CHANGES IN THE STRUCTURE O F EARNINGS DURING A PERIOD OF RAPI D TECHNOLOGICAL CHANGE :

Transcription

1 CHANGES IN THE STRUCTURE O F EARNINGS DURING A PERIOD OF RAPI D TECHNOLOGICAL CHANGE : EVIDENCE FROM THE POLISH TRANSITIO N Michael P. Keane Yale University Eswar S. Prasad International Monetary Fund The National Council for Eurasian and East European Researc h th Street, N.W. Suite 300 Washington, D.C TITLE VIII PROGRAM

2 Project Information* Sponsoring Institution : Principal Investigator: University of Minnesot a Michael P. Keane Council Contract Number: g Date : September 3, Copyright Informatio n Individual researchers retain the copyright on their work products derived from research funde d through a contract or grant from the National Council for Eurasian and East European Research (NCEEER). However, the NCEEER and the United States Government have the right to duplicat e and disseminate, in written and electronic form, reports submitted to NCEEER to fulfill Contract o r Grant Agreements either (a) for NCEEER's own internal use, or (b) for use by the United State s Government, and as follows : (1) for further dissemination to domestic, international, and foreig n governments, entities and/or individuals to serve official United States Government purposes or (2) for dissemination in accordance with the Freedom of Information Act or other law or policy of th e United States Government granting the public access to documents held by the United State s Government. Additionally, NCEEER may forward copies of papers to individuals in response to specific requests. Neither NCEEER nor the United States Government nor any recipient of this Report may use it for commercial sale. The work leading to this report was supported in part by contract or grant funds provided by the National Council fo r Eurasian and East European Research, funds which were made available by the U.S. Department of State under Title VIII (The Soviet-East European Research and Training Act of 1983, as amended). The analysis and interpretations contained herein are those of the author. ii

3 Abstract Poland experienced a dramatic structural change in that has become known as the "bi g bang." The subsequent economic transition was a period of massive changes in production structures an d technologies, providing an interesting testing ground for theories of the relationship between technica l change and wage inequality. Using micro data on individual workers' earnings over the period , this paper documents various aspects of changes in cross-sectional earnings inequality during the Polis h transition. While education premia rose sharply during the transition. experience premia declined. especially in the early years of transition. These findings are consistent with the notion of technologica l obsolescence driving down the returns to specific human capital and increasing the returns to genera l human capital. As in the case of industrial countries such as the U.S. and the U.K. over the last two decades, changes in within-group inequality account for about two-thirds of the increase in overall wag e inequality. However, in contrast to those countries, increases in within-group inequality in Poland ar e very different across skill groups. In particular, increases in within-group inequality appear to have bee n larger for more highly-educated workers. Refinements to existing theoretical models are proposed t o explain these findings. We present a simple model under which the limited information conveyed by employment histories can keep the returns to unobserved skill attributes relatively compressed in a transition economy. iii

4 Introduction Poland has experienced a dramatic change in its political and economic structures over the las t decade. This transition was launched with a radical set of reforms in that has become known a s the "big bang." The private sector's contribution to GDP rose sharply after the transition commenced and state enterprises. which were progressively cut off from subsidies, were forced to change their productio n processes and technologies. Enterprises that were unable to attract new capital to finance suc h modernization rapidly became uncompetitive and were forced to shut down. Thus, the process o f transition to a market economy has been accompanied by marked changes in the structure of productio n and also by massive technological change. The sharp rise in wage inequality m countries such as the United States and the United Kingdo m in the 1980s and early 1990s has produced a spate of theoretical models that have attempted to explai n this phenomenon as the outcome of skill-biased technical change, coupled with capital-skill complementarity. Since it involved a period of rapid technical change, the transition in Poland provide s an interesting testing ground for such theories of the relationship between technological change and wag e inequality. In this paper, we examine the evolution of the structure of labor earnings in Poland over th e period using micro data from the Polish Household Budget Surveys. The relatively long span of the dataset allows us to trace out changes in the wage structure for an extended period both leading u p to and following the "big bang." We find that overall wage inequality rose significantly during th e transition period of For instance, we estimate that the percentile ratio for individua l labor earnings increased steadily from 2.64 in 1988 to 3.07 in 1996 (using a sample of individuals aged for whom labor earnings is the primary source of income). The increase in male wage inequality was even greater, with the percentile ratio rising from 2.56 in 1988 to 3.16 in We also conduct a more detailed examination of the sources of the increase in wage inequality. Prior to the transition, the wage structure in Poland was highly compacted_ with wages of college - educated white-collar workers little different from those of manual workers. We find that income 1

5 differentials by education level increased rapidly during the transition, reflecting sharp increases in education premia. But the premium for labor market experience declined in the early years of transitio n and the position of older workers deteriorated relative to younger workers, consistent with the notion o f rapid obsolescence of industry- and firm-specific skills of older workers during a period of industria l restructurin g We find that a substantial fraction of the increase in overall earnings inequality is attributable to changes in within-group inequality. Interestingly, however. increases in within-group inequality are mor e concentrated among workers with higher levels of formal education. This is quite different from th e patterns documented in the U.S. and the U.K. of sharp increases over the last two decades in between - group inequality as well as within-group inequality for virtually all broadly-defined groups ( see. e.g., Juhn. Murphy and Pierce, Machin. 1996, Machin and van Reenen, 1998). In these countries, in contrast to our observations for Poland, both education and experience premia increased, and inequalit y within all education categories and groups with different levels of experience increased as well, particularly during the 1980s. A popular explanation for the increase in education and experience premia is skill-biased technical progress. But this cannot explain increases in inequality within skill groups. To rationalize th e U.S. and U.K. evidence, Violante (1996) proposes a model in which technology is embodied in capital of different vintages. If there are frictions in the matching of workers with the new technology, an acceleration of the rate of technical progress then leads to increased wage dispersion among identica l workers. Aghion and Commander (1999) have argued that transition can be viewed as generating a technology gap between the old state sector and the new entrepreneurial sectors. Growth in within-group inequality can then be explained by frictions in the matching of workers with the new technology, as i n Violante's (1996) framework. Clearly, however, this framework needs to be modified to account for th e fact that increases in within-group inequality in Poland are correlated with education levels. We develop a simple model that allows for capital-skill complementarity but where employmen t histories play an important role in signaling the quality of unobserved worker attributes. Given the

6 limited length of workers` relevant employment histories in the new market economy, the returns to unobserved attributes are. therefore. relatively compressed in a transition economy. The correlation between observed and unobserved skill attributes leads to greater increases in within-group inequality among groups with better observed skill attributes, such as higher education levels. The model is very similar in spirit and structure to Galor and Moav's (2000) model of ability-biased technological change, which those authors use to explain the concomitant rise in between- and within-group inequality in the U.S. The next section of the paper provides a brief overview of existing literature on wage inequalit y in Poland and other transition economies. Section III describes the dataset and Section IV contains th e main empirical results. Section V presents the theoretical model and concluding remarks are in Sectio n VI. Review of prior research Most earlier work on the Polish earnings distribution has relied on aggregate statistics that ar e released annually by the Polish Central Statistical Office (CSO). These aggregate statistics are described in detail in Atkinson and Micklewright (1992). Each September, starting in 1981, the CSO conducted a census of enterprises, and "the information requested of the enterprise was the total persons in a numbe r of discrete earnings bands." The CSO then published aggregate statistics such as total numbers o f employees in various earnings bands. and deciles of the earnings distribution. From this data, it i s possible to construct approximate measures on earnings inequality, such as approximate Gini coefficients. A number of other countries, such as Hungary, the former Czechoslovakia, and the former U.S.S.R., had similar data collection and reporting procedures for earnings. Using such data, Atkinson and Micklewright (1992) compare the degree of earnings inequalit y across several communist countries in They obtain Gini values for Czechoslovakia (1987), Hungary, Poland and the U.S.S.R. of 0.197, 0.221, and 0276, respectively. And they report 9-1 3

7 decile ratios of 2.45, 2.64, 2.77 and 3.28, respectively. ' Thus, there was a clear ranking of inequality (consistent across both measures), with Czechoslovakia being the most equal. the U.S.S.R. the least equal. and Poland in the middle. As a point of comparison. these authors report a Gini coefficient and 9-1 decil e ratio of and respectively, for the U.K. in Thus, the earnings distribution in Poland prio r to the transition was noticeably more compact than in the U.K. Based on the same data Atkinson and Micklewright (1992) also calculate that earnings inequalit y in Poland fell significantly over the period They report Gini values of , and for these years. and decile ratios of and Rutkowski (1996a) uses the same September earnings distribution survey to examine changes i n the Polish earnings distribution during the transition. His calculations indicate that earnings inequality jumped dramatically in the early phase of the transition, with the Gini and decile ratio rising to and 2.86 in By the last year of his study, these had risen further, to and respectively. Rutkowski also reports that earnings inequality was much greater in the private sector than the publi c sector in 1993, that the ratio of white collar to blue collar wages rose substantially in the transition, and that this ratio was much higher in the private sector. Rutkowski (1998) extends this analysis to 1995, by which time the Gini for earnings increased to and the decile ratio to 3.38, a large increase in inequality over Recall that. for Atkinson and Micklewright (1992) reported a Gini of and a decile ratio of So the increase in (gross) earnings inequality for Poland from implie d by these data is a bit greater than what they report for the U.K. in the 1980s. 2 Rutkowski (1996b) presents a cross-country analysis of changes in earnings inequality usin g similar data sources for several transition economies. These results indicate that, by 1993, the earning s distribution for Czechoslovakia had become very similar to that for Poland, while the Gini and decil e ratio for Hungary had risen to and 3.67 (more unequal than Poland). Thus, given the baseline The 9-1 decile ratio is based on the average earnings of individuals in the top and bottom deciles of the earning s distribution. 2 Note that the data are gross income for both Poland and the U.K. 4

8 figures from Atkinson and Micklewright (1992), it appears that both Czechoslovakia and Hungar y experienced larger increases in earnings inequality from 1986 to 1993 than did Poland. As both Atkinson and Micklewright (1992) and Rutkowski (1996) describe, there are a number of limitations of the September earnings survey data. First, the aggregate nature of the data may lead t o approximation errors in inequality measures, and limits the type of analysis that can be performed. Second. the coverage of establishments is incomplete because small firms (i.e., less than 6 employees) are not sampled. This is especially a problem for the transition in Poland because, according to OECD (1998. p. 107), "Poland's recent growth performance rests on a strong entrepreneurial basis, with many dynami c small and medium-sized enterprises (SMEs) and creations of new firms... SMEs make up the bulk of Poland's 2.2 million registered non-agricultural enterprises... almost 90 percent [are] micro-enterprises (employing 1 to 5 persons)." Third, this data does not account for in-kind payments, which have been important in Poland. Fourth, this survey reports gross earnings. This creates comparability problem s over time, because a progressive income tax (with rates up to 45 percent) was introduced in Failur e to account for this will tend to exaggerate the measured increase in inequality of net earnings. Another survey that has been used by some authors to examine recent changes in the Polish wag e structure is the Polish Labor Force Survey that was introduced in 1992 (see, e.g., Newell and Socha, 1998). However, this survey is clearly not useful for understanding changes in wage inequality in th e crucial early years of transition or changes relative to the pre-transition wage structure. Existing work on other transition economies has mostly focused on the early years of transition. For instance, Orazem and Vodopivec (1995) analyze micro data from Slovenia and report that, from to 1991, wage inequality increased significantly, with returns to both education and experience rising ove r this period. Using grouped data, Flanagan (1993) finds that the returns to education rose while the return s to experience declined in the Czech Republic in the initial phase of transition. Based on her analysis of Russian data, Brainerd (1998) reports that, from J991 to 1994, the marginal return to a year of educatio n rose from 3.1% to 6.7% for men and from 5.4% to 9.6% for women. Using data from the ILO's seldom- 5

9 used Occupational Wages around the World file. Freeman and Schettkat (2000) also find that overal l earnings inequality and skill differentials increased in transition economies during the 1980s and 1990s. The datase t The CSO has been collecting detailed micro data on household income and consumption at leas t since using fairly sophisticated sampling techniques. In the HBS, the primary sampling unit is the household. A two-stage geographically stratified sampling scheme is used, where the first-stage samplin g units are the area survey units and the second-stage units are individual households. The typical sample size is about 25,000 households per year. The CSO uses the data obtained from these household survey s to create aggregate tabulations that are then presented in their annual Statistical Bulletins. or Surveys. The HBS contains detailed information on sources and amounts of income both for households and individuals within each household. Total income is broken down into four main categories : labo r income (including wages, salaries and nonwage compensation) ; pensions; social benefits and other transfers; and other income. Social benefits include income from unemployment benefits, which were introduced in late A key point is that the data include measures of the value of in-kind payment s from employers to workers. which have been an important part of workers' compensation in Poland an d other transition economies. There were no taxes on personal income until After that year, we us e net incomes in the analysis. The HBS also contains information on demographic characteristics of al l household members and on labor earnings of all employed individuals in each household. Unfortunately, data on individual workers' earnings were not collected in Hence, the dataset in fact goes fro m and then from The HBS does include a panel element a part of the dataset contains households that are surveyed for four successive years before being rotated out of the sample. However, attrition bias in th e panel is significant and, in addition, the panel was changed completely a couple of times over the period 3 Until 1992, some firms were levied an "excess wage tax," essentially a payroll tax imposed on part of a firm's tota l wage bill. The actual incidence of this tax is, of course, a complicated matter. 6

10 To maintain the representativeness of the sample and to use all of the information in the dataset, we treat the data as a repeated set of cross-sections. The structure of the survey instrument and the sampling scheme were both kept essentially unchanged after the transition commenced. However, one major change was introduced in 1993 that ha s important implications for analyzing cross-sectional inequality. In order to improve survey response rates, in 1993 the CSO switched from quarterly to monthly data collection for the HBS. Since earning s are more variable at the monthly than the quarterly frequency, this shift alone would have created a substantial increase in measures of cross-sectional earnings inequality. Indeed as we demonstrate in th e next section, failure to account for this change in survey frequency has quantitatively important effects o n measures of earnings inequality. In the Appendix, we develop a technique for adjusting the earnings data for th e increased variability that may be attributable to the shift from quarterly to monthly reporting. Our approach models earnings as the sum of a permanent or predictable component (determined by workers ' education, age and other observable characteristics) and a mean zero idiosyncratic component. We then assume that the variance of the idiosyncratic component would not have jumped abruptly after the fourt h quarter of Rather, we assume that the variance of idiosyncratic earnings varied smoothly over tim e (measured in months) according to a polynomial time trend. We estimate this polynomial trend alon g with a dummy for post-i992 that captures the discrete jump in variance that occurred with the change to monthly earnings reporting. Then, at the individual level, we scale down the idiosyncratic component o f the post-1992 earnings statistics to eliminate this jump in variance. Our procedure for adjusting for the spurious increase in inequality stemming for the switch to th e monthly reporting interval relies on access to the HBS micro data. In particular, the variance correctio n requires access to the data for an extended period of time. Our study is unique in that it is based on th e HBS micro data for a long sample period extending from 4 years prior to the "big bang" to 7 years after. 7

11 To our knowledge, no prior study of earnings inequality in Poland has adjusted for the change in surve y design in We restrict our wage analysis sample to individuals between the ages of 18 and 60 who repor t that labor income is their principal income source. Nominal monthly earnings are deflated by th e aggregate CPI (1992Q4=100). The results we report in the paper are mostly annual. But. to avoid any distortions arising from the sharp changes in price levels that occurred in the early years of transition, w e deflate nominal wages using CPI data for the survey quarter until the end of 1992 and for the survey month thereafter. Prior to 1993, there were seven education categories reported for individuals in the survey. Beginning in 1993, two of these categories (basic vocational training and some high school) were combined into a single category; for consistency, we combine these categories in a similar manner for the period. We also combined the two categories primary school and less than primary school into a single base category (since the less than primary school group among workers is quite small), thereb y yielding a total of five educational categories over the full sample. Table 1 (at the end of this paper) reports sample means for some of the variables used extensivel y in our analysis of wage inequality. 5 The demographic characteristics of the cross-sectional sample s remain quite stable during and after the transition. The means of the education dummies indicate a small increase in avera ge levels of educational attainment in the 1990s. The distribution of employment among men and women is relatively stable, although there is a slight increase in the share of women in tota l 4 At the time we began our study, the Polish CSO had never before released the HBS micro data. Subsequently, th e micro data for the first half of 1993 was released to the World Bank, and this data is used in World Bank (1995) an d Milanovic (1998). More recently, data for have been obtained by researchers at the World Bank. A subsample of the HBS is also now available in the Luxembourg Income Survey (LIS) for 1987, 1990 and Thus, no prior researchers have had access to the micro data for the entirety of the extended period that we examine. Note that the sample size falls in In that year, half of the total sample was used to test the new monthl y survey; these data were considered unreliable and not made available to us. In addition, a number of observations i n that year had to be left out of the analysis due to incomplete data. Cross-sectional sampling weights (at th e household level) are used in all of the analysis to ensure representativeness of the sample in each year, including

12 employment after Private enterprises accounted for less than 10 percent of total employment befor e the transition but this proportion had grown to about 40 percent by Earnings Inequality In this section, we examine the evolution of earnings inequality for individual workers. For the years we use earnings measures that are adjusted (using the procedure described in th e Appendix) for the increase in idiosyncratic variance that occurred with the shift to a monthly reportin g period..measures of overall inequalit v The first panel of Table 2 reports and percentile differentials of log earnings for al workers. Wage inequality is quite stable in the pre-transition years Between 1988 and 1989, the differential increases sharply, going from 0.97 to J.04. A further significant increase occurs from 1991 to 1992, followed by a moderate increase through The total increase in inequality from to 1996, as measured by the differential, is about 15 percent_ a sizeable increase over an 8-year period. The increase in the percentile differential is also quite significant_ from 0.50 in 1988 t o 0.59 in Even by 1996, however. the wage structure is considerably more compressed than in the United States. For instance, in 1991, the differential for full-time workers in the U.S. was close to (Gottschalk and Smeeding, 1997). Interestingly, however, wage differentials are greater in Poland than i n certain continental European countries such as Germany in the 1990s (see Prasad, 2000). In Table Al, we present an alternative measure of inequality the Gini coefficient. We prefer not to focus on this coefficient, despite its usefulness as a summary measure of inequality, since it tend s to be more sensitive to outliers than the percentile differentials and is also sensitive to changes around the median/mode of the distribution. Nevertheless, since this measure is used by numerous authors, we sho w the estimated Gini coefficients to enable comparisons with other studies. It will be evident that th e 9

13 patterns of changes in inequality revealed by the evolutions of the percentile differentials are, in general, quite similar to those indicated by the coefficients. For instance, the Gini coefficient for the ful l sample rises sharply from in 1988 to in 1994 and then remains flat through As discussed earlier, we adjusted the earnings data for to account for the change i n survey frequency. In the second panel of Table 1, we show the percentile differentials for wit h unadjusted data. Clearly, the adjustment makes a significant difference to the absolute level of inequality, although the profile of little change in inequality over the period is not affected by whether or not the data are adjusted. In terms of understanding changes in inequality over the full sample, however, adjusting for the change in survey frequency is clearly quantitatively important. The first four columns in the lower panels of Table 2 show percentile differentials separately for workers employed in the public and private sectors. In absolute terms, inequality is higher in the privat e sector than in the public sector. The change in inequality is also much greater in the private sector. For instance, from 1988 to 1996, the differential rises from 0.96 to l.05 in the public sector and fro m 1.04 to l.19 in the private sector. However, the change in the differential is similar across the tw o sectors, suggesting that increased inequality at the upper and lower ends of the distribution, rather than i n the middle of the distribution, is responsible for the greater increase in inequality in the private sector. To obtain a visual representation of changes in the shape and features of the entire distribution, we also examine kernel density estimates of the overall earnings distribution as well as the wage distributions in the public and private sectors. Figure 1 presents kernel density estimates for real earnings for the years 1988, 1992, 1994 and Adopting an approach used by Fortin and Lemieux (2000), we construct weighted kernel density estimates for the two sectors, where the weights are the respectiv e sectoral shares in total employment for each year. Based on their analysis of data from the Labor Force Survey, Newell and Socha (1998) also find little change i n wage inequality between 1992 and An Epanechnikov kernel with a bandwidth of 4000 was used for the kernel density estimation. 1 0

14 This figure indicates the rising importance of the private sector in determining the shape of th e overall wage distribution. In addition, it appears that the mean of the private sector wage distribution is t o the right of that of the public sector wage distribution in This position seems to be reversed b y Figure 2 (left panel), which plots the differential between (unconditional) median private an d public sector wages, confirms this point. Interestingly, the private-public sector differential was positive before the transition but falls sharply and turns negative after the early years of transition. s To sum up, although the increase in wage inequality is higher in the private sector, workers i n public sector enterprises have also experienced significant increases in wage inequality during th e transition. In other words, the rapid reallocation of labor from the public sector to the private sector can not by itself explain much of the increase in wage inequality during the transition in Poland. Within-group wage Inequalit y Next, we examine changes in within-group inequality. In the U.S. and the U.K., within-group inequality has been documented to have risen sharply within virtually all broadly-defined educationa l groups. The relative importance of between- and within-group inequality in accounting for changes i n overall inequality also has important implications for different sets of theoretical models. In Table 3, we report percentile differentials for workers in each of five educational groups. One feature of these results is that. prior to the transition. both the 90-I0 and percentile differentials were not too dissimilar across these groups. Interestingly, there is a significant and sustained increase i n wage inequality from 1988 to 1996 for two groups those with college degrees and those with hig h school degrees. Workers with college degrees experience by far the greatest increase in inequality durin g the transition, with the percentile differential rising from 0.96 in 1988 to 1.14 in 1996 and the differential going from 0.50 to 0.62 over the same period. For workers with only a primary school degree Earnings data from the Labor Force Survey reveal a similar pattern of a sharp decline in this differential during th e transition. Newell and Socha (1998) report a mean private sector wage premium of 5 percent in 1992, falling to a negative differential of 10 percent by

15 or with basic vocational training, the increases in within-group inequality are much more modest. The changes in the differentials for workers with vocational training or a primary school degree are les s than half of the corresponding differential for workers with a college degree. The difference is eve n greater for the differential. Thus, increases in within-group inequality seem to be a prominen t feature of the transition mainly for highly-educated workers. A possible explanation for these results is that workers with higher education levels are mor e likely to be employed in the private sector, where inequality is higher. In fact, the private secto r employment shares of all education groups have risen sharply during the transition. Somewhat surprisingly, the group-specific share of employment in the private sector is even higher for workers wit h lower levels of education. Hence, differential patterns of reallocation of labor across the public an d private sectors can not explain observed differences in increases of within-group inequality. Next, we examine the evolution of inequality within broadly-defined experience groups. Table 4 reports percentile differentials for groups of workers with different experience levels. There are fairl y significant increases in inequality for all groups. For instance, from 1988 to 1996, the differential rises by about 0.2 for all experience levels. Over the same period. the differential rises by about 0. l for all experience groups. We also examined cohort effects on the evolution of overall inequality. Typically, inequality tends to rise over the life cycle within a given cohort as employment histories, cumulative effects o f individual productivity shocks and other factors drive up within-cohort wage dispersion over time. Thus, differences in cohort sizes could result in changes in overall inequality even if within-cohort inequality were to evolve in an identical pattern across cohorts. The lower panel of Table 4 reports percentile ratios for cohorts defined on the basis of birth year. All cohorts experienced fairly substantial increases in inequality during the transition. The patterns o f 9 The group-specific shares of private sector employment in 1988 and 1996 for each of the education groups are a s follows: college degree (2.9,19.4) ; some college (5.3,20.3) ; high school (3.4,33.9) ; vocational training (6.6,48.3) ; and primary school (3.8,44.7). 1 2

16 changes are, however, quite similar across cohorts, but for one notable exception. The group of younger workers, born between the years of 1966 and have lower levels of inequality before the transitio n compared to other cohorts but this cohort experiences a much greater rise in inequality during th e transition, with the percentile differential rising from 0.75 in 1988 to I.07 in I996. One possibility is that this relatively young cohort experienced the greatest dispersion of employment opportunitie s during the transition. Consistent with this notion. we find that the share of private sector employmen t (relative to total employment) within the cohort is much greater for this cohort than for the older ones. 1 0 Since, as noted earlier, wage inequality tends to be greater in the private sector, it is not surprising tha t this cohort experiences a much greater increase in inequality during the transition. The cohort that experiences the next largest rise in inequality in the early years of transition (through 1992) is the birth-year cohort, which includes workers who were relatively old by the time the transition began. Thus, inequality increases during the transition seem to have been largest at the upper and lower ends of the ag e and experience distributions. Another approach to examining within-group wage inequality, suggested by Juhn. Murphy an d Pierce (1993), is to regress earnings on observed attributes such as gender, education. and experience and to examine the dispersion of the wage residuals. These wage residuals arguably control for between - group differences across many different group characteristics and indicate the evolution of inequalit y within narrowly-defined groups. Percentile differentials for log wage residuals are reported in the last (top right) panel of Tabl e 2. 1 k These differentials indicate that about two-thirds of the level of overall wage inequality is accounte d for by within-group inequality. Changes in within-group inequality also account for a substantial fractio n ' The share of total employment in each cohort accounted for by private sector employment in (average) i s as follows : 0.21 ; : 0.24; : 0.30; : 0.34; : 0.48; We do no t report results for the birth-year cohort before 1992 since the cell sizes are too small (less than 10 0 observations). ' Wage residuals are from annual OLS regressions of earnings on a constant, four education dummies, experienc e and its square, and dummies for gender, urban residence, and employment in the private sector. These are identical to the specifications examined in greater detail in the next section. 1 3

17 of the increase in overall inequality. For instance, the percentile differential based on the wag e residuals goes from 0.82 in 1988 to 0.91 in which accounts for about two-thirds of the increase i n total wage inequality over that period. This is similar to the experience of the U.S., where the change in residual inequality accounts for more than two-thirds of the total increase in wage inequality during the 1970s and 1980s (Juhn, Murphy and Pierce. 1993). To summarize. the evidence in this sub-section indicates that similar to the experiences of the U.S. and other industrial economies, within-group inequality accounts for only a small portion of the total change in earnings inequality duri n g the Polish transition. However, there are some important difference s across groups. In particular. within-group inequality did rise significantly among highly-educated workers but increased far less for workers with lower levels of formal education. Increases in inequality seem to have been quite similar within broadly-defined experience groups. Wage inequalitv among men and women Next, we examine the evolution of wage inequality among men and women. This could have important implications for interpreting labor market developments. Much of the literature on wag e inequality has traditionally tended to focus on male wage inequality. This has largely been driven by th e desire to carefully examine a homogeneous sample of workers and to abstract from the analytical complications resulting from the relatively more tenuous labor market attachment of women in industria l countries. But, in a country like Poland, where women have traditionally had very high employment an d participation rates, it is of considerable interest to analyze these two groups separately and also t o examine their respective impacts on the overall earnings distribution. The bottom right panels of Table 2 show percentile differentials separately for men and women. As in many other countries, the increase in male wage inequality is significantly greater than the increas e in inequality among women. For instance, from 1988 to 1996, the percentile differential goes fro m 0.94 to l.15 for men and from 0.85 to l.00 for women. Similarly, the increase in the differential from is 0.12 for men and 0.07 for women. It is worth noting that the increase in male wag e 1 4

18 inequality is greater than the increase in overall inequality, a phenomenon that has been documented fo r numerous other countries as well during the 1980s and 1990s (see. e.g., Fortin and Lemieux. 2000, fo r U.S. evidence). Figure 3 presents employment-share weighted kernel density estimates for the real earnings o f men and women for the years 1988, and We defer a more formal analysis of th e gender wage gap until the next section but it is evident from this figure that, throughout the period the female wage distribution is to the left of the male wage distribution at virtually all percentil e points of the respective distributions. It is also hard to discern any tendency for the two distributions to conver ge over time. Figure 2 (right panel) shows that the median wage differential between men and women fell during the early stage s of transition but stabilized at around 25 percent after A similar pattern in true in the private sector. This is in sharp contrast to the experiences of many industrial countries, where the gender wag e differential has declined and the wage distribution for women has converged markedly towards the wag e distribution for men over the last decade or two. Evidence from other transition economies is more in line with that for industrial countries. Fo r instance, Hunt (2001) reports that German unification resulted in a 10 percentage point decline in th e gender wage gap in the former East Germany but notes that this was accompanied by a much greate r decline in female employment than in male employment. She also finds that low earners were mor e likely to leave employment after unification and that such persons were disproportionately female. In Poland, by contrast, our evidence suggests that the relative share of women in total employment actuall y increased after the transition (see Table I). It is also worth noting that the relative employment share of women, compared to many industrial countries, was quite high even before the transition. Thus, there appear to be very different forces at work in Poland in terms of influences on the gender wage gap. This issue is explored further below. 12 An Epanechnikov kernel with a bandwidth of 4000 was used for the kernel density estimation. 1 5

19 Effects of changes in structure of employment on wage inequalit y Since resource allocation in a command economy is typically not determined by market forces, the opening up of the transition economies to competitive pressures. both internally and externall y (through the expansion of trade with other market economies), has resulted in marked changes in th e structure of production and employment. An important question at this juncture is whether such change s in the industry shares of employment could be contributing to changes in overall wage inequality. This could occur due to changes in both the within-industry and between-industry components of inequality. Following Juhn. Murphy and Pierce (1993), consider the following decomposition: Q,` E S Jr QJr + E s, (w/r W, ) z (1 ) J where 6r, is the cross-sectional variance of log hourly wages, s it is the employment share of sector f, 2 j, is the within-industry variance of wages, w j, is the mean sectoral wage, w, is the mean wage in the sample and the subscript t is a time index. Using this formula, the change in variance over time can b e decomposed into changes attributable to within- and between-industry variance as well as composition effects within and between industries. Table 5 presents the results of this decomposition for 13 broadly - defined sectors of the economy. ' The total increase of 7.00 in overall wage variance from 1988 to 1996 is the result of sharp increases in wage variance from 1988 to 1992 and from 1992 to 1994, with a subsequent moderate increase from 1994 to The key result from this table is that virtually all of the increase in overal l wage variance is attributable to within-industry changes in wage variance. The between-industry 'j The disaggregation used here is relatively coarse. Unfortunately, a change in industry classification in vitiated our attempts to derive a consistent set of codes at a finer level of disaggregation for the full sample. '4 Note that the numbers in the table are multiplied by I00 ; the absolute increase in variance over the full sample i s actually quite small. 1 6

20 component of the change in variance is significantly positive in all three sub-periods, but is roughly offse t by within- and between-industry composition effects. Table 5 also presents (in the footnotes) employment shares of each of the 13 sectors in 1988 an d Clearly, there have been significant changes in industry employment shares over this period. Nevertheless, these shifts in sectoral employment do not seem to have played much of a role i n influencing patterns of overall wage dispersion. Within-industry wage variation appears to dominat e overall wage variation and both appear to have evolved in a smaller pattern. Earnings inequality at the household leve l We have characterized various aspects of the evolution of individual earnings inequality thus far. A related question is how earnings inequality changed at the household level. This is a topic of interest in its own right (see, e.g., Burtless, 1999). Although a full-fledged analysis of the determinants of household earnings inequality is beyond the scope of this paper, it is nevertheless interesting to examin e the dispersion of both labor income and total income at the household level for households that repor t labor income as their primary source of income. To make these measures comparable across households, we adjust both income variables by a simple equivalence scale. 1 5 Table 6 reports Gini coefficients and percentile differentials for equivalence-scale adjuste d measures of total household income and household labor earnings. Note that income data at th e household level are available for 1993 and 1997 as well. The level of household labor earnings inequality is greater than that of individual earnings (compare with Table 2), which is to be expected given th e cross-sectional variability in number of persons employed per household. Interestingly, however, th e increase in earnings inequality, as measured by the Gini coefficient or percentile differentials, is not tha t much greater at the household level than at the individual level. 15 The household head gets a weight of l.0; each additional adult gets a weight of 0.5 ; and each child below the ag e of 17 gets a weight of 0.3. We experimented with various other equivalence scales described in Keane and Prasad (2000). These made little difference to the time profile of changes in inequality, although the levels of inequality were marginally affected. 1 7

21 A decomposition of the variance of household earnings into the components attributable to th e variances of employment (number of persons employed) and of mean household earnings (results no t reported) indicated that cross-sectional employment variance was more important before the transition. During the transition, much of the increase in overall household earnings variance is attributed to the increase in mean earnings rather than employment. with both components accounting for a roughly equa l contribution to total variance by the 1990s. The covariance component is relatively small throughout th e sample. Social transfers (including unemployment insurance benefits) and other income sources have a clearly moderating effect on total income inequality. as shown in the second panel of Table 6. An interesting feature of the results is that. from 1988 to the increase in the dispersion of tota l household income is much more muted than the increase in household earnings. From 1992 to 1997, however, the position is exactly reversed. This corresponds to the pattern of changes in social transfer s that we documented in earlier work (Keane and Prasad. 2000). For political reasons, transfers as a percent of GDP were almost doubled from 1988 to 1992 (from 10 percent to 20 percent of GDP). But, in light of significant budgetary pressures, cash transfers as a percent of GDP were then gradually reduce d through Consequently. the effects of the increases in labor earnings inequality on total incom e inequality, which had been masked during by the surge in transfers. quickly become evident after As noted above. we leave a more detailed analysis of household earnings inequality for future work. 1 8

22 Regression analysis of the structure of earning s In this section, we conduct a regression analysis of the structure of earnings in order to gain mor e insight into the sources of changing inequality among workers during the transition. In Table 7, we present estimates of standard human capital earnings functions (see Willis, 1986) for a selected set o f years during the period. Prior to 1993, the survey contains a question that can be used to determine if the responden t worked for a privately-owned firm. Staring in 1993, this question is refined, so we can determine if th e respondent works for one of three types of privately-owned firms : small firms (including the selfemployed); large privately-held firms: and large mixed-ownership firms with majority ownership by th e private sector. So, in the later years, we include separate dummies these three categories. The coefficients on the educational category dummies show that the premia for educationa l qualifications have increased substantially during the transition. For instance, the college degre e coefficient increases from in 1988 to in 1992 and further to in Since primary school is the omitted group, this implies that the wage premium for a college degree relative to a primar y school education was approximately 47 percent in 1987, 69 percent in 1992 and 98 percent in Th e high school premia relative to primary school for the same three years are 23, 30 and 41 percent, respectively. Note that this implies a widening of the college-high school premium as well. These results also imply that, by 1996, the high school and college premia in Poland were toward the high end o f estimates typically obtained using data from Western countries. One interesting question is whether academic qualifications acquired in the communist era hav e the same value in the labor market as more recently acquired qualifications. Unfortunately, it is difficul t to answer this question decisively with our sample, since it ends in 1996, seven years after the big bang. As a first pass, we constructed a dummy variable for individuals who, based on their age and impute d years of education, attained their highest degree on after This dummy, and its interactions with the 1 9

23 education dummies. did not show any strong evidence that recent degrees command a premium relative to older degrees. '' Since (potential) labor market experience (age-years of education-6) enters the regressions as a quadratic, the returns to experience need to be evaluated at specific levels of experience. We investigate changes in the returns to experience further below. but it is worth noting the significant decline in th e returns to experience implied by the OLS estimates for the early years of the transition. For instance, for workers with 25 years of potential labor market experience, the marginal return to an additional year o f experience is 0.41 percent in 1988, drops to 0.36 percent in 1990 and then rises to 0.48 percent by For younger workers with five years of potential experience. this return drops from 2.64 percent in J988 to 2.46 percent in 1990 and then partially recovers to 2.52 percent by Consistent with the evidence from the kernel density plots in the previous section. the wag e premium for men relative to women fluctuates around 30 percent and shows no clear trend, although there is a temporary decline in this premium in the period The wage premium for workers in urban areas relative to those in rural areas, on the other hand, almost doubles by the mid-1990s relative to pre - transition levels. It is also worth noting the differences in values of the private sector dummies across small and large firms for I Wages for those who work in small private firms (this includes the self employed) are actually below those who work for the government or state-owned firms the exclude d dummy covers all workers in the public sector). The wage premium for workers in lar ge private firm s (including large foreign-owned firms) is approximately 20 percent, while the premium for people wh o work for large firms with mixed but majority private ownership is around 10 percent. A separate dummy for employment in mixed-ownership firms with majority state ownership was small and insignificant. We also ran regressions separately for workers in the public and private sectors. Some of th e main results, and a comparison with the results for the full sample, are summarized in Figure 4. Two 06 It has been suggested to us by certain Polish observers that major curriculum changes in the post-communist era have taken place mainly at the university level. There also appears to be some debate about whether the level of rigor in school and university programs has improved, with some actually arguing the opposite. 20