimating all eight provinces separately, in 1979 the three northern provinces plus Gelderland (which lies more close to the South than Overijssel) produce positive coefficients, while the South plus Overijssel produce negative coefficients. As none of the single provinces comes up with a significant coefficient on its own, it was decided to make two meaningful clusters, keeping Gelderland and Overijssel separately.
135
Table 5.4 Means of the explanatory variables in the samples Variable gender (male) married # children Protestant Reformed Roman Catholic humanist other religions age years of education political party # hours worked manager (# subordinates) agricultural sector industrial sector service sector transport finance sector public sector net household income (per month) fringe benefits (per week) northern provinces southern provinces province Overijssel province Gelderland western provinces urbanization # observations % union members
1979
1987
0.71 0.70 1.54 0.19 0.08 0.38 0.01 0.04 34.73 11.48 0.09 37.48 4.18 0.04 0.35 0.10 0.06 0.04 0.41 2498.20 5.87 0.10 0.26 0.07 0.11 0.46 0.55
0.66 0.67 1.23 0.17 0.08 0.34 0.01 0.04 35.34 12.89 0.05 34.91 4.44 0.05 0.25 0.17 0.06 0.10 0.37 3261.58 6.32 0.11 0.25 0.09 0.12 0.43 0.55
5642 31.35
5350 26.45
In the following paragraphs, the most important changes of the explanatory variables in the two samples will be compared with some actual socio-economic developments in the period under consideration. In addition, we shall formulate several expectations with regard to the effect of these developments upon the union membership rates. The subsequent empirical analysis in the sections 5.6 and 5.7 will show to what extent these expectations are confirmed by the estimation results.
136
Personal characteristics
All empirical studies, including our own preceding analysis, find that being a woman reduces the probability of union membership. Between 1979 and 1987 the share of women in the Dutch workforce shows a steady rise, from 27.9% to 35.1% (cf. Ministry of Social Affairs and Employment (SAE), 1986-1990). Their share in union membership also increased, but at a lower pace, from 13.3% to 15.9% (cf. CBS 1987). Hence, the proportion of women in unions is far below their proportion in the labour force, and this gap has only widened in the period under consideration. From table 5.5 it can be deduced that in the samples as well the share of women in the employed workforce has increased. This development is expected to have contributed to the decreasing union density rate. Although secularization has gone quite far in the Netherlands, statistics (CBS 1989b) show that nominally more than 60% of the population belongs to one of the three main denominations. Further, the main Christian political party (and its predecessors) has been in government since the end of World War I. This may indicate that, in spite of the secularization, religion might still play a role in the decision to join a union, which we already found in the preceding analysis. Dutch studies all show that younger workers are less inclined to join a union (see our preceding analysis; Schippers 1986) or have a higher propensity to leave the union (Van der Veen & Klandermans 1988; Van Rij et al. 1990). Their share in the total trade union membership has declined rapidly (from 15% to 10%; cf. CBS 1987), so that the share of younger workers in unions is far less than their share in the labour force. In the AVO-samples, their share has remained almost constant; but given the declining appeal of union membership to the youth this could contribute to the explanation of union decline. Concerning the level of education, the theory assumes that higher educated workers are less likely to join a trade union. The educational level of the labour force has increased steadily in the last decade, which is also reflected in the two samples. This would imply a negative impact upon unionism. However, in our previous analysis it was found that in the Netherlands more schooling has a significant positive effect on union membership, which might be explained by the fact that in this country several unions operate especially for the promotion of interests of the higher personnel. Contrary to time-series studies, cross-sectional analyses do not pay much attention to the relationship between political preferences and union membership (Kornhauser 1961; Schippers 1986). Our data set only gives information about whether one is a member of a political party at all. It could be argued that people who are more socially active in general, are more inclined to become a union member than "socially passive persons". In the period 1979-1987 all the larger political parties in Parliament lost members, especially after 1981/1982 (cf. Koole 1988). This result can also be found in the two AVO-samples. This would imply that, if our assumption is correct, the decrease in membership of political parties has contributed to the decline of unions.
137
Occupational and industrial characteristics
The number of part-time jobs increased tremendously in a very short period. In 1979 17% of all jobs were part-time jobs (with a working week of less than 35 hours), and in 1987 this had risen to 31.1% (cf. Ministry of SAE, 1985, 1989). This development can also be detected in the two samples in table 5.5, where the average working week declines from 37.5 to 35 hours. This is regarded as unfavourable to union growth, since part-time jobs lead to a lower labour force attachment and hence to a lower propensity to join a union. The effects of the sectoral shift in employment have received a lot of attention in the literature. Van Ours (1992) finds no impact of changes in the industry structure on Dutch union membership. However, Visser (1987) calculates that about one third of the union decline in the Netherlands is caused by the changing sectoral structure. Traditionally, the union movement had its stronghold in the industrial sector. However, during the past two decades employment in this sector has been dwindling. Table 5.5 denotes which proportion of the total working population is employed in each of six sectors of industry, set against the proportional distribution of the working union members in these sectors.58) From this we can draw conclusions on union 'over'- and 'under'representation. For example, it shows that in 1979 36.8% of all employees worked in the industrial sector, while 41.8% of all working union members had a job in this sector. This indicates that the trade union density rate of employees in the industrial sector is higher than average. Sectors with an overrepresentation of unionized workers are: the industrial sector, transport sector and the public sector. Unions are underrepresented in the service sector at large, which is here subdivided into trade, hotel and catering industry ('service sector') and the banking and insurance sector ('finance sector'). In addition, table 5.5 shows the very fast rising share of non-working union members, who are however excluded from our analyses. Comparing these actual shifts in employment with those of the samples in table 5.4, about the same trends can be detected: the share of people working in the industrial sector decreased by no fewer than 10 percentage-points between 1979 and 1987, while the service and finance sector have greatly expanded. The proportion of people working in the public sector slightly declined however, but on the other hand the sample share of 1987 (37%) is in accordance with the real share of 34%. The strong employment decline in the overrepresented industrial sector, in combination with employment growth in the underrepresented sectors is assumed to explain a considerable part of the decline in the union participation rate.
58)Retired, unemployed and occupational disabled workers and union members are excluded.
138
Table 5.5 Percentage o the working population and of the working union members in each branch of industry, 1979-1987 1979
1987
Agricultural sector- % working persons - % working union members
1.59 1.85
1.44 1.58
Industrial sector - % working persons - % working union members
36.79 41.82
28.61 33.19
Service sector - % working persons - % working union members
16.75 4.62
17.29 5.97
Transport sector - % working persons - % working union members
6.83 8.98
6.99 8.86
Finance sector - % working persons - % working union members
9.56 2.40
11.31 3.02
Public sector - % working persons - % working union members
28.46 40.34
34.35 47.40
% non-working union members
11.16
18.09
Source: Own calculations, based on CBS 1987, and Ministry of SAE, 1987-1990
Wage related variables
The effect of the level of earnings on union membership has been studied a great deal. Although the findings are somewhat mixed, one could argue that lower-income workers have more need of a union than high-income workers. In the previous analysis we found a parabolic relationship: the probability of unionization first increases with earnings, and after a certain level it decreases again. In the Netherlands, the rise in the average level of real earnings has slowed down between 1979 and 1983, but has subsequently recovered (cf. CPB, 1980-1988). In the AVO-samples, the average net nominal household income rose about 30% over the whole 8-year period, but if we correct these figures for price developments (according to the CBS price index of household expenditures), real income has only increased with 7%. If real earnings rise one could theoretically expect a positive (credit) effect on unionization, but given the practise of extending a collective labour agreement to all workers in an industry, the impact of an income improvement on union membership is expected to be modest in the Netherlands; this result was not only found in chapter 4, but was also indicated by the rather small values of the mean probability derivatives with respect to the wage variables of the OSA-sample in table 5.3.
Regional characteristics
It is often stated in American reports, that the chance of unionization rises with the degree of urbanization. A higher concentration of the work force not only lowers the costs of union organizing, but also promotes the so-called 'proximity influence' (i.e., workers stimulate each other to join the labour movement) and enlarges the bargaining power of the trade unions. In the Netherlands, large areas are heavily urbanized. If we look at actual developments (CBS 1979-1988), we see that not much has changed between 1979 and 1987: the share of the smallest and the largest towns in the total population decreased by 0.3% and 1.3% respectively, in favour of the
139
medium sized towns. This precludes giving any meaningful predictions. We do know however, that our previous analysis came up with a surprisingly negative sign of the urbanization coefficient, rejecting the theoretical assumptions. A second group of variables in this category concerns regional differences between provinces. In the Netherlands, several regions with their own specific features can be distinguished. Statistics (CBS 1979-1988) show, that the distribution of the Dutch over the provinces has not changed at all in the period under consideration. This is confirmed by the sample means of 1979 and 1987. A priori, it is not possible to predict the impact of living in a certain area on the union membership probability. 5.6 Econometric framework Changes in union membership can be caused by changes in observable characteristics of employees and by changes in behaviour (i.e. the propensity to join a union). Assume that the relation between (a vector of) observable characteristics of workers xi and the propensity to become a union member y in year i can be expressed as: y = βixi + εi (5.11) where βi is a vector of coefficients and εi is an identically and independently distributed error term. Changes over time in y can be due to changes in the characteristics xi and/or to changes in the associated parameters βi. In practice, y is unobservable. What we observe is a dummy variable y defined by y = 1 if y > 0 and y = 0 otherwise (see Maddala 1983, p. 22). Stoker (1985) has developed a framework for decomposing a change of an aggregate variable into a change of the behavioural model and a change in distribution of the characteristics. Gomulka and Stern (1990) apply this method for decomposing a binary dependent variable. Assume that εi is normally distributed. The probability of being a union member in year i (yi = 1) conditional on the observed characteristics xi is: (5.12) P(yi = 1_ xi) = P(εi > -βixi) 59) This is the probit model . The change in the average probability of union membership between year i and year j can be decomposed as (cf. Gomulka and Stern 1990): Pa(βjxj) - Pa(βixi) = {Pa(βjxj) - Pa(βixj)} + {Pa(βixj) - Pa(βixi)} (5.13) where Pa(βixj) is the average predicted probability of union membership using characteristics xj and coefficients βi. The first term in brackets in equation (5.13) describes the change in the average probability due to changes in the coefficients. The second term gives the contribution of a change in the distribution of the characteristics to the change in the average probability of union membership. To find the contribution to the change in average union membership probability of each coefficient separately, we can further decompose the first term in equation (5.13): Pa(βjxj) - Pa(βixj) = {Pa(βjxj) - Pa(βxj)} + {Pa(βxj) - Pa(βixj)} (5.14) th th where β is the vector βj, with the k element replaced by the k element of βi. First we predict union membership using the sample and the coefficients of years j. This 59)The probit model is just like the logit model, a discrete choice model. The two models only differ with respect to the distribution of the errors (see for instance Maddala (1983) for a comparison between logit and probit models).
140
yields Pa(βjxj). Next, we take all coefficients of year j except for the kth coefficient for which we take the year i value. With this set of coefficients we can again predict membership. This yields Pa(βxj). Taking the difference between Pa(βjxj) and Pa(βxj) gives the contribution of the kth coefficient in year i to the change in union membership. Besides decomposing the changes in the average union membership probability into a 'coefficient effect' and a 'variable effect', we shall also test for parameter equality between year i and year j. This comes down to a test on βi = βj. Rejection of this test indicates a structural change. By performing a test on parameter equality for each of the coefficients separately, structural changes in the marginal contribution to the probability of union membership of each of the variables can be established. Parameter equality is tested by likelihood ratio tests. 5.7 Empirical results Table 5.6 contains the parameter estimates of the probit model (5.12), which mostly have the expected signs. In both years men and those who are member of a political party have a higher probability of being a union member than women and those who are not a member of a political party. Even though these probit estimates can not be interpreted in the usual (OLS) way60), the fact that the gender coefficient substantially declines between 1979 and 1987 does suggest that the positive behavioural effect on union membership of being a male has decreased. Whether this constitutes a significant difference, will be tested further on in this section. Anyhow, given the positive sign and the fact that the share of males in the samples has declined (cf. table 5.4), this shows that the change in the distribution of the working population in this respect has been unfavourable to union growth. Being a Catholic or of another religious conviction lowers the chance of union membership. The same holds for the number of children, contrary to what theory predicts. It was assumed that having greater family responsibilities increases the chance of membership. We cannot really come up with a satisfactory explanation, other then the possibility that workers with children have more expenses, and might therefore decide to refrain from the costs of union membership. Further, the probability of union membership rises with age, education, and the number of hours of work in both samples. The increase of the age coefficient suggests that the propensity of younger workers has indeed become smaller, as was assumed. With respect to the hours variable, the fact that the share of part-timers in the samples has increased (cf. table 5.4), confirms that the composition of the labour force in this respect has become less advantageous for the union density rate. The industry dummies all are highly significant and show that workers in the transport and public sector are more highly organized, while workers in the agricultural, commercial services, and banking sector are less frequently organized than workers in the industrial sector. The changes in the coefficient values between 1979 and 1987 suggest that that the negative effects of working in the agriculture and of working in the service sector have diminished, while the positive effects of a job in transport and in the civil service have gained in weight. Finally, living in one of the northern provinces increases the probability of union membership. 60)The estimated coefficients are in fact the vector β of equation (5.11), and describe changes in the latent endogenous variable y , which is a continuous variable as opposed to the observed variable y, which can only take on the value 0 or 1. i
*
141
Some variables have a significant effect in one year but not in the other. For 1987 we find a significant negative effect of being Protestant, but not for 1979. The variable 'urbanization' only has an effect in 1979. The sign of this effect is negative, indicating that workers living in large cities are less organized. This refutes the so-called 'proximity' hypothesis, and confirms our findings of the OSA-model. Living in the province Gelderland only has a significantly positive effect in 1987. In table 5.6 it can be seen that the coefficients of all provinces become positive in 1987, which suggests that living in the West (i.e., the reference group) has ever more decreased the chance of being a union member; this also confirms the negative sign found in the OSAmodel. Overall, the wage-related variables performed rather poorly. This last result might be ascribed to the suspicion that workers determine their choice whether or not to join a union on the basis of their own salary rather than on the basis of the total household income. As has been explained in footnote 56, we were not able to include net individual wages in this analysis.
142
Table 5.6 Parameter estimates of the explanatory variables (t-values in brackets) 1979
1987
intercept
-2.247***
(11.912)
-2.299***
(11.492)
personal characteristics
gender (male) married # children Protestant Reformed Roman Catholic humanist other religions age years of education political party
0.465*** 0.090* -0.039*** -0.070 -0.032 -0.182*** -0.042 -0.256*** 0.016*** 0.028*** 0.365***
(8.522) (1.842) (2.754) (1.281) (0.456) (3.632) (0.233) (2.579) (9.132) (4.588) (5.840)
0.341*** 0.075 -0.038** -0.184*** 0.017 -0.136*** -0.011 -0.320*** 0.019*** 0.017** 0.458***
(6.394) (1.533) (2.124) (3.176) (0.219) (2.669) (0.061) (2.942) (8.984) (2.014) (5.774)
occupational characteristics
# hours worked manager agricultural sector service sector transport finance sector public sector
0.014*** 0.001 -0.429*** -0.545*** 0.179** -0.388*** 0.092**
(4.811) (1.078) (3.810) (7.137) (2.295) (3.484) (2.042)
0.012*** -0.000 -0.209** -0.324*** 0.255*** -0.358*** 0.152***
(4.942) (0.312) (2.241) (5.019) (3.036) (4.671) (2.834)
wage related variables
household inc./100 household inc.2/1000 fringe benefits/100
0.012 -0.000 -0.055*
(1.242) (1.530) (1.701)
0.011 -0.000* 0.026
(1.480) (1.879) (0.758)
regional characteristics
urbanization northern provinces southern provinces province Overijssel province Gelderland
-0.120*** 0.132** -0.090* -0.047 0.095
(2.952) (2.046) (1.701) (0.624) (1.523)
-0.029 0.153** 0.027 0.019 0.132**
(0.682) (2.248) (0.484) (0.250) (2.014)
Loglikelihood R2 correctly predicted
-3147.46 0.12 0.70
-2840.75 0.09 0.74
* significant at 10% level; ** significant at 5% level; *** significant at 1% level
143
On the basis of the results of table 5.6, it is possible to apply the decomposition method announced in the preceding section. In that way it can be determined to what extent the decline of the trade union density rate between 1979 and 1987 is caused by changes in the characteristics of workers (referred to as the 'variable effect') on the one hand, and by changes in the unionization behaviour of workers (the 'coefficient effect') on the other. Before this decomposition was executed, we have performed likelihood ratio tests on parameter stability concerning the probit results of 1979 and 1987. First, we have tested for overall stability of the coefficients. For this purpose we re-estimated our model on the pooled 1979/1987 data. The value of the log likelihood function for this pooled sample is - 6006.51, while the value of the likelihood ratio test is 36.6, which indicates that the null-hypothesis of total parameter stability between 1979 and 1987 cannot be rejected. This would imply that there is no structural change between the coefficients of 1979 and those of 1987, which in turn suggests that the decomposition method, unfortunately, might not have so much meaning. Still, it might be the case that within the total group of coefficients, some of the single coefficients have changed significantly. Therefore, we also looked at the stability of the parameters separately. In table 5.7 the results of the likelihood ratio tests are presented. Some variables have been reduced to the same denominator. Marital status and number of children is 'family structure', the five religious variables are 'religion', the five employment sectors are 'industry', the three income variables are 'earnings' and finally the five regional characteristics are 'region'. Table 5.7 Loglikelihood ratio tests for parameter stability 61) 1979/87 1987/79 intercept
0.06
0.06
5.24**
5.46**
family structure
0.10
0.08
religion
9.98*
7.94
age
2.36*
1.68
education
3.50*
1.92
political party
2.22
1.38
hours worked
0.26
0.36
manager
1.96
1.82
industry
11.56**
15.18***
earnings
2.82
11.62***
region
8.20
7.46
male
61)In each column the results of the test statistic 2[loglikelihood(u) - loglikelihood(r)] are shown. Here (u) stands for the unrestricted model of 1979 in the first column, and for the unrestricted 1987 model in the second column, while (r) stands for the same model except for the coefficient value of the particular variable denoted in each row, which is replaced by its 1987 value in the first column, and by its 1979 value in the second one.
144
The likelihood ratio test statistics show that especially the coefficients of gender (male) and of the industry variables have changed significantly over the period, next to the coefficients of religion, age, education, and earnings. From this we may conclude that the decomposition of the change in union membership into a change in the distribution of the population and a change in behaviour is not that senseless; although this was prompted by the absense of a structural change in the coefficients with regard to the whole model, the results of table 5.7 indicate that some of the separate parameters did change significantly. That is why the decomposition method as set out in section 5.6 will be executed after all, beginning with an analysis of the whole model. In table 5.8 the results for the decomposition of the change of the union membership probability into a variable effect and a coefficient effect are presented, based on equation (5.13). The diagonal elements give the average predicted probability using coefficients and sample of the same year. The mean predicted probabilities correspond closely to the actual proportion of union members in each year (cf. table 5.4). The columns of table 5.8 give the variable effect, while the rows give the coefficient effect. This table shows for example that if the characteristics of the workers in 1987 had been those of 1979, the behavioural model in 1987 would have predicted union density to be 28.7%. Alternatively, if the behaviour of workers in 1987 had been the same as in 1979, the change in their characteristics would have caused union density to be 28.0% in 1987, instead of 26.4%. The total predicted decline in trade union density between 1979 and 1987 was 4.9 percentage-points. If we look along the rows of table 5.8, we can get an estimate for the decline in union density due to changes in the coefficients, reflecting changes in unionization behaviour. When concentrating on the situation in 1987, we see that 1.6 percentage-points of the decline in union density are due to changes in the model structure. The columns show the change in union density due to changes in the distribution of characteristics in the sample. It follows that 3.3 percentage-points of the decline in union density can be attributed to changes in the workforce. So, this suggests that the contribution of changes in behaviour to the decline in union density is smaller than the contribution of changes in the characteristics of workers: approximately 67% of the 4.9 percentage-points decline in union density can be said to be caused by changes in the population, while 33% is due to changes in unionization behaviour. Table 5.8 Mean union membership probability using coefficients for year j and sample for year i (t-tests in brackets) Coefficients sample
1979
1987
1979
31.3
1987
28.0 (23.6) 0.30
skewness mean probability year i and sample i
28.7 (39.3) 26.4 0.64
145
In table 5.8 the null-hypothesis of the t-test for paired samples is that the mean predicted probability of union membership in the sample of year i using the coefficients of year j is equal to the mean predicted probability using both the sample and the coefficients of year i. The t-values are significant at the 1% level, indicating significant differences in the mean predicted probability of union membership. Contrary to the joint stability test mentioned on page 144, this result indicates that differences in the coefficients between 1979 and 1987 do have a significant impact on the union decline. The last row of the table gives the skewness of the distribution of the predicted probabilities of union membership using the sample and coefficients of the same year. In both years the distribution of predicted probabilities is skewed to the right. So, more than 50% of the employees in the sample have a predicted probability of being a union member that is less than the average probability. This suggests that a large fraction of the employees has a very small chance of being a union member. The right skewness increases over the years, which implies that the fraction of employees with a low probability of being a union member has been rising. To determine which variables show the largest differences in behaviour, we estimated the mean predicted probabilities of a change in the coefficients separately, in the way that was formulated in equation (5.14). The most interesting results can be found in table 5.9. With respect to each particular variable, the upper left and lower right segments give the mean predicted probabilities of union membership using both the sample and the coefficients of the same year. In each row the characteristics remain unchanged, while only the coefficient of the variable at issue varies: in the upper right segment the 1987 coefficient of one specific variable is imposed on the model outcome of 1979 (hence, the coefficients of all other variables keep their 1979 values), while in the lower left segment the 1979 coefficient of the variable at issue is imposed on the model outcome of 1987. We only report results for variables that show a marked effect. Table 5.9 Mean predicted probabilities of varying coefficients coefficients
male
age
earnings
sample
1979
1987
1979
31.3
28.2
1987
29.2
26.4
1979
31.3
34.4
1987
23.5
26.4
1979
31.3
32.2
1987
24.9
26.4
coefficients
education
industry
region
sample
1979
1987
1979
31.3
27.2
1987
31.1
26.4
1979
31.3
33.0
1987
24.4
26.4
1979
31.3
35.2
1987
23.8
26.4
146
The changes in the model structure (i.e., the coefficient effects) are not all in the same direction, as can be seen if we look along the rows of table 5.9. Some effects are positive to the union membership decision, others are negative. Responsible for a decline in membership are the coefficients for gender and for education. From this we may conclude that the propensity of males and the propensity by educational level to join a union has declined. The effects of age, earnings, industry and region on union membership have become more positive. Within these last two categories, the variables that contribute most to this positive trend are: the service and public sector on the one hand, and the southern provinces and the urbanization dummy on the other. The results with respect to all six (groups of) coefficients confirm the outcome of table 5.6, in which the probit estimates of gender and education became smaller, and those of the other four variables larger (or less negative). From this it can be inferred that the differences in unionization behaviour between men and women on the one hand, and between lower and higher educated workers on the other hand, have diminished. On the contrary, the differences in the propensity to join a union between younger and older workers have become larger, and so have the differences be-tween higher and lower salaried workers. One last notable development can be detected with regard to the industry dummies: it would appear that the industrial sector, the (former) stronghold of the trade union movement, is ever more losing ground: the differences in unionization behaviour between civil servants and manufacturing workers have become larger at the expense of the industrial sector, while the differences in the propensity to join between workers in the service sector and those in the industrial sector have become significantly smaller. 5.8 Conclusions The cross-sectional analyses of this chapter have indeed produced many more determinants of unionization, than we were able to show and test by means of the time-series analyses of chapter 4. Both the OSA and the AVO data sets contain a large number of variables that are brought forward by the theories expounded in chapter 2. As far as possible, the validity of these theories have been tested with respect to the determination of Dutch union membership. The OSA-survey includes the most variables suitable for our purposes, while the second advantage of this data set lies in the fact that we were able to make a distinction between the factors that influence the decision to join a union at all, and the factors that affect the choice for a specific federation. The great benefit of the AVO-survey is the availability of two different observations, 1979 and 1987, which cover the period of union decline in the Netherlands exactly. In that way, it has become possible to shed some light on the determining factors of the downfall in the union membership. From the analysis executed on the OSA-data it is clear that a considerable number of variables are significant, most of them with the expected sign. If a variable has an effect contrary to its expectation this can for a large part be attributed to specific Dutch circumstances, namely with respect to education, urbanization and the densely populated West. This demonstrates that a general theory on the determinants of trade union membership can never be wholly universal, but will always have to allow for nationally determined exceptions. The OSA-analysis further shows the importance of careful specification of the structure of the union membership decision process. The comparison of the results of the multinomial logit and the sequential logit model shows that misspecification of
147
the decision structure can lead to some false conclusions. For the majority of the Dutch workers the decision to become a trade union member is taken prior to the decision which particular union to join. Important positive factors to join a union in the Netherlands are: gender (male), age, tenure, educational level, an educational profession, working in a large company, being a civil servant, doing heavy, irregular or dirty work, the judging that unemployment in one's region is high and the expectation to become unemployed oneself. Significantly detrimental effects on unionization are radiated by practising a commercial profession, having a part-time or a temporary job, and living in a larger city. Only a few variables have a significant impact on the decision which union to join. Civil servants more often choose for the CNV or for one of the other unions than for the FNV. Workers with a higher educational or job level are more often a member of one of the other unions, as expected: employees with those characteristics more often join a union which especially promotes the interests of higher personnel. Strikingly enough, the choice for a specific confederation is for a large part determined by the religious persuasions. So, in spite of an ongoing secularization, religion still seems to play an important role in the unionization decision. The results of the analyses executed on the AVO-data mostly confirm the conclusions drawn from the OSA-study, with respect to the signs and levels of significance of the coefficients. Very useful additional information can be derived from the comparison of the sample and the parameters of 1979 against those of 1987, because it enables us to analyze the fall in Dutch union membership in this period. We expect and find that this decline can partly be explained by the rising share of female and part-time employment, the waning interest of younger workers to join a union, and the sectoral shifts in employment from the industrial towards the service sector. To examine these, and other, effects in more detail, we applied a decomposition method by which it becomes possible to ascertain to what extent the decline in union density is caused by an increase in the share in the employed labour force of employees with a low propensity to join and to what extent by a diminishing appeal of unions to workers. The empirical analysis shows that two-third of the decline in union density is caused by changes in the working population, while one-third is the result of changes in unionization behaviour. The mixed results of the parameter stability tests indicate that we should interpret the impact of behavioural changes with caution. This conclusion will be closed with an overall evaluation of the cross-sectional results in terms of the demand and supply framework, about which we can be brief. The determinants included in the models are grouped in the categories personal and occupational characteristics, wage related variables and regional characteristics. This implies that all variables employed are linked to the demand-side of the theoretical framework. As was set forth in chapter 2, we can also distinguish another group of determinants, connected with the impact of union leadership and policies on trade union growth. In that section (2.3.6) it is argued that the degree to which unions succeed in improving the terms of employment, and the degree to which they are able to remain attractive for individual members, influence the perceptions about union effectiveness, which in turn affect the decision whether or not to be (or stay) a union member. In other words, measures for perceptions about union policies also reflect the demand-side of our framework, but they are in fact closely related to the supply-side
148
as well, because they form a reaction to the supply of union services. Unfortunately, the OSA- nor the AVO-data contain any variables that measure the valuation of workers with respect to the unions' performance. From this we come to the conclusion that the cross-sectional analyses of this chapter have contributed a great deal to the explanation of Dutch unionization. The demandside of our framework has been elaborated to a large extent, but the supply-side has remained underexposed due to a lack of the required information in the data sets. To fill this gap, in the next chapter attention will be paid to theories and empirical work of Dutch social scientists, who partly lay stress on other determinants of unionization, and who have the disposal of other data sets, which contain more information about the role of unions themselves in the explanation of union membership.
149
APPENDIX Table 5.10 Parameter estimates of the unrestricted union membership sequential model (t-values in brackets) union intercept personal characteristics gender nationality age tenure/100 0.337** (2.379) tenure squared/1000-0.007*-0.008-0.003 (1.925)(0.980)(0.429) general human capital educational level marital status # children church Roman Catholic Reformed Protestant father employee
member of union member CNV *** -2.319 -5.344 (9.458) (1.517)
OTHER -3.520*** (2.618)
0.533*** (3.332) -0.539** (2.069) 0.038*** (5.508) 0.205 (0.575)
0.129 (0.322) -1.701 (1.514) 0.019 (1.109) -0.003 (0.886)
0.215 (0.683) 0.927* (1.754) 0.002 (0.110)
-0.171 (1.134) 0.197*** (3.775) -0.066 (0.437) 0.003 (0.066) -0.311* (1.770) 0.036 (0.278) 0.134 (0.750) 0.117 (0.747) 0.293** (2.498)
-0.359 (1.042) -0.088 (0.682) 0.276 (0.655) -0.010 (0.085) 0.659* (1.760) 0.994*** (2.818) 2.671*** (6.482) 1.679*** (4.317) -0.443 (1.545)
-0.455 (1.527) 0.205* (1.830) 0.083 (0.260) -0.101 (0.947) 0.722* (1.951) 0.641*** (2.623) 0.820** (2.082) 0.420 (1.269) -0.324 (1.313)
-0.230 (0.660) -0.899 (0.706) -0.499 (0.919) -0.115 (0.256) 0.514 (0.735) 0.534 (0.965) -0.300 (0.519) -0.627 (0.868)
-0.126 (0.398) 0.573 (0.649) 0.258 (0.699) -0.513 (1.257) 0.372 (0.660) 0.476 (1.064) -0.611 (1.348) 0.866* (1.668)
occupational and industrial characteristics profession 1 0.118 (0.810) profession 2 -0.947** (2.471) profession 3 -0.155 (0.885) profession 4 0.229 (1.210) profession 5 -0.181 (0.676) profession 6 0.856*** (3.351) profession 7 -0.147 (0.641) company size/1000 0.535** (2.012)
150
manager civil servant job characteristic 1 job characteristic 2 job characteristic 3 job characteristic 4 part-time temporary job unemployment expects unemployment job level a-d hours a-d contract
wage related variables wage rate wage rate squared fringe benefits/1000 job satisfaction wage satisfaction
regional characteristics region urbanization
# observations loglikelihood pseudo R2 2[logbase - loglikelihood] percentage union member percentage confederation member
0.015 (0.354) 0.654*** (5.783) -0.276** (2.394) 0.370*** (2.835) 0.407*** (3.775) 0.439*** (3.349) -0.424** (2.327) -0.627** (2.246) 0.245** (2.026) 0.725** (2.340) -0.034 (0.715) 0.002 (0.277) 0.159 (0.801)
0.002 (0.023) 0.452 (1.474) -0.355 (1.206) -0.121 (0.388) 0.072 (0.261) 0.297 (0.908) -0.675 (1.342) 0.219 (0.297) 0.258 (0.872) -0.028 (0.038) 0.111 (0.909) -0.014 (0.719) 0.242 (0.442)
0.024 (0.290) 1.082*** (4.486) -0.302 (1.165) -0.044 (0.154) 0.245 (1.088) 0.231 (0.801) 0.051 (0.146) -0.172 (0.248) -0.522** (1.973) -0.015 (0.023) 0.194** (1.988) -0.033** (2.033) 0.230 (0.517)
0.196*** (4.291) -0.005*** (4.076) 0.093 (1.040) -0.160 (0.992) -0.208** (1.994)
-0.104 (0.798) 0.004 (1.034) -0.087 (0.393) 0.020 (0.050) -0.076 (0.292)
-0.207 (0.176) 0.003 (0.975) -0.006 (0.035) 0.172 (0.502) -0.081 (0.377)
-0.202* (1.808) -0.041** (2.356)
-0.032 (0.113) 0.004 (0.086)
0.078 (0.338) 0.014 (0.388)
2589 -1262.18 0.173 471.54 0.265 0.162
687 -544.65 0.296 241.55 0.269
* significant at 10% level ** significant at 5% level *** significant at 1% level
151
Table 5.11 Parameter estimates of the union membership simultaneous model (t-values in brackets) member of union FNV -6.259*** (7.947)
CNV -7.127*** (5.853)
OTHER -8.686*** (8.845)
0.478** (2.380) nationality -0.529 (1.644) age 0.033*** (3.943) (3.784) tenure/100 0.349** (2.092) tenure squared/1000 -0.005 (1.352) general human capital -0.026 (0.144) educational level 0.175*** (2.768) marital status -0.119 (0.649) # children 0.025 (0.396) church -0.602** (2.430) Roman Catholic -0.239* (1.688) Reformed -0.645** (2.515) Protestant -0.201 (1.051) father employee 0.401*** (2.727) occupational and industrial characteristics profession 1 0.176 (1.018) profession 2 -0.997** (2.040) profession 3 -0.074 (0.356) profession 4 0.362 (1.602) profession 5 -0.398 (1.059) profession 6 0.494 (1.452) profession 7 0.075 (0.266) company size/1000 0.405 (1.262)
0.402 (1.116) -2.164** (2.072) 0.053*** (3.007) 0.494 (1.606) -0.013* (1.719) -0.401 (1.367) 0.100 (0.922) 0.077 (0.210) 0.017 (0.173) -0.174 (0.608) 0.808** (2.436) 2.129*** (6.061) 1.449*** (4.181) 0.081 (0.340)
0.629** (2.383) -0.083 (0.206) 0.036*** 0.222 (0.912) -0.006 (0.953) -0.343 (1.372) 0.306*** (3.360) -0.029 (0.114) -0.101 (1.186) -0.007 (0.023) 0.514** (2.490) 0.197 (0.626) 0.293 (1.080) 0.205 (1.057)
-0.122 (0.402) -1.586 (1.479) -0.923* (1.925) 0.295 (0.788) -0.106 (0.196) 0.864* (1.777) -0.444 (0.866) -0.280 (0.446)
0.125 (0.461) -0.536 (0.835) 0.004 (0.014) -0.134 (0.384) -0.044 (0.107) 0.847** (2.306) -0.491 (1.304) 1.109*** (2.625)
intercept personal characteristics gender
152
manager civil servant job characteristic 1 job characteristic 2 job characteristic 3 job characteristic 4 part-time temporary job unemployment expects unemployment job level a-d hours a-d contract
wage related variables wage rate wage rate squared fringe benefits/1000 job satisfaction wage satisfaction
regional characteristics region urbanization
# observations loglikelihood pseudo R2 2[logbase - loglikelihood] percentage union member
0.012 (0.220) 0.311** (2.262) -0.153 (1.120) 0.420*** (2.733) 0.328** (2.534) 0.392** (2.522) -0.405* (1.742) -0.662* (1.912) 0.317** (2.212) 0.625* (1.672) -0.089 (1.571) 0.009 (0.927) 0.243 (1.037)
-0.018 (0.198) 0.773*** (3.072) 0.517** (2.096) 0.263 (0.980) 0.417* (1.801) 0.554** (2.072) -1.219*** (2.759) -0.238 (0.396) 0.469* (1.890) 0.554 (0.891) -0.010 (0.091) 0.002 (0.087) 0.083 (0.174)
0.047 (0.696) 1.357*** (6.802) -0.449** (2.096) 0.299 (1.244) 0.615*** (3.325) 0.447* (1.862) -0.101 (0.351) -0.813 (1.526) -0.019 (0.090) 0.858 (1.625) 0.109 (1.294) -0.013 (0.918) 0.122 (0.329)
0.351*** (4.261) -0.010*** (4.159) 0.100 (0.948) -0.182 (0.969) -0.207* (1.654)
0.141 (1.466) -0.003 (1.394) 0.031 (0.160) -0.245 (0.697) -0.262 (1.170)
0.210*** (3.128) -0.004*** (2.772) 0.149 (1.012) -0.101 (0.356) -0.184 (1.035)
-0.243* (1.793) -0.041** (1.992)
-0.228 (0.939) -0.058 (1.539)
-0.082 (0.432) -0.031 (1.061)
2589 -1805.96 0.241 714.820 0.151
0.043
0.071
* significant at 10% level ** significant at 5% level *** significant at 1% level
153
154
6. SUPPLEMENTARY LINES OF APPROACH FROM THE SOCIAL SCIENCES 6.1 Introduction In the preceding two chapters emphasis has been laid on economic modelling of union growth and decline. Within these models several variables could be analyzed that were not purely economic by nature, as economic theory does allow for such determinants to enter the models. For example, the time-series analysis can include variables that proxy the political climate, such as anti- or pro-union legislation and the share of leftist parties in Parliament. In chapter 4 we found that the rise of the welfare state has had a negative impact on Dutch union growth. In the cross-sectional analyses of chapter 5 we measured the influence of religion, as well as tradition. Depending on the available data, cross-sectional models can capture many more variables that are emphasized by the social sciences, and they provide some important additional explanations of union membership decision making. Social scientists argue, that economic theory can explain the conditions under which individuals decide whether or not to join a union, but that it fails to show how the actual decision to become a member is determined. Given the limited scope of this thesis, it is not possible to give a comprehensive explanation of all the reasons why people join a union or not. However, it would mean a shortcoming if the contribution from the social sciences were neglected altogether. Therefore this last chapter intends to provide a short overview of meaningful social theories, without the pretention of exhaustiveness. In section 6.2 the point of departure will be a survey by Klandermans (1986), who discusses three lines of approach to union participation within the social psychology. In addition, several sociologic theories will be touched upon. Consequent on these theories, social scientists have performed empirical studies with respect to the Netherlands, the most interesting of which will be discussed in section 6.3. We are especially interested in the question to what extent these results can provide a valuable contribution to the (economic) explanation of Dutch trade union growth and decline. The chapter will be closed with a small empirical exercise (in section 6.4), in which we have tried to incorporate several sociological insights into a socio-economic model. After the conclusion in section 6.5, an extensive appendix is taken up containing all the data as well as the mathematical proof of the assumptions underlying the simultaneous models in section 6.4. 6.2 Sociologic and psychologic theories of union membership Within the social psychology, Klandermans (1986) distinguishes three theoretical lines of approach to trade union participation: the frustration-aggression approach, the rational-choice approach, and the interactionist approach. Frustration-aggression theory explains union membership as a result of some kind of dissatisfaction with someone's working conditions. In our cross-section analyses in chapter 5 we already allowed for several of such 'dissatisfaction variables' with respect to working hours and wages. Furthermore, frustration partly finds expression in the form of strikes, the impact of which we examined in chapter 4. Alternatively, frustration-aggression theory can also explain membership resignation in terms of frustration with union policies. Klandermans argues, however, that dissatisfaction in
155
itself is neither a necessary, nor a sufficient condition for union participation: frustration only affects the union membership decision through cost-benefits deliberations or through the social context. In the rational-choice approach union participation is the outcome of a process in which the costs of joining are weighed against the benefits. This approach also underlies economic theories of union membership, and forms the basis of our empirical studies in chapter 4 and 5. However, whereas economic theories stress individual, selective costs and benefits (in a financial sense), the social sciences pay attention to other rational deliberations as well. The decision whether to join or not is also influenced by collective, ideological and social motives. These motives are more difficult to measure and to operationalize, but have nonetheless come to light via questionnaires and the like. Hence, attitude towards union membership is determined by balancing all the positive and negative consequences of participation. The actual decision-making is then determined by this balance of costs and benefits, combined with expectations about the degree in which the union is able to realize these motives; this refers to the so-called value-expectancy theory (Van Teeffelen & Klandermans 1989, p. 4). The interactionist approach ascribes union membership decisions to the influence of an individual's social context. Here a distinction is made between the living and the working environment. Concerning the living environment: on the one hand tradition plays a role (has someone's father also been a union member, or not), and on the other the prevailing opinions of someone's surroundings (group culture, social custom to be a member, or not). The social custom theory with regard to union membership has been explained by Booth (1985), and Naylor & Cripps (1989), respectively: they show that, even in absence of a closed shop or of union services exclusively for members, people join a union anyway, as long as it is the social custom to do so (or else one will loose his/her reputation). Here it is clear that interaction and rational-choice theory overlap. Concerning the working environment: on the one hand the union density rate of someone's company/industry plays a role, and on the other hand the degree in which employees come into contact with the union on the workfloor. The first line of approach is also in use in economic literature, and has been elaborated in chapter 5 and 6; the second line of approach is far less developed by economists. However, this is an important factor when it comes to commitment; the big problem of (Dutch) unions is not so much the recruitment of new members, but trying to hold on to their existing members. The turnover rate of union members is very large indeed, as has been established in chapter 3. To keep members committed, the unions have to make sure two things: members should agree to the union policies, and members should feel at home within the union's culture. Either way, unions must create facilities to establish and maintain good communication between members and leaders. That is why the organizational structure of unions plays an important role, analysis of which is typically the sphere of work of social scientists. From the discussion of the three approaches a few conclusions can be drawn. First, although there are clearly interfaces and overlaps with the arguments brought forth by economic theories, some new points of view are added. People are not merely seen as rational, selfish individuals who only consider their own monetary gains with respect to the union membership-decision. In addition, motives stemming from dissatisfaction, collective and ideological deliberations, and motives influenced by the social environment are drawn into the analysis. Furthermore, all three approaches in
156
a greater or lesser degree implicate the role of unions themselves as an explanatory variable. In our survey of the economic literature we have noticed in chapter 2.3.6 that economists are divided among themselves whether, and to what extent, union leadership and policies affect union growth and decline. One problem involved is the difficulty of collecting information on perceptions about union effectiveness, and subsequently the operationalization of variables. Anyway, the role of unions (i.e., the supply-side in our theoretical framework of chapter 1) has remained underexposed sofar. Second, the three approaches supplement each other. Frustration with one's working conditions or with union policies seldom leads directly to joining or resigning; these feelings of dissatisfaction are influenced by social interactions, and/or need to be amplified by rational deliberations. Alternatively, if the membership decision-making is based upon rational choice, this may very well include the balancing of social costs against benefits. Hence, in theory three approaches are distinguished, but in practice they are more difficult to treat separately. Therefore the next section discusses several empirical studies, which are rooted both in rational choice and interaction theory. 6.3 Empirical findings of social scientists In this section we discern two types of research; on the one hand predominantly descriptive studies, based on opinion polls, and on the other hand more analytically substantiated studies, using estimation techniques. The first category gives us insight into all the motives that people adduce with respect to joining and leaving the union. We shall discuss the results of two questionnaires, one dating from the late 1950's and one from 1992. In this way we hope to learn more about (changes in) attitudes towards unions. The second category pays more attention to interaction processes, the effects of which are estimated by means of statistical modelling.
6.3.1 Questionnaires: results and discussion The motives of union members to join and to resign have been investigated by Van de Vall (1963). Because the questionnaire was held in the period 1956-1958, the results are somewhat biased: precisely in those years the union movement suffered from a temporary decline, as a result of the restriction on spending in 1957, to which the unions had agreed voluntarily (see chapter 3.5). Many workers criticed this union policy by discontinuing their membership. Nevertheless, this study by Van de Vall has been very influential and for a long time unique of its kind in the Netherlands. The main findings follow below. With respect to joining, the single motives are categorized in three main groups. These are: 1) individual motives: workers become a union member because they strive after personal benefits; 2) ideal-collective motives: the decision to join is based on ideological beliefs and collective interests; 3) social control motives: people join a union under the influence of the social environment. Each of these three groups consists of five to six different single motives, only one of which will be explained below. The number of respondents is 204. A distinction is made between blue- and whitecollar workers (which was a custom in those days), counting 136 and 68 persons, respectively, and between three age groups. The most important results are shown in table 6.1.
157
Table 6.1 Most important motives to join a union (in 1956-58) bluecollars
white collars
age ≤ 30
age 3150
age ≥ 51
1. Individual motives
44%
30%
39%
42%
32%
2. Ideal-collective motives
24%
32%
17%
38%
37%
3. Social control motives
32%
38%
44%
20%
31%
Source: M. van de Vall, De vakbeweging in de welvaartsstaat (1963), pp. 147-149.
Within the category of individual motives the most important single motive to join of both blue-collar workers and white-collar workers can be found, namely for reasons of protection (the so-called 'conflict insurance'); this accounts for 40% and 22%, respectively. The weight of the motives differs quite strongly per age group. Social control motives play an important role for young and white-collar workers. Idealistic motives play a relatively minor role, especially for the workers under 31. Translating the percentages to absolute numbers, it appears that the largest group (80 workers) indicated individual motives to be the most important reason to join, followed by the social control motives (69), while only 55 people put down collective motives to be the decisive reason. From this it can be inferred that already in the late 1950's individual motives started to dominate over ideological motives. With respect to resigning, also three main categories are distinguished (Van de Vall, p. 207). These are: 1) structural motives: doubts about the raison d'être of the union movement, often generally formulated; 2) functional motives: discord about the functioning of the union, mostly based on concrete experience; 3) idealistic motives: objections to union policies. Again, each category contains several distinct motives, and in this case people could give more than one answer, so that the total does not add up to 100%. Leaving aside the third category (which plays a very minor role), it is found that 70% of all the respondents mentioned structural motives to leave (no use; too expensive), and 46% mentioned functional motives (Van de Vall, p. 239). Hence, within this second group former members can be found who did not question the usefulness of the union without due consideration, but they had been disappointed in the services rendered by the union; this mainly concerns older workers, while the youths are predominantly in the first group. This image of motives to resign links up with the conclusions concerning the motives to join. A much more recent opinion poll among the labour force (n=2000) has been executed by social scientists (Van Rij et al.) of the two universities in Amsterdam (UvA and VU) in 1992. With respect to joining, the same motives are distinguished as by Van de Vall, albeit they are now grouped into four categories:
158
Table 6.2 Most important motives to join a union (in 1992) members
ex-members
1. individual motives
36.0%
29.1%
2. collective motives
40.9%
28.2%
3. idealistic motives
9.9%
10.9%
4. social motives
3.2%
19.5%
Source: UvA/VU, Participatie in vakbonden (1992), p. 23.
A distinction is made between motives of members and former members. As can be seen from the table, both groups rank individual and collective motives much higher than the other two categories. When asked, if there had been a special reason to join a union, the majority of the interviewees (66.5% and 70.5%, respectively) responded negatively. If there was an immediate cause, this usually concerned problems at the work floor. Only 1.5% of the respondents indicated that they had become a member on account of a strike. This last result corresponds with the findings of the post-war time-series analysis in chapter 4. With respect to resigning, the motives are grouped into four categories, which, unfortunately, are different from the tripartition made by Van de Vall. Furthermore, a distinction is made between motives of former members and those of members who are considering to resign. Here the motives are: 1) product motives: discord about the functioning of the union and/or doubts about the use of union membership (hence, a combination of the categories functional and structural motives distinguished by Van de Vall); 2) price motive: the membership fee is considered to be too high; 3) change of status: into unemployment or retirement, or from one job to another; 4) policy motives: objections to union policies or actions. Table 6.3 Motives to leave a union (in 1992) ex-members
members (considering)
1. product motives
25.9%
35.5%
2. price motive
15.5%
17.7%
3. change of status
58.2%
17.7%
4. policy motives
15.5%
46.8%
Source: UvA/VU, Participatie in vakbonden (1992), pp. 26, 28.
The most important reason to leave a union is formed by a change of status; remarkably enough, within this category the loss of one's job was brought forward just as many times as the change to another job. This is explained by the possibility that at the new job the union is less clearly present, or there is less social pressure. A second noticeable result concerns the category policy motives. As can be seen from the table, only a minority of the former members resigned out of dissent with union policies. However, of the members who think of resigning, no less than 46.8% blame that on
159
union policies. From this it could be concluded, that in the end the policy motive is not decisive, at least not in most cases. The last interesting finding to be mentioned relates to the question whether unions should only serve the interests of their own members, or the interests of the whole labour force. From the report (p. 33) it becomes clear, that among the members the number of advocates of a broad union movement is about as large as the number of adversaries: 35% against 37.7%. This result corresponds with the almost equal weight that is ascribed to the collective and individual motives to join. The two questionnaires both reveal a tension between the rendering of service in a narrow sense (to members only) and in a broad sense (to all workers, even to the whole society). In his thesis Van de Vall concludes, that the extension of the welfare state has led up to two contrary developments (pp. 102-104). The risen prosperity caused a mentality change towards more materialism and individualism among the workers. However, among the (recognized) unions the reconstruction of the Dutch economy caused a deep sense of responsibility towards the whole society. Hence, a discrepancy was arising between the growing need of workers for union services directed towards the individual member, and the increasing emphasis of the union movement on its external, collective functions. This gap between opposite objectives has only widened in the following decades. Research by Teulings (1984) shows that unions have only become more active in all kinds of national advisory and consultation boards, while Van Ham et al. (1985) conclude that ever more workers are driven by pragmatism: they only join a union when they personally benefit from it. As the union movement often appears in the headlines with opinions concerning complex societal problems, it makes the average member wonder if his or her own interests are served by these union policies. An attendant difficulty lies in the fact, that the majority of Dutch union members are organized in unions that cover a whole branch of industry, or even several industries. Hence, within one union, groups of members have divergent interests. It is the task of the unions to reduce all these different interests to the same denominator; to that end an appeal has to be made to the mutual solidarity of members. Because the unions also call on the solidarity of their members towards nonmembers, in particular the non-active people in the society, this does not increase their appeal to a lot of workers. Teulings (1984, pp. 729-730) argues that it is necessary for unions to build in 'social learning-processes', which should enhance feelings of solidarity. With this the discussion turns to the interaction theory, which will be commented on in the next subsection.
160
6.3.2 Statistical modelling: results and discussion The disadvantage of questionnaires lies in the fact, that this kind of research is rather subjective by nature; the findings should be interpreted with caution, and preferably be checked by other methods. For instance, the results of Van de Vall suggest that social control motives played an important role in the decision to join, whereas in the UvA/VU research the influence of the social environment does not seem to exert a large impact (anymore). This discovery might not be reliable: social pressure may be of importance, without the consciousness of the respondents. In fact, several social scientists have shown by means of statistical analyses that the social environment plays an important role in the decision whether or not to join (or leave), through interaction processes. The most interesting results of a few recent empirical studies will be presented below. In Van Rij & Van der Putte (1992) the so-called 'event-history analysis' is applied, which is a more dynamic approach to union membership. The authors not only ask the question why people decide to join a union, but also when they make this decision. In short, a hazard rate is defined which gives the probability that an event takes place within a certain amount of time; in this case the event concerns the individual decision to join a union given that this person is not yet a member. Next, the Cox model62) is applied which estimates the impact of several sociological and economic variables on this rate of change (of nonmember to member). Van Rij and Van der Putte find the following results. The three interactionist variables (density rate at one's first job > 50%; number of trade union members one speaks with; number of contacts with the union) show a significantly positive sign. However, the personal characteristics that were significant in our own analyses of chapter 5, turn out to be insignificant in theirs: this concerns gender, education, and age at one's first job (which is different from our age variable); a possible explanation for this deviating result may be found in the fact that different control variables are used. Furthermore, a few macro-economic variables have been tested, such as unemployment, inflation, number of strikes and wages. All these variables are insignificant as well, but if they are removed it significantly reduces the fit of the model. This suggests that macro-economic factors (i.e., the business cycle) play a role of importance in creating better or worse circumstances in which the unions have to operate (recruit). Hence, the auth-ors conclude that economic determinants constitute the framework within which the membership decision-making takes place, which in turn is foremost influenced by interactionist variables. Thereupon, a survival function, related to the hazard rate, is defined as the probability that an event does not take place within a certain amount of time; i.e., how long does somebody survive as a nonmember. The authors find that the number of contacts with other union members, and especially the level of the density rate significantly influence the probability of surviving. This result confirms the strong impact of the work environment on union membership. In Van Rij & Saris (1993) event-history analysis is applied to the process of joining as well as leaving a union. As the same data are used as in the above mentioned study, 62)The Cox model is a variant of a category of models based on survival and hazard functions, which emanates from biometrics; originally, these models were designed to calculate life-tables. Later on, there have been applications in econometrics, and even more recently in the social sciences.
161
the results with respect to becoming a member are identical: the three interactionist variables have a significantly positive impact. Concerning the factors that affect the probability of leaving, again the number of contacts with other union members and with the union, respectively, are significant, together with the variable age at the first job. By comparing different model specifications, the authors attempt to find the best fit for the process of joining and leaving, respectively. They show that the process of joining can best be characterized by a decreasing rate: the probability that someone will become a member is largest when (s)he just enters the labour market, and declines as time elapses. The process of leaving is described by a log-logistic distribution, implying that the probability of leaving first increases, and after a while decreases. This means that new workers have a high propensity to join, but at the same time young members have a higher propensity to resign, which is owed to a commitment problem. With several co-authors, Klandermans has paid a lot of attention to this phenomenon of commitment. On the basis of three Dutch case-studies, Van der Veen & Klandermans (1989) reach the conclusion that the most important reasons for terminating union membership are doubts about the usefulness of the membership, and a lack of positive experiences with union activities. They explain exit-behaviour in trade unions with the aid of cost/benefit theory and commitment theory, respectively. The first theory refers to the rational-choice approach, which assumes that people continuously balance the costs of union membership against the benefits: not only at the moment they decide to join, but also during their membership. In this approach it is assumed that motives of joining can also be used to explain leaving. A distinction is made between selective motives, referring to individual advantages and disadvantages of participation, and collective motives, related to the expectation that participation will help the movement to realize its goals. Given the high turnover rate in the Netherlands (see table 3.7), it is suspected that many people consider this balance of costs and gains to be just positive, or to be just negative. A small change in circumstances is enough to alter the balance. Based on three enquiries, the authors conclude that about 90% of the former members resigned because the positive balance of benefits over costs that had incited them to join, had become negative. The majority (76.5%) left due to a negative balance of selective motives, while 13.5% left due to a negative collective motivation. This result confirms the finding of much research, that selective motives dominate over ideological ones. The resignation of the remaining 10% of the former members cannot be explained in terms of the rational-choice theory; they left in spite of a positive valuation of selective and collective gains. Their departure could be ascribed to a waning commitment to the union. During the period of membership, new dimensions of affiliation develop owing to interaction processes within the organization. This commitment finds expression in feelings of loyalty and responsibility towards the union, willingness to participate in it, and belief in the utility of unionism. According to Hirschman (1970), dissatisfied loyal members usually choose for the 'voice-option' (i.e., to exercise critique without resigning), whereas non-committed members sooner opt for the 'exit-option'. From the three case-studies it appears that the majority of the ex-members never had been active in the union and were not enrolled for very long, which indicates that the socialization process leading to
162
commitment could not even come about. According to Van der Veen & Klandermans, the two theories which explain union membership do not conflict with each other, but they are complementary: "When little or no integration has been achieved, the cost/benefit theory offers an adequate explanation for exit. When commitment has developed, the decision to leave involves more than a mere calculation of costs and benefits. Leaving the organization, then, is due to factors influencing commitment" (p. 192). They mention two reasons for this change in attitude, namely alterations in the employment status, and in the union organization. The first explanation covers a change of job (to a management position) as well as the loss of a job. In both cases people alienate from their former social environment, leading to a loss of contact with the union. The second explanation puts the blame on the unions themselves, for becoming so bureaucratic that the distance between the organization and its members has become too large to develop or maintain commitment. In Van Teefelen & Klandermans (1989), the rational-choice and the interactionist approach are used to explain exit-behaviour by means of statistical modelling. The authors argue that commitment to the union is the key concept in explaining why union members resign, and they link this concept to both theories. Commitment is not only the result of socialization processes, but also arises from the benefits that people derive from the union membership. They formulate a model which incorporates these notions. The underlying idea is that rational deliberations concerning the use of the union (value-expectancy) lead to a stronger commitment if people are satisfied in their expectations; this in turn may lead to a more active effort for the union, which stimulates interaction processes, and hence commitment. This means that the chance of resigning decreases. Of course, this pattern can also develop the other way around if one is not satisfied about the functioning of the union. With the aid of stepwise regression, a path analysis is conducted which indicates whether variables influence the chance of resigning directly or indirectly. Three categories are distinguished: rational-choice variables (divided in instrumental and idealistic motives), interactionist variables, and commitment (composed of five topics63)). The results are as follows: the strongest impact on resignation comes from 'commitment'. Although some of the rational-choice variables also exert a direct influence on the degree of resigning, most variables correlate stronger with commitment. Hence, the rational-choice variables (especially 'serving collective interests' and 'effect of the union on the shop floor') and foremost the interactionist variables (notably 'the degree of member activity for/within the union') affect exit behaviour indirectly, through the commitment variable. Turning back to Van der Veen & Klandermans (1989), three dimensions are distinguished which influence the inclination of members to participate in their union: 1) the organizational structure; 2) mobilization potential; and 3) union strategy. The lower the degree of participation, the lower the degree of commitment will be, hence the higher the chance of quitting the union. With respect to organization, it has been established that a decentralized structure presents more possibilities for communication, and hence for more commitment. In the Netherlands, unions are characterized by a very centralized organizational struc63)To measure the degree of commitment, five questions were asked about people's attitude towards their union and its achievements. The answers were transformed to a scale ranging from 1 to 5, with value 1 indicating a strong commitment, and value 5 indicating none.
163
ture: decisions are taken at the top, while ordinary members are hardly consulted. Union representatives are hardly visible at the work floor, so there are few opportunities for interaction between the union and its members, which impedes the development of commitment to the union. The effect of the absence of union representation in the workplace is also commented on by Visser (1987, pp. 91-92). At the company level, the services that unions render are most personalized. "It is through workplace negotiations over worker grievances, lay-offs, etc. that benefits which accrue from union bargaining tend to be related to membership". This is referred to as bargaining depth or plant effectiveness, which has been found to play an important role in the explanation of the level of unionization: in countries where workplace representation is not unionbased, union density levels are low. Concerning the mobilization potential, Van der Veen and Klandermans argue that organizations which levy membership fees, do not expect their members to act out of idealism, but out of cost-benefit calculations. Most union members only participate passively, by paying their dues. Consequently, the mobilization potential of unions is low. Finally, with respect to union strategy, Dutch unions strive after three goals: emancipation of the labour force, protection of the collective interests of the labour force, and protection of the individual interests of union members. Ever since the end of World War II, most unions have stressed collective bargaining, thereby serving the interests of the whole labour force. After Olson (1965), this implies that unions are at risk of losing members or fail to attract new ones, because they offer a collective good which can be obtained without becoming a union member. In order to prevent this from happening, unions should stress their importance in individual services, available to members only. According to Van der Veen & Klandermans, unions today already adapt to the increasing tendency of workers to join merely on pragmatic grounds, by putting more and more emphasis on their facilities for individual members. As to how far this notion is right, can be questioned; after all, at the end of section 6.3.1 we concluded that there is still a strong emphasis on the broad union movement. 6.4 Incorporation of sociologic notions into economic modelling It can be concluded that social scientists partly advance other determinants of membership fluctuations, or at least emphasize different factors. Especially the interactionist approach with commitment in its wake finds a lot of response, and has delivered strong empirical results. Further, much more attention is paid to the role of unions themselves: how visible is a union on the work floor, what can a union offer to its (potential) members. Apart from the fact that economists partly stress different determinants, one other reason why they have included so little sociologic tinged variables is the poor availability of suitable data. The Dutch Central Bureau of Statistics (CBS), which provides most of the union data, has only published a number of key variables since approximately 1965. Before that year, the union statistics were much more elaborate. A few of those statistics, with respect to union finances, might be useful additions to the already employed series. This concerns expenditures on several sorts of benefits payments to the members (variables with regard to union services) and receipts from
164
union dues, over the period 1920-1962 (except 1921, 1940-1945). The statistics give 64)
information about benefits payments and membership fees per trade union federation. In total 4 useful series remain, which have enough observations. These are listed in appendix 6.I. Since these extra data are time-series, the first option would be to incorporate them in the two models of the time-series analysis in chapter 4. However, there are two strong disadvantages: first, many observations get lost (in the time-series, the period 1912-1986, except for 1940-1946 is available), which would be at the expense of the reliability. Second, in an attempt to include them anyway, the explanatory variables had to be transformed first. The endogenous variable stands for the percentage aggregate membership growth, while the original exogenous variables give information per federation. Therefore the different benefits payments and the membership fees of each separate federation have been added up. The estimation results lead to the conclusion that these new aggregate variables have no explanatory power whatsoever. That is why it is decided to pose a new question, and hence to construct a new model, in which the additional data can be included in a meaningful way. The next paragraph sets out which findings and assumptions form the basis of this new model, which in turn will be discussed thereafter. The results of the time-series analyses in chapter 4 have led to the conclusion that fluctuations in the aggregate membership are mainly determined by social-economic developments on the macro level. It has become clear however, that this research method has left many questions unanswered. Next, the cross-sectional analyses in chapter 5 have shown that the individual decision to become a union member is influenced by a wide range of predominantly personal and job-related characteristics. It was also demonstrated that the decision-making process is a sequential one. For most people the decision to become a union member is taken prior to the decision which particular federation to join. Factors which determine this last choice could mainly be found in the religious persuasions. Finally, from the research of social scientists that is discussed in this chapter it emerges that the motives of workers to join a union have become ever more materialistic. Reasons to remain in the union are strongly influenced by the degree of commitment, which in turn for a considerable part depends on the extent to which members are satisfied in their expectations with respect to the union's efforts and achievements. This leads to the following question: Given the fluctuations in the overall membership, which factors thereupon determine which share each of the separate federations has in the total union membership? In other words: what determines the choice for a specific federation? According to the results of chapter 5, religious background plays a prominent role; the increased importance of pragmatic deliberations, as set out in chapter 6, could however not be checked in the cross-sectional analyses because the right variables were not available in the data-set. By means of the new model it is analyzed in what degree the share of a specific federation in the total can be explained by the following variables. Firstly, the degree in which such a federation involves more or less than average costs for the members (= dues per member of that federation, compared to the average dues per member of all federations together). Secondly, the degree by which such a federation more or less than average endows to the members (= expenditures on three different sorts of 64)Sources: CBS (1920-1940), Overzicht van den omvang der vakbeweging in Nederland; CBS (1945-1964), Omvang der vakbeweging in Nederland.
165
benefits payments of the federation, compared to the average expenditures on these items of all federations together). In addition, it is tested to what degree the percentage development of the number of non-religious people of the community plays a negative or positive role. Given the strong result in chapter 5, a variable related to the pillarization of the Dutch union movement may not be left out. The expectation is that when a federation pays more than on average to its members, its share in the total membership will be higher; also, when a federation levies a subscription fee higher than the average, its share in the membership will be lower. This idea is primarily based on the rational-choice approach, applied by the economic as well as the social sciences. The CBS data set does contain two sort of interactionist variables, namely 'money payments from the unions to their local branches' (which proxies the degree of decentralization), and 'number of paid officials in the service of the unions' (which proxies the number of contacts with the union). Unfortunately, neither series can be used. The first one has too few observations, while the incorporation of the second one is questionable because of doubts about the direction of the causality: it is very likely that the number of functionaries follows on union growth, rather than precedes it.65) Hence, four models have been constructed, each before as well as after World War II, because of the structural break which was detected in chapter 4. The eight models subsequently explain the share of NVV, CNV, NKV and categorial unions in the total membership (adding up to 100%) using the following exogenous variables:
65)The following remarks can be made with respect to these two series. The degree of decentralization is measured by the share of contributions to the local branches in the total expenditures of all unions of one federation; information is only available for the period 1953-1962 and 1966. From these data it can be seen that at the NVV unions this share declined from 15.8% in 1953 to 10.2% in 1966, while it declined only a little at the CNV and NKV, from 13.7% and 14.1% in 1953 to 12.2% and 13.1% in 1966, respectively. This percentage constituted a stable 5% at the categorial unions. This low figure could be explained by the fact that these unions had far less means, but probably also less local branches. The number of contacts with the union is measured as the number of members per union official, between 1925 and 1970. Before the war, no real trends can be detected; there were large fluctuations within each pillar, but no large differences between the three major federations. On average, the unions had about 500 members per official. In the 1920's the categorial unions had far more members per official, but in the 1930's their numbers were almost equal to those of the three pillars. After the war, the differences became larger. The NVV unions had the least members per official, declining to about 450 in 1960, but after that rising to 560 in 1970. The CNV and NKV showed the same trend, only with higher numbers (in 1970 ample 600). The categorial unions employed far less officials: in 1970 the number of members per official amounted 1290.
166
the log of the federation's expenditures on strike-pay per member relative to those of the total union movement P2 the log of the federation's expenditures on sickness benefits per member relative to those of the total movement P3 the log of the federation's expenditures on death benefits per member relative to those of the total movement P4 the log of the federation's subscription costs per member relative to the subscription level of the total movement %NR percentage of the non-religious in Dutch society Dummy51 dummy with value 1 for the period 1951-1962 P1
After the war P1 en P2 are not included anymore due to too many missings; a dummy is added with value 1 for the period 1951-1962, because in 1951 the largest categorial union EVC was suddenly left out of the statistics, causing a huge structural decline in the share of the categorial unions in the total membership in the subsequent years. An example of a pre-war model is the following: Mj = α0 + α1 ⋅ P + α2 ⋅ P + α3 ⋅ P + α4 ⋅ P + α5 ⋅ %NR j with Qi j Pi = ln NVV CNV NKV CAT (Qi • Qi • Qi • Qi )1/4 where Mj stands for the membership share of the relevant federation (j = NVV, CNV, NKV66), Categorial unions) in the total membership, and Qi stands for the absolute amount of money that is involved in the relevant benefits payment (i = 1, 2, 3) or contribution (i = 4) per member of the relevant federation. The actual data which are used can be found in the appendix. In the first instance, all eight models are estimated separately by means of OLS. Before the war, the period 1922-1939 is estimated. This means we dispose of 18 observations, with 5 exogenous variables (excluding the intercept). This implies that the D.W. value will always be situated in the inconclusive region. After the war the period 1946-1962 is estimated, with a missing in 1948. This means the disposal of 16 observations with 4 exogenous variables, which makes the model only really reliable when the D.W. value lies between 1.93 and 2.07. The results must therefore be interpreted with a lot of caution. Looking at the tables on the next page it is clear that the results differ per model. None of the variables (except for the dummy) is significant with the right sign in all models. One variable is not significant in any model; this concerns the relative share in expenditures on strikes. The explanatory power is high everywhere, with an adjusted R2 between 0.80 and 0.95. The D.W. value is always in the inconclusive region, except for the postwar model of the NVV. Due to the small number of observations, the possible presence of an autocorrelation structure is difficult to detect. However, because we are dealing here with a so-called 'distribution model', in which the aspect of time is of secondary importance, the eventual correlation of the errors in time is not so relevant. Therefore no more attention is paid to the autocorrelation issue. 66)The NKV is the most recent name of the Roman Catholic federation, and stands here for the RKWV regarding the period 1922-1939, and for the KAB regarding the period 1946-1962.
167
The results of the pre-war models are shown in table 6.4, and lead to the following conclusions. In model 1, P2 and %NR show the expected significantly positive sign: according as the NVV pays more than average on sickness benefits, its share increases; in the same way a percentage growth of the number of non-religious people is in the advantage of the socialistic, non-religious union federation. In model 2, P2, P3 en P4 show the expected significant sign: a relative increase in the expenditures on sickness and decease has a positive effect on the share of the CNV, a relative increase of the subscription costs a negative effect. The last variable, %NR is very significant, but unfortunately has the wrong sign: here the suggestion is made that the share of the Christian CNV increases thanks to the percentage growth of the number of non-religious Dutchmen, which is not logical. In model 3, P3 shows the expected significantly positive sign, but %NR again has the wrong sign: for the Roman Catholic NKV, we would expect a negative effect from the secularization on its membership share. In model 4, P3 en P4 show the expected sign; %NR is here significantly negative, which was not expected because categorial unions are usually only organized by profession, and not by religion. With respect to the results of the religion variable, the following explanation can be given for the unexpected findings. It can be argued that people with strong religious beliefs are less likely to be a member of any union, including the CNV and NKV. This result was found in the preceding chapter, where it was established that regular church visitors have a significantly lower probability of becoming a union member. A growth in the proportion of non-religious people indicates a development towards a society with a smaller disposition against progressive or even radical points of view. In that sense, an increase in the percentage of the non-religious can also contribute to the membership share of the CNV and the NKV. The negative sign of %NR in the fourth model of the categorial unions can then be explained as well. Since the exogenous variable %NR has a positive sign in all three preceding equations, and the four endogenous variables add up to 1 (100%), it automatically implies that the coefficient of %NR in model 4 becomes negative. If the rise of the percentage of non-religious people increases the share of the NVV, CNV and NKV in the total membership, it must by definition decrease the share of the remaining group. This does not mean that secularization has a negative impact on the absolute membership figure of categorial unions, only that they profited less from the secularization in relative terms.
168
Table 6.4 Determinants of federation shares 1922-1939 (t-values in brackets) 1 share NVV 2
2 share CNV
3 share NKV
4 share CAT
adj. R
0.802
0.928
0.824
0.944
D.W.
1.543
2.342
1.265
2.238
0.246***
0.011
0.168***
0.528***
(8.725)
(0.730)
(6.732)
(11.654)
0.006
0.000
-0.005
0.006
(1.285)
(0.064)
(1.079)
(1.559)
**
**
0.000
0.013
(2.360)
(2.537)
(0.044)
(1.373)
-0.011
**
***
0.016**
Intercept P1 (strikes) P2 (sickness)
0.023
P3 (death) P4 (dues) %NR
0.019 0.014
0.022
(0.870)
(1.999)
(3.473)
(2.584)
0.075
***
0.094
-0.145**
(1.085)
(5.065)
(1.314)
(2.357)
***
***
***
-0.026***
(3.489)
(8.669)
0.010
(7.432)
-0.279
0.011
(10.609)
0.005
Table 6.5 Determinants of federation shares 1946-1962 (t-values in brackets) 5 share NVV 2
adj. R
P3 (death)
%NR
8 share CAT
0.946
0.926
0.949
+
1.026
1.654
1.526
0.600***
0.133**
0.321***
-0.219
(8.771)
(2.620)
(4.693)
(1.274)
-0.007**
-0.003*
0.001
-0.003
(2.528)
(1.778)
(0.376)
(1.023)
-0.095
-0.021
-0.104
*
-0.149**
(1.284)
(0.616)
(2.001)
(2.454)
***
***
***
-0.127***
1.942
P4 (dues) Dummy51
7 share NKV
0.905
D.W. Intercept
6 share CNV
0.052
0.031
0.048
(7.093)
(10.618)
(9.355)
(10.157)
***
0.000
-0.003
0.025***
(3.728)
(0.056)
(0.870)
(2.872)
-0.015
+null-hypothesis of serial independence is accepted *significant at the 10%-level **significant at the 5%-level ***significant at the 1%-level
169
The following conclusions can be drawn on the basis of the post-war models, shown in table 6.5. In model 5, in which the share of the NVV is explained, only the dummy has the expected positive sign: the structural omission of the EVC from the share of the categorial unions had a very significantly positive effect on the shares of all other federations, hence also of the NVV. P3 suddenly becomes significantly negative. The same holds for %NR, suggesting that a percentage growth of the number of non-religious people after the war counted against the NVV. This could maybe be explained by the fact that after the war the NVV profiled itself as a federation which pointedly opened up to religious workers, but this is not a strong argument. Following the argument on the pre-war models, the interpretation could be reformulated as follows: even though the NVV might have gained many new members owing to the secularization, relatively speaking other federations gained even more, as a result of which the share of the NVV decreased (other things being equal). In model 6, only the dummy shows the expected positive sign, while again P3 shows the wrong (significantly negative) sign. A possible explanation for this deviant finding in models 5 and 6, is the following. The variable P3 is deduced from the total expenditures on death benefits per federation, divided by the total number of members of that federation. The numerator is the product of the level of the death benefit and the number of deceased members. Now, if this benefit level of one particular federation rises (other things being equal), it will have the expected positive effect on this federation's membership share. However, if a federation has more members dying than on average its total payments on death benefits will rise, but its membership necessarily decreases and hence its membership share will also decline. In model 7, P4 shows the expected significantly negative sign, confirming the assumption that a relative increase of the subscription fee reduces the membership share of the federation in question. The dummy is again significantly positive. In model 8, P4 shows the expected significantly negative sign, while %NR has the expected significantly positive sign. Finally, the dummy now becomes significantly negative, reflecting the strong adverse effect of the omission of the EVC on the membership share of the categorial unions. Evaluating the overall result of the estimations, the following remarks can be made. The fact that many results of the pre-war models differ from those of the post-war models, was to be expected on the basis of prior results; already in the time-series analysis it was clear that a structural break around World War II necessitated the construction of two separate models, as the pre-war (working) conditions differed widely from those after the war. One strong result is the high significant value of P4 in both categorial models: the negative sign implies that categorial members were highly sensitive to the dues level, which supports the notion that many people joined these categorial unions especially because of the relatively low subscription fee. Finding the wrong sign and/or insignificant parameters in several models implies that (part of) the theory is not supported by empirical results; however, this in itself is an important result as well. On the other hand, it might also be that the theory is correct, but is not confirmed by the empirical findings due to the fact that so few observations, and such a small number of variables could be used. One way to obtain more reliable results with these data, is to model all four equations simultaneously. Since the membership shares of the four federations as well as the explanatory variables P1 to P4 (as values relative to their mean) are interdependent, simultaneous modelling offers two advantages. In the first place, it will lead to a
170
higher degree of precision and efficiency of the estimates. Secondly, it precludes the occurrence of inconsistent coefficient values, because of the restrictions that are imposed on the models.67) Using Two-Stage Least Squares (2SLS), the estimations results are shown in the tables 6.6 and 6.7.68) The values of the D.W. statistic are not interpretable anymore, as a direct result of the simultaneous modelling: since all endogenous and most exogenous variables are interdependent, the estimated errors of the four models will also be related to each other. We shall not dwell upon this issue any further. From the tables 6.6 and 6.7 it can be concluded that simultaneous modelling indeed leads to more efficient parameter estimates than when separate models are estimated. This can be inferred from the t-values of the coefficients, which are almost everywhere higher in tables 6.6 and 6.7 than in the tables 6.4 and 6.5 (with the notable exception of the variable P1, but this one remains insignificant in all models anyway). The four P-variables in the pre-war models all come up with the expected sign, while P2, P3 and P4 are significant as well. The variable %NR shows the same results as in the separate modelling. With respect to the post-war models, almost all variables obtain significant coefficients, except for %NR in the models 6 and 7. The variable P3, concerning the relative expenditures on death benefits, again shows the deviant negative sign, while P4 (relative level of union dues) obtains the expected negative sign. From the system of the four equations it could be deduced that all Pi-variables have the same coefficient value. This implies that all workers are supposed to have the same reaction coefficient with respect to a change in one of these variables, which represent the relative level of expenditures or receipts of each federation, adding up to zero. For instance, the coefficient value of P4 in the post-war models is -0.067. To measure the effect of this variable on the separate membership shares in a certain year, -0.067 has to be multiplied by the four values of P, respectively. Using table 6.12 of the appendix, it turns out that the NVV, NKV, and CNV levied a subscription fee which was higher than the average during the whole period 1946-1962, while the categorial unions levied a relatively lower subscription fee. This implies that the positive value of P4 for the NVV, NKV and CNV, multiplied by the negative coefficient value, produces a negative effect on their membership shares, while the negative value of P4 for the categorial unions produces a positive effect on their membership share.
67)In the appendix it will be proven that the following restrictions can be deduced from the system of the four equations, based on the knowledge that the values of the endogenous variables add up to 1, and the sum over j = 1,2,3,4 of P equals zero for all i = 1,2,3,4. With respect to the values of the intercept, these have to add up to 1; the sum of the coefficients of %NR and of Dummy51, respectively, equals zero. Finally, with respect to each of the P-variables, their coefficients must equal each other. 68)It should be noted that the model computes a different adjusted R (biased towards 1) due to technical reasons, but it is not relevant to discuss this point any further, since this information is not used in evaluating the results anyway. 2
171
Table 6.6 Simultaneous models of federation shares 1922-1939 (t-values in brackets) 1 share NVV
2 share CNV
3 share NKV
4 share CAT
adj. R2
0.999
0.998
0.999
0.998
D.W.
0.840
0.691
0.842
1.851
***
*
***
0.525***
Intercept
0.326
P3 (death) P4 (dues) %NR
0.127
(20.370)
(2.006)
(8.224)
(20.307)
0.001
0.001
0.001
0.001
(0.509)
(0.509)
(0.509)
(0.509)
***
***
***
0.013***
P1 (strikes) P2 (sickness)
0.022
0.013
0.013
0.013
(2.992)
(2.992)
(2.992)
(2.992)
***
**
**
0.015**
0.015
0.015
0.015
(4.370)
(4.370)
(4.370)
(4.370)
-0.075**
-0.075**
-0.075**
-0.075**
(2.435)
(2.435)
(2.435)
(2.435)
0.008***
0.009***
0.008***
-0.025***
(8.642)
(11.685)
(7.113)
(14.993)
Table 6.7 Simultaneous models of federation shares 1946-1962 (t-values in brackets) 5 share NVV 2
6 share CNV
7 share NKV
8 share CAT
adj. R
0.999
0.999
0.999
0.998
D.W.
1.065
0.989
1.719
1.323
***
***
***
-0.079
Intercept P3 (death) P4 (dues) Dummy51 %NR
0.593
0.186
0.300
(9.695)
(4.530)
(5.207)
(0.887)
-0.003**
-0.003**
-0.003**
-0.003**
(2.251)
(2.251)
(2.251)
(2.251)
-0.067**
-0.067**
-0.067**
-0.067**
(2.747)
(2.747)
(2.747)
(2.747)
***
***
***
-0.136***
0.055
0.029
0.051
(10.511)
(10.770)
(10.998)
(19.789)
***
-0.002
-0.003
0.020***
(4.139)
(1.097)
(0.778)
(4.127)
-0.015
*significant at the 10%-level **significant at the 5%-level ***significant at the 1%-level
172
To give one concrete example: between 1946 and 1962 the dues level of the NVV lay about 20% higher than the average. Given the negative coefficient value of -0.067 this means that the relatively high subscription fee of the NVV decreased its membership share with circa 0.067 x 0.20 = 1.3% points. In other words, the NVV share could have been 1.3% points higher if its dues level would have been equal to the average dues level. In the same way, it can be calculated that the membership share of the categorial unions would have become approximately 0.067 x 0.45 = 3% points lower if their dues level had been equal to the average level. 6.5 Conclusion Although the lines of approach from the social sciences to the explanation of union membership decisions could not be dealt with extensively, this chapter has made clear that their contribution to this subject constitutes a very useful addition to the insights gathered in the chapters 4 and 5. It has been shown that individuals base their decisions on a wide range of motives, prompted by individual and collective incentives, which in turn are influenced by economic circumstances, ideological deliberations, and not least by social interaction processes. In addition, much more than in the preceding chapters, emphasis has been laid on the role that unions themselves play in the membership decision. Due to a lack of data at our disposal, it has not been possible to test all these additional insights ourselves. In an attempt to employ a few 'supply-side variables' in a new economic model, especially the simultaneous modelling comes up with quite satisfactory results, because all but one of the variables turn out to be significant. Whether the additional information and insights are derived from the social-scientific literature, or from our own empirical analysis, it has been possible to shed some more light on the supply-side of our theoretical framework. From the work of social scientists, as well as from the analysis in section 6.4 it has been shown that the way in which trade unions function (with respect to services rendered to their own members, their emphasis on a broad union movement, their presence or absence on the workfloor, etc.) has a significant impact on the unionization decision of workers. At the end of this chapter, the final question then becomes to what degree the findings of the chapters 3 through 6 enable us to answer all the questions that were posed in chapter 1, with respect to the determination of Dutch trade union membership from its early history until now. This will be the subject of the next and last chapter.
173
APPENDIX 6.IData used in section 6.4 Table 6.8 Values of the endogenous variables (membership shares per federation) Year share NVV share CNV share NKV share categorial ----------------------------------------------------------------------------------1909 0.21988 0.03566 0.06321 0.68125 1910 0.22595 0.03817 0.07888 0.65700 1911 0.24530 0.03665 0.07707 0.64098 1912 0.26580 0.03420 0.09134 0.60866 1913 0.32564 0.04198 0.11120 0.52118 1914 0.33001 0.04582 0.13045 0.49372 1915 0.34423 0.05070 0.14043 0.46464 1916 0.37521 0.05861 0.16489 0.40129 1917 0.39001 0.06754 0.17284 0.36961 1918 0.38263 0.09164 0.18787 0.33786 1919 0.38738 0.10501 0.22524 0.28237 1920 0.36056 0.12243 0.25449 0.26252 1921 0.35563 0.11664 0.23979 0.28794 1922 0.36798 0.11455 0.22470 0.29277 1923 0.36601 0.10783 0.20208 0.32408 1924 0.40363 0.10943 0.20376 0.28319 1925 0.41054 0.10412 0.19762 0.28772 1926 0.41128 0.10742 0.20574 0.27556 1927 0.41188 0.10722 0.20964 0.27125 1928 0.41294 0.10959 0.21489 0.26258 1929 0.42536 0.12132 0.22733 0.22599 1930 0.42460 0.12621 0.23505 0.21414 1931 0.42764 0.13792 0.24446 0.18999 1932 0.42899 0.14745 0.24987 0.17369 1933 0.42840 0.15401 0.25659 0.16100 1934 0.42159 0.16004 0.25649 0.16187 1935 0.41719 0.15866 0.25533 0.16882 1936 0.41656 0.15929 0.24967 0.17448 1937 0.42016 0.15827 0.24599 0.17559 1938 0.42202 0.15720 0.24857 0.17222 1939 0.42514 0.15869 0.25287 0.16330 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 0.29869 0.11812 0.22971 0.35348 1946 0.31359 0.12396 0.23408 0.32837 1947 0.31671 0.12573 0.24028 0.31728 1948 0.33255 0.13331 0.24350 0.29064 1949 0.32925 0.13460 0.25574 0.28041 1950 0.33421 0.13685 0.25640 0.27254
174
Table 6.8 (continued) Year share NVV share CNV share NKV share categorial ----------------------------------------------------------------------------------1951 0.38299 0.15985 0.29401 0.16315 1952 0.38353 0.16095 0.29599 0.15953 1953 0.38470 0.16189 0.29422 0.15918 1954 0.37930 0.16342 0.29542 0.16187 1955 0.37090 0.16374 0.30296 0.16239 1956 0.37434 0.16162 0.30827 0.15578 1957 0.36482 0.16395 0.29639 0.17483 1958 0.35717 0.16357 0.29651 0.18274 1959 0.35927 0.16166 0.29556 0.18351 1960 0.36222 0.15989 0.29421 0.18368 1961 0.35738 0.15847 0.29439 0.18975 1962 0.35669 0.15843 0.29143 0.19345 1963 0.35952 0.15629 0.28484 0.19935 1964 0.36147 0.15722 0.27953 0.20178 1965 0.36157 0.15827 0.27808 0.20208 1966 0.36314 0.15713 0.27763 0.20210 1967 0.36583 0.15635 0.27827 0.19955 1968 0.36184 0.15491 0.26425 0.21900 1969 0.36535 0.15408 0.25888 0.22168 1970 0.38564 0.15025 0.25344 0.21067 1971 0.38847 0.14877 0.24891 0.21385 1972 0.39083 0.14556 0.24409 0.21952 1973 0.39833 0.13796 0.23866 0.22505 1974 0.40210 0.13332 0.21094 0.25364 1975 0.40739 0.13114 0.20686 0.25461 1976 0.41089 0.15020 0.20079 0.23812 1977 0.41794 0.16897 0.19564 0.21744 1978 0.42375 0.17114 0.19076 0.21435 1979 0.42410 0.17181 0.18488 0.21921 1980 0.42745 0.17366 0.17808 0.22082
175
Table 6.9 Values of variable P1 (relative expenditures on strike-pay) Year P1NVV P1CNV P1NKV P1CAT --------------------------------------------------------------1920 2.06652 0.76816 0.71267 -3.54736 1921 . . . . 1922 1.07095 0.55132 1.18553 -2.80780 1923 1.14961 0.25124 0.74464 -2.14550 1924 1.87971 0.89010 1.71715 -4.48696 1925 1.92283 1.49884 1.56166 -4.98333 1926 0.58719 0.14349 0.54483 -1.27551 1927 1.14147 0.48401 0.76159 -2.38706 1928 1.85702 0.00108 0.90609 -2.76420 1929 2.34650 1.21206 0.80641 -4.36497 1930 1.23300 0.98301 1.41118 -3.62719 1931 1.45928 0.66425 0.49934 -2.62286 1932 1.24577 0.62167 0.65859 -2.52603 1933 1.75223 1.11824 -0.40401 -2.46646 1934 1.36568 0.29556 1.28974 -2.95097 1935 0.37210 0.20026 1.00178 -1.57414 1936 0.95349 0.12575 1.01541 -2.09466 1937 0.70771 -0.79286 1.07029 -0.98513 1938 1.54315 0.82295 0.48427 -2.85037 1939 0.76306 0.70974 1.03394 -2.50674 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 1.91572 0.97658 0.14624 -3.03854 1947 . . . . 1948 . . . . 1949 . . . . 1950 0.53735 0.65933 0.91561 -2.11230 1951 . . . . 1952 2.54196 0.87379 -1.53894 -1.87680 1953 . . . . 1954 . . . . 1955 3.32061 0.69826 -0.94642 -3.07245 1956 0.85548 0.37216 0.15271 -1.38035 1957 0.57357 -2.02477 1.93854 -0.48735 1958 -0.92014 -0.89730 -0.47280 2.29024 1959 . . . . 1960 0.85803 1.35120 1.38656 -3.59579
176
Table 6.10 Values of variable P2 (relative expenditures on sickness benefits) Year P2NVV P2CNV P2NKV P2CAT ------------------------------------------------------------------1920 0.94328 0.13191 0.55992 -1.63511 1921 . . . . 1922 0.42811 0.26697 0.58958 -1.28467 1923 0.39942 0.27977 0.51862 -1.19782 1924 0.45771 0.14923 0.56105 -1.16799 1925 0.40383 0.07213 0.54164 -1.01760 1926 0.42621 0.06338 0.62394 -1.11353 1927 0.42585 0.00496 0.63236 -1.06318 1928 0.15440 -0.09533 0.59405 -0.65312 1929 0.52103 0.31024 -0.37958 -0.45170 1930 0.57141 -0.18822 -0.14475 -0.23844 1931 0.18837 -0.37613 0.14069 0.04706 1932 0.19503 -0.30394 0.24600 -0.13708 1933 0.52735 0.32322 0.51321 -1.36378 1934 -0.09252 0.38053 0.43335 -0.72135 1935 -0.02523 0.20007 0.46324 -0.63807 1936 -0.04307 0.16293 0.22808 -0.34793 1937 -0.20176 -0.11935 0.23338 0.08773 1938 -0.29066 -0.05980 0.17789 0.17257 1939 -0.47898 -0.02260 0.20218 0.29940 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 1.19613 1.88391 1.56489 -4.64493 1947 0.38053 1.08769 -3.32327 1.85505 1948 . . . . 1949 . . . . 1950 . . . . 1951 1.69666 -1.61250 -4.57625 4.49209 1952 . . . . 1953 . . . . 1954 . . . . 1955 . . . . 1956 . . . . 1957 . . . . 1958 0.93407 -2.09147 -0.47620 1.63360 1959 1.91990 -5.18975 0.64500 2.62485 1960 1.96892 -4.93116 0.61247 2.34976
177
Table 6.11 Values of variable P3 (relative expenditures on death benefits) Year P3NVV P3CNV P3NKV P3CAT ------------------------------------------------------------------1920 0.29551 -0.78005 -0.64553 1.13006 1921 . . . . 1922 0.13387 -0.57350 0.00999 0.42964 1923 -0.26269 -0.64973 -0.26495 1.17737 1924 -0.31674 -0.68570 -0.57046 1.57290 1925 -0.11131 -0.61063 -0.51201 1.23395 1926 -0.19411 -0.62669 -0.88145 1.70225 1927 -0.25172 -0.76760 -0.63220 1.65152 1928 -0.17463 -0.71180 -0.53298 1.41942 1929 -0.02860 -0.59397 -0.26958 0.89216 1930 0.01280 -0.73517 -0.40201 1.12438 1931 0.04075 -0.68270 -0.38618 1.02813 1932 -0.18655 -0.71411 -0.18721 1.08787 1933 0.42715 -0.19592 0.11342 -0.34465 1934 0.13941 -0.80049 0.38302 0.27806 1935 -0.01919 -0.34439 0.08774 0.27584 1936 -0.19455 -0.02023 0.30896 -0.09419 1937 -0.27745 -0.13679 0.32723 0.08702 1938 -0.25299 -0.14372 0.33577 0.06094 1939 -0.22081 -0.06113 0.76043 -0.47848 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 0.46084 0.33707 0.76681 -1.56472 1947 0.27514 1.65015 -3.37040 1.44511 1948 . . . . 1949 -0.46968 -0.30086 -0.27291 1.04345 1950 -0.45746 0.17677 -0.60361 0.88431 1951 -1.51656 0.41877 -0.16401 1.26181 1952 -0.69548 0.14432 -0.37583 0.92699 1953 -0.91550 0.41916 -0.11032 0.60666 1954 -0.93646 0.63473 -0.26555 0.56728 1955 -0.75441 0.56877 -0.31278 0.49841 1956 0.76155 1.25481 0.13493 -2.15129 1957 0.72812 1.23403 -0.36235 -1.59980 1958 -0.27215 1.00147 -0.43264 -0.29668 1959 -0.29481 1.16618 -0.52930 -0.34207 1960 -0.19286 1.08005 -0.41828 -0.46892 1961 -0.41230 1.23385 -0.47235 -0.34920 1962 -1.27633 1.49136 -0.16111 -0.05393
178
Table 6.12 Values of variable P4 (relative subscription costs)
Table 6.13 Values of %NR* (% of the non-religious)
Year P4NVV P4CNV P4NKV P4CAT ---------------------------------------------------------------1920 0.38795 0.05021 0.11275 -0.55091 1921 . . . . 1922 0.29696 0.01563 -0.01076 -0.30183 1923 0.27236 0.03559 -0.04193 -0.26601 1924 0.29234 0.04070 0.02993 -0.36297 1925 0.32942 0.09490 -0.00672 -0.41760 1926 0.31976 0.08854 -0.05131 -0.35700 1927 0.33868 0.15077 -0.00860 -0.48085 1928 0.32061 0.09595 -0.02528 -0.39128 1929 0.33345 0.14555 0.00003 -0.47903 1930 0.30357 0.10492 0.03262 -0.44111 1931 0.28903 0.09729 0.01256 -0.39887 1932 0.23497 0.06702 -0.03150 -0.27049 1933 0.27032 0.08337 0.06885 -0.42255 1934 0.25440 0.07826 0.00801 -0.34067 1935 0.25508 0.10679 -0.01630 -0.34556 1936 0.26969 0.10864 -0.04714 -0.33119 1937 0.30603 0.08110 -0.03674 -0.35039 1938 0.31686 0.10168 -0.02853 -0.39000 1939 0.34137 0.12553 -0.01231 -0.45460 1940 . . . . 1941 . . . . . 1942 . . . . . 1943 . . . . . 1944 . . . . . 1945 . . . . . 1946 0.30101 0.18183 0.24343 -0.72627 1947 0.25966 0.19625 0.19729 -0.65321 1948 0.32255 0.16339 0.23233 -0.71828 1949 0.23282 0.19428 0.10546 -0.53256 1950 0.23322 0.12779 0.12858 -0.48960 1951 0.21941 0.14499 0.10947 -0.47387 1952 0.21812 0.08044 0.13647 -0.43503 1953 0.21929 0.11620 0.13351 -0.46900 1954 0.19187 0.14686 0.14684 -0.48557 1955 0.15939 0.13162 0.16367 -0.45467 1956 0.19991 0.11908 0.14428 -0.46327 1957 0.15188 0.05857 0.08709 -0.29754 1958 0.20980 0.07339 0.13014 -0.41333 1959 0.20866 0.06733 0.13137 -0.40735 1960 0.21055 0.06524 0.13559 -0.41138 1961 0.22674 0.06041 0.11690 -0.40405 1962 0.22581 0.05657 0.09679 -0.37917 *
7.80 8.46 9.12 9.78 10.44 11.10 11.76 12.42 13.08 13.74 14.40 14.56 14.72 14.88 15.04 15.19 15.35 15.51 15.67 15.83 15.99
16.94 17.10 17.20 17.30 17.40 17.50 17.60 17.70 17.80 17.90 18.00 18.10 18.20 18.30 18.40 18.87 19.35
Interpolated on the basis of CBS (1989), Negentig jaren statistiek in tijdreeksen
179
Table 6.14 Values of variable Q1 (absolute amount of expenditures on strike-pay per member, in guilders per year) Year Q1NVV Q1CNV Q1NKV Q1CAT -------------------------------------------------------------------------------1920 9.98481 2.72564 2.57852 0.03641 1921 . . . . 1922 5.53837 3.29390 6.21077 0.11452 1923 4.72650 1.92478 3.15256 0.17518 1924 7.03280 2.61425 5.97767 0.01208 1925 4.85537 3.17748 3.38351 0.00486 1926 1.67138 1.07245 1.60207 0.25949 1927 1.33625 0.69241 0.91392 0.03922 1928 4.56399 0.71338 1.76343 0.04491 1929 5.47733 1.76151 1.17413 0.00667 1930 1.12506 0.87621 1.34449 0.00872 1931 3.79187 1.71228 1.45197 0.06397 1932 7.78173 4.16900 4.32581 0.17906 1933 3.40575 1.80664 0.39425 0.05013 1934 0.41485 0.14228 0.38451 0.00554 1935 0.50489 0.42518 0.94768 0.07210 1936 0.30755 0.13441 0.32720 0.01459 1937 0.15636 0.03487 0.22469 0.02877 1938 0.79501 0.38689 0.27574 0.00982 1939 0.17957 0.17024 0.23543 0.00683 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 1.46749 0.57374 0.25009 0.01035 1947 0.12931 0.03720 0.00000 0.01034 1948 0.04569 0.03054 0.00000 0.08487 1949 0.09584 0.05068 0.07089 0.00000 1950 0.06017 0.06798 0.08784 0.00425 1951 0.03188 0.02721 0.01395 0.00000 1952 0.36417 0.06868 0.00615 0.00439 1953 0.37688 0.00990 0.01025 0.00000 1954 0.05988 0.02605 0.02090 0.00000 1955 1.25180 0.09092 0.01755 0.00209 1956 0.06712 0.04139 0.03324 0.00718 1957 0.01336 0.00099 0.05231 0.00462 1958 0.00920 0.00941 0.01439 0.22803 1959 0.03053 0.00000 0.00365 0.01197 1960 1.45574 2.38378 2.46957 0.01694 1961 0.44379 0.00000 0.02169 0.00000 1962 0.35294 0.00000 0.00478 0.00000
180
Table 6.15 Values of variable Q2 (absolute amount of expenditures on sickness benefits per member, in guilders per year) Year Q2NVV Q2CNV Q2NKV Q2CAT -------------------------------------------------------------1920 2.49372 1.10784 1.69964 0.18926 1921 . . . . 1922 1.99676 1.69958 2.34666 0.36014 1923 1.60787 1.42654 1.81142 0.32552 1924 1.79893 1.32143 1.99477 0.35398 1925 1.89849 1.36255 2.17900 0.45823 1926 1.88653 1.31247 2.29901 0.40454 1927 1.97519 1.29664 2.42827 0.44559 1928 1.71318 1.33457 2.65912 0.76401 1929 2.09942 1.70041 0.85304 0.79369 1930 1.36268 0.63752 0.66585 0.60630 1931 1.22778 0.69817 1.17061 1.06599 1932 1.13018 0.68620 1.18927 0.81080 1933 0.81402 0.66373 0.80260 0.12284 1934 0.42038 0.67466 0.71125 0.22415 1935 0.44131 0.55283 0.71926 0.23911 1936 0.40376 0.49613 0.52953 0.29767 1937 0.39640 0.43045 0.61251 0.52949 1938 0.33204 0.41826 0.53049 0.52767 1939 0.30016 0.47377 0.59318 0.65374 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 0.14731 0.29305 0.21301 0.00043 1947 0.15776 0.31997 0.00389 0.68924 1948 0.10528 0.21245 0.00000 1.20774 1949 0.15348 0.11163 0.00000 1.41231 1950 0.10745 0.00000 0.00049 1.57835 1951 0.10091 0.00369 0.00019 1.65178 1952 0.13715 0.00000 0.00622 1.15284 1953 0.37688 0.00990 0.01025 0.00000 1954 0.14798 0.00000 0.00006 0.98546 1955 0.12281 0.00000 0.00145 1.02075 1956 0.21318 0.000000 0.00000 0.51069 1957 0.20674 0.000230 0.00000 0.44673 1958 0.23191 0.011255 0.05660 0.46680 1959 0.27973 0.000229 0.07817 0.56610 1960 0.27410 0.000276 0.07060 0.40114 1961 0.23866 0.000000 0.07952 0.26471 1962 0.23529 0.000000 0.08134 0.20661
181
Table 6.16 Value of variable Q3 (absolute amount of expenditures on death benefits per member, in guilders per year) Year Q3NVV Q3CNV Q3NKV Q3CAT -------------------------------------------------------------1920 0.24229 0.08265 0.09455 0.55818 1921 . . . . 1922 0.20080 0.09898 0.17740 0.26991 1923 0.20071 0.13630 0.20026 0.84721 1924 0.20216 0.13978 0.15686 1.33769 1925 0.18575 0.11274 0.12443 0.71314 1926 0.17027 0.11048 0.08563 1.13431 1927 0.17192 0.10263 0.11751 1.15315 1928 0.18161 0.10613 0.12692 0.89420 1929 0.16678 0.09476 0.13107 0.41883 1930 0.16233 0.07683 0.10721 0.49333 1931 0.15472 0.07505 0.10096 0.41531 1932 0.11258 0.06643 0.11250 0.40265 1933 0.13926 0.07469 0.10176 0.06436 1934 0.12779 0.04992 0.16304 0.14679 1935 0.11868 0.08573 0.13208 0.15941 1936 0.09974 0.11874 0.16502 0.11027 1937 0.09426 0.10849 0.17255 0.13571 1938 0.08845 0.09867 0.15937 0.12107 1939 0.09562 0.11218 0.25509 0.07390 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 0.17386 0.15362 0.23609 0.02294 1947 0.06141 0.24287 0.00160 0.19785 1948 0.03977 0.03888 0.00000 0.30650 1949 0.06328 0.07492 0.07704 0.28735 1950 0.06197 0.11686 0.05355 0.23710 1951 0.02062 0.14283 0.07975 0.33185 1952 0.06354 0.14716 0.08748 0.32188 1953 0.06293 0.23904 0.14077 0.28834 1954 0.06088 0.29296 0.11907 0.27385 1955 0.07189 0.26997 0.11180 0.25162 1956 0.19087 0.31257 0.10200 0.01037 1957 0.18349 0.30431 0.06166 0.01789 1958 0.13224 0.47260 0.11263 0.12904 1959 0.11502 0.49577 0.09098 0.10971 1960 0.13523 0.48294 0.10794 0.10261 1961 0.09467 0.49107 0.08916 0.10084 1962 0.03529 0.56195 0.10766 0.11983
182
Table 6.17 Value of variable Q4 (absolute amount of dues per member, in guilders per year) Year Q4NVV Q4CNV Q4NKV Q4CAT -------------------------------------------------------------1920 26.4985 18.9036 20.1235 10.3629 1921 . . . . 1922 27.1336 20.4798 19.9463 14.9092 1923 24.5179 19.3488 17.9056 14.3111 1924 24.9570 19.4047 19.1968 12.9598 1925 24.9079 19.7009 17.7972 11.8008 1926 25.3385 20.1078 17.4834 12.8786 1927 25.5299 21.1563 18.0396 11.2494 1928 25.3548 20.2531 17.9409 12.4420 1929 25.4469 21.0879 18.2320 11.2923 1930 24.4600 20.0532 18.6546 11.6158 1931 23.4385 19.3490 17.7771 11.7809 1932 19.9914 16.9006 15.3150 12.0594 1933 19.6418 16.2927 16.0578 9.8237 1934 18.5654 15.5671 14.5110 10.2393 1935 17.7113 15.2705 13.5018 9.7140 1936 17.5938 14.9767 12.8162 9.6471 1937 18.4300 14.7177 13.0817 9.5598 1938 18.9865 15.3107 13.4414 9.3640 1939 18.9335 15.2578 13.2932 8.5417 1940 . . . . 1941 . . . . 1942 . . . . 1943 . . . . 1944 . . . . 1945 . . . . 1946 30.9661 27.4870 29.2333 11.0851 1947 32.6414 30.6359 30.6676 13.1013 1948 34.2878 29.2428 31.3297 12.1092 1949 34.2578 32.9626 30.1612 15.9353 1950 34.8756 31.3859 31.4107 16.9280 1951 37.4702 34.7831 33.5693 18.7327 1952 37.7106 32.8600 34.7539 19.6248 1953 38.2802 34.5307 35.1336 19.2333 1954 39.1844 37.4596 37.4588 19.9023 1955 40.0755 38.9779 40.2476 21.6868 1956 43.7997 40.3991 41.4300 22.5661 1957 44.8412 40.8464 42.0282 28.6087 1958 47.4307 41.3825 43.7991 25.4353 1959 48.6642 42.2504 45.0446 26.2832 1960 51.1352 44.2196 47.4421 27.4549 1961 53.5858 45.3750 48.0120 28.5168 1962 55.6255 46.9646 48.8923 30.3760
183
6.II
Simultaneous modelling: derivation of the restrictions from the system of four equations
a)The pre-war system of four equations Define NVV = a, CNV = b, NKV = c, CAT = d. The four equations can then be written as: Ma Mb Mc Md
= α0 = β0 = γ0 = δ0
+ α1P + α2P + α3P + α4P + β1P + β2P + β3P + β4P + γ1P + γ2P + γ3P + γ4P + δ1 P + δ2P + δ3P + δ4 P
+ α5%NR + β5%NR + γ5%NR + δ5 %NR
(6.1) (6.2) (6.3) (6.4)
Since the four membership shares Ma to Md add up to 1 for all possible values of the variables on the right-hand side of the equation, Md can be rewritten: Md = 1 - Ma - Mb - Mc (6.5) Given the formula of P on page 167, each of the four Pi-variables can be rewritten as: P = ln Q - ¼ln (Q⋅Q⋅Q⋅Q) =+¾lnQ - ¼lnQ - ¼lnQ - ¼lnQ P = -¼lnQ + ¾lnQ - ¼lnQ - ¼lnQ P = -¼lnQ ¼lnQ + ¾lnQ - ¼lnQ P = -¼lnQ ¼lnQ - ¼ lnQ + ¾lnQ
(6.6) (6.7) (6.8) (6.9)
Adding up these last four equations leads to the following result: P+P+P+P=0 (6.10) This proves that the values of the P-variables add up to 0. Rewriting (6.10) gives: P = -P - P - P Substituting (6.5) and (6.10)' in (6.4) gives: (1-Ma-Mb-Mc) =δ0 + δ1(-P - P - P) + δ2(-P - P - P) + δ3(-P - P - P) + δ4(-P - P - P) + δ5%NR Rewriting this equation gives: Ma+Mb+Mc =(1-δ0) + δ1(P + P + P) + δ2(P + P + P) + δ3(P + P + P) + δ4(P + P + P) - δ5%NR
(6.10)'
(6.11)
(6.12)
Comparing (6.12) with the sum of (6.1), (6.2) and (6.3), the following three restrictions can be derived: 1e)(1 - δ0) = α0 + β0 + γ0 ⇒ α0 + β0 + γ0 + δ0 = 1 2e)-δ5 = α5 + β5 + γ5 ⇒ α5 + β5 + γ5 + δ5 = 0
Finally: δi(P + P + P) = αiP + βiP + γiP This equation has to hold for all values of P, P, and P. We therefore have: 3e)αi = βi = γi = δi(i = 1,2,3,4)
(6.13) (6.14) (6.15)
(6.16)
184
b)The post-war system of four equations The four equations can be written as: Ma Mb Mc Md
= α0 = β0 = γ0 = δ0
+ α3P + α4P + α5%NR + α6Dummy51 + β3P + β4P + β5%NR + β6Dummy51 + γ3P + γ4P + γ5%NR + γ6Dummy51 + δ3P + δ4 P + δ5 %NR + δ 6Dummy51
(6.17) (6.18) (6.19) (6.20)
The equations (6.5) up to and including (6.10)' also hold true for the post-war system of equations. Substituting (6.5) and (6.10)' in (6.20) gives: (1-Ma-Mb-Mc) =δ0 + δ3(-P - P - P) + δ4(-P - P - P) + δ5%NR + δ6Dummy51
(6.21)
Rewriting this equation gives: Ma+Mb+Mc =(1-δ0) + δ3(P + P + P) + δ4(P + P + P) - δ5%NR - δ6Dummy51
(6.22)
Comparing (6.22) with the sum of (6.17), (6.18) and (6.19), the following four restrictions can be derived: 1e)(1 - δ0) = α0 + β0 + γ0 ⇒ α0 + β0 + γ0 + δ0 = 1 2e)-δ5 = α5 + β5 + γ5 ⇒ α5 + β5 + γ5 + δ5 + 0 3e)-δ6 = α6 + β6 + γ6 ⇒ α6 + β6 + γ6 + δ6 = 0 And finally: δi(P + P + P) = αiP + βiP + γiP ⇒ 4e)αi = βi = γi = δi(i = 3,4)
(6.23) (6.24) (6.25) (6.26) (6.27)
185
186
7.SUMMARY AND CONCLUSIONS In chapter 1 the object of this study is introduced. We set out to explain trade union growth and decline in the Netherlands from a broad point of view, taking into account the contribution to this subject of the economic, historical, and social sciences. In an attempt to integrate these different lines of approach, the theoretical framework of demand and supply by Hirsch & Addison (1986) is utilized. They formulate two functions, representing the demand for union services and the supply of union services. The demand for union services is determined by the subscription fee, income or wage, the relative union compensation gain, the costs of substitute services, the nonpecuniary benefits of a union work environment, and the worker tastes towards unionism. The supply of union services is determined by the revenue of union services, the costs of organizing new members, the servicing costs, and union goals. This economic framework forms the starting point of this research, but it had to be adjusted and extended in order to allow for specific Dutch circumstances, and to incorporate the determinants brought to the fore by other disciplines. The basic questions concerning the demand for unionism are: what determines the probability whether or not to become or remain a union member, and what kind of services are demanded? The basic questions concerning the supply of unionism are: which factors determine the supply of unions, and what kind of services are offered? These questions run through this thesis like a continuous thread. Actually, union services cannot be quantified properly, so that we study fluctuations in union membership instead, which adequately reflects trade union growth and decline. Trying to explain this phenomenon through time, we make use of several research methods, depending on the scientific lines of approach and the data at hand. Chapter 2 explores the explanations of union growth and decline, the way that they are advanced in the international economic literature. The findings of this chapter, supplemented with the findings of chapter 3, in which the historical development of the Dutch union movement is described, serve as guidance for our own empirical analyses in the chapters 4 and 5. Finally, in chapter 6 attention is paid to the supplementary lines of approach from the social sciences. The results of each of these chapters are summarized in some more detail below, in addition to which we shall evaluate to what extent the questions of this thesis have been answered. The survey of economic theories and empiricism in chapter 2 distinguishes between studies based on time-series analyses, and on cross-sectional analyses, respectively. Both research methods have their pros and cons. The advantage of time-series studies lies in the fact that (annual) fluctuations in (aggregate) union membership can be explained over a long period of time, but the disadvantage is formed by the small number of suitable explanatory variables at disposal. Apart from discontinuity problems in many series, several structural determinants of trade union growth and decline, such as gradual changes in the value system of a society, cannot be measured at all.
187
Conversely, cross-sectional analyses are based on often large enquiries, from which a great many suitable variables emanate which explain why individuals choose to be a member or not, but the disadvantage lies in the fact that such questionnaires are mostly conducted only once. Panel data could solve this problem, but to our knowledge the existing ones do not contain information on Dutch trade union membership. In order to acquire an overall picture of (fluctuations in) union membership, we therefore need both approaches as they complement each other. In chapter 2 an overview is presented of all the relevant theories and empirical findings with respect to both fields. Next, chapter 3 sets out how the trade union movement in the Netherlands has evolved from the 1860's, the time in which the first unions emerged, until approximately 1990. It is important to gain insight into the features of the Dutch industrial relations and institutions before we can build our own time-series - and cross-sectional models, because allowance has to be made for country-specific determinants of trade union membership. Moreover, this historical analysis also stresses developments and events which cannot be quantified properly or not at all, but which did have a significant impact on Dutch unionization. This argument especially applies to the early history of the first trade unions in the Netherlands, a period for which hardly any statistics exist. From the description of this genesis of the Dutch labour movement it becomes clear that its size remained very small until the beginning of the 20th century, not only due to economic circumstances (the relative late industrialization in the Netherlands, and the large dispersion of economic activity), but to a large extent also owing to the prevailing mental climate, which was heavily influenced by religious beliefs that preached obedience (to the employer and to the Church) and condemned labour disputes. With the foundation of the socialist federation NVV in 1906, this provoked the Protestants and the Roman Catholics to establish their own labour centres in 1909, which was the start of a long-lasting tripartite division within the trade union movement, consequent on the ideological fractionalization of the whole Dutch society. The Dutch unions pursued three goals: emancipation of the workers, protection of the collective interests of the labour force, and protection of the individual interests of their members. Before World War II, the trade union movement accomplished a lot with respect to all of these goals. Supported by the political parties of the same signature, the three major federations managed to improve the social and legal position of the working class considerably. Because the state hardly provided any social benefits payments yet, the unions offered all kinds of social insurances, which also served as a means to hold on to their members. In chapter 3 it is concluded that the supply of union services linked up with the demand in the pre-war period, but this does not imply that the trade union density rate was very high. Partly due to the pillarization, which forestalled unity in the union movement, individual unions were often not able to offer resistance to the joint actions of employers' associations. In spite of the Ghent system, through which unemployed union members receiv-ed a (higher) unemploy-
188
ment benefit for 6 or 13 weeks at the most, during the severe downturns of the business cycle the Dutch workers were very vulnerable, and the unions powerless. Only very gradually, unions became an accepted phenomenon by employers, government and church, which paved the way for unionization on a larger scale. After the second World War the three major federations were recognized as official advisors to the government, which expressed itself in the unions' participation in several consultative boards jointly with the employers' associations. In order to rebuild the heavily damaged economy, and to prevent a recurrence of the great miseries of the Great Depression, almost all parties (initially) worked closely together. Only the communist EVC organized many strikes in the early post-war years and at first attracted a lot of disappointed workers, but it lost strength quickly from the 1950's. The three recognized pillars held the view that conflicts were to be avoided for the common good. Thanks to this approach the recovery programmes quickly yielded rewards, and upon solid economic foundations the Dutch welfare state developed. In exchange for their enhanced influence on the central level, unions had renounced the claim of worker participation on the plant level. In the course of the post-war period, the union movement on the whole (except for the, mostly small-scale, categorial unions), kept emphasizing the collective interests of the entire labour force, and failed to become really visible on the work floor. This neglect of their members' individual interests, combined with their contribution to better working conditions and to a very good social security system provided by the state, partly explains why the union density rate in the Netherlands remained rather low. It is also argued that the law on extension of collective labour agreements stimulates free rider conduct. The strong decline in union density from 1979 can partly be ascribed to the recession, but with the upturn of the business cycle this density rate did not really recover. There are structural developments in progress, obstructing union growth. The most important determinants appear to be the changing composition of the labour force, and the ongoing decentralization of the bargaining process to the level of individual firms, for which unions are (still) not fully equipped to handle. In addition, the Dutch union movement has remained divided, which also weakens its (recruitment) power. Chapters 2 and 3 form the basis of the empirical analyses conducted in the chapters 4, 5, and 6. The second chapter contains the general theories of union membership determination, while the third chapter provides us with insights into the Dutch trade union movement, leading to the formulation of extra hypotheses in addition to those of the preceding chapter. In chapter 4 the fluctuations in the aggregate trade union membership are explained with the aid of time-series analyses. This research starts with an attempt to replicate the influential models of Ashenfelter & Pencavel (1969) and of Bain & Elsheikh (1976) on the Dutch data, but without success. Because there is a structural break around World War II, it is not possible to estimate these models on one continuous period only. But even if two different subperiods are analyzed, the specific
189
combinations of variables that could explain American and British union growth, respectively, both turn out to be unsuitable for the Dutch case. Next, a new pre-war and a new post-war model are constructed, employing other combinations of variables, which are much better able to explain the developments of Dutch union membership. Before World War II, union growth was predominantly stimulated by price and wage increases (reflecting a threat and a credit effect, respectively), and obstructed by the number of strikes in the preceding year, probably owing to the fact that many workers were unfavourably disposed towards strikes out of fear of losing their jobs, or out of religious conviction. Finally, the level of unemployment initially advanced the membership rates, but had a negative effect on unionization in the long term. This confirms the theoretical notions and the findings in chapter 3, that a downturn in the business cycle (often) at first incited people to join or to hang on to their union in hopes of protection, and because they received an unemployment benefit through the union. When unemployment persisted, unions lost power and workers lost confidence, so that they decided not to join or to quit the union after all. After the second World War, due to laws and institutional arrangements, price and wage changes no longer had a significant impact on Dutch unionization. Unemployment remained an important determinant, but had a negative effect only. Union growth was also obstructed by the union density rate, which exerted a saturation effect. This can be ascribed to the rather small recruitment efforts of the unions because of their official recognition as negotiating and consultation partners. One last detrimental impact on Dutch unionism has been the large extension of social benefits payments by the state, which exerted a sizeable substitution effect. The influence of left-wing parties in Parliament favoured the union movement, and, unlike the pre-war period, strikes now induced workers to join a union. However, when the EVC-members are excluded from the analysis, the positive impact of strikes becomes much smaller, confirming the notion that the increase of strike activities right after the war especially appealed to workers choosing for the radical EVC. Evaluation of the results of chapter 4 leads to the conclusion, that only a relatively small number of variables can be taken into account. It is impossible to test all the explanations brought to the fore in the chapters 2 and 3. In chapter 4 the emphasis lies on cyclical determinants of union growth and decline, supplemented with institutional variables. From the findings of the pre- and post-war models we may conclude that the ups and downs of the business cycle especially had a significant impact before the second World War, reflected in the development of wages, inflation, strikes and unemployment. After the war, the impact of these business cycle effects became partly suppressed by institutional arrangements, but the level of unemployment remained an important determinant. In chapter 5 we have two large data sets at our disposal, the OSA and the AVO, which we employ in two separate cross-sectional analyses. The information in the OSA-survey of 1988 enables us to analyze a great many characteristics of workers which influence the probability whether or not to become a union member. In addition, we can distinguish between the factors that influence
190
the decision whether to join a union at all, and the factors that affect the choice for a specific federation. The econometric analysis shows that these two decisions are not taken simultaneously, but that most workers first decide whether or not to become a union member and only after that decide which federation to join. Therefore we use a so-called sequential logit model in our estimations. The results of this model mostly confirm the theories and empirical findings discussed in chapter 2, with a few notable exceptions which can usually be attributed to specific Dutch circumstances. With respect to personal characteristics, it is found that the likelihood of union membership rises with being a male, age, tenure, and education. This last result deviates from international empirical findings, and is typical for the Netherlands, where a lot of unions operate which especially promote the interests of the higher personnel. Regarding occupational characteristics, the membership probability is negatively influenced by having a commercial job, and a part-time or a temporary job. A positive effect emanates from having an educational profession, working in a large company, being a civil servant, doing unpleasant work and the expectation to become unemployed. In all these cases, workers expect to improve their situation or terms of employment through union membership, contrary to the first three job related characteristics mentioned above. We find a significant parabolic effect of net wages, indicating that the likelihood of union membership first rises with wage, but after a certain income level decreases again. This result suggests a credit effect of pay rises until a certain break point, after which the benefits of trade union membership decrease: better paid employees are often less in need of union protection. However, despite its significance, we find the actual effect of a wage increase on unionization to be very small, which is in accordance with the results obtained in the post-war time-series analysis of chapter 4. Finally, the two regional characteristics employed both show a significantly negative effect, contrary to the expectations. Living in the densely populated West, as well as the degree of urbanization, are included to test the assumption that when a lot of workers are concentrated in one area or city, this stimulates unionization. This so-called proximity influence does not occur in the Netherlands. One last remarkable finding of the OSA-survey concerns the choice for a specific federation. This is influenced by just a small number of variables, of which the religious denominations are the most important ones. In spite of an ongoing secularization, religion still seems to have a considerable impact on the unionization decision. Turning to the AVO-data, a somewhat smaller number of variables could be used, and those are not all identical to the ones employed in the OSA-analysis. As far as the variables are comparable, the AVO-analysis shows the same results. Relevant new variables are five dummies indicating in which industrial sector the individual works. The outcome with respect to these industry variables confirms the theoretical and empirical findings reported in chapter 2: workers in agriculture and in services are significantly less organized than workers in the manufacturing
191
industry, while workers in the transport and the public sector are significantly higher organized. The availability of two AVO-samples, of 1979 and 1987, makes it possible to analyze the changes between these two years, which coincide exactly with the period of union decline in the Netherlands. On the basis of actual socio-economic developments between 1979 and 1987, we formulate several assumptions with respect to which variables contribute most to the fall in the union density rate. Most of our expectations are confirmed by the estimation results: union membership rates have been negatively affected by the rising share of females and part-timers in the labour force, the waning interest of younger workers to join a union, and the shifts in employment from the industrial towards the service sector. By means of a decomposition method we can establish that the decline in union density is largely caused by changes in the working population, and to a lesser extent by changes in unionization behaviour. The evaluation of the overall results of chapter 5 leads to the conclusion that the basic questions of this thesis regarding the demand for unionism, have been extensively dealt with, thanks to the abundance of variables in the OSA and AVO data sets. However, the cross-sectional analyses nor the time-series analyses are able to sufficiently answer the basic questions regarding the supply of unionism, due to a lack of relevant information in the OSA- and the AVO-data. The findings of chapter 6 partly supply this deficiency, as we turn to the contribution of the social sciences to the explication of Dutch union growth and decline. In the first part of this chapter we review the most interesting theories and empirical results of social scientists, from which it becomes clear that they partly stress other determinants of unionism, and that they use other data sets, which contain more information about the role of unions themselves. Within the social psychology, three theoretical lines of approach to trade union participation are distinguished. The first approach views the unionization decision as a result of feelings of frustration and aggression, but is considered to be subordinate to the other two theories, the rational-choice and the interactionist approach. Rational choice closely resembles the costs-benefits approach of economists, but social scientists have more eye for collective, ideological and social motives, next to individual (material) motives. The workers' attitude towards unionism is determined by balancing all the positive and negative consequences of participation. Next to viewing unionization as a deliberate rational choice, much attention is paid to the influence of someone's social environment in the so-called interactionist approach. People may choose to become a union member out of tradition or social custom, or they are put up to join as a result of a high union density rate in their own firm, or because of their contacts with the union on the work floor. The importance of this last aspect is touched upon in the historical analysis of chapter 3, but has not received much attention in the economic literature. The short survey of empirical results obtained by social scientists shows that in the Netherlands ever less people join a union out of idealistic motives. The majority chooses to become members for personal gains, and easily leaves again if the costs
192
outweigh the benefits. This is reflected in a very high turnover rate, already noted in chapter 3. Apparently, unions have great difficulties in holding on to their members, which is really not so strange given the discrepancy between the growing need of workers for union services directed towards the individual member, and the ongoing emphasis of the Dutch trade union movement on its external, collective functions. One way or another, unions must try to keep their existing members committed to the organization, to prevent their leaving. Several statistical studies have shown that the probability of union membership is significantly and foremost influenced by the union density rate of one's company, the number of contacts with other union members, and the number of contacts with the union itself. These interaction processes lead to a certain degree of commitment to the union, which is found to be an important determinant of the decision whether or not to quit union membership. It is argued that the key concept of commitment combines the rational-choice approach with the interactionist approach. Rational choice and/or social pressure may induce people to join the union; if they are satisfied in their expectations concerning the use of their membership, this leads to more commitment, which in turn induces them to participate more actively, which again stimulates interaction processes, and hence commitment, so that the likelihood of resigning decreases. In the second part of chapter 6 an attempt is made to incorporate some of these sociologic notions into economic modelling, but because of a lack of sufficient data at our disposal, only those variables are included that reflect the rational-choice approach, and not the interactionist approach. The surplus value of this new way of modelling, lies in the fact that it investigates the effects of trade union policies on membership. More specifically, it is analyzed to what extent the share of a federation in the total Dutch union membership can be explained by the degree in which such a confederation involves more or less than average costs for the union members (referring to its membership dues), and by the degree in which such a federation more or less endows to the members (referring to its benefits payments). Four 'federations' are distinguished, namely the NVV, CNV, NKV and the categorial unions. In the first instance, we estimate all four equations separately over two subperiods, before and after World War II. This does not lead to very satisfactory results, because the models come up with a lot of insignificant coefficients and/or variables obtaining the wrong sign. In a second analysis, we incorporate the four equations in a simultaneous system, leading to more efficient estimates; almost all variables included now obtain significant coefficients. Overall, the empirical results of these models confirm the theoretical notion that trade union members at least partly base their decision which union (federation) to join on costs-benefits considerations: for the majority of the unions' benefits payments it is found, that when a federation pays more than the average to its members, its share in the total union membership will be higher; and when a federation levies a subscription fee higher than the average, its share will be lower.
193
Suggestions for further research All in all, the findings of this thesis enable us to fill in the framework of demand and supply of unionism reasonably well, but given the broadness of the subject a lot of questions still remain unanswered. For instance, not much attention is paid to the large differences in trade union density between the different branches of industry. It should be possible, at least for the post-war period, to collect more sectoral data and analyze these differences between industries in more detail. When given the opportunity, a new questionnaire could be developed and conducted, which contains all the relevant questions that do justice to all the theoretical notions discussed in the chapters 2, 3, and 6. Ideally, such a survey should take the form of a panel data enquiry. Unfortunately, regarding the pre-war era no workable questionnaire exists at all, so that the unionization behaviour in that period cannot be empirically analyzed, other than with the aid of time-series analysis, of which we already established that only a few variables could be employed. Finally, on the basis of the most recent statistics, more should be said about the development of the Dutch trade union movement during the last few years, between 1990 and 1995. Partly due to the period in which the analyses of the chapters 4 and 5 were conducted (between 1989 and 1992), these analyses do not cover the developments of the most recent years. By the end of 1992 the AVO 1991 became available, while the AVO 1995 comes out next year. This will enable us the analyze the changes in union membership between 1979 and 1995. The biennial CBSsurvey on Dutch trade unions was the last time published in 1993. From this edition it can be seen that despite an overall growth in the number of union members from 1986 onwards, the density rate has remained low (about 24% to 25% between 1987 and 1993). This suggests that trade unions in the Netherlands still have structural problems in organizing a greater part of the labour force.
194
REFERENCES Albeda, W. (1971), Vakbeweging en onderneming (Universitaire Pers, Rotterdam) Albeda, W. & W.J. Dercksen (1989, 4th edition), Arbeidsverhoudingen in Nederland (Samson, Alphen aan den Rijn) Andriessen, J.E., S. Miedema & C.J. Oort (1970, 6th edition), De sociaaleconomische besturing van Nederland (Wolters-Noordhoff, Groningen) Arnhem, C. van & P. Knegtmans (1976), 'Vakorganisaties in Nederland, 1837-1893' (Working paper, University of Amsterdam) Arts, W.A. (1983), 'Vakbondsdemocratie en ledenverlies', Economisch Statistische Berichten 68, pp. 605-608. Arts, W.A. (1985), 'Vakbeweging, conjunctuur en ledenbinding', Economisch Statistische Berichten 70, pp. 857-859. Ashenfelter, O. & G. Johnson (1972), 'Unionism, relative wages and labor quality in U.S. manufacturing industries', International Economic Review 13, pp. 488-507. Ashenfelter, O. & J.H. Pencavel (1969), 'American trade union growth: 1900-1960', Quarterly Journal of Economics 83, pp. 434-448. Bain, G.S (1970), The growth of white-collar unionism (Clarendon Press, Oxford) Bain, G.S. & P. Elias (1985), 'Trade union membership in Great Britain: an individual level analysis', British Journal of Industrial Relations 23, pp. 71-92. Bain, G.S. & F. Elsheikh (1976), Union growth and the business cycle. An econometric analysis (Basil Blackwell, Oxford) Bain, G.S. & F. Elsheikh (1979), 'An inter-industry analysis of unionisation in Britain', British Journal of Industrial Relations 17, pp. 137-157. Bain, G.S. & R.J. Price (1980), Profiles of union growth: a comparative statistical portrait of eight countries (Basil Blackwell, Oxford) Bain, G.S. & R.J. Price (1983), 'Union growth: dimensions, determinants, and destiny', Industrial Relations in Britain (Basil Blackwell, Oxford) Bakker, G.P. den & W. van Sorge (1991), 'Het onbenut arbeidsvolume in het Interbellum', Economisch- en Sociaal Historisch Jaarboek 54, pp. 212-228 (NEHA, Amsterdam) Barnett, G.E. (1916), 'Growth of labor organization in the United States, 1897-1914', Quarterly Journal of Economics 30, pp. 780-795. Berg, A. van den (1994), 'Determinanten van vakbondslidmaatschap in Nederland, 1900-1990' in: H. Flap & M.H.D. van Leeuwen (eds.), Op lange termijn. Verklaringen van trends in de geschiedenis van samenlevingen (Verloren, Hilversum) Berg, A. van den & W. Groot (1992), 'Union membership in the Netherlands: a cross-sectional analysis' Empirical Economics 17, 537-564. Bernstein, I. (1954), 'The growth of American unions', American Economic Review 44, pp. 301-318. Bernstein, I. (1961), 'The growth of American unions, 1945-1960', Labor History 2, pp. 131-157. Blackburn (1967), R.M., Union character and social class (Batsford, London) Blum, A.A. (1968), 'Why unions grow', Labor History 9, pp. 39-72.
195
Boomgaard, P. (1973), 'Stakingen, vakbondslidmaatschap en conjunctuur gedurende het interbellum in Nederland', in: J. van Herwaarden (ed.), Lof der historie, pp. 337-374 (Universitaire Pers Rotterdam) Booth, A.L. (1983), 'A reconsideration of trade union growth in the United Kingdom', British Journal of Industrial Relations 21, pp. 377-391. Booth, A.L. (1985), 'The free rider problem and a social custom model of trade union membership', Quarterly Journal of Economics 99, pp. 253-261. Booth, A.L. (1986), 'Estimating the probability of trade union membership: a study of men and women in Britain', Econometrica 53, pp. 41-62. Brugmans, I.J. (1975, 10th edition), De arbeidende klasse in Nederland in de 19e eeuw 1813-1870 (Het Spectrum, Utrecht/Antwerpen) Centraal Bureau voor de Statistiek (CBS 1979-1988), Statistisch Zakboek Centraal Bureau voor de Statistiek (CBS 1987), Statistiek van de vakbeweging 1987 Centraal Bureau voor de Statistiek (CBS 1989a), 'Leden van de vakbeweging 1989', Sociaal-Economische Maandstatistiek Supplement no.4, pp. 11-24. Centraal Bureau voor de Statistiek (CBS 1989b), Negentig jaren statistiek in tijdreek-
sen
Centraal Bureau voor de Statistiek (CBS 1990), Sociaal-Economische Maandstatistiek 7, no. 3. Centraal Bureau voor de Statistiek (CBS 1991), Sociaal-Economische Maandstatistiek 8, no. 1. Centraal Planbureau (CPB 1980-1988), Centraal Economisch Plan Commons, J.R. et al. (1918), History of labour in the United States I (Macmillan, New York) Conen, G.J.M. (1977), 'Ontwikkeling en samenstelling ledental vakorganisaties', in: T. Akkermans (ed.), Facetten van vakbondsbeleid, pp. 22-57 (Samsom, Alphen aan den Rijn) Cornfield, D.B. (1986), 'Declining membership in the post-war II era: the United Furniture Workers of America, 1939-1982', American Journal of Sociology 91, pp. 1112-1153. Davis, H.B. (1941), 'The theory of union growth', Quarterly Journal of Economics 55, pp. 611-637. Duncan, G.M. & D.E. Leigh (1980), 'Wage determination in the union and nonunion sectors: a sample selectivity approach', Industrial and Labor Relations Review 34, pp. 24-34. Dunlop, J.T. (1949), 'The development of labor organization: a theoretical framework' in: R.A. Lester & J. Shister (eds.), Insights into labor issues (Macmillan, New York) Ebbinghaus, B.O. & J. Visser (1993), Trade union systems in Western Europe; data handbook vol. 1 (Campus Verlag, Frankfurt/New York) Eickhof, N. (1973), Eine Theorie der Gewerkschaftsentwicklung. Entstehung, Stabilität und Befestigung (J.C.B. Mohr, Tübingen) Ellwood, D.T. & G. Fine (1987), 'The impact of right-to-work laws on union organizing', Journal of Political Economy 95, pp. 250-273.
196
Elsheikh, F. & G.S. Bain (1978), 'Trade union growth: a reply', British Journal of Industrial Relations 16, pp. 99-102. Elsheikh, F. & G.S. Bain (1980), 'Unionisation in Britain: an inter-establishment analysis based on survey data', British Journal of Industrial Relations 18, pp. 169178. European Industrial Relations Review (EIRR) (1991), Collective bargaining, trade unions and employers' organisations in Europe (Eclipse Group Ltd, London) Farber, H.S. (1983a), 'Worker preferences for union representation', in: J.D. Reid, Jr (ed.), New approaches to labor unions (JAI Press, Greenwich, CT), pp. 171-206. Farber, H.S. (1983b), 'The determination of the union status of workers', Econometrica 51, pp. 1417-1438. Farber, H.S. & D.H. Saks (1980), 'Why workers want unions: the role of relative wages and job characteristics', Journal of Political Economy 88, pp. 349-369. Fase, W.J.P.M. (1980), Vijfendertig jaar loonbeleid in Nederland (Samson, Alphen aan den Rijn) Fiorito, J. & Ch. R. Greer (1982), 'Determinants of U.S. unionism: past research and future needs', Industrial Relations 21, pp. 1-33. Flanagan, R.J., J. Hartog & J. Theeuwes (1993), 'Institutions and the labour market: many questions, some answers' in: J. Hartog & J. Theeuwes (eds.), Labour market contracts and institutions. A cross-national comparison, pp. 415-446 (North-Holland, Amsterdam) Freeman, R.B. (1976), 'Individual mobility and union voice in the labor market', American Economic Review Papers and Proceedings 66, pp. 361-368. Freeman, R.B. & J.L. Medoff (1984), What do unions do ? (Basic Books, New York) Giele, J. (1978), 'Socialisme en vakbeweging. De opkomst van socialistische vakorganisaties in Nederland (1878-1890) deel 1', Jaarboek voor de geschiedenis van socialisme en arbeidersbeweging in Nederland (Socialistische Uitgeverij Nijmegen) Godfrey, L.G. (1990), Misspecification tests in econometrics (Cambridge University Press) Gomulka, J. & N. Stern (1990), 'The employment of married women in the United Kingdom 1970-83', Economica 57, 171-199. Goudriaan, F.G.W. (1986), 'Geeft ons Nederlanders toch werk'. Een literatuurstudie naar de bestrijding van de werkloosheid in de jaren dertig (Staatsuitgeverij, Den Haag) Groot, W. & A. van den Berg (1994), 'Why union density has declined', European Journal of Political Economy 10, pp. 749-763. Hagoort, R. (1933), De christelijk-sociale beweging (Edeca, Hoorn) Ham, J.C. van, J. Paauwe & A.R.T. Williams (1985), 'De vakbeweging in Nederland: van leden naar klanten', Economisch Statistische Berichten, pp. 468-473. Hamermesh, D.S & A. Rees (1988), The economics of work and pay (Harper & Row, New York)
197
Harmsen, G. (1971), Historisch overzicht van socialisme en vakbeweging in Neder-
land. Deel 1: van de begintijd tot het uitbreken van de eerste wereldoorlog
(Socialistische Uitgeverij Nijmegen) Harmsen, G. & B. Reinalda (1975), Voor de bevrijding van de arbeid; beknopte geschienis van de Nederlandse vakbeweging (Socialistische Uitgeverij Nijmegen) Harnett, D.L. (1982, 3rd edition), Statistical methods (Addison-Wesley, Reading, Massachusetts) Heldt, B.H. (1896), Algemeen Nederlandsch Werklieden-Verbond 1871-1896 (Coöperatieve Handelsdrukkerij, Leeuwarden) Hendriks, J.A.T.P. (1986), De conjunctuur en de ontwikkeling van de Nederlandse vakbeweging. Een tijdreeksanalyse (M.A. thesis sociology, University of Amsterdam) Hines, A.G. (1964), 'Trade unions and wage inflation in the United Kingdom, 18931961', Review of Economic Studies 31, pp. 221-252. Hirsch, B.T. & J.T. Addison (1986), The economic analysis of unions. New approaches and evidence (Allen & Unwin, Boston) Hirsch, B.T. & M.C. Berger (1984), 'Union membership determination and industry characteristics', Southern Economic Journal 50, pp. 665-679. Hirschman, A.O. (1970), Exit, voice and loyalty: responses to declines in firms, organizations and states (Harvard University Press, Cambridge Mass.) Horssen, P. van & D. Rietveld (1977), 'De Sociaal Democratische Bond', Tijdschrift voor Geschiedenis 1, pp. 5-71. Hudig, D. (1904), De vakbeweging in Nederland 1866-1878 (Scheltema & Holkema, Amsterdam) Hueting, E., F. de Jong Edz. & R. Neij (1982), Naar groter eenheid. De geschiedenis van het Nederlands Verbond van Vakverenigingen 1906-1981 (Van Gennep, Amsterdam) Hundley, G. (1988), 'Education and union membership', British Journal of Industrial Relations 26, pp. 195-201. Jong Edz., F. de (1956), Om de plaats van de arbeid. Een geschiedkundig overzicht
van ontstaan en ontwikkeling van het Nederlands Verbond van Vakverenigingen
(De Arbeiderspers, Amsterdam) Judge, G.G., R.C. Hill, W.E. Griffiths, H. Lütkepohl, T.-C. Lee (1988, 2nd edition), Introduction to the theory and practice of econometrics (John Wiley & Sons, New York) Klandermans, P.G. (1986), 'Participatie in de vakbond: een overzicht van theorie en onderzoek', Tijdschrift voor Arbeidsvraagstukken 2, pp. 14-29. Koole, R.A. (1988), 'Politieke partijen', Compendium voor Politiek en Samenleving 6 Kornhauser. R. (1961), 'Some social determinants and consequences of union membership', Labor History 2, pp. 30-61. Kruithof, B. (1981), 'Trouw aan beginsel. De christelijk-sociale beweging in Nederland van ± 1875 tot 1909', Tijdschrift voor Geschiedenis 7, pp. 347-373. Lammers, C. (1951, 2nd edition), De vakbeweging en haar problemen (De Arbeiderspers, Amsterdam)
198
Lee, L.-F. (1978), 'Unionism and wage rates: a simultaneous equations model with qualitative and limited dependent variables', International Economic Review 19, pp. 415-443. Levenbach, M.G. (1964), Sociale verzekering in Nederlands bestuursrecht III (Samson, Alphen aan den Rijn) Lulofs, J.G. (1968), 'Het categoralisme in de Nederlandse vakbeweging', Sociaal Maandblad Arbeid, pp. 514-540. Maddala, G. (1983), Limited-dependent and qualitative variables in econometrics (Cambridge University Press, Cambridge) Mancke, R.B. (1971), 'American trade union growth, 1900-1960: a comment', Quarterly Journal of Economics 85, pp. 187-193. Mekkelholt, E., W. Praat, & E. Brouwer (1989), 'Werkloosheid in geografische beroepsdeelmarkten: creatie en verantwoording' (Research Memorandum 8928, Department of Economics, University of Amsterdam) Ministry of Social Affairs and Employment (SAE) (1985-1990), Rapportage
Arbeidsmarkt
Mitchell, B.R. (1975), European historical statistics 1750-1970 (MacMillan, London) Moore, W.J. & R.J. Newman (1975), 'On the prospects for American trade union growth: a cross-section analysis', The Review of Economics and Statistics 57, pp. 435-445. Moore, W.J. & D.K. Pearce (1976), 'Union growth: a test of the AshenfelterPencavel model' in: Industrial Relations 15, pp. 244-247. Naylor, R. & M. Cripps (1989), 'Reputation, social custom and trade union membership' (paper presented at the Econometric Society European Meeting, Munich). Neumann, G.R., P.J. Pedersen & Westergard-Nielsen (1989), 'Long-run international trends in aggregate unionization' (working paper, University of Aarhus) Neumann, G.R. & E.R. Rissman (1984), 'Where have all the union members gone?', Journal of Labor Economics 2, pp. 175-193. Olson. M. (1965), The logic of collective action (Harvard University Press, Cambridge Mass.) Ours, J. van (1992), 'Union growth in the Netherlands 1961-1989', Applied Economics 24, pp. 1059-1066. Owen Smith, E. (ed.) (1981), Trade unions in the developed economies (Martin's Press, New York) Pedersen, P.J. (1978), 'Union growth and the business cycle: a note on the Bain-Elsheikh model', British Journal of Industrial Relations 16, pp. 373-377. Pedersen, P.J. (1982), 'Union growth in Denmark, 1911-1939', Scandinavian Journal of Economics 84, pp. 583-592. Pencavel, J.H. (1971), 'The demand for union services: an exercise', Industrial and Labor Relations Review 24, pp. 180-190. Peper, A. (1979), 'Coöperatie, conflict en coalitie. De Nederlandse arbeidsverhoudingen 1970-1980: veranderingen, constanten en perspectieven', Economisch Statistische Berichten, pp. 1359-1372.
199
Perry, J. (1978), 'De Katholieke arbeidersbeweging en haar geschiedschrijving', Jaar-
boek voor de geschiedenis van socialisme en arbeidersbeweging in Nederland
(Socialistische Uitgeverij Nijmegen) Purdy, D.L. & G. Zis (1973), 'Trade unions and wage inflation in the UK: a reappraisal' in: M. Parkin & A.R. Nobay (eds.), Essays in modern economics (Longman, London) Reynaerts, W.H.J. & A.G. Nagelkerke (1986), Arbeidsverhoudingen, theorie en praktijk (Stenfert Kroese, Leiden-Antwerpen) Rezler, J. (1961), Union growth reconsidered: a critical analysis of recent growth theories (Kossuth Foundation, New York) Richardson, R. (1977), 'Trade union growth' (review article), British Journal of Industrial Relations 15, pp. 279-282. Richardson, R. (1978), 'Trade union growth: a rejoinder', British Journal of Industrial Relations 16, pp. 103-105. Rij, C. van, A. Kersten, W. Saris and J. Visser (1990), 'Trade union membership and age' (Paper, presented at the 12th. World Congress of Sociology, Madrid july 1990) Rij, C. van & B. van den Putte (1992), 'Wanneer en waarom worden werknemers lid van een vakbond?', Tijdschrift voor Arbeidsvraagstukken 8, pp. 265-274. Rij, C. van, B. van den Putte, W. Saris, J. Visser, S. Nandram, G. van der Veen, B. Klandermans (1992), Participatie in vakbonden. Een opiniepeiling onder de beroepsbevolking (UvA/VU, Amsterdam) Rij, C. van & W. Saris (1993), 'Time dependency of trade union membership. An event-history analysis', Quality and Quantity 27, pp. 73-93. Roes, J. (1985), 'Katholieke arbeidersbeweging in historische banen', in: J. Roes (ed.), Katholieke arbeidersbeweging (Ambo, Baarn) Rüter, A.J.C. (1935), De spoorwegstakingen van 1903. Een spiegel der arbeidersbeweging in Nederland (Brill, Leiden) Schippers, J.J. (1986), 'Determinanten van vakbondslidmaatschap', Sociaal Maandblad Arbeid 41, pp. 790-799. Schnabel, C. (1989a), 'Determinants of trade union growth and decline in the Federal Republic of Germany', European Sociological Review 5, pp. 133-146. Schnabel, C. (1989b), Zur ökonomischen Analyse der Gewerkschaften in der Bundesrepublik Deutschland (Peter Lang, Frankfurt/Bern/New York/Paris) Schram, A. & H. van Ophem (1989), 'Sequential or simultaneous logit models: a wald specification test' (Research Memorandum 89.10, Center for Research in Public Economics, Leiden University) Scoville, J.S. (1971), 'Influences on unionization in the U.S. in 1966', Industrial Relations 10, pp. 354-361. Sheflin, N., L. Troy & T. Koeller (1981), 'Structural stability in models of American trade union growth', Quarterly Journal of Economics 96, pp. 77-88. Shister, J. (1953), 'The logic of union growth', Journal of Political Economy 55, pp. 413-433. Sharpe, I.G. (1971), 'The growth of Australian trade unions: 1907-1969', Journal of Industrial Relations 13.
200
Stoker, T.M. (1985), 'Aggregation, structural change, and cross-section estimation', Journal of the American Statistical Association 80, 720-729. Teeffelen, A. van & P.G. Klandermans (1989), 'Tussen rationele afweging en interactie. Ledenbinding en het opzeggen van vakbondslidmaatschap', Tijdschrift voor Arbeidsvraagstukken 5, pp. 4-15. Teulings, A. (1984), 'Ontwikkelingen binnen werknemersorganisaties, een organisatie-sociologische benadering' Sociaal Maandschrift Arbeid 39, pp. 720-736. Teulings, C. (1988), 'De grenzen van beroepsdeelmarkten', Tijdschrift voor Arbeidsvraagstukken 4, 46-61. Top, W. (1965), 'Het dilemma van de vakbeweging', Economisch Statistische Berichten, pp. 452-454, 484-486. Top, W. (1970), Werknemer en vakbeweging. Over de verhouding tussen het NVV en het categoralisme in Nederland (NVV report) Thomas, R.L. & P.J.M. Stoney (1970), 'A note on the dynamic properties of the Hines inflation model', Review of Economic Studies 37, pp. 286-294. Troy, L. (1986), 'The rise and fall of American trade unions: the labor movement from FDR to RR' in: S.M. Lipset (ed.), Unions in transition. Entering the second century (Institute for Contemporary Studies, San Francisco) Vall, M. van de (1963), De vakbeweging in de welvaartsstaat (Boom & Zn, Meppel) Veen, G. van der & B. Klandermans (1989), '"Exit" behavior in social movement organizations', in B. Klandermans (ed.), Organizing for change. Social movement organizations in Europe and the United States, pp. 179-198 (Jai Press, Greenwich, Conn.) Verberne, L.G.J. (1940), De Nederlandsche arbeidersbeweging in de 19e eeuw (Van Kampen, Amsterdam) Visser, J. (1985), 'Vakbondsgroei en vakbondsmacht in West-Europa', Tijdschrift voor Arbeidsvraagstukken 1, pp. 18-38. Visser, J. (1986), 'De afkalving van een verouderd verzorgingsarrangement. De Nederlandse vakbeweging verliest haar leden', Namens 1, pp. 133-142. Visser, J. (1987), In search of inclusive unionism. A comparative analysis (PhD thesis, University of Amsterdam) Visser, J. (1988), 'Trade unionism in Western Europe: present situation and prospects', Labour and Society 13, pp. 125-182. Visser, J. (1989a), European trade unions in figures (Kluwer, Deventer/Boston) Visser, J. (1989b), 'New working time arrangements in the Netherlands', in: A. Gladstone et al. (eds.), Current issues in labour relations: an international perspective, pp. 229-250 (Walter de Gruyter, Berlin/New York) Visser, J. (1990), In search of inclusive unionism (Kluwer, Deventer/Boston) (amended version of his 1987 publication) Voos, P.B. (1984), 'Trends in union organizing expenditures, 1953-1977', Industrial and Labor Relations Review 38, pp. 52-63. Vuurde, R.E.M. van (1986), 'Het arbeidsethos in de vroege Nederlandse arbeidersbeweging, 1868-1885', Economisch- en sociaal historisch jaarboek 49, pp. 100146 (NEHA, Amsterdam)
201
Waarden, F. van (1977), 'Veranderingen in de strategie van de Nederlandse vakbeweging' in: T. Akkermans (ed.), Facetten van vakbondsbeleid, pp. 71-97 (Samsom, Alphen aan den Rijn) Wal, F. van der (1915), De oudste vakbond van ons land 1866-1916. Ontstaan en
vijftigjarige werkzaamheid van den Algemeenen Nederlandsche Typografenbond
(Thieme, Nijmegen) Wallerstein, M. (1989), 'Union organization in advanced industrial democracies', American Political Science Review 83, pp. 481-501. Western, B. (1992), 'Institutional mechanisms for class formation: unionization in 16 OECD-countries' (working paper, University of California, Los Angeles) Weyforth, W.O. (1917), The organizability of labor (Johns Hopkins University Press, Baltimore) Windmuller, J.P., C. de Galan & A.F. van Zweeden (1987, 6th edition), Arbeidsverhoudingen in Nederland (Het Spectrum, Utrecht) Wolman, L. (1924), The growth of American trade unions, 1880-1923 (National Bureau of Economic Research, New York) Wolman, L. (1936), Ebb and flow in trade unionism (National Bureau of Economic Research, New York) Zanden, J.L. van & R.T. Griffiths (1989), Economische geschiedenis van Nederland in de 20e eeuw (Het Spectrum, Utrecht)
202
SAMENVATTING (SUMMARY IN DUTCH) De Nederlandse vakbeweging heeft sinds haar ontstaan perioden van groei en terugval doorgemaakt. Verklaringen waarom mensen lid worden van een vakbond of weer opzeggen, worden aangedragen vanuit meerdere vakgebieden, die elk een eigen benaderingswijze kennen. Enigszins chargerend kan worden gezegd dat economen vooral kijken naar (materiële) kosten en baten, terwijl sociologen meer oog hebben voor de invloed van iemands sociale omgeving. Historici dragen aan het onderwerp b_ door hun aandacht voor de rol van maatschappelijke en institutionele ontwikkelingen en gebeurtenissen. Deze verschillende invalshoeken zijn in dit proefschrift samengebracht in een analytisch kader, dat van de vraag naar en het aanbod van vakbondsdienstverlening. Het uitgangspunt van dit theoretisch raamwerk wordt gevormd door een studie van de economen Hirsch & Addison (1986). Vanuit de veronderstelling dat mensen hun beslissingen baseren op kosten-batenafwegingen, formuleren z_ een vraag- en een aanbodfunctie. De vraag naar vakbondsdienstverlening wordt in hun model bepaald door de contributiehoogte, het inkomen, het relatieve loonverschil tussen georganiseerden en niet-georganiseerden, de kosten van alternatieve dienstverlening, de immateriële voordelen van vakbondslidmaatschap op de werkplek, en de attitude van werknemers jegens de vakbeweging. Het aanbod van vakbondsdienstverlening wordt bepaald door de ontvangsten uit contributies, de wervingskosten, de kosten van de dienstverlening en de doelstellingen van de vakbond. Het dient vermeld te worden dat het hier een theoretisch model betreft, dat niet empirisch wordt getoetst. Enerzijds kan een aantal opgevoerde determinanten niet (goed) worden waargenomen, zodat ze benaderd moeten worden met behulp van andere variabelen. Anderzijds worden veel van de hier genoemde determinanten op hun beurt weer beïnvloed door allerlei omgevingsfactoren, die niet rechtstreeks in het model zijn terug te vinden. Het bovenstaande theoretische kader is bovendien in dit onderzoek aangepast en uitgebreid om rekening te kunnen houden met specifieke Nederlandse omstandigheden en om de verklaringen van andere disciplines te kunnen inpassen. Aldus zijn de volgende vraagstellingen geformuleerd. Met betrekking tot de vraagzijde: wat bepaalt de kans om al dan niet vakbondslid te worden of te blijven, en wat voor diensten worden er gevraagd? Wat betreft de aanbodzijde: welke factoren bepalen het bestaan van vakbonden, en wat voor diensten worden aangeboden? Aan beide vragen wordt nog toegevoegd: hoe en waarom veranderen deze zaken in de loop der tijd? Om fluctuaties in het vakbondslidmaatschap te kunnen verklaren wordt er gebruik gemaakt van diverse onderzoekmethoden, die afhankelijk zijn van de gekozen wetenschappelijke invalshoek en van de beschikbare data. Na het inleidende hoofdstuk 1 wordt in hoofdstuk 2 een overzicht gegeven van de verklaringen voor vakbondsgroei en verval die in de economische literatuur naar voren worden gebracht. Hoofdstuk 3 vervolgt met een uiteenzetting van de geschiedenis van de Nederlandse vakbeweging, waarin de nadruk ligt op de ontwikkeling van de ledentallen. Aan de hand van de algemene theoretische
203
verbanden zoals beschreven in hoofdstuk 2, en de specifieke hypothesen omtrent de Nederlandse vakbonden, voortvloeiend uit hoofdstuk 3, zijn data verzameld en modellen opgesteld die achtereenvolgens in hoofdstuk 4 (tijdreeksanalyse) en hoofdstuk 5 (cross-sectie-analyse) empirisch worden getoetst. In hoofdstuk 6 tenslotte komt de benaderingswijze van de sociale wetenschappen aan bod. In het onderstaande worden de belangrijkste bevindingen van elk hoofdstuk samengevat en geëvalueerd. In hoofdstuk 2 wordt onderscheid gemaakt tussen theoretische en empirische studies die gebaseerd zijn op tijdreeksanalyse enerzijds, en cross-sectionele analyse anderzijds. Beide onderzoekmethoden hebben hun voor- en nadelen. Het voordeel van tijdreeksanalyse is de mogelijkheid om (jaarlijkse) fluctuaties in het (geaggregeerde) vakbondslidmaatschap te verklaren over een lang tijdsverloop, maar de keerzijde van deze methode is gelegen in het feit dat slechts een klein aantal variabelen daarvoor geschikt is. Vele tijdreeksen bestaan uit te weinig waarnemingen, of variabelen hebben definitiewijzigingen ondergaan, wat de consistentie van de reeksen aantast. Daarnaast kunnen verscheidene structurele determinanten van vakbondsgroei, zoals geleidelijke veranderingen in maatschappelijke opvattingen, niet (goed) worden gemeten. Cross-sectionele analyses daarentegen zijn gebaseerd op vaak uitgebreide enquêtes en kunnen daardoor gebruik maken van een groot aantal geschikte variabelen ter verklaring van de vraag waarom individuen al dan niet opteren voor vakbondslidmaatschap. Het nadeel van deze methode is dat deze enquêtes meestal eenmalig worden uitgevoerd, zodat de onderzoekresultaten slechts een momentopname weerspiegelen. Paneldata zouden dit probleem kunnen oplossen, maar de bestaande paneldata bevatten tot op heden geen vraag aangaande vakbondslidmaatschap. Zowel de tijdreeks- als de cross-sectie-analyse worden in dit proefschrift uitgevoerd, omdat ze elkaar goed aanvullen. Hoofdstuk 2 bespreekt de relevante theorieën en empirische resultaten die in de economische literatuur over beide onderzoekmethoden naar voren komen. Vervolgens wordt in hoofdstuk 3 uiteengezet hoe de Nederlandse vakbeweging zich heeft ontwikkeld sedert haar ontstaan in de jaren zestig van de negentiende eeuw. Het belang van deze historische analyse is gelegen in het feit dat hierdoor inzicht wordt verkregen in de typische kenmerken van de Nederlandse arbeidsverhoudingen en instituties, wat van pas komt in de formulering van de economische modellen in de volgende hoofdstukken. Bovendien brengt deze geschiedkundige studie een aantal ontwikkelingen en bijzondere gebeurtenissen onder de aandacht, die niet kwantificeerbaar zijn, maar wèl een belangrijke uitwerking hebben gehad op het Nederlandse vakbondswezen. Vooral over de beginperiode van de vakbeweging zijn weinig statistische gegevens voorhanden, maar uit de historische literatuur wordt duidelijk dat de omvang van de Nederlandse vakbeweging tot het begin van de twintigste eeuw klein bleef. Enerzijds was dit het gevolg van economische omstandigheden, zoals de relatief late industrialisatie en de grote spreiding van economische activiteiten. Anderzijds werden vakbonden in hun groei belemmerd door de heersende mentaliteit, die erg onder de
204
invloed stond van religieuze opvattingen: het geloof preekte gehoorzaamheid (aan de werkgever en aan de kerk) en veroordeelde arbeidsconflicten. De oprichting van de socialistische vakcentrale NVV in 1906 zette de protestanten en rooms-katholieken ertoe aan in 1909 hun eigen vakorganisaties in het leven te roepen, wat het begin vormde van een langdurige driedeling binnen de vakbeweging, in navolging van de verzuiling van de hele Nederlandse samenleving. De Nederlandse vakbonden streefden drie doeleinden na: emancipatie van de arbeiders, collectieve belangenbehartiging van de beroepsbevolking, en individuele belangenbehartiging van hun leden. Op elk van deze terreinen werd voor de Tweede Wereldoorlog veel bereikt. Met steun van politieke partijen van dezelfde kleur, wisten de drie belangrijkste vakcentrales de sociale en rechtspositie van de arbeidersklasse aanzienlijk te verbeteren. Aangezien de overheid toen nog nauwelijks uitkeringen verstrekte, boden de vakbonden allerlei sociale verzekeringen aan, die ook fungeerden als middel om de leden aan zich te binden. De bestudering van de vooroorlogse periode leert dat het aanbod van vakbondsdienstverlening aansloot op de vraag, maar dat dit nog geen hoge organisatiegraad tot gevolg had. Mede vanwege de verzuiling bleef de vakbeweging verdeeld, waardoor individuele bonden vaak zwak stonden tegenover de georganiseerde werkgevers. Vooral tijdens recessies waren de arbeiders erg kwetsbaar, wat zich meestal vertaalde in een dalend vakbondslidmaatschap. Pas nadat vakbonden geleidelijk op grotere schaal werden erkend door werkgevers, overheid en de kerken, waren de voorwaarden geschapen die een een grotere vakbondsorganisatie mogelijk maakten. Na de Tweede Wereldoorlog werden de drie toonaangevende vakcentrales erkend als overheidsadviseurs, wat tot uitdrukking kwam in vakbondsdeelname in diverse adviesraden, gezamenlijk met de werkgeversorganisaties. De overheid en de sociale partners waren het erover eens dat zij moesten samenwerken om de zwaar gehavende economie te kunnen herstellen, en om een herhaling van de ontberingen van tijdens de grote depressie te voorkomen. Alleen de communistisch geöriënteerde EVC wilde hieraan niet meewerken en organiseerde als enige in de eerste na-oorlogse jaren een groot aantal stakingen; hiermee werden in eerste instantie veel teleurgestelde arbeiders geworven, maar vanaf de jaren vijftig verloor de EVC snel aan invloed. Het coöperatieve gedrag van de drie erkende vakbondszuilen droeg inderdaad bij aan een spoedig herstel van de Nederlandse economie, waardoor de ontwikkeling en uitbreiding van de welvaartsstaat mogelijk werd. In ruil voor hun toegenomen invloed op het centrale niveau, zagen de bonden voorlopig af van enige aanspraak op medezeggenschap van werknemers op ondernemingsniveau. Gedurende het hele na-oorlogse tijdvak bleven de vakbonden, met uitzondering van de categorale bonden, nagenoeg onzichtbaar op de werkplek. De vakbeweging is altijd grote nadruk bl_ven leggen op de collectieve belangenbehartiging van de hele beroepsbevolking, en boekte daarin ook veel resultaten. Mede dankzij haar inspanningen werden wetten ingevoerd of verbeterd ten aanzien van sociale zekerheid, werkomstandigheden en arbeidsvoorwaarden. De wetgeving inzake het algemeen verbindend verklaren van collectieve arbeidsovereenkomsten maakte dat alle werknemers in een bedrijfstak, of ze nu wel of niet vakbondslid waren, meeprofiteerden van de door vakbonden bereikte onderhandelings-
205
resultaten. Dit werkte het zogenaamde 'liftersgedrag' in de hand. Door de sterk verbeterde sociale zekerheid en arbeidsvoorwaarden voor de hele beroepsbevolking werden de individuele voordelen van het vakbondslidmaatschap steeds kleiner. Dit verklaart vermoedelijk waarom de organisatiegraad sinds de jaren vijftig al niet meer toenam. Vanaf 1979 is de organisatiegraad sterk afgenomen, wat slechts deels kan worden verklaard uit conjuncturele ontwikkelingen; immers, met het herstel van de economie in de tweede helft van de jaren tachtig ging het ledental wel weer stijgen, maar de organisatiegraad is laag gebleven. Dit wordt veroorzaakt door diverse structurele ontwikkelingen, die een verdere groei van de vakbeweging in de weg staan. Een belangrijke oorzaak is te vinden in de veranderende samenstelling van de beroepsbevolking, welke steeds meer moeilijk te organiseren werknemers bevat zoals vrouwen, part-timers en mensen in de dienstensector. Een andere nadelige ontwikkeling is de doorzettende decentralisatie van het arbeidsvoorwaardenoverleg in de richting van het ondernemingsniveau, waar vakbonden (nog) niet goed op ingesteld z_n. Tenslotte draagt de blijvende verdeeldheid binnen de vakbeweging ook niet b_ aan haar (wervings)kracht. Op basis van de verkregen inzichten uit de hoofdstukken 2 en 3 is in hoofdstuk 4 een tijdreeksanalyse uitgevoerd, met als doel de fluctuaties in het geaggregeerde Nederlandse vakbondslidmaatschap te verklaren. Om te beginnen worden twee invloedrijke buitenlandse modellen toegepast op de Nederlandse data. De aanwezigheid van een structurele breuk rond de Tweede Wereldoorlog verhindert een replicatie voor één ononderbroken periode, maar zelfs als twee subperioden worden geschat, blijkt dat noch het Amerikaanse model van Ashenfelter & Pencavel (1969), noch het Britse model van Bain & Elsheikh (1976) de Nederlandse ledentalontwikkelingen kunnen verklaren. Vervolgens worden een nieuw voor- en na-oorlogs model geconstrueerd, die een grotere verklaringskracht hebben. Vóór de Tweede Wereldoorlog werd vakbondsgroei vooral gestimuleerd door prijs- en loonstijgingen, ook wel aangeduid als respectievelijk een 'threat'- en een 'credit'-effect. De ledentoename werd gehinderd door het aantal stakingen, vermoedelijk omdat vele werknemers negatief tegenover stakingen stonden uit angst om ontslagen te worden of vanwege hun religieuze opvattingen. Tenslotte toont de analyse aan dat werkloosheid in eerste instantie het lidmaatschap bevorderde, maar uiteindelijk toch een negatieve uitwerking had. Dit bevestigt het beeld uit de voorgaande hoofdstukken, dat een recessie mensen vaak eerst aanzet om lid te worden of te blijven in de hoop op die manier beschermd te worden en omdat men in die tijd een (hogere) werkloosheidsuitkering kreeg via de vakbond. Wanneer de werkloosheid aanhield, verloren de bonden hun macht en de werknemers hun vertrouwen, waardoor men besloot om geen lid te worden of op te zeggen. Na de Tweede Wereldoorlog blijkt dat loon- en prijsveranderingen niet langer een significante invloed uitoefenden op de omvang van de vakbeweging. De werkloosheid bleef een belangrijke determinant, maar hiervan ging nu alleen nog maar een negatief effect uit. Verder werd de ledengroei belemmerd door de hoogte van de organisatiegraad, wat wordt aangeduid met de term 'verzadigingseffect'; dit
206
wordt toegeschreven aan het feit dat vakbonden relatief weinig moeite deden om nieuwe leden te werven aangezien ze, ongeacht hun ledental, officieel werden erkend als onderhandelings- en overlegpartner van de werkgevers en de overheid. Tenslotte ging er een ongunstig, zogenaamd 'substitutie'-effect uit van de sociale zekerheidsuitgaven door de overheid, die op dit terrein de vooroorlogse rol van de vakbonden geheel overnam. Een positieve uitwerking op het lidmaatschap kwam van het zetelaandeel van linkse partijen in de Tweede Kamer, alsmede van stakingen. Als echter de EVC-leden van de analyse worden uitgesloten, vermindert het effect van de stakingsvariabele aanzienlijk, wat beter overeenstemt met het beeld van Nederland dat gekenmerkt wordt door een grote mate van arbeidsrust; deze uitkomst bevestigt dat de kortstondige opleving in stakingsactiviteiten in de eerste na-oorlogse jaren vooral een aantrekkingskracht uitoefende op werknemers die voor de radicale EVC kozen. Een evaluatie van de schattingsresultaten van hoofdstuk 4 leert dat slechts een relatief klein aantal variabelen een rol speelt in de tijdreeksanalyse; lang niet alle theorieën die in hoofdstuk 2 en 3 zijn genoemd, kunnen met behulp van deze methode worden getest. De nadruk ligt op cyclische determinanten van vakbondsgroei en verval, aangevuld met institutionele verklaringen. Veel meer theorieën worden belicht in hoofdstuk 5, met behulp van twee crosssectie-analyses op respectievelijk het arbeidsaanbodbestand van de Organisatie voor Strategisch Arbeidsmarktonderzoek (OSA) uit 1988 en het Aanvullend Voorzieningengebruikonderzoek (AVO) uit 1979 en 1987. Met name de OSA-enquête stelt ons in staat om vele karakteristieken van werknemers te analyseren die de kans beïnvloeden dat iemand wel of geen lid is van een vakbond. Bovendien kan via deze dataset onderscheid worden gemaakt tussen de factoren die de algemene keuze voor het vakbondslidmaatschap beïnvloeden en de factoren die de keuze voor een specifieke vakcentrale beïnvloeden. Uit de analyse komt naar voren dat deze twee beslissingen niet gelijktijdig worden genomen; de meeste werknemers bl_ken eerst te besluiten om al dan niet vakbondslid te worden, om daarná pas te besluiten b_ welke vakcentrale men zich zal aansluiten. Daarom wordt in de schattingen een zogenaamd sequentieel logitmodel gebruikt. De resultaten van dit model bevestigen in de meeste gevallen de theorieën en empirische bevindingen zoals die in hoofdstuk 2 zijn besproken, met een paar uitzonderingen die meestal zijn toe te schrijven aan bijzondere Nederlandse omstandigheden. Hieronder volgen de belangwekkendste uitkomsten, onderverdeeld in vier categorieën. Met betrekking tot de persoonlijke karakteristieken wordt de kans dat iemand vakbondslid is groter indien men: man is, ouder is, een langer dienstverband heeft, en meer opleiding genoten heeft. Dit laatste resultaat wijkt af van internationale bevindingen en is typisch voor Nederland, waar vele vakbonden actief zijn die zich toeleggen op de belangenbehartiging van het hogere personeel. Wat betreft baankarakteristieken worden de lidmaatschapskansen negatief beïnvloed door het uitoefenen van een commerciële functie, een part-time of een tijdelijke functie. Een positief effect gaat uit van het hebben van een functie in het onderwijs, in
207
een grote onderneming, en als ambtenaar, en van het doen van onplezierig werk en van de verwachting werkloos te worden. Uit de analyse blijkt verder dat de kans om lid te zijn eerst stijgt met het nettoloon, maar na een bepaald niveau weer daalt. Dit wijst op een 'credit-effect' van loonstijgingen tot een zeker omslagpunt, waarna de voordelen van het lidmaatschap afnemen: hoger betaalde werknemers hebben meestal minder behoefte aan vakbondsbescherming. Ondanks de significantie is het daadwerkelijke effect slechts klein, wat overeenstemt met de bevinding in hoofdstuk 4 dat loonstijgingen in de naoorlogse periode weinig invloed uitoefenen op de ledentalontwikkelingen. In de laatste categorie, regionale karakteristieken, zijn de urbanisatiegraad en 'wonend in de Randstad' in de analyse opgenomen om na te gaan of een hoge concentratie van de (beroeps)bevolking inderdaad vakbondsgroei stimuleert. In weerwil van de theoretische verwachtingen blijken beide variabelen echter een significant negatief effect te hebben op de lidmaatschapskansen. Tenslotte komt uit dit deelonderzoek naar voren dat de keuze voor een specifieke vakcentrale slechts door een klein aantal factoren wordt bepaald. De belangrijkste determinanten blijken te worden gevormd door de kerkelijke gezindte, wat zeer opmerkelijk is gezien de voortschrijdende secularisatie in Nederland. Het tweede deelonderzoek betreft het AVO. Voor zover dezelfde variabelen als bij de OSA-data konden worden gebruikt, zijn de schattingsresultaten vergelijkbaar. Een relevante toevoeging is te vinden in de dummy-variabelen die aangeven in welke bedrijfstak het individu werkzaam is. De uitkomsten bevestigen de theoretische en empirische bevindingen uit de economische literatuur, namelijk dat werknemers in de landbouw en in de dienstensector significant lager zijn georganiseerd dan werknemers in de industrie, terwijl werknemers in de transport- en publieke sector juist hoger zijn georganiseerd. Kort gezegd is dit te verklaren uit allerlei sectorale verschillen wat betreft onder meer de concentratiegraad, werkomstandigheden, salariëring en sociale positie van werknemers, en de aan- of afwezigheid van een vakbondstraditie. Dankzij de beschikbaarheid van twee steekproeven in de AVO-data, is het mogelijk om de veranderingen tussen 1979 en 1987 nader te analyseren. Deze twee jaren markeren precies de periode dat de Nederlandse vakbeweging een sterke terugval doormaakte. De schattingen brengen aan het licht dat de daling in het ledental toe te schrijven is aan het groeiende aandeel van vrouwen en part-timers in de beroepsbevolking, de afnemende belangstelling onder jongeren om lid te worden en de verschuivingen in werkgelegenheid van de industriële naar de dienstensector. Via een decompositiemethode is vastgesteld dat de afname van de organisatiegraad met name het gevolg is geweest van wijzigingen in de samenstelling van de beroepsbevolking, en in veel mindere mate is toe te schrijven aan gedragsveranderingen. Op basis van de uitkomsten van hoofdstuk 5 wordt geconcludeerd dat dankzij de grote hoeveelheid bruikbare variabelen in de twee bestanden, uitgebreid kan worden ingegaan op de vraagstelling aangaande de determinanten van de behoefte aan vakbonden. Tegelijkertijd moet worden geconstateerd dat de aanbodzijde in de
208
hoofdstukken 4 en 5 nauwelijks ter sprake komt, omdat daarvoor de relevante informatie in de beschikbare data ontbreekt. De bijdrage van de sociale wetenschappen aan de verklaring van vakbondsgroei en terugval in hoofdstuk 6 vult, in ieder geval gedeeltelijk, deze lacune op. In het eerste deel van dit hoofdstuk worden de meest belangwekkende theorieën en empirische resultaten van Nederlandse sociale wetenschappers besproken, waaruit blijkt dat zij deels andere determinanten van vakbondslidmaatschap benadrukken; bovendien hebben zij andere datasets ter beschikking, die meer informatie bevatten over de rol van vakbonden zèlf. Binnen de sociale psychologie worden drie theoretische benaderingen van vakbondsparticipatie onderscheiden. De eerste invalshoek beschouwt de beslissing om al dan niet lid te worden als een resultante van frustratie- en agressiegevoelens, maar deze zienswijze wordt ondergeschikt verondersteld aan de andere twee benaderingen, te weten de rationele-keuzetheorie en de interactionistiche theorie. De rationele-keuzebenadering lijkt sterk op de kosten-baten benadering van economen, hoewel sociale wetenschappers in het algemeen meer aandacht schenken aan collectieve, ideologische en sociale motieven. De keus voor vakbondsdeelname wordt bepaald door alle positieve en negatieve consequenties van het lidmaatschap tegen elkaar af te wegen. De keus voor een vakbond hoeft niet (geheel) voort te vloeien uit een weloverwogen rationele overweging, en kan ook worden beïnvloed door de sociale omgeving. De interactionistische theorie veronderstelt dat mensen (al dan niet) lid worden uit traditie of conventie, of hiertoe worden aangezet door een hoge organisatiegraad in hun bedrijf, of door contacten met de vakbond op de werkvloer. Uit de empirische bevindingen van sociale wetenschappers blijkt dat in Nederland steeds minder mensen uit idealistische motieven voor een bond kiezen. De meerderheid wil lid worden om er persoonlijk beter van te worden en zegt gemakkelijk weer op zodra de kosten de baten overtreffen. Dit gedrag komt tot uitdrukking in een hoog lidmaatschapsverloop. Blijkbaar hebben de vakbonden de grootste moeite om hun leden aan zich te binden, wat niet zo verwonderlijk is gezien de grote discrepantie tussen de groeiende behoefte van leden aan individuele dienstverlening en de nadruk die (een groot deel van) de Nederlandse vakbeweging nog steeds legt op haar externe, collectieve functies. Om hun vertrek te verhinderen, moeten vakbonden ervoor zorgen dat leden zich bij de zaak betrokken gaan voelen. Statistische analyses wijzen uit dat de kans op lidmaatschap significant wordt beïnvloed door de organisatiegraad van iemands bedrijf, het aantal contacten met andere leden, en met de vakbond zelf. Deze interactieprocessen leiden tot een zekere mate van betrokkenheid bij de bond (ledenbinding), wat een belangrijke determinant vormt voor de beslissing wel of niet lid te blijven. Het sleutelbegrip ledenbinding ('commitment') legt tevens de link tussen de interactionistische en de rationele-keuzebenadering. Zowel sociale druk als rationele keuze kan mensen ertoe bewegen om zich bij een vakbond aan te sluiten. Wanneer (rationele-keuze)verwachtingen omtrent het nut van het lidmaatschap worden bevestigd, zal dit tot meer ledenbinding leiden, wat vervolgens mensen aanzet tot
209
actievere participatie; dit stimuleert weer interactieprocessen en die versterken op hun beurt de ledenbinding. Aldus neemt de kans dat men opzegt af. In het tweede deel van hoofdstuk 6 wordt een poging ondernomen om enkele van bovenstaande sociologische inzichten te incorporeren in een eigen statistisch model. Door een gebrek aan beschikbare data, is echter alleen een aantal variabelen opgenomen die een weerspiegeling vormen van de rationele-keuzetheorie, en niet van de interactionistische benadering. De meerwaarde van deze nieuwe modellering is gelegen in het feit dat hier het effect van vakbondsbeleid op het lidmaatschap kan worden geanalyseerd, waardoor de aanbodzijde van de hoofdvraagstelling wat beter over het voetlicht komt. Onderzocht is tot op welke hoogte het aandeel van een vakcentrale in het totale vakbondslidmaatschap kan worden verklaard door de mate waarin zo'n centrale meer of minder dan gemiddeld kost voor leden, en de mate waarin de centrale meer of minder dan gemiddeld sociale uitkeringen verstrekt aan leden. Hierbij zijn vier 'zuilen' onderscheiden, te weten NVV, CNV, NKV en de categorale bonden. In eerste instantie worden vier vergelijkingen ter verklaring van het ledenaandeel, zowel voor als na de Tweede Wereldoorlog, afzonderlijk geschat. Dit leidt echter niet tot een bevredigend resultaat, omdat er veel insignificante coëfficiënten worden gevonden en enkele variabelen het verkeerde teken aannemen. In een tweede procedure zijn de vier vergelijkingen in een simultaan stelsel opgenomen, wat efficiënte schattingen oplevert; bijna alle variabelen verkrijgen nu significante coëfficiënten. De uitkomsten van deze modellen bevestigen het theoretische inzicht dat vakbondsleden tenminste gedeeltelijk hun keuze voor een bepaalde bond/centrale baseren op kosten-batenafwegingen. Als een vakcentrale een hoger bedrag aan sociale uitkeringen aan haar leden uitbetaalt dan het gemiddelde, zal haar aandeel in het totale lidmaatschap hoger zijn en wanneer bij een vakcentrale een hogere contributie dan gemiddeld wordt geheven, zal haar aandeel juist lager zijn. Al met al luidt de eindconclusie dat met de resultaten in dit proefschrift het raamwerk van vraag en aanbod van vakbonden redelijk is ingevuld, alhoewel de breedheid van het onderwerp en de beperkte beschikking over gegevensbestanden een volledige beantwoording van de vraagstelling niet mogelijk maken. In vervolgonderzoek zou meer aandacht kunnen worden besteed aan de grote verschillen in organisatiegraad tussen de diverse bedrijfstakken en met het beschikbaar komen van enkele onlangs verschenen statistieken en datasets kan nader worden ingegaan op de ontwikkelingen in de jaren negentig.
210
