Soon after the publication of the Jobs Study, Richard Freeman nicely summarized both the ..... measure for sickpay generosity is also drawn from Scruggsâ data.
Welfare State Regimes and Women‟s Employment
Moira Nelson and John D. Stephens University of North Carolina, Chapel Hill
Paper prepared for delivery at the workshop on “The Political Economy of the Service Transition,“ Institute for International Integration Studies, Trinity College, Dublin, May 1617, 2008.
1
Introduction High levels of employment have been recognized as the key to the viability of generous welfare states (Huber & Stephens 2001; Scharpf 2001). When levels of employment are high and correspondingly levels of non-employment (unemployment, early retirement, housewives out of the labor force, etc.) are low, many people contribute to the welfare state and fewer draw on it, making any given level of benefit generosity more affordable. Under these conditions, even the very generous Nordic welfare states are viable. By contrast, where employment levels are low as in the continental welfare states, generous benefits generate large budget deficits and thus pressure to retrench the welfare state. A decade ago, the conventional wisdom on employment in OECD countries was shaped by neo-liberal economic thinking and then reinforced by the findings of the original OECD Jobs Study (1994). Soon after the publication of the Jobs Study, Richard Freeman nicely summarized both the conventional wisdom and The OECD Jobs Study (1994) on the role of “labor market rigidities” in accounting for the high and persistent unemployment in Europe: Throughout the 1980s, the superior performance of the US in job creation compared to OECD-Europe suggested that labour-market flexibility, US-style, was the panacea for European unemployment problems. Remove labour-market regulations, eliminate jobprotection laws, reduce unemployment benefits, weaken unions, decentralize wagesetting, and presto! European unemployment would vanish. That, at least is the crude version of the conventional wisdom of the decade. In more sophisticated form, this is the message of the OECD Jobs Study. (Freeman 1995, p. 63) This was plausible enough at the time; Sweden and Finland were suffering record levels of unemployment and the turnaround in the Danish economy was in its initial stages. A decade later, the neo-liberal analysis appeared much more problematic: Though unemployment was still high in Finland, all four Nordic economies had very high levels of employment and they all enjoyed budget surpluses and growth above the OECD average. In the academic and political debate on employment, the economic neo-liberalism of the original Jobs Study was challenged by the institutionalist paradigm implicitly underlying the European Employment Strategy (Casey 2004). Based on extensive data collection by the OECD since the publication of the Jobs Study, OECD analysts themselves had retreated from at least some of the recommendations of the 1994 report. Recently, based largely on these same data, Bradley and Stephens (2007) presented a pooled time series analysis which shows that a number of aspects of generous welfare states are positively related with employment rates. We have extended this line of analysis, showing that the welfare state features associated with social democratic government, such as generous short term replacement rates and active labor market policy, are associated with positive employment performance, while welfare state features associated with Christian democratic government; such as high long term replacement rates, high payroll taxes, and high levels of employment protection; are associated with negative employment performance (Huo et al 2008) In this paper, we extend this line of analysis by examining policy determinants of women‟s employment and service sector employment. The variations in total employment analyzed by Bradley and Stephens (2007) are largely a product of differences in women‟s employment. Variations in women‟s employment are, in turn, largely a product of service sector employment (Daly 2000). By further breaking down service employment into its public and private employment components, we are able to clearly distinguish different paths to high employment and high women‟s employment, a social democratic one and a liberal one, something that was obscured by Bradley and Stephens‟ (2007) way of conducting the analysis. Literature and Hypotheses
2
We take Bradley and Stephens (2007) and our previous work (Huo et al 2008) as a point of departure (see Table 1 below). We begin by briefly summarizing the logic behind the hypotheses in these two works and reviewing the results. We then move on to the literature on women‟s employment to derive additional hypotheses that pertain specifically to women. Table 1. Research Summary Consequences Related to Social Related to Christian for Employment Democracy Democracy Unemployment Replacement rate, 1 year
++
++
0
Unemployment Replacement rate, 5 year
––
0
+
Social security taxes
––
0
++
–
0
++
++
++
0
EPL ALMP
0: not significant; + / –: significant / not at the .05 level; ++/– –: significant/ not at .01 level or below
With respect to labor market and social policy, the results from Bradley and Stephens (2007) and Huo, et. al. (2008) demonstrate the employment-supporting effects of Social Democratic policy and the employment-damaging effects of Christian Democratic policy. First, the Bradley and Stephens (2007) study provides evidence that the type of policy matters for employment levels. Policies such as high short-term replacement rates and active labor market policies sustain high employment levels. This supportive role ceases to exist, however, when social policies lead to prolonged labor market exit, as is the case with generous long-term unemployment insurance, or inhibit employers from hiring new workers, as is the case when social security taxes and employment protection are high. Though the neoliberal agenda correctly identifies the employment-damaging effects of Christian-Democratic social policy, their broad denuciation of labor market regulation and generous social policy writ large overlooks the ways in which social policy supports labor market participation. The assumption of efficient allocation of labor ignores the high costs individuals face in finding employment suited to their skill level. Without income support, unemployment can be „scarring,‟ by forcing individuals to accept jobs with lower relative wages and skill requirements. The length of the unemployment spell makes scarring worse. Generous short-term income support can mediate the scarring effects of unemployment by increasing the intensity of individuals‟ job search efforts (Gangl 2004; DiPrete and McManus 2000; DiPrete 2002). This effect, however, weakens with the length of income support, such that generous long-term income support reduces job search intensity. In short, contrary to the neoliberal agenda, policies that support human capital investments, such as short-term unemployment insurance and active labor market policies, lead to higher relative employment rates. The Huo, Nelson, and Stephens (2008) research finds that the structure and employment effects of various labor market and social policies vary strongly with political incumbency. States with strong Social Democratic incumbency demonstrate high levels of short-term unemployment insurance and active labor market policies. States with strong Christian Democratic incumbency show relatively high levels of long-term unemployment insurance, EPL and social security taxation. Liberal welfare states display lower levels of labor market regulation and social policy generosity than both Social Democratic and Christian Democratic welfare states. In this paper, we extend these lines of argumentation to understand determinants of women‟s employment and, relatedly, public and private service sector employment. This focus speaks to questions of gender equality in welfare state policy, on the one hand, and the future viability of the
3
welfare state, on the other. Thus far, we have discussed total employment levels, without differentiating by characteristics of labor market participants. Individuals‟ employment chances, however, vary quite strongly according to such factors as gender, age, race, etc (Browne & Misra 2003; Elliott & Smith 2004). The enduring nature of these biases creates structural inequalities, the labor market consequences of which find recognition within the framework of the “insider-outsider” theory (Lindbeck & Snower 1988; 2002; Rueda 2005a; 2005b). The „outsider‟ status of women in the labor force becomes a salient issue in light of the expansion of the service sector, because many service occupations are dominated by women. Extending service sector employment not only jobs typically filled by women, but also remains the only path towards aggregate employment growth. Therefore, to the extent that funding social policy relies on high active to passive ratios, extending service sector employment sustains the future viability of the welfare state. The growth in women‟s labor force participation represents a fundamental shift in labor markets across advanced industrialized countries. Though the onset of this shift remains debated in the literature (Ginn et al 1996; Hakim 1995), an argument can be made that the labor force participation rates of women began a strong upward trend since the 1970s albeit with marked cross-national differences. The premise of this paper is that various labor market and social policies contribute to the increased levels of women‟s employment across advanced industrial economies. Since potential job growth must happen in services, we are primarily interested in factors that influence women‟s job growth in public and private services. Data on public and private sector employment, however, does not control for gender. Therefore, we will look at the determinants of both total women‟s employment and at private and public sector employment growth, respectively, then drawing links between the two. For the most part, we expect that policies that facilitate overall employment will also help women. This view rests on the assumption that women‟s work is in many ways indistinguishable from men‟s work and the empirical fact that expansion in employment in many countries is driven largely by increases in women‟s employment (Burniaux et al 2003; Daly 2000; Esping-Andersen & Regini 2000, p. 107). Our hypotheses on the determinants women‟s employment are shown in Table 2. With two exceptions, the hypotheses on general welfare state regime characteristics (that is “non-gendered” characteristics) follow the results of Bradley and Stephens (2007). In the case of wage dispersion, Bradley and Stephens found negative but insignificant effects. We expect positive effects because women are employed disproportionately in low wage service occupations, such as retail sales, hotel and restaurant, and personal services. We expect wage compression to price many of these jobs out of the market. Sickpay was not included in the Bradley and Stephens study. Our subsequent work with their data has shown that sickpay generosity has a positive effect on employment. As in the case of short term replacement rates, this can be seen as a result of investment in human capital. At the same time, we recognize that women‟s decision to seek paid employment is often mitigated by other care obligations within the family. Therefore, while some employment-enhancing policies help women and men alike, countries differ in how they negotiate the nexus between unpaid caring responsibilities and labor market participation. In other words, countries differ with respect to the degree to which the state provides for risks for which the family typically provided, i.e. the degree of „de-familialization‟ (Esping-Andersen 1999). Assessing causes of women‟s employment levels across advanced industrialized economies provokes questions of why more women seek paid employment today than in the past and what factors support the creation of jobs that women hold. On the one hand, the expansion of women‟s jobs is straightforward – factors that increase jobs increase women‟s jobs – but on the other women‟s role within the family, whether biologically or socially defined, generates trade-offs between unpaid domestic and care responsibilities and labor market participation.
4
The analysis here focuses on employment levels and, in doing so, cannot control for or otherwise address a number of characteristics that tend to characterize women‟s employment. For example, compared to male work, potential differences exist in terms of working hours, pay, mobility and continuity, all of which may influence women‟s incentives to join and remain in the labor market over time and between countries. These factors may mitigate the effect of policies discussed in this paper; for example, the incentives to enter the labor market created by day care policies may be tempered by the pay that mothers stand to gain by returning to work. However, while employment rates may be a blunt instrument, they allow us to grasp factors that influence women‟s decision to enter the labor market at all. We choose this dependent variable over others, such as the total volume of women‟s work, because the focus lies on understanding factors that induce women to enter employment, regardless of the extent of that commitment. Caring Roles: The Influence of Parental Leave and Day Care Changes in the family‟s capacity to internalize social risks receive perhaps the most attention in explaining increases in women‟s employment. Compared to the pre-industrial family, the average family unit today no longer provides the stability or capacity to absorb various risks (Esping-Andersen 1999, p. 47). Individuals tend to be more mobile, weakening the connection to the nuclear family. Couples cohabitate or marry later, and the longevity of these units has weakened. Moreover, the ability of a “male bread-winner” to be the sole provider for his family has decreased due in part to waning growth rates. All of these changes increase the economic necessity for women to work. At the same time, the role of women in bearing children and the social norm attached to women‟s role as the primary care-giver to both children and ageing parents create additional challenges for the labor market participation for women (see Daly 2002). Increasing women‟s employment levels involves revising expectations about women‟s responsibility to stay at home as well as providing alternative care facilities. In the literature, „family policies‟ have such a goal, because they “contribute to the health and wellbeing of families with children through direct cash transfers (such as child allowances and maternity benefits), tax expenditures (such as dependent tax credits), in-kind benefits (such as housing) and direct services (such as public education and child care)” (Gornick et al 1997, p. 46). From this range of policies, we look at parental leave and day care, for which there are comparable data across countries. Parental leave policies facilitate employment among parents, and mothers in particular (Gornick et al 1998; Ruhm 1998). Empirical evidence also shows that the provision of day care facilities increases aggregate employment levels (EspingAndersen 1999; Gornick & Meyers 2003; Gornick et al 1998; Korpi 2000; Stier et al 2001); microlevel analyses support the logic of these studies by demonstrating that the cost of day care influences take-up rates (Anderson & Levine 1999; Kimmel 1998). Facilitating return to work may also support women‟s long-term labor market chances, because long periods of absence are seen to cause skill atrophy and signal low commitment to work (see Estevez-Abe 2005; Stier et al 2001). Providing day care facilities, therefore, should increase women‟s employment patterns by supporting their capacity to enter the labor market as well as the value of their human capital in the labor market. Viewed from a broader context, the influence of day care policies on women‟s employment patterns depends in part on the financial benefit of labor market participation. Following the Welfare State Regimes categorization (Esping-Andersen 1990), one can identify quite different patterns in the state‟s role in providing support to new families and the consequences for women‟s employment (Daly 2000; Stier et al 2001). In the Social Democratic regime, the state subsidizes childcare in the public sector, which in turn creates jobs mainly filled by women. Christian Democratic states also provide childcare albeit at a lower level than their Scandinavian neighbors, but household taxation and high non-wage labor costs reduce the incentive to take on paid work and increase the costs of childcare (Daly 2000, p. 491), respectively. Liberal welfare states do not subsidize day care to the extent of the other welfare states, but the cheap price of care services means that private day care centers have flourished (Morgan 2005). In sum, we may expect day care policies to increase public service sector employment relatively more in Social Democratic welfare states and private service sector employment relatively more in Liberal welfare states.
5
Increased Educational Attainment Women‟s chances of finding stable employment have on average increased due to higher levels of educational attainment. On the micro-level, education has been seen to generate returns in the form of higher wages (Harmon et al 2003). While keeping in mind that women continue to earn less than men on average (Sainsbury 1996), the return on education for women is actually larger than for men (Psacharopoulos 1985), which creates incentives for women to gain formal education. On the macro-level, education levels of the population also increase economic growth (Sianesi & Reenen 2003), which may increase aggregate employment. Empirical evidence supports the theoretical link between higher educational attainment and women‟s employment levels (Hicks & Kenworthy 2006; Jaumotte 2003). We therefore expect a positive relation between higher levels of educational attainment and women‟s employment. However, the relationship between women‟s education and women‟s employment is at least partly endogenous: If women‟s employment is high in a country, women have a greater incentive to acquire more education in anticipation of being employed. Public and Private Services Since the growth of women‟s employment depends on expansion of service sector jobs, we pay particular attention to the various factors that influence growth in these areas. The central difficulty in expanding service employment lies in the low productivity of many service sector jobs. Known as Baumol‟s cost disease, competition for employees between the high productivity industrial sector and low productivity service sector inflates wages in the latter. Expanding service sector employment has therefore been recognized to follow one of two paths: either letting wage levels sink or subsidizing service employment by expanding government-funded employment (Esping-Andersen 1993). The former path implies private service sector growth and growing wage differentials, whereas the latter path points to high public sector job growth. Understanding the capacity for growth in public and private services means explaining the factors influencing wage inequality as well as the government‟s willingness to expand public employment. As previous work has shown (Bradley and Stephens 2007), neocorporatism and wage coordination are positively related to total employment levels, but this aggregrate variable may mask sectoral effects of encompassing interest group structures. In encompassing systems, preferences for solidaristic wage policies (Iversen 2005) may lead to compressed wage structures, which have damaging effects on the employment of workers with high demand elasticity ( Bertola et al 2007; Kahn 2000). As a result, to the extent that labor market institutions keep the price of labor in the service sector above productivity levels, expanding private sector employment is difficult. The existence of high social security taxes only exacerbates this problem. In the presence of compressed wage structures, a possible alternative to private service sector employment includes the expansion of government-subsidized public jobs. Where Liberal states are characterized by high levels of wage dispersion which facilitates the expansion private services, Social Democratic states resisted wage dispersion and expanded service sector jobs through the public sector. Only Christian Democratic states appear unable to follow either one of these paths, unwilling to allow for wage dispersion or public sector expansion. Our hypotheses for public and private service sector employment are summarized in the last two columns of Table 2. Our hypothesized determinants of employment in the two sectors are almost the mirror image of each other. We expect similar outcome only for active labor market policy (ALMP) and daycare spending and welfare state characteristics which Bradley and Stephens found to be negatively related to employment, social security and payroll taxes and long term unemployment replacement rates. Otherwise we expect the public sector path to be a social democratic path and the private sector path to be a (neo)liberal path. Measurement
6
We confine our analysis to the period 1972-2001 for 17 advanced industrial democracies because our hypotheses focus on employment performance in the post Bretton Woods, post-industrial period and not in the Golden Age period of rapid growth based the expansion of manufacturing.1 Our main data sources are the OECD and the Huber et al. (2004) Comparative Welfare States Data Set which in turn relies heavily on OECD data (see Table 2).2 The dependent variable in the analysis is women‟s employment as a percentage of the working age population (15-64 years of age). For our purposes the best indicators of welfare state generosity as it relates directly to the labor market are replacement rates and duration of unemployment insurance. The we use the OECD summary indicator of “benefit generosity” at 2/3 the average earning level. We use the 2/3 level rather then the overall summary measure used by Bradley and Stephens (2007) because it would appear to be the most important for womens„ employment, especially in the private service sector. Our independent variables are the one year replacement rate and the replacement rate in the fourth and fifth year averaged across three different family situations (single, married with dependent spouse, and married with spouse in work). The OECD data are gross replacement rates: Both the income and the transfer are pre tax. Net replacement rate is clearly the preferable measure. Scruggs (2005) has recently released data on net replacement rates and duration of benefits, but there it is not possible to calculate the replacement rate for bouts of long duration from the data, which is essential for our purposes. To check the validity of the gross replacement rates data, we calculated a net replacement rate for a bout of unemployment one year long from the Scruggs data. The one year gross and net replacement rate series are highly correlated (.85) which increases our confidence in the analysis using the OECD measure. Our measures of social security taxes and total taxes are those taxes as a percent of GDP. Our measure of wage coordination is Kenworthy‟s indicator. This measure is preferable to measures of bargaining centralization because it taps institutionalized practices such as pattern setting, tacit coordination, and government intervention which are missed by measures of bargaining centralization. Our measure of union strength is net union membership as a percent of wage and salary workers. Our measure for sickpay generosity is also drawn from Scruggs„ data. He provides measures for replacement rates, coverage, waiting days, and qualifying conditions. We standarized each measure then added the standard scores of replacement rates and coverage and subtracted the standard scores of waiting days and qualifying conditions. As control variables, we include two measures of economic openness or “globalization”. Following Bradley et al. (2001), we use the Quinn/Inclan (1997) measure of capital and current account controls as our measure of capital market openness. As a general measure of capital market openness, we favored the control measures over the flow measures (inward and outward FDI as a percent of GDP) because, as Simmons (1999) and others have argued, it is the possibility of easy exit that changes the behavior of actors not variations in actual flows. In the Quinn/Inclan measure, the maximum score indicates no capital controls. For these same reasons, our preferred measure of trade openness would be a measure of tariff and non-tariff barriers to trade. Unfortunately, no such time series exists, so we use the conventional measure of trade flows, imports plus exports as a percent of GDP. The measure for women‟s education attainment comes from Barro and Lee (2001). We use their measure of average years of schooling for women aged 25 and over. Total years of schooling is preferred over attainment of a given level of schooling (primary, secondary, etc.) in order to recognize all forms of education participation regardless of whether or not it culminated in an official qualification. Educational attainment of women over 25 is included rather than women aged 15 and over in order to capture the 1
Unless otherwise specified, the data for variables in the analysis are annual time series from 1972 to 2001 for all seventeen countries. 2 The countries included in this model are: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Italy, Ireland, Japan, Netherlands, Norway, Sweden, Switzerland, the United Kingdom, and the United States. The classification scheme in Table 2 follows Kitschelt H, Lange P, Marks G, Stephens J. 1999. Convergence and Divergence in Advanced Capitalist Economies. Continuity and Change in Contemporary Capitalism Eds H Kitschelt, P Lange, M Gary, JD Stephens (Cambridge University Press, Cambridge) pp:42760.
7
higher end of the spectrum; measuring educational attainment for women aged 15 and over would underestimate the educational attainment of women who were simply too young to have completed higher forms of education. For the remaining variables, data coverage is more limited and the time series are unbalanced with varying time points per country but without gaps in the time series. The wage dispersion measure is the 50-10 ratio, the ratio of the wage/salary of the median full time employee to the wage/salary of the full time employee at the 10th percentile. We reasoned that wages at the bottom of the distribution would be most relevant for the development of a large private service sector.3 For this variable, we have 314 observations from 1972 to 1999. Our data on employment protection laws (EPL) are the OECD‟s (2004) recently released annual time series. The summary index summarizes a number of sub indices measuring the difficulty of layoff (notice, severance pay, etc.) and regulations restricting the use of temporary work. The data are available for 16 of the 17 countries from 1985 to 1991 and all 17 from 1992 to 1999. Following Nickell (1997), we operationalize active labor market policy (ALMP) effort as active labor market spending as a percentage of GDP divided by the unemployed portion of the population. This OECD series is available for 10 countries from 1980 to 1984 and all of them from 1985 to 1999. Mean values for the dependent variable and selected independent variables are displayed in Table 2. [NB: We need to update the time series in the data set for epl, almp, and wage dispersion!] The data measuring parental leave and benefits come from Gauthier and Bortnik‟s (2001) dataset. Our summary variable to measure overall parental leave generosity is the average replacement rate for 52 weeks, which is calculated from maternity/parental leave duration (in weeks) and maternity/parental leave benefits (expressed as a percentage of women‟s wages in manufacturing). Where the leave period was more than one year, as in Finland in 1991-1992 and in Sweden 1990-1999 (54 and 64 weeks, respectively), the replacement rate was recalibrated to fit a scale of 52 weeks. These data are available from 1972 to 2000. Daycare generosity is measured by public daycare spending as a percentage of GDP. These data are drawn from the OECD and Jaumotte (2003) series on daycare spending. Based on Gornick and Meyers, Chapter 7, we use Jaumotte for Belgium and and Norway because OECD appears to be too low, as well as for Canada, for which OECD had no data. For the UK and France, we use OECD and fill in missing data from Jaumotte. For the U.S., Jaumotte is used with reservations. Based on Gornick and Meyers, Jaumotte appears to be too high but OECD appears to be much too low. There are 338 observations in this series for the period 1980-2001 but of varying length depending on the country. Finally, we develop a composite measure of work and family policy generosity, which is a combination of the standardized maternity/parental generosity variable and the standardized daycare spending variable. We use this composite variable because we cannot put both leave and daycare variables into the regressions at the same time, due to the problem of multicollinearity (correlation = .89). Therefore, we have three separate models using daycare spending, parental leave, and the composite work and family policy variables, respectively. The means of the variables in the analysis are displayed in Table 3. Analytic Techniques Hicks (1994) notes that "errors for regression equations estimated from pooled data using OLS [ordinary least squares regression] procedures tend to be (1) temporally autoregressive, (2) crosssectionally heteroskedastic, and (3) cross-sectionally correlated as well as (4) conceal unit and period effects and (5) reflect some causal heterogeneity across space, time, or both" (p.172). We follow Beck and Katz's (1995) recommended procedure, using panel-corrected standard errors, corrections for firstorder auto-regressiveness, and imposition of a common rho for all cross-sections. This procedure is implemented in version 8.0 of the STATA econometrics program. Since there is some trend in our data, we do not include a lagged dependent variable as recommended by Beck and Katz (1996) because in this situation the lagged dependent variable inappropriately suppresses the power of other independent 3
We got similar results for the 90-10 ratio.
8
variables, as Achen (2000) has shown.4 Beck and Katz (Beck & Katz 2004, p. 16-17) have shown that correcting for first order auto-regressiveness actually does include a lagged dependent variable on the right hand side of the equation (known as Prais Winsten estimations). Thus, it does deal with the problem of serial correlation but without, as our results show, suppressing the power of other independent variables. Beck and Katz (1996) and others have argued for the inclusion of country dummies in order to deal with omitted variable bias. Plümper et al. (2005, p. 330-334) have countered that inclusion of country dummies does much more than eliminate omitted variable bias. It also (1) eliminates any variation in the dependent variable which is due to time invariant factors such as difference in constitutional structures, (2) greatly reduces the coefficients of factors that vary mainly between countries, (3) eliminates any differences in the dependent variable due to differences at t1 in the time series, and (4) “completely absorb(s) differences in the level of the independent variables across the units” (p.331, emphasis in the original). Elaborating on this last point, they argue that if one hypothesizes that the level of the independent variable has an effect on the level of the dependent variables (e.g. the party incumbency and level of replacement rates), “a fixed effects specification is not the model at hand. If a theory predicts level effects, one should not include unit dummies. In these cases, allowing for a mild bias resulting from omitted variables is less harmful than running a fixed effects specification.” (p. 334). We do hypothesize (#3) effects in the levels of our independent variables prior to t1 on the level of the dependent variable at t1, and (#4) effects of levels of the independent variables on levels of the dependent variable. In addition, variation in several of our independent variables (e.g. EPL) is primarily cross sectional (#2). Thus, it is clear that fixed effects estimation or the inclusion of country dummies is not appropriate in this case. As we mentioned, the relationship between women‟s education and women‟s employment is arguably partly endogenous. The standard procedure for dealing with endogeneity is two stage least squares which requires the development of an instrument variable for women‟s education. We were unable to find good predictors of women‟s education which were not related to the dependent variable, so we have opted to run all the models with and without women‟s education. Results The results of the regressions on women‟s employment are shown in Table 4. Consistent with the institutionalist hypotheses derived from our previous work on the determinants of overall (men and women) employment, the one year unemployment replacement rate, active labor market policy, and sickpay generosity are positively related to women‟s employment, while long term unemployment replacement rates, social security payroll taxes, and employment protection legislation are negatively related to women‟s employment. As in the case of overall employment; total taxes, wage dispersion, and union density are not related to women‟s employment. Wage coordination is correctly signed and significant in four of the ten models but insignificant and not consistently signed in the other six. To compare the magnitude of the effects of the independent variables we can examine the change in the dependent variable associated with a two standard deviation change in the independent variable. The increase in women‟s employment associated with an increase of this size in the one year unemployment replacement rate is 4.6%, in sickpay, 6.3%, and in active labor market policy, 15.8%. The decrease in employment associate with a two standard deviation increase in social security and payroll taxes is 4.2%, in employment protection legislation, 6.7%, and in long term unemployment replacement rates, 1.1. Except for the long term replacement rates (which is in any case not consistently significant), these effects are very large. Daycare was not included in Bradley and Stephens (2007) due to the absence of time series data at the time that the article was accepted for publication. Daycare spending is positively related to women‟s employment and highly significant (models 9 and 10). A two standard deviation increase in daycare spending results in a 7.7% increase in women‟s employment. We ran equations 9 and 10 alternatively substituting parental leave and our combined indicator of work and family policies (not 4
In these data, the lagged dependent variable explains 98% of the variation in the dependent variable.
9
shown). The parental leave variable was correctly signed but fell just short of significance. The work and family policy indicator was positive and highly significant but its effect on women‟s employment was not as large as daycare spending alone: A two standard deviation increase in work and family policies is associated with a 4.6% increase in women‟s employment. We also re-ran equation 9 and substituted total employment and male employment as the dependent variable. Daycare policy was significant in the regressions on total employment but not men‟s employment. The even numbered models in Table 4 include women‟s education. The coefficients are positive and highly significant in every case. They are also very large: even the smallest coefficient (model 6) indicates that two standard deviation increase in women‟s education results in a 11.6% increase in women‟s employment. This is almost certainly an overestimate due to endogeneity. But given an effect of half this magnitude would still be in the range of the effects of active labor market policy or social security and payroll taxes, it seems likely that the actual causal effect of women‟s education on women‟s employment is quite large. The addition of women‟s education only affects one of our other independent variables: employment protection legislation falls to insignificance. However, given the endogeneity problem, we still consider the results for EPL to be robust. The results of our analysis of public services sector employment are shown in Table 5. It is most useful to compare and contrast these results with the results of our analysis of women‟s employment shown in Table 4 (which were very similar to Bradley and Stephens results for total employment5) and with the results of our analysis of private service sector employment shown in Table 6. In regressions on all four dependent variables, sickpay and the one year unemployment replacement rate are almost always significant and positive and the long term unemployment replacement rate and payroll taxes are negative. As we hypothesized, total taxes and union density have different effects on the dependent variables: Negative on private service employment, positive on public service employment, and none on total employment and women‟s employment. EPL is negative and ALMP is postive for total employment, women‟s employment, and private service employment but neutral to public employment. Daycare spending is positive for total employment, women‟s employment, public service employment, and total private service employment. In our previous work, we regressed social democratic and Christian democratic government and ten control variables on the policy variables which Bradley and Stephens found to be related to total employment (see Table 1). Based on similar analyses in Huber and Stephens (2001) and Iversen and Stephens (2008), we extend the observations we made in Huo, Nelson, and Stephens (2008) to state that social democratic government is positively related to every policy or institutional (e.g union density )variable found to be positively related to women’s employment, total employment, and public service employment and is not significantly related to the variables which were negatively related to employment. One finds almost precisely the reverse profile with regard to Christian democratic government and private service employment: Christian democratic government is positively related to private service employment impeding policies (long term replacement rates, payroll taxes, total taxes, and EPL) and not related to policies that promote service employment (ALMP and one year unemployment replacement rates). On the other hand, one cannot clearly postively associate the private service employment independent variables with either of the two remaining welfare state types (liberal and social democratic) and underlying political tendencies (secular center-right government and social democratic government). High private service employment is associated with some social democratic features (high short term unemployment replacement rates and active labor market policy) and some liberal features (low union density and low levels of total taxes). Thus, high private service employment is not a uniquely liberal path to high employment, as we had assumed (Huber and Stephens 2001: xxx).
5
Sickpay was not included in Bradley and Stephens (2007). We replicated Bradley and Stephens adding sickpay and reran the odd number models in Table 4 with their dependent variable (male and female employment). In all cases, sickpay was positive and highly significant.
10
That private service employment is not uniquely liberal is reinforced by a rather surprising result not discussed up to this point: contrary to our hypothesis, wage dispersion is negatively and significantly related to private service employment. Based on the discussion and scatterplots in Scharpf (2000), we expect a strong positive relation, so we went to some effort to account for this unexpected result. The zero order correlation between wage dispersion and private service employment is positive but quite modest, r=.20. Noting that both union density and wage coordination figured among the determinants of wage dispersion (Pontusson et al 2002; Wallerstein 1999)6, we dropped these two variables from the model (not shown). Wage dispersion was still wrongly signed but now insignificant. We tentatively conclude that union density„s effect on private service sector employment operates primarily through mechanisms other than raising wages at the bottom of the distribution, such as union influence on part time work and opening hours and other regulations in the sector. In her orienting paper for the workshop, Wren (2008) emphasizes the heterogeneity of the service sector. We reasoned that if we examined a clearly low wage private service sector, such as ISIC 6 (wholesale and retail trade, hotels and restaurant), we might find a positive effect of wage dispersion and more generally might find that the purely neo-liberal set of policies would be associated with size of the sector. Table 7 shows the results of our analysis. Wage dispersion is positively signed but still not significant, despite the fact that zero order correlation between wage disperion and ISIC 6 employment is moderately high r=.59. Dropping union density and wage coordination increases the coefficient, making it just barely significant. Aside from the change in the coefficient of wage dispersion, there is one notable difference between the analysis of total private service sector employment and ISIC 6 employment, namely sickpay is now negative. While short term unemployment replacement rates and ALMP are still positive and significant, otherwise the policies and institutional arrangements associated with high ISIC 6 employment are those that are associated with the liberal welfare state and neo-liberal policy prescriptions. Conclusions Given the ordering of the welfare state - production regimes on employment levels and women‟s employment levels, with the social democratic welfare states on the high end, the Christian democratic welfare states on the low end and that the liberal welfare states in the middle, our findings in this paper and related work are not surprising. Contrary to neo-liberal orthodoxy, generous welfare states and high employment are not simply compatible, generous social policies are actually supportive of high employment provided they are configured properly. The social democratic pattern of high short term unemployment replacement rates, high ALMP, moderate EPL, high daycare spending, low long term unemployment replacement rates, and coordinated wage bargaining increases employment and women‟s employment. By contrast, the Christian democratic pattern is negative for total employment and women‟s employment. We explored this further in the analysis service sector employment. This analysis did result in some surprising findings. We expected to find a social democratic path to high employment via high public employment and low private service sector employment and a liberal path with the opposite characteristics. We did find a uniquely social democratic path to high public service employment. In the private services, we found that Christian democratic welfare state features were associated with low private service employment. The welfare state features which were associated with high service sector employment were an apparently contradictory mix of social democratic ones (high short term unemployment replacement rates, high sickpay, high ALMP, low wage dispersion) and liberal ones (low union density, low taxes) and ones associated with both regimes (modest to low EPL, low long term unemployment replacement, modest to low payroll taxes). We reasoned that this package is seemingly contradictory because of the heterogeneity of the service sector pointed to by Wren in her 6
Both of these works actually find that bargaining centralization and union density are strongly related to wage dispersion, but since wage coordination is highly correlated with bargaining centralization, it still makes sense to investigate whether dropping these two variables from the model might affect the finding.
11
orienting paper: Different mixes of these policies were associated with the employment performance of different segments of the private service sector. To explore this possibility, we focused on a low wage private service sector, ISIC 6, wholesale and retail sales, hotel and restaurant. We did find that the policy and institutional profile that was associated with good employment performance of this sector was much more unambiguously liberal. We would expect a different, more social democratic mix to be associated with employment in other sub-sectors of private services.
12
Bibliography
Achen CH. 2000. Why Lagged Dependent Variables Can Suppress the Explanatory Power of Other Independent Variables. Annual Meeting of the Political Methodology Section of the American Political Science Association, UCLA, July:20-2 Anderson PM, Levine PB. 1999. Child Care and Mothers' Employment Decisions. NBER Beck N, Katz J. 1995. What to Do (And Not to Do) With Time Series Cross Sectional Data. American Political Science Review 89:634-47 Beck N, Katz JN. 1996. Nuisance vs. Substance: Specifying and Estimating Time-SeriesCross-Section Models. Political Analysis 6:1-36 Beck N, Katz JN. 2004. Times-series-cross-section issues: Dynamics, 2004. In Unpublished manuscript. New York, NYU Bertola G, Blau FD, Kahn LM. 2007. Labor market institutions and demographic employment patterns. Journal of Population Economics 20:833-67 Bradley D, Stephens J. 2007. Employment Performance in OECD Countries: A Test of NeoLiberal and Institutionalist Hypotheses. Comparative Political Studies 40 Browne I, Misra J. 2003. The Intersection of Gender and Race in the Labor Market. Annual Review of Sociology 29:487-514 Burniaux J, Duval MR, Jaumotte F. 2003. Coping with ageing: a dynamic approach to quantify the impact of alternative policy options on future labour supply in OECD countries. In Documents de travail du Département des affaires économiques de l’OCDE Casey BH. 2004. The OECD jobs strategy and the European employment strategy: Two views of the labour market and the welfare state. European Journal of Industrial Relations 10 Daly M. 2000. A Fine Balance: Women's Labor Market Participation in International Comparison. In Welfare and Work in the Open Economy Volume II: Diverse Responses to Common Challenges in Twelve Countries, ed. FW Scharpf, VA Schmidt. Oxford: Oxford University Press Daly M. 2002. Care as a Good for Social Policy. Journal of Social Policy 31:251-70 Elliott JR, Smith RA. 2004. Race, gender, and workplace power. American Sociological Review 69:365-86 Esping-Andersen G. 1990. The three worlds of welfare capitalism. Princeton, N.J.: Princeton University Press. xi, 248 pp. Esping-Andersen G. 1993. Changing Classes: Stratification and Mobility in Post-Industrial Societies: Sage Publications Esping-Andersen G. 1999. Social Foundations of Postindustrial Economies: Oxford University Press Esping-Andersen G, Regini M. 2000. Why deregulate labour markets? Oxford UK ; New York: Oxford University Press. xiv, 357 pp. Estevez-Abe M. 2005. Gender Bias in Skills and Social Policies: The Varieties of Capitalism Perspective on Sex Segregation. Social Politics: International Studies in Gender, State & Society 12:180-215 Freeman R. 1995. THE LIMITS OF WAGE FLEXIBILITY TO CURING UNEMPLOYMENT. Oxford Review of Economic Policy 11:63-72 Gauthier AH, Bortnik A. 2001. Comparative Maternity, Parental, and Childcare Database, Version 2. University of Calgary Ginn J, Arber S, Brannen J, Dale A, Dex S, et al. 1996. Feminist Fallacies: A Reply to Hakim on Women's Employment. The British Journal of Sociology 47:167-74 Gornick JC, Meyers MK. 2003. Families that Work: Policies for Reconciling Parenthood and Employment. New York
13
Gornick JC, Meyers MK, Ross KE. 1997. Supporting the Employment of Mothers: Policy Variation Across Fourteen Welfare States. Journal of European Social Policy 7:45 Gornick JC, Meyers MK, Ross KE. 1998. Public policies and the employment of mothers: A cross-national study. Social science quarterly 79:35-54 Hakim C. 1995. Five Feminist Myths about Women's Employment. The British Journal of Sociology 46:429-55 Harmon C, Oosterbeek H, Walker I. 2003. The Returns to Education: Microeconomics. Journal of Economic Surveys 17:115-56 Hicks A, Kenworthy L. 2006. Family Policies and Women's Employment: A Regression Analysis Hicks AM. 1994. Introduction to Pooling. The Comparative Political Economy of the Welfare State:169–88 Huber E, Ragin C, Jd S. 2004. Comparable Welfare State Dataset. Northwestern University and University of North Carolina Huber E, Stephens JD. 2001. Development and crisis of the welfare state : parties and policies in global markets. Chicago: University of Chicago Press. 416 pp. Huo J, Nelson M, Stephens J. 2008. Decommodification and Activation in Social Democratic Policy: Resolving the Paradox. European Journal of Social Policy Jaumotte F. 2003. Female Labour Force Participation: Past Trends and Main Determinants in OECD Countries, OECD, Paris Kahn LM. 2000. Wage Inequality, Collective Bargaining, and Relative Employment from 1985 to 1994: Evidence from Fifteen OECD Countries. The Review of Economics and Statistics 82:564-79 Kimmel J. 1998. Child Care Costs as a Barrier to Employment for Single and Married Mothers. The Review of Economics and Statistics 80:287-99 Kitschelt H, Lange P, Marks G, Stephens J. 1999. Convergence and Divergence in Advanced Capitalist Economies. Continuity and Change in Contemporary Capitalism Eds H Kitschelt, P Lange, M Gary, JD Stephens (Cambridge University Press, Cambridge) pp:427-60 Korpi T. 2000. Faces of Inequality: Gender, Class, and Patterns of Inequalities in Different Types of Welfare States. Social Politics Summer Lee JW, Barro RJ. 2001. International Measures of Schooling Years and Schooling Quality Dataset. World Bank Lindbeck A, Snower DJ. 1988. The insider-outsider theory of employment and unemployment. Cambridge, Mass.: MIT Press. xii, 285 p. pp. Lindbeck A, Snower DJ. 2002. The Insider-Outsider Theory: A Survey. In IZA Discussion Paper Series. Bonn Morgan KJ. 2005. „The Production of Child Care: How Labor Markets Shape Social Policy and Vice Versa”. Social Politics 12:243-63 Nickell S. 1997. Unemployment and Labor Market Rigidities: Europe versus North America. Journal of Economic Perspectives 11:55-74 OECD. 1994. The OECD Jobs Strategy: Facts, Analyses, Strategies. Paris Plumper T, Troeger VE, Manow P. 2005. Panel data analysis in comparative politics: Linking method to theory. European Journal of Political Research 44:327-54 Pontusson J, Rueda D, Way C. 2002. Comparative Political Economy of Wage Distribution: The Role of Partisanship and Labour Market Institutions. British Journal of Industrial Relations 32:281-308 Psacharopoulos G. 1985. Returns to Education: A Further International Update and Implications. The Journal of Human Resources 20:583-604
14
Quinn DP, Inclan C. 1997. The Origins of Financial Openness: A Study of Current and Capital Account Liberalization. American Journal of Political Science 41:771-813 Rueda D. 2005a. Insider-Outsider Politics in Advanced Industrialized Democracies: The Challenge to Social Democratic Parties. American Political Science Review 99:61-74 Rueda D. 2005b. Social Democracy and Active Labour Market Policies: Insiders, Outsiders, and the Politics of Employment Promotion. British Journal of Industrial Relations Forthcoming Ruhm CJ. 1998. The Economic Consequences of Parental Leave Mandates: Lessons From Europe*. Quarterly Journal of Economics 113:285-317 Sainsbury D. 1996. Gender, Equality, and Welfare States: Cambridge University Press Scharpf FW. 2001. 7 The viability of advanced welfare states in the international economy Vulnerabilities and options. Welfare State Futures 8:399-425 Scruggs L. 2005. Comparative Welfare Entitlements Dataset. Department of Political Science, University of Connecticut Sianesi B, Reenen JV. 2003. The Returns to Education: Macroeconomics. Journal of Economic Surveys 17:157-200 Simmons B. 1999. The Internationalization of Capital. Continuity and Change in Contemporary Capitalism:36–69 Stier H, Lewin-Epstein N, Braun M. 2001. Welfare Regimes, Family-Supportive Policies, and Women‟s Employment along the Life-Course Americal Journal of Sociology 106 Wallerstein M. 1999. Wage-Setting Institutions and Pay Inequality in Advanced Industrial Societies. American Journal of Political Science 43:649-80
Table 2. Variable Descriptions, Data Sources and Hypothesized Effects on Employment level
Variable
Description
Dependent Variables Women's employment
Percentage of the women aged 15-64 employed.a,b
Public service employment Private service employment Private service employment II Independent Variables Gross replacement rate, 1 year Gross replacement rate, 4-5 year Social security taxes Government Revenue Wage coordination Union density Wage dispersion Employment protection ALMP
Hypothesized impact on employment levels Women Public Private service service
Civilian government employmenta Service sector employment minus civilian government employment ISIC 6
Replacement rate for an unemployment spell of one year for 2/3 median payb Replacement rate for an unemployment spell of four and five years for 2/3 median payb Social security payroll taxes as a percent of GDP.a,b Total taxes as percent of GDPa,b Degree of coordination of wage bargaining.c Union members as percent of wage and salary workers.d Ratio of median wage and salary worker to wage and salary worker at the 10th percentile (full time work)b Index of employment protection legislationb Spending on active labor market policy measures divided by the unemployed populationa,b
+ 0 +
+ + +
-
0 +
+ 0
+
+
0 +
+
Sickness Insurance Day care spending Women's educational attainment Capital market openness
Index of replacement rates, coverage, and qualify conditions (see text) Spending on daycare as a percent of GDP (see text) Barro and Lee Index of controls and international agreements on capital and currency accounts, high values indicate fewer controls.a,e
+ + + -/+
+ +
-/+ +
-/+
-/+
Trade openness
Exports plus imports as a percent of GDP.a
-/+
-/+
-/+
Sources: aHuber et al. (2004); b OCED cKenworthy (2001); dEbbinghaus and Visser (2000); eQuinn
16
Table 3. Mean Values of Key Variables by Country Female Empl. Public Private Gross Gross % working age Service Service replacement replacement population Employment Employment rate, 1 year rate, 4-5 year Nordic CMEs Sweden 70.6 22.1 27.2 85.0 0.0 Norway 63.5 17.9 29.2 48.3 5.7 Denmark 65.9 19.9 22.3 87.8 37.3 Finland 65.0 11.9 26.6 51.8 24.6 Mean
Sickness Insurance
Social security taxes
Government Wage Revenue Coordination
3.2 3.4 2.6 1.0
13.1 11.9 1.3 6.8
56.8 50.2 50.9 45.8
Union Density
Capital Market Openness
Trade Openness
Wage Employment dispersion protection 50-10 ratio legislation
Active labor market policy
Daycare Women's Spending Education Attainment
4.1 4.8 3.9 3.8
80.6 56.2 75.1 71.1
2.3 1.0 1.1 1.7
65.1 74.2 66.9 58.9
1.3 1.4 1.4 1.5
3.0 2.8 2.0 2.2
0.5 0.2 0.2 0.2
1.7 1.4 1.8 1.0
8.8 8.0 8.8 7.4
66.3
18.2
26.3
68.2
16.9
2.5
8.3
50.9
4.1
70.8
1.5
66.3
1.4
2.5
0.3
1.5
8.2
Continental CMEs Austria 51.7 Belgium 43.3 Netherlands 43.4 Germany 51.4 France 50.5 Italy 35.9 Switzerland 60.3
11.1 8.9 6.1 8.2 12.4 7.7 7.6
22.1 26.5 28.3 21.3 23.5 20.3 37.6
34.9 57.9 68.1 39.4 58.4 10.3 51.6
24.3 45.5 30.0 23.8 14.8 0.0 0.0
1.5 1.8 2.2 3.2 0.2 0.3 1.0
13.1 14.4 18.0 15.3 18.8 12.6 10.2
45.9 45.6 50.8 44.0 44.4 38.4 32.7
4.4 4.3 3.9 4.0 2.0 2.8 4.0
50.3 52.9 30.1 32.7 14.8 42.7 27.0
0.5 2.3 3.6 1.1 1.4 0.4 2.5
76.6 129.2 101.1 55.1 43.7 43.6 68.3
1.9 1.4 1.6 1.6 1.6 1.6 1.6
2.2 3.0 2.7 3.0 2.9 3.5 1.1
0.1 0.2 0.2 0.2 0.1 0.0 0.2
0.5 0.6 0.4 0.3 0.6 0.1 0.1
6.3 7.7 7.0 8.2 6.5 5.0 8.4
Mean
48.1
8.9
25.7
45.8
19.8
1.4
14.6
43.2
3.6
35.8
1.7
73.9
1.6
2.6
0.1
0.4
7.0
LMEs Australia Canada Ireland UK USA
52.5 55.8 37.1 57.3 59.3
9.9 13.5 7.0 12.4 9.5
32.8 31.1 20.1 29.5 36.7
28.3 57.5 47.2 32.0 30.8
28.8 0.0 22.6 21.0 6.1
-3.0 -2.2 -5.9 -1.8
0.0 4.5 4.8 6.3 7.6
30.8 39.5 38.0 39.0 30.8
3.0 1.4 3.6 1.8 1.3
44.8 34.8 52.1 42.8 18.8
0.9 1.2 0.6 3.2 0.7
34.6 58.0 121.6 53.5 19.6
1.6 2.3 2.0 1.8 2.0
1.0 0.8 0.9 0.6 0.2
0.0 0.1 0.1 0.1 0.0
0.1 0.3 0.1 0.1 0.3
9.6 9.7 7.5 8.3 10.4
Mean
52.4
10.5
30.1
39.1
15.7
-3.1
4.6
35.7
2.2
38.7
1.3
57.5
1.9
0.7
0.1
0.2
9.1
Japan
56.4
5.8
33.8
32.2
0.0
-0.8
8.4
28.7
5.0
28.4
0.6
21.4
1.7
2.1
0.1
0.2
7.6
17
Table 4. Prais-Winsten Models of Women's Employment Levels Independent Variables 1 2 3 4 Replacement rate, 1 year .116 *** .065 *** .203 *** .134 *** Replacement rate, 5 year -.034 -.004 -.066 -.009 Sickness Insurance .963 *** 1.138 *** .533 .967 ** Government Revenue .029 -.069 -.101 -.128 Wage coordination .746 *** .812 *** .853 *** .899 *** Union Density -.055 -.040 .007 .000 Social security taxes -.383 *** -.236 ** -.333 *** -.234 ** Capital market openness .284 *** .186 * .381 ** .286 * Trade openness -.097 *** -.083 *** -.123 *** -.123 *** Wage dispersion -.199 .800 EPL ALMP Daycare spending Women's education 3.642 *** 3.077****** Constant Common rho R2 N
56.054 *** 28.420 ***
57.051 ***
31.912 ***
5 .133 -.083 1.421 .028 .069 .041 -.329 .222 -.122
*** *** ***
6 .086 -.040 1.337 .006 .101 .020 -.289 .175 -.099
-3.288 ***
-1.012
*** *** ***
*** ***
** ** ***
7 .131 -.062 .808 -.065 -.181 .030 -.465 .223 -.121
*** *** ***
*** *** ***
13.772 *** 3.274 *** 65.631
8 .066 -.031 1.051 -.001 -.120 .011 -.289 .124 -.091
*** ***
*** * ***
9 .107 -.091 .936 -.080 -.007 -.076 -.522 .343 -.098
*** *** ***
* *** *** ***
62.921 ***
*** *** ***
6.681 ***
3.740 *** 3.633 ***
31.104 *** 66.383 ***
31.851 ***
.90
.93
.90
.93
.87
.88
.84
.89
.84
.84
.77 442
.76 442
.75 305
.74 305
.88 242
.90 242
.88 287
.91 287
.89 272
.92 272
Level of significance: ***=.001, **=.01, *=.05, ^=significant in opposite direction of hypothesis
** *** ***
11.943 *** 3.277 ***
33.462 ***
10 .057 -.062 1.255 -.039 -.002 -.025 -.341 .234 -.086
18
Table 5. Prais-Winsten Models of Public Service Employment Levels Independent Variables 1 2 3 Replacement rate, 1 year .038 *** .067 *** .040 *** Replacement rate, 5 year -.009 -.047 *** -.012 Sickness Insurance .118 ** -.041 .159 ** Government Revenue .134 *** .141 *** .119 *** Wage coordination -.014 -.054 -.065 Union Density .103 *** .111 *** .112 *** Social security taxes -.081 *** -.082 ** -.080 ** Capital market openness .048 ** .047 .039 * Trade openness -.029 *** -.031 *** -.039 *** Wage dispersion .017 EPL .068 ALMP Daycare spending Constant Common rho R2 N
1.820 **
1.288
3.319 ***
4 .037 -.012 .149 .122 -.059 .116 -.080 .033 -.039
*** ** *** *** *** ***
5 .032 -.019 .142 .137 -.064 .091 -.087 .062 -.036
*** ** ** *** *** *** ** ***
1.076 1.940 *** 3.175 ***
3.075 ***
.92
.88
.93
.93
.90
.79 430
.85 297
.82 227
.83 272
.87 257
Level of significance: ***=.001, **=.01, *=.05, ^=significant in opposite direction of hypothesis
19
Table 6. Prais-Winsten models of Private Services Employment Levels Independent Variables 1 2 3 Replacement rate, 1 year .020 .084 ^ .121 ^ Replacement rate, 5 year -.024 -.087 *** -.024 Sickness Insurance .087 -.077 .792 ^ Government Revenue .009 -.121 -.295 *** Wage coordination -.066 .037 .324 Union Density -.187 *** -.190 *** -.124 *** Social security taxes -.147 -.376 *** -.117 Capital market openness .329 ** .380 ** .340 *** Trade openness -.008 .025 -.040 *** Wage dispersion -5.016 ^ EPL -2.737 *** ALMP Daycare spending Constant Common rho R2 N
4 .114 .003 .355 -.317 .168 -.167 -.296 .356 -.036
^ ^ *** *** *** ** ***
5 .113 -.017 .383 -.343 .230 -.209 -.382 .736 -.023
^ ^ *** *** *** *** *
4.796 ** 2.621 *
36.649 ***
50.166 ***
51.097 ***
48.954 ***
50.724 ***
.84 .65 374
.79 .75 267
.71 .82 171
.73 .80 216
.66 .78 201
Level of significance: ***=.001, **=.01, *=.05, ^=significant in opposite direction of hypothesis
20
Table 7. Prais-Winsten models of Employment Levels in ISIC 6 Independent Variables 1 2 3 Replacement rate, 1 year .009 ^ .028 ^ .031 ^ Replacement rate, 5 year -.025 *** -.039 *** -.045 *** Sickness Insurance -.327 *** -.259 *** .034 Government Revenue .056 ^ .025 -.014 Wage coordination .000 .050 .191 Union Density -.039 *** -.040 *** -.043 *** Social security taxes -.199 *** -.240 *** -.143 *** Capital market openness .049 .069 * .038 Trade openness -.005 -.003 -.007 * Wage dispersion .552 EPL -1.564 *** ALMP Daycare spending Constant Common rho R2 N
4 .030 -.024 -.293 -.030 .039 -.049 -.215 .012 -.007
^ *** ***
*** ***
5 .049 -.027 -.352 -.015 .123 -.037 -.210 .070 -.012
^ ** *** * *** *** * **
2.731 *** -.455
11.164 ***
11.258 ***
16.412 ***
14.456 ***
12.932 ***
.83 .82 318
.74 .80 252
.79 .89 150
.82 .86 191
.76 .85 186
Level of significance: ***=.001, **=.01, *=.05, ^=significant in opposite direction of hypothesis
