Jobs 72 - 88 - ... participants at Berkeley, Essen, Mannheim, at the EEA Prague und the. Verein für Socialpolitik, Linz annual meetings provided helpful comments.
Eastern European Trade and the Austrian Labour Market
Karl Aiginger Austrian Institute for Economic Research, Vienna, University of Linz, Austria
Rudolf Winter-Ebmer Institute of Industrial Relations, University of California, Berkeley University of Linz, Austria Centre for Economic Policy Research, London
Josef Zweimüller University of Linz, Austria Institute for Advanced Studies, Vienna
This research was supported by a grant from the Austrian Central Bank (Jubiläumsfonds #4674) to the second and third author. A Eymann, C. Martin, W. Kohler, K. Kraft, F. Pfeiffer, R. Schnabel, V. Steiner and L. Ulman as well as seminar participants at Berkeley, Essen, Mannheim, at the EEA Prague und the Verein für Socialpolitik, Linz annual meetings provided helpful comments.
Abstract The last years saw a major structural break in trade relations between Western European countries and the former eastern bloc. Austria experienced a disproportionately large bilateral trade creation with these countries. In this paper we take a closer look at the impact this trade growth had on the Austrian labour market. To differentiate as far as possible between different segments of the labour market, we concentrate on unemployment experience and wage growth for a panel of individual workers in Austrian industry. The results show rather small employment effects, the impact on wage growth is more pronounced with interesting modifications for mobile and immobile workers.
JEL: F14, J23, J31, J64. Keywords: CEEC trade, unemployment, wage growth.
3 1. Introduction The breakdown of the communist regimes in Eastern Europe resulted in a major structural break in the international economic relations of most Central European countries. This is particularly the case for Austria. Due to its geographical situation and its strong historical ties, especially with former Czecho-Slovakia, Hungary and Poland, Austria received disproportionally more immigrants than other countries1 and its trade flows reacted more strongly. Already in 1989, Austria had the largest export volume to the Central and Eastern European Countries (CEECs) all over Europe, except for Germany and Italy. Furthermore, Baldwin (1994, p. 90) estimated a very high future export potential into those countries using a gravitation model, matched only by Germany, Italy and France. In this paper we take a look at the consequences of this trade creation for the Austrian labour market. Unlike in the U.S., where recently many empirical studies dealt with labour market effects of the NAFTA treaty, almost no comparable studies2 exist for Europe. The Austrian example may therefore serve as a benchmark case of a country which is most heavily influenced in relative terms. Clearly, the short period of transition prevents up to now time-series studies. Therefore, we concentrate on the experience of individual workers who were employed in the manufacturing sector in 1988. We follow these persons over the period 1988-91 and analyse the impact of trade creation on their labour market experience - in terms of i) employment rates and ii) wage changes. Looking at aggregated industry employment data would give a flawed picture due to the high mobility of labour: more than 40 % of individuals working in a specific industry in 1988 have left this industry by 1991. For this reason we also take pains to distinguish properly between mobile and immobile workers - taking the issue of selectivity in the mobility decision into account. Using this framework, we get a picture of a cohort of workers, that of 1988, but not necessarily one of an industry. The main innovations of this paper are thus twofold: (i) first evidence on the impact of trade with the CEECs on Western Europe and (ii) possible reactions of workers are modelled in more detail than in previous studies. The paper is organised as follows. Section 2 reviews the existing literature on trade and the labour market and discusses some theoretical problems. Section 3 turns to the Austrian trade relations with the CEECs. Section 4 discusses the data, whereas in Sections 5 and 6 econometric evidence on the trade effects on Austrian employment and wage growth is presented. Section 7 concludes.
Winter-Ebmer and Zweimüller (forthcoming, and 1994) for the impact of migration on the Austrian labour market. 2Cadot and de Melo (1994) provide simulation results for the regional distribution of possible job creation and destruction caused by CEEC-trade with France.
4 2. Trade and the Labour Market Assessing the impact of international trade on the labour market - in line with the equilibrium character of trade theory - would ideally call for the use of a computable general equilibrium model. See e.g. Kohler (1991) for an application to trade liberalisation in Austria.3 These studies typically analyse these effects ex ante rather than ex post. They are, however, not particularly well suited for our purpose because we want to consider the labour market effects in more detail than is usually done. Some recent very influential studies in the U.S. used simulation techniques. First, the labour content of trade flows is calculated and hypothetically added to or subtracted from domestic labour supply4. Wage effects are then deduced by using estimates of demand and supply elasticities (Borjas, Freeman and Katz, 1991, Murphy and Welch, 1991). Owing to the large U.S. trade deficits in the eighties, big wage impacts are generally found. Regression analyses either use trade flows or prices as explanatory variables. Using trade flows, Katz and Revenga (1989) look at bilateral U.S.-Japan trade, MacPherson and Stewart (1990) as well as Freeman and Katz (1991) distinguish between union and non-union sectors in the U.S. economy. With respect to unionisation, they find a greater sensitivity of wages to sales (exports, imports) in unionised than in non-unionised sectors. This is explained by rent sharing.5 If international prices are used to predict domestic wages or employment, Grossman (1986, 1987) finds less significant evidence. In a similar study Revenga (1992) employs instrumental variables techniques and concludes that wages react to a smaller extent to import prices as employment levels do. Finally, Lawrence and Slaughter (1993) test if the observed pattern of consumer prices can account for the pattern of U.S. factor prices and find it inconsistent with basic premises of the Stolper-Samuelson theorem. The notion that freer trade can have adverse impacts on factors of production is based on the Factor Price Equalisation theorem or, more specifically, on the StolperSamuelson theorem: free trade will harm the factor which is scarce at home relative to the trading partners, i.e. presumably unskilled labour in the case of Western Europe versus the CEECs, relative to the other factors of production such as capital. Bhagwati and Dehajia (1993) - in a defence of Free Trade positions - challenge the inescapability of the Stolper-Samuelson conclusions. Given the demanding assumptions that underlie this theorem, a more realistic set-up would easily change the prediction of immiseration of
Brown (1992) for a survey of CGE models applied to the consequences of NAFTA. Wood( 1991, 1995) for a critique on labour content studies. Wood attributes most of the relative decline in the wages of low-skilled workers to trade with developing countries. 5Gaston and Trefler (1994a) are unable to find significant effects of import or export flows on Canadian employment as well as wage changes. This may be due to multicollinearity, because they also include tariffs and the exchange rate in those regressions. 4See
5 unskilled labour; e.g. the factor intensity reversal phenomenon, increasing returns to scale, the specific factors model or complete specialisation. A similar case can be made if more trade leads to more competition and technical progress in the economy. Deardorff and Hakura (1993) criticise the aforementioned labour studies as not being based on trade theory. They argue if only the labour market was studied, the essentially general equilibrium character of international trade would be lost. Trade could not be seen as a cause of changing factor prices, instead trade would only arise because of factor price differentials, „ ... in any sensible model, the volume of trade and the level of wages are simultaneously determined.“ (p.3). Furthermore, they plead for the use of goods prices instead of trade flows in regression analysis. Although Deardorff and Hakura (1993) are rather agnostic about the possibility of any empirical estimation of the trade-wage relationship, they concede the importance of one specific question: How do exogenous changes in economic conditions abroad, transmitted to the domestic economy (only) through trade, affect wages? This is exactly the environment of trade creation between Austria and the CEECs after the fall of the communist regimes, that we are going to study. We would argue, therefore, that this specific historical episode allows us to compare income distributions before and after the removal of this "trade barrier". Notwithstanding these criticisms, in the following empirical part we take a pronounced labour market stance, basically to be able to give a broad and differentiated picture of possible labour market impacts. Whereas most quoted studies use aggregate industry data,6 we use individual panel data to be able to model different threats and options of workers. We do, however, try to properly incorporate the issue of endogeneity of trade volumes and factor prices. The use of goods prices instead of trade volumes is impossible because of a lack of reliable price data.7 3. Austria as a trading partner of the CEECs Austria is a small open economy which experienced a rapid and smooth economic development after World War II. Its currency moved parallel with the German Mark and therefore steadily appreciated against Austria’s other trade partners during the last 20 years. Exports reach 32% of GNP, imports 39%, the trade balance is negative. The extent of the deficit is more or less stable over time. The deficit in the trade balance is compensated by a surplus in the service balance (mainly by tourism). The current balance is not seen as a severe restriction of economic policy, in the long run total exports and total imports are well balanced. Before the opening of the borders, exports were heavily concentrated on the three western neighbour countries. Germany, Switzerland and Italy combined to make up 54% 6The
exception being Kruse (1988), Freeman and Katz (1991) and Gaston and Trefler (1994). unit values for exports and imports were available.
6 of the exports of goods and 58% of imports (1988). Trade with overseas countries is relatively low if compared, for example, to Switzerland. The trade volume with socialist countries was high relative to other Western European economies, but very low if compared with pre-war ratios and if evaluated from the perspective of location and neighbourhood. The former CSFR and Hungary were Austria’s 16th and 15th largest export partners in 1988. Switzerland with similar size and position was number three, its volume was four times respectively six times higher than the bilateral volumes between Austria and CSFR. Total Austrian exports to the three CEECs (former CSFR, Hungary and Poland) was 15.2 bn ATS, imports amounted to 16.7 bn ATS in 1988, the same figures with Switzerland were 27.6 resp 19.9 bn ATS (Table 1). Table 1 Before the transition started, the bilateral trade balance was in approximate equilibrium. Austria had a slight surplus with Hungary and a small deficit with former CSFR and Poland, adding up to a deficit of 1.4 bn ATS with the area as a whole in 1988. Between 1998 and 1991 Austria’s exports expanded by 105%, imports increased by 47%, giving a surplus for Austria of 6.6 bn ATS in 1991. The development continued in the years since, so that for the total period 1988 - 1993, Austria’s export rose by 154%, imports by 67%. The export share of the three countries add up to 8% of Austria’s exports (1993) after 4% in 1988. The overall impact of the opening of the borders on the economic development in Austria had been a hotly debated political issue since the start of „Ostöffnung“. Many industries and firms were confronted with increasing competition because wages in these countries were between 5% (Poland) and 10% (Hungary) of the comparative Austrian labour costs (Peneder, 1993, for year 1991).8 These huge wage differentials lead to fears of serious detrimental impacts, given the structural problems of Austrian manufacturing: a large sector of undifferentiated basic goods (steel, non ferrous metals, fertilisers) and of consumer products relying on cheap labour (wearing apparel, shoe industry); as well as an underdeveloped sector of sophisticated engineering industries and high tech firms and a deficit in the active internationalisation, be it via direct investments or strategic partnerships. A study on the prospects of the opening of the borders by the Austrian Institute of Economic Research (Aiginger 1993) mentioned three channels for possible employment effects for Austria: i) the substitution of low quality production by imports, ii) the increase in exports due to the increasing market, and iii) cheaper inputs (via imports or
8 For a rather pessimistic assessment see Beirat (1992), for a discussion of structural impacts of the opening of borders on OECD countries in general see Landesmann, Snell (1993) and Holzmann et al. (1994)
7 subsidiary firms producing in the reform countries) for manufacturing firms. The first effect should decrease employment, effect two and three should increase it. The balance of the three effects was estimated to be slightly positive (10,000 - 15,000 employees). Aggregate figures on trade and employment between 1988 and 1991 and then further to 1994 show that the opening of the borders aggravated existing structural problems, but on balance could not be assessed as negative. The largest decline in employment has been experienced in the mining sector and in the steel industry by approximately one third, as compared to overall employment decrease in manufacturing by 12% (1988-94), but both sectors had experienced a similar drop between 1980 and 1988. The other problem area is the textile sector: it lost 26% of its employment between 1988 and 1993, following a decline of 28% in the period before. On the other hand, employment in construction-based industries (stone & quarrying, glass, wood products) and the engineering industries experienced rather stable employment in the transition period and before. Bilateral exports and imports of Austria with CEECs only partially correspond with the employment pattern. Steel exports of Austria rose faster than imports transforming a deficit into a surplus. Exports of the textile industry are five times and imports are ten times higher in 1993 than in the pre-transition period. An analysis of the bilateral trade and its changing structure shows (Aiginger, Peneder, Stankovsky, 1994) that the Austrian position had improved in sectors with high capital intensity and slightly deteriorated in sectors with labour and skill intensity. This result is not too surprising if we are reminded that capital investment had a high priority in socialist economies and skill intensive goods were largely required to be interchanged with raw material in intra CMEA trade. Now capital is dear and labour became cheap in the transforming countries. However, explaining trade between former socialist countries and its change by Heckscher Ohlin theory alone cannot be expected to be a complete description of reality. The overall structural impact of the „Ostöffnung“ therefore seems to be the following: A strong low-cost competitor entered the Austrian market for undifferentiated products creating losses in employment and production. This is, however, also a chance for firms to increase direct investments and combine low cost elements with more sophisticated factor inputs including management and service, the latter being produced in Austria and increasing employment additional to that created by higher exports. The shifting trade balance (from a slight deficit to a large surplus) indicates a positive aggregate impact on employment.
8 4. Data We use a 2% sample of workers in the Austrian manufacturing industry employed on May 31, 1988. To eschew problems with early retirement we concentrate on male persons below the age of 57. This subsample is part of a bigger representative sample of workers from social security records. For these persons, official information is available on all economic aspects - relevant for the calculation of old age pensions - for the period 1972 to 1991. As the data have been collected mainly for social security purposes, several drawbacks exist. First, there is no information on family affiliation. Second, the level of schooling can only be calculated for a subset of persons. The information on work experience and tenure with the actual firm relates only to the period 1972-91. Furthermore, our wage indicator, monthly gross earnings, is right censored, because earnings are only available up to the social security contribution ceiling9. No information on hours worked is available. For the period May 1988 to May 1991 we can observe the labour market status of the individuals on each single day, including unemployment entry, job changes or the move into another industry. To assess the impact of trade with the CEECs we impute additional data from trade statistics. As CEECs we define former CSFR, Hungary and Poland, these countries account for the bulk of trade with Eastern Europe. For 25 industries in the manufacturing sector we construct two trade variables: growth of exports (imports) Austria-CEECs 1988-91 as a percentage of output in the respective industry 1988 (∆ X/Q, ∆ I/Q)10. As standard trade theory gives strong indication for endogeneity of trade flows, we test for exogeneity in the regressions. As instruments for the change of exports (imports) we use the level of exports (imports) in 1988, four industry factor shares as well as some further industry characteristics: unionisation, share of blue-collar and immigrant workers. 5. Employment effects As an indicator for individual employment performance we define as unemployment rate (U) in the period 1988-91 the number of days not in employment divided by the potential maximum (365x3). This could be seen as a simple complement to the individual employment rate, but we also count „unemployment days“ if the worker is in transition between two jobs but not registered as unemployed. Measuring unemployment experience in such a way encompasses the usual distinction between unemployment risk
censoring applies only to 8.5% of the individuals. trade data use the SITC nomenclature, while the labor data apply the ISIC code, we had to use a bridge which was developed at the Austrian Institute of Economic Research (WIFO), this bridge defines which ISIC code is nearest to an SITC code (at the three digit level).
9 and unemployment duration.11 As the unemployment rate has a lower bound of 0.0, we have to use a Tobit estimator: U*i = βSi + γ1(∆X / Q)i + γ 2 ( ∆I / Q)i + εi (1)
Ui = U*i
for U*i ≥ 0
Ui = 0
for Ui* < 0
where U* is the underlying latent variable, S are control variables, mainly from human capital theory and β, γ 1 and γ 2 are parameters to be estimated. Table 2 Results of the Tobit regression for unemployment rates are given in Table 2. We test the goodness of fit of the model with the usual likelihood-ratio test and a Pseudo R 2 − R 2MZ1 - which is found to be closest to the usual OLS-R2 in Monte-Carlo studies by Veall and Zimmermann (1994). Exogeneity of the trade variables is rejected using a Wald test at the 6% level (Smith and Blundell, 1986). We therefore present only results with instrumented trade volumes. The trade parameters indicate that growing export shares to the CEECs in the period 1988-91 significantly reduced unemployment, whereas no significant impact of imports could be found. Table 3
As the impact of trade may be different for socio-economic groups, Table 3 presents results for various subsamples: blue-collar workers, young and elderly as well as lowincome workers. In contrast to the results for the full sample, here all trade coefficients are significant. The results suggest that the ability to gain from positive export performance and the threat to be hurt by rising import penetration may be different across demographic groups. As far as the impact of export-growth is concerned, Table 3 reveals that with the exception of low-income earners, the impact on unemployment risk is comparable in size to the full sample. Workers earning less than ATS 15,000 benefit much less from increased exports. All groups listed in Table 3 are heavily affected by CEEC-import growth. In all cases the effect is larger in size than for a comparable rise in exports. It is particularly strong for elderly and low-income earners. As the Tobit coefficients cannot be easily interpreted, we report in Table 4 simulated effects of an increase in trade volumes on unemployment. We have to distinguish be11Kruse
(1988) looks only at the impact of trade on unemployment duration of displaced workers.
10 tween three effects: i) the expected probability of unemployment entry in the period 8891 (Pr(U>O)), ii) the mean unemployment rate for those experiencing at least one unemployment spell (E(U|U>O)) and iii) the expected unemployment rate for the whole sample (E(U)). Table 4 The quantitative impacts of CEEC trade are only of minor importance. A worker in an industry with exports one standard deviation above average faces a lower unemployment rate of 0.8 percentage points. The corresponding effect for higher imports is 0.1 percentage points higher unemployment (insignificant coefficient). For blue-collar workers a rise in imports by one standard deviation would give rise to higher unemployment by 0.4 percentage points. These simulations are calculated from a cross-section. Ignoring general equilibrium effects which might disturb these simulations, we might calculate an employment balance of the increase in CEEC trade over the period 1988-91: The export ratio increased by 0.020 and imports by 0.008 percentage points calling for an increase in employment of approximately 0.6 percentage points for all workers and an increase of 0.4 percentage points for blue-collar workers. This corresponds remarkably well to aggregate figures (10,000-15,000 workers) estimated by the Austrian Institute of Economic Research (Aiginger, 1993). Furthermore, our results resemble closely those available for U.S. microdata. Kruse (1988) calculates the impact of import growth on unemployment duration of displaced workers using the U.S. Displaced Workers Survey. This corresponds to our term E(U| U>O) in Table 4. A one percent rise in industry import shares is associated with one further week of joblessness over the period 1979-83.12 Using panels of aggregate industry data e.g. Freeman and Katz (1991) as well as Revenga (1992) find significantly larger impacts of import volumes respectively import prices on industry employment. The industry employment figures, however, cannot grasp individual unemployment experience; workers may be lucky to find a new position immediately but they may also remain unemployed for a long time. Note that we consider only the unemployment experience of workers already employed in the manufacturing sector in May 1988 - irrespective of their labour market states later on. We do not consider the recruitment of new workers in manufacturing, who have come from other sectors or have been unemployed or out of the labour force in 1988. Additional exports may ease new hires of the unemployment, whereas imports may impede them.
coefficient is +0.003 for a rise in import shares of 0.6 percentage points. Standardizing the Kruse estimate for the four year period and our chosen change in the independent variable results also in a rise in unemployment rates of 0.3 percentage points (0.006/(52x4)=0.003).
11 Almost all control variables in the Tobit regression are highly significant. As we can only observe job experience back until 1972, we also included age as a regressor. Work experience and job tenure have the expected negative effect on unemployment. On the other hand, older persons have a higher risk of unemployment, which can in part be explained by lower mobility and partly by special and more generous social security legislation for the elderly.13 Persons holding more different jobs in the recent past face a higher risk of joblessness. Whereas blue-collar workers have a longer unemployment experience, the reverse applies to foreigners. This may be caused by a very restrictive employment regulation for immigrants, who may lose the right of residence in Austria in case of long-term unemployment. Industry structure is captured by two variables, firm size and the eight-firm concentration ratio, both measured in 1988. Both firm size as well as industry concentration reduce unemployment risk, output growth is insignificant. Partly to remedy some problems with our measurement of human capital - and partly to cover other aspects unassociated with human capital - we also included monthly earnings in 1988 in the regression. Because of the upper censoring due to social security contribution ceilings we included also an additional dummy. Both variables reveal that high-income earners tend to become less unemployed. 6. Wage Growth Foreign trade shocks are likely to influence not only job opportunities, but also the wage structure. In economies where employees can extract (or share) rents from (with) employers, high growth in exports increases the distributable pie, at least in the short run. So, we would expect that workers in industries with a larger increase in exports will enjoy a higher growth in their earnings. Similarly, higher imports may foster competition in the concerned industries, with adverse consequences for wage increases.14 These arguments should be of particular importance in Austria. The labour force is highly organised and wage determination is a result of industry-wide collective bargaining. Subsequent negotiations at the firm level are possible. This is common in larger firms (which are more likely to be exposed to international competition), and takes the current firm performance into account. Table 5 In order to test these hypotheses, we will analyse wage increases of male, private sector employees between May 1988 and May 1991. Because of the design of the sample, as well as the way earnings data are collected, information on wage growth is only 13Similar 14For
results on displacement risk of the elderly are obtained in Winter-Ebmer (1994). a more rigorous theoretical discussion on this subject, see Freeman and Katz (1991).
12 available for a subgroup of the total sample. On the one hand, employees who had been working in May 1988 may be out of work in May 1991. On the other hand, social security records provide information on wage levels only if an individual earns below the social security contribution ceiling.15 A further problem arises from individual’s behaviour. A worker expecting a wage cut because of negative trade shocks may move to another industry, where labour market prospects are more favourable. This suggests the importance of distinguishing between industry-stayers and industry-changers. Table 5 summarises the various states in which an individual can be found as well as the empirical frequencies in the sample. 6.1. The Sample Selectivity Problem From a methodological point of view, there is a multidimensional selection problem. Confining regression analyses to workers for whom data on earnings growth are available, may lead to biased estimates if workers select themselves (or are selected) non-randomly into the various groups. To account for potential selectivity problems we adopt Lee’s (1983) correction methodology, which is especially appropriate for multiple selectivity criteria. This is a two-stage estimation method similar to Heckman’s (1979) model for the case of dichotomous choice models. This involves the estimation of a multinomial logit model in the first stage,16 which estimates the probability that an individual is in one of several possible states. In a second step, least-squares regressions for subgroups can be run, including a selectivity correction term calculated from the multinomial logit, which accounts for non-randomness in the respective samples. Denote by Zi a vector of variables which determines the selection of individual i into one of the observed categories. The multinomial logit specifies the probability that individual i is found is state j, Pij , by (2) Pij =
exp( Zi b j ) 4
1 + ∑ exp( Zi b j ) j=1
where bj (j=1, ... , 4) is a vector of coefficients determining selection into state j. bo=0 is taken as normalisation. Wage growth of individual i, ∆ ln Wi , is assumed to be determined by a vector of independent variables Yi , and our trade variables in the following way (3) ∆ ln Wij = Ya i j + b1 j ( ∆X / Q) i + b2 j ( ∆I / Q) i + eij
upper limit was ATS 27,147 in 1988 and ATS 29,598 in 1991 for monthly earnings. model is more general and applies also to conditional logit models.
The left-hand-side of equation (3) can be observed only for individuals found in states 1 and 3 (Table 5), whereas information for the remaining states is unavailable. The problem of sample selectivity comes from the fact that the conditional mean of the error term eij of subgroup j may not be equal to zero. If eij is normally distributed then we may write
(4) E(eij j) = ρjσj
( ) +u Φ( H ) φ H ij
= ρjσjλij + uij
where φ and Φ are the density and cumulative density of the standard normal distribution. The selectivity correction term λij is calculated from the results of the multinomial logit with H ij = Φ−1 (Pij ). σj and ρj are the variance of eij and the correlation of eij with unobservable factors determining selection into state j. The remaining error term uij has mean zero but is heteroscedastic with a covariance matrix similar to the binary choice selection case.17 Table 6 Table 6 presents the results from the multinomial logit model. The reference category refers to state 0, that is being out of the labour force in May 1991. For appropriate interpretation, note that from equation (2) we may write
[ ] ∂ log [ P / P ] / ∂Z = b − b . ∂ log Pij / Pio / ∂Zi = b j and ij
This means that a positive coefficient increases the likelihood to be found in state j rather than in the reference state 0 (out of the labour force). To evaluate the impact of some variable on the probability ratio between any two states j and k, the difference between the coefficients is relevant. The results in Table 6 are in line with a priori expectations. So we will comment on them only briefly. Age significantly increases the probability of being out of work. The effect is weaker for high-wage earners (censored wage), but concerns workers irrespective of their mobility. A continuous and stable work history between 1972 and 1988 makes it more likely that an individual is at work in 1991. This is shown by the variables experience and # of previous jobs. Given that a person is at work in 1991, higher job stability in the past also increases the likelihood of holding a job in the same industry. 17For
a detailed description of the estimation procedure, see Greene (1991), pp. 618-622.
14 Compared to experience, the impact of tenure is marginal for less mobile workers. However, low-tenured workers are more likely to have changed industry until 1991. Illness and unemployment in the past raise the probability of being out of work, but strongly reduce mobility. Blue-collar workers and foreigners tend to hold a job paying below the social security limit. Finally, the industry’s mean wage level relative to the economy-wide average is an important determinant of the mobility decision.
15 6.2 Determinants of Wage Growth We now turn to the results of our primary interest, the rate of growth in labour earnings between May 1988 and May 1991 (Table 7). Average (nominal) wage growth over this three-year period accounted for close to 20%, with only a marginal difference between industry-movers and industry-stayers. However, the standard deviation in the dependent variable is about twice as high for changers compared to those remaining within the same industry. This is a first hint of the importance of treating the two groups separately. Table 7 Both groups are significantly affected by the change in CEEC-trade. Among immobile workers, the negative relation between the growth of wages and the CEEC-import ratio is particularly strong whereas the corresponding export-growth effect is small. This may be due to the fact that CEEC-imports took place in formerly protected domestic markets. The fall of trade barriers allowed low-cost CEEC-producers to enter and successively compete with domestic producers. While Austrian exporters were able to penetrate newly opened CEEC markets, competition with producers from other Western countries might have been the reason why the gain in exports did not translate into wage growth at a comparable amount. In terms of absolute magnitude, an individual employed in an industry with a growth in the CEEC-import share of 1 percentage point is expected to have had a growth in earnings 2.9 percentage point (1% p.a.) lower compared to other worker employed in an industry with no change over the period 1988-91. A parallel increase in the export ratio of 1 percentage point leads only to a wage increase of 0.4 percentage point over 1988-91 (0.13% p.a.) Also for workers who took a job in another industry, CEEC-trade growth had a significant impact on the growth in earnings. The picture, however, is somewhat different whereas the growth in CEEC-imports (in the industry of destination) resulted in lower wage growth, the deduction is smaller than the corresponding effect for less mobile workers. An additional 1 percentage point increase in the CEEC-import ratio accounted only for about 1.8 percentage points lower growth in wages between 1988 and 1991. The export effect, on the other hand, turned out to be stronger for more mobile workers. A 1 percentage point-increase in the export-ratio translated into about 0.9 percentage point higher growth in earnings over the three-year period. It is worth noting that these results are in accordance with a priori expectations: Industry changers were able to escape from jobs more vulnerable to import competition, whereas on the average they could take advantage of improved exports. Comparing the above results to the evidence obtained in other studies turns out to be problematic. Most
16 estimates are concerned with U.S. experience and differ substantially from our analysis both with respect to the type of data as well as to methodology; most importantly, they do not make a difference between mobile and immobile workers. Nevertheless, two observations can be made. First, most studies find a significant positive (negative) impact of the export (import-)ratio. This is in conformity with our findings. Second, our results indicate a relatively modest impact of trade on wages compared to U.S. evidence. For instance, according to Gaston and Trefler (1994b) the impact of a 1% growth in the export-(import-)ratio would result in a change in wages of about 4-8% (0-4%). For imports, Macpherson and Stewart (1990) reach a corresponding estimate of up to 1.3% for unionised and up to 0.4% for non-unionised workers. This effect decreases with the degree of unionisation within the industry. Finally, Katz and Freeman (1991) estimate export-(import-) effects between 0 and 7.1% (0 and 5.9%).18 The results for control variables are as expected. Older individuals have lower wage increases. An additional year of age accounts for about a 0.2% smaller rise in earnings. Continuity of work within the period 1988-1991 had a strong impact on the increase in earnings. Industry stayers who were suffering from unemployment or illness were confronted with a reduction in earnings growth of about 8% per year of absence. Interestingly, other reasons of work interruptions did not result in lower wage growth. One reason might be absence due to military service, or temporary layoffs, which allow individuals to return to the same employer at conditions equal to their continuously employed colleagues. More mobile workers with unemployment, or other work interruptions had to accept a significantly lower pay increase. However, absence for health reasons had no significant impact on earnings growth, possibly reflecting asymmetric information in the new employment relationship. Frequent job changes result in lower pay. This is especially true for industry changers, who are the group with the higher turnover. For stayers within the industry, job changes are rare and do not seem to affect earnings growth significantly. One of the most important variables to explain wage growth is the initial wage level. The impact is significantly negative for both groups: there is regression towards the mean. Interestingly this effect is much stronger in absolute terms for more mobile workers. Those who were initially at a very low (high) position in the earnings hierarchy gain most (least) from moving. Wage differentials between white- and blue-collar workers, as well as between natives and aliens have widened. This effect is particularly strong among industry movers. Being employed in a growing firm, or moving to a larger establishment turns out to be favourable for wage growth. Over the period 1988-91 employees in medium-size cities have enjoyed higher pay increases compared to Vienna or more rural
two former studies use individual wage levels as the dependent variable. Freeman and Katz (1991) study the growth of aggregate industry wages. They find markedly different impacts for annual and hourly wages.
17 areas. Firm concentration within industries did not have a strong impact on wage changes. Only among less mobile workers, individuals in highly concentrated sectors had a significantly smaller, but quantitatively minor earnings increase. For mobile workers, the sector of destination matters. Those who joined the service industry fell back in the earnings hierarchy. The average industry wage turns out to be important. Moving from a low- to a high-wage industry leads to a pay premium equal to about 15% of the aggregate industry wage differential. This is further evidence for the hypothesis of non-competitive inter-industry wage differentials (Krueger and Summers, 1988). 7. Conclusions This paper in one of the first to assess detailed labour market consequences of the new trade structures that emerged between Western Europe and the former communist countries in the east. We focus our analysis on Austria. In relative terms, Austria’s trade was affected most heavily compared to other countries by the opening of the Eastern European borders. Extrapolating our results to other Western European Countries who have less trade with the east would thus provide an upper bound of potential labour market reaction. For the Austrian manufacturing sector, we found that the increase in the CEEC-export output ratio has a significantly negative impact on unemployment risk. Whereas there is no significant relation between the increase in the import-ratio for the sample as a whole, blue-collar workers, the elderly, and low-income earners experienced significantly higher risk of unemployment. In all cases, however, the quantitative impact is not very large. Turning to the effects on wage growth we have to distinguish between mobile and immobile workers. Likewise, growing export ratios tend to raise wage growth. Workers who manage to change from one industry to another do better: they can prevent the most serious detrimental consequences of import competition, on the other hand they gain the most from an extraordinary good export performance in the industry. The larger reaction of wages than unemployment to trade changes would Austria put closer to the "American model" than to the "European model" in the characterization of Krugman (1995). To summarise, our results support the view that a further opening of EC markets to the Central and Eastern European Countries would have no disastrous consequences on labour markets in the west. This conclusion has to be qualified, because up to now free trade is limited by restrictions from agriculture to textiles, from deterrent EC antidumping duties to „voluntary export restraints“ (Hamilton and Winters, 1992).
18 Table 1: Austria-CEEC-trade (bn ATS)
4.7 6.0 -1.4
5.0 6.7 -1.7
8.6 6.4 2.2
9.2 7.4 1.7
13.8 11.1 2.7
15.4 12.3 3.1
6.8 6.4 0.5
8.7 7.8 0.8
10.5 8.7 1.7
14.5 11.5 3.0
15.6 12.0 3.6
16.5 10.8 5.7
3.7 4.2 -0.5
5.2 4.4 0.9
4.4 5.0 -0.6
7.5 5.7 1.8
7.1 5.0 2.0
6.4 4.7 1.8
15.2 16.7 -1.4
18.9 18.9 0.0
23.5 20.2 3.3
31.2 24.6 6.6
36.4 28.0 8.4
38.4 27.8 10.6
27.6 19.9 7.7
31.1 21.3 9.8
32.4 23.7 8.7
30.6 24.7 5.9
28.9 23.8 5.1
29.8 23.1 6.7
Czechoslovakia Export Import Balance
Hungary Export Import Balance
Poland Export Import Balance
Central Eastern Europe Export Import Balance
For comparison: AustriaSwitzerland Export Import Balance
Source: Austrian Institute of Economic Research
19 Table 2: Individual Unemployment Rates 1988-91 Results
(asymptotic std. error)
Dependent variable: unemployment rate (%)
(∆X / Q)a,c (∆I / Q) b,c
Number of Previous Jobs 72-88
Foreign Worker (Dummy)
Blue-Collar Worker (Dummy)
log (Firm Size 1988)
City Size ≥ 100.000 100,000 and ≤ 1,000,000 (Dummy)
8 firm concentration ratio in industry 1988
25 Moved into Service Industry (Dummy)
∆ log (Mean Industry
( ∆Q / Q) e
F (10, N-10) e
*** / ** / * significant at 1 / 5 / 10 %-level. a change in exports to CEEC's 1988-1991 as a percentage of output 1988; two digit industry b instrumented by the following 1988 industry variables: union density, share of blue collar workers, share of immigrant worker, CEEC's export (import) share, industry wage level, share of skilled workers in employment, Investment-output ratio, share of energy costs in output, share of R&D expenditures in output. c change in imports from CEEC's 1988-1991 as a percentage of output 1988; two digit industry d growth in output, two-digit industry, 1988-1991 e Hausman-test of exogeneity of import and export shares. The null-hypothesis (exogeneity) is rejected at the 1%-level.
References Aiginger, Karl: Chancen und Gefährdungspotentiale der Ostöffnung, Austrian Institute for Economic Research, Wien, 1993. Aiginger, Karl, Peneder, Michael and Jan Stankovsky: The explanatory power of market based trade theories for the trade between market economies and reform countries, Empirica 21, 1994, 197 - 220. Baldwin, Richard E.: Towards an Integrated Europe, London (Centre for Economic Policy Research), 1994. Beirat für Wirtschafts- und Sozialfragen: Ostöffnung, Wien, 1992 Bhagwati, Jagdish and Vivek Dehajia: Free Trade and Wages of the Unskilled, in: Bhagwati, Jagdish and Marvin H. Kosters (eds.), Trade and Wages, Washington (American Enterprise Institute), 1993. Borjas, George J., Freeman, Richard B. and Lawrence H. Katz: On the Labor Market Effects of Immigration and Trade, in: Borjas, George J. and Richard B. Freeman (eds.), Immigration and the Work Force, Chicago (University of Chicago Press), 1992, 213-244. Brown, Drusilla K.: The Impact of North American Free Trade Agreement: Applied General Equilibrium Models, in: Nora Lustig et al (eds.): North American Free Trade: Assessing the Impact, Washington, D.C. (The Brookings Institution). Cadot, Olivier and Jaime de Melo: France and the CEECs: Adjusting to another Enlargements, mimeo, INSEAD, paper presented to a CEPR Conference, Brussels, 1994. Deardorff, Alan and Dalia Hakura: Trade and Wages: What are the Questions, in: Bhagwati, Jagdish and Marvin H. Kosters (eds.), Trade and Wages, Washington (American Enterprise Institute), 1993. Freeman, Richard B. and Lawrence F. Katz: Industrial Wage and Employment Determination in an Open Economy, in: Abowd, John M. and Richard B. Freeman (eds.), Immigration, Trade and the Labor Market, Chicago (University of Chicago Press), 1991, 235-260. Gaston, Noel and Daniel Trefler: The Role of International Trade and Trade Policy in the Labour Markets of Canada and the U.S., The World Economy 17, 1994a, 45-62. Gaston, Noel and Daniel Trefler: Protection, Trade and Wages: Evidence from U.S. Manufacturing, Industrial and Labor Relations Review 47, 1994b, 574-593. Greene, William H., LIMDEP Version 6.0, A User’s Manual, Bellport NY (Econometric Software Inc.) Grossman, Gene M.: Imports as a Cause of Injury: The Case of the U.S. Steel Industry, Journal of International Economics 20, 1986, 201-223. Grossman, Gene M.: The Employment and Wage Effects of Import Competition in the United States, Journal of International Economic Integration 2, 1987, 1-23. Hamilton, Carl B. and Alan L. Winters: Opening up International Trade with Eastern Europe, Economic Policy 14, 1992, 77-115. Heckman, James J., Sample Selection Bias as a Specification Error, Econometrica 47, 1979, 153-161. Holzmann Robert, Thimann Christian, and Angela Petz: Pressure to adjust: Consequences for the OECD Countries from Reforms in Eastern Europe, Empirica 21, 1994, pp141 - 196
27 Katz, Lawrence F. and Ana Revenga: Changes in the Structure of Wages: The United States vs. Japan, Journal of the Japanese and International Economies 3, 1989, 522-523. Kohler, Wilhelm: Multilateral Trade Liberalization: Some General Equilibrium Simulation Results for Austria, Empirica 18, 1991, 167-200. Krueger, Alan B. and Lawrence H. Summers: Efficiency Wages and the Inter-Industry wage Structure, Econometrica 56, 1988, 259-293. Krugman, Paul: Growing World Trade: Causes and Consequences, Brookings Papres on Economic Activity 1995, 327-362. Kruse, Douglas L.: International Trade and the Labor Market Experience of Displaced Workers, Industrial and Labor Market Review 41, 1988, 402-417. Landesmann, Michael, Snell, Andrew, Structural shifts in the manufacturing export performance of OECD economies, Journal of Applied Econometrics 8, 1993, 149 162. Lawrence, Robert Z. and Matthew J. Slaughter: International Trade and American Wages in the 1980s: Giant Sucking Sound or Small Hickup, Brookings Papers on Economic Activity Microeconomics 1993, 161-226. Lee, Lung-Fei, Generalized Econometric Models with Selectivity, Econometrica 51, 1983, 507-512. MacPherson, David A. and James B. Stewart: The Effect of International Competition on Union and Nonunion Wages, Industrial and Labor Relations Review 43, 1990, 434-446. McKelvey, R. and W. Zavoina: A Statistical Model for the Analysis of Ordinal Level Dependent Variables, Journal of Mathematical Sociology 4, 1975, 103-120. Murphy, Kevin M. and Finis Welch: The Role of International Trade in Wage Differentials, in: Kosters, Marvin H. (ed.), Workers and their Wages, Washington D.C. (American Enterprise Institute), 1991. Peneder, M., Kosten- und Produktivitätsvergleich der Industrie in den Ländern OstMitteleuropas, in: Aiginger, 1993. Revenga, Ana: Exporting Jobs: The Impact of Import Competition on Employment and Wages in U.S. Manufacturing, Quarterly Journal of Economics 107, 1992, 225284. Smith, Richard I. and Richard W. Blundell: An Exogeneity Test for a Simultaneous Equations Tobit Model, Econometrica 54, 1986, 679-685. Veall, Michael R. and Klaus F. Zimmermann: Goodness of Fit Measures in the Tobit Model, Oxford Bulletin of Economics and Statistics 56, 1994, 485-499. Winter-Ebmer, Rudolf: Firm Size, Earnings, and Displacement Risk, WP #9418, University of Linz, 1994. Winter-Ebmer, Rudolf and Josef Zweimüller: Immigration and the Earnings of Young Native Workers, Oxford Economic Papers, forthcoming. Winter-Ebmer, Rudolf and Josef Zweimüller: Do Immigrants Displace Native Workers? The Austrian Experience, CEPR WP 991, London, 1994. Wood, Adrian: How Trade Hurt Unskilled Workers, Journal of Economic Perspectives 9, 1995, 57-80. Wood, Adrian: The Factor Content of North-South Trade in Manufactures Reconsidered, Weltwirtschaftliches Archiv 127, 1991, 719-743.