The gender earnings differential in the Russian transition economy.
On the threshold of the Russian economic transition, researchers attempted to analyze its labor market implications for women. Among the few published papers that discuss gender differences in earnings are those by Standing (1994) and Linz (1996). Most of these studies are based on data sources that may be unrepresentative. The only publication that examines the gender pay gap using a nationally representative sample of Russian workers - drawn from the first round of the Russia Longitudinal Monitoring Survey (RLMS) - is that by Newell and Reilly (1996). However, given the relatively short time period between the introduction of the economic reform program in Russia (the beginning of 1992) and the timing of the first round of the RLMS (July-November 1992), the wage structure observed in that study does not yet reflect the impact of the transition, but rather provides a baseline against which subsequent effects of the transition can be measured.
The present study uses data from the second phase of the RLMS (Rounds 5-7), which reflect the situation in the Russian labor market in 1994-96. By the beginning of this period, the initial measures of the radical economic reform - the price and wage liberalization and the mass privatization - had been implemented long enough to show effects. I attempt to find and explain key determinants of the gender earnings differential in the Russian transition economy. The evolution of the gender pay gap's institutional context is discussed, starting from the Soviet era. I apply the Oaxaca-Blinder-Neumark method of gender earnings differential decomposition to quantify the roles of different factors, including job segregation by gender and gender differences in hours worked and human capital endowments, in the formation of the gender pay gap.
To estimate the earnings equations, I first adopt a basic human capital regression model, then augment it with variables that reflect gender job segregation. Due to the high incidence of wage arrears in present-day Russia, a specific problem faced in the study is that the OLS estimates obtained from a regression model based on observed earnings are subject to selectivity bias. To obtain consistent earnings equation estimates, I perform a regression analysis with a selectivity bias correction. I then suggest an approach for decomposing the gender earnings differential based on the regression model with selectivity bias correction.
The idiosyncrasies of the present-day Russian labor market stem from the institutional legacy of the Soviet past. In Soviet Russia, women's full integration into the "social sphere of production" was an ideological principle, which concurred with the demand of the extensively growing economy for labor resources (Lapidus 1993:138-41). Policies predicated on this principle resulted in an almost 90% female labor force participation rate, which was exceptionally high by international standards and almost as high as the male participation rate (McAuley 1981:36-37; Newell and Reilly 1996:341). Women were given educational opportunities equal with men and, by the 1970s, their average educational attainment exceeded men's (McAuley 1981:142-46).
The principle of equal pay for equal work regardless of gender was institutionalized through the Soviet Constitutions and Labor Codes. The wage-setting system based on a highly centralized and rigid "tariff" system and on centrally planned wage funds gave Soviet enterprises and organizations little discretion in terms of wage rates and wage differentials.(1) Nonetheless, both the official Soviet data, published in 1990 (Goskomstat SSSR 1990:71-72), and earlier Western estimates (McAuley 1981; Ofer and Vinokur 1992) suggest that a gender earnings differential of about 30% in favor of men existed in the Soviet economy.
It is generally agreed in the literature that this differential was due mainly to occupational segregation and discriminatory promotion practices. The patterns of occupational segregation by gender in the Soviet Russia resembled those in the Western economies, since they were formed in the common patriarchal social context for all modern societies.
However, an essential feature of the Soviet-type occupational segregation by gender that distinguished it from the Western patterns was that the patriarchal stereotypes were institutionalized and sustained through official attitudes and policies. In spite of proclaimed gender equality, Soviet labor legislation always regarded women as a "specific labor force" because of their maternity and childcare function. Accordingly, along with the provision of long maternity leave and the reduction of working time for women with small children, this legislation restricted women's employment in occupations considered unsuitable for them or harmful to them and encouraged their entry into occupations that corresponded to the "biological and psychological peculiarities" of women and their "moral ethical temperament" (Lapidus 1993:145; Posadskaya 1993). As a result, women were heavily concentrated in healthcare, education, food and light industries, trade, and personal services and were inclined toward white-collar occupations, while men dominated such economic sectors as engineering, heavy industry, mining, and construction and gravitated to blue-collar jobs.
This pattern had direct consequences for the gender pay gap in the Soviet economy. Under the centralized wage system, earnings differentials between sectors of the economy and between occupations were shaped by an administrative perception of their laboriousness, productivity, and social usefulness. Judging by these criteria, the system generally rated "women's" jobs lower then "men's." Here the patriarchal approach quite remarkably coincided with the principles of Marxist political economy, according to which the production of means of production is superior to the production of consumer goods, and the so-called "nonproductive sphere" (healthcare, education, personal services) has the lowest priority. In addition, occupational wage differentials reflected a heavy ideological emphasis on the virtues of the "proletariat" (working class occupied with industrial labor) as opposed to the "intelligentsia" (teachers, doctors, scientists). Consequently, white-collar wages were remarkably close to blue-collar wages, and sometimes even lower.
The patriarchal social context also influenced promotion patterns. Since household and family responsibilities were explicitly treated as women's domain, women often chose to sacrifice career interests to family responsibilities. Furthermore, since both creativity and authority were identified with men, women who tried to pursue managerial or professional careers encountered subtle but effective resistance to their promotion (Lapidus 1993:144-46).(2) Consequently, although a majority of doctors, teachers, and research assistants were women, most head doctors, principals of schools, and senior academics were men, as were most managers and senior officials in general (McAuley 1981:72-73).
During the perestroyka period (1987-91), despite the intellectual and political liberation unleashed by Gorbachev's reforms, the patriarchal order of society and official attitude toward women as a specific part of the labor force remained largely intact. Fundamental change in gender roles and in the gender division of labor was never seriously proposed (Posadskaya 1994).(3)
The perestroyka measures did relax the rigidity of the centralized wage system and weakened the ideological determinants of the gender pay differences. At the same time, however, it was during this period that the first steps were made toward wage-setting conditions in which economic forces are an important factor determining the earnings gap between "male" and "female" jobs. The Law on the State Enterprise, passed in 1987, allowed enterprises in the "productive sphere" to be self-financing units that meet wage payments and other obligations with their own revenues, while the female-dominated "nonproductive sphere" continued to be financed from the state budget. Due to the growing budget deficit and the state's continuing neglect of the "nonproductive" sector, wages and salaries in the latter were set at a level significantly lower than that set by the self-financed enterprises.
The gender wage differential in this period was analyzed by Katz (1994, 1997). Using household data collected in 1989 in Taganrog - a middle-sized industrial city - she calculated female/male wage-ratios of 65% for monthly and 73% for hourly wages and found that occupational segregation was an important determinant of this gender pay gap. Linz (1996), using Taganrog data from 1992, found that gender was one of the most significant factors affecting workers' income. Gender differences in earnings were partly attributable, she argued, to occupational segregation by gender and promotion patterns discriminating against women.(4)
Newell and Reilly (1996) presented estimates for the gender wage gap in Russia based on a nationally representative sample drawn from the first round of the Russia Longitudinal Monitoring Survey (RLMS) carried out between July and November 1992. Given the relatively short time period between the introduction of the economic reform program in Russia (the beginning of 1992) and the first round of the RLMS, the wage structure observed in that study does not yet reflect the impact of the transition but rather provides a baseline against which subsequent effects of the transition can be measured. The average logarithmic gender wage gap computed in the study suggests a male wage advantage of about 30%, and most of the differential is found to be attributable to gender differences in returns to worker characteristics and not to differences in the characteristics themselves. The authors ascribe most of the gender pay gap to unequal gender treatment within the one-digit occupational groups rather than across these groups.
The radical economic reform that started in 1992 with price and wage liberalization resulted in a wage system that Standing (1996:113) characterized as the most flexible conceivable. Several facts allow us to form hypotheses about the transition's potential effects on the gender earnings differential. First, the Soviet system of centrally determined pay virtually ceased to exist. The tariff system, although still formally in use, had essentially no standardizing effect on wages in the enterprise sector.(5) In the unstable economic environment that characterized the first years of transition, available resources and adjustment for inflation became key wage-setting factors for enterprises.(6) The managers' influence on wage decisions increased greatly, while workers were largely passive.(7)
International evidence (Gunderson 1994:13) suggests that the earnings gap tends to be smallest in countries with centralized collective bargaining that emphasize egalitarian wage policies in general (for example, Sweden, Norway, and Australia), and it tends to be largest in countries with decentralized wage determination and enterprise-level bargaining (for example, the United States and Canada). As noted by Blau and Kahn (1996:S37), decentralized wage setting entails generally greater pay differences across industries and firms than those determined by centralized wage systems; that is, wage-setting decentralization is likely to strengthen the effect of job segregation by gender on the gender pay differential. In Russia, this general effect of wage-setting decentralization has been combined with a specific effect caused by differences in wage determination between the female-dominated budgetary sphere and the enterprise sector of the economy. In the former, salaries have been relatively strongly linked to the statutory minimum wage,(8) which has been held fixed for months, while wages in the latter have effectively adjusted for inflation.
A greater gender earnings gap in countries with decentralized wage-setting systems may also be explained by a greater degree of wage inequality in general. Since women tend to be at the low end of the wage distribution, growing general inequality is likely to decrease women's relative wages disproportionately and hence to increase the gender earnings gap (Blau and Kahn 1996:S37). All available evidence suggests a sharp increase in earnings inequality in Russia during the transition.(9) In the Soviet era, wages were compressed because, in particular, the official minimum wage was close to the average wage. By contrast, in the transition economy, the statutory minimum wage has been held far below the average wage and has served mainly as an anchor of the wage system in the budgetary sector rather than a common wage floor.(10) Hence, the official minimum wage could no longer reduce the male-female wage differential by truncating the wage distribution from below.
A potential effect of the transition that might, to some extent, offset these consequences of the wage-system decentralization and reduce the gender earnings gap is an increased role of the emerging market forces in wage determination. According to the competitive theory of discrimination (Becker 1971), as the market becomes more competitive, discriminating employers cannot maximize profits and are driven out of business. However, as Gunderson (1994) noted, even in the developed market economies relative wages, especially for jobs within the internal labor market, are determined by administrative procedures and historical relationships, rather than by market forces. The market forces are important for determining the wage rates for certain "port-of-entry" jobs, which are then used as "benchmarks" for establishing the relative pay of other jobs within the firms' internal labor market. Furthermore, in Russia, despite the profound change in the economic system, the competitive market forces are still rather weak and no significant change in social attitudes and policies toward women is apparent. It may be hypothesized that the ideologically defined and administratively enforced Soviet occupational and earnings patterns are sustained in the employment and wage-setting practices that, although newly introduced, are based on the same historical/cultural background and on largely unchanged social attitudes, and that occupational and sectoral segregation by gender remains a key determinant of the gender earnings gap.
Before we test this hypothesis empirically, it is important to note that nominal wages to which Russian enterprises and organizations were committed were not necessarily the wages actually paid. Wage arrears became a major problem during the transition. In the last quarter of 1996, back wages in the economy were equal to 114% of the total monthly wage bill (Russian Economic Trends). As explained by Clarke (1997), the wage arrears problem arose as a part of the problem of the general monetary shortage, which resulted from the demonetization of the Russian economy due to the restrictive financial and monetary policies used by the government to combat inflation. Enterprises experience difficulties in paying wages not because they cannot afford to pay, but because they currently do not have cash to pay. In this situation, employers are better off not paying wages than cutting employment. Since organized worker opposition is generally absent, the administration is under very little pressure to pay wages, and it costs the enterprises nothing to just keep their workers on the books, unpaid, instead of laying them off and paying severance.
This study is based on data drawn from the Russia Longitudinal Monitoring Survey (RLMS), a household-based survey designed to measure systematically the effects of Russian reforms on households and individuals. The data are unique since the RLMS is the only representative random household survey covering the entire Russian Federation. Carefully designed by an interdisciplinary partnership of leading Russian and American experts, the survey is of exceptionally high quality for a country undergoing such dramatic upheaval.(11)
Data from Phase II of the RLMS (Rounds 5 through 7) are used in the study.(12) In this phase, a multi-stage probability sample was employed. First, a list of 1,850 consolidated rayons (administrative subdivisions similar to counties in the United States) was used to serve as primary sampling units (PSUs). Three very large population units - Moscow city, Moscow oblast (province), and St. Petersburg city - constituted self-representing (SR) strata. The remaining non-self-representing (NSR) rayons were allocated to 35 equal-sized strata. One rayon was then selected from each NSR stratum using the method "probability proportional to size" (PPS). That is, the probability that a rayon in a given NSR stratum was selected was directly proportional to its measure of population size. The NSR strata all have approximately equal sizes because they were purposefully designed that way to improve the efficiency of estimates.
The number of households drawn into the sample was 4,718. Within each selected PSU the population was stratified into urban and rural substrata, and the target sample size was allocated proportionately to the two substrata. In both urban and rural substrata, interviewers were required to visit each selected dwelling up to three times to secure the interviews. They were not allowed to make substitutions of any sort. The interviewer then conducted interviews with as many adults as possible, acquiring data about their individual activities. The response rate exceeded 80%. The multivariate distribution of the sample by sex, age, and urban-rural location compares quite well with the corresponding multivariate distribution of the 1989 census.
The data used in the study reflect the situation in the country after the initial stage of the radical economic reform, when price liberalization, mass privatization, and macroeconomic stabilization had been virtually completed. The data were collected in November 1994-January 1995 (Round 5), October-November 1995 (Round 6), and October-December 1996 (Round 7). Throughout this period, no significant changes in labor market institutions and policies that might influence the gender earnings differential occurred. My preliminary research indicated no notable trends in the wage structure and the gender gap. Hence, this study uses pooled data from Rounds 5-7. The study focuses on working individuals, women aged 18-55 years old and men aged 18-60 (which are considered the normal working ages for women and men in Russia).
Women in the Russian Labor Market in 1994-96: An Overview
The characteristics of labor force participation and employment by gender in 199496 calculated from the RLMS data are presented in Table 1. The female participation rate shown in the table conforms with the historical pattern: it is very high and close to the male participation rate.(13) The high unemployment rates in the table reflect the effect of the economic decline and restructuring associated with the transition. The gender differential in rates of unemployment, however, is relatively small. Moreover, since self-employment is more common for men than for women, the gender differential in wage-employment rates is even smaller. Thus, gender differences in rate of employment do not appear to be a key determinant of gender inequality in the Russian transition economy.
The RLMS data analysis also has shown that despite the presence of self-employment, secondary employment, and incidental employment, wage employment remains by far most important for both men and women, accounting for, respectively, 96% and 98% of regularly employed workers. Further, 85% of the wage-employed women and 91% of the wage-employed men consider wage-employment their main occupation, and the difference is due mainly to those women who viewed maternity, childcare, and household responsibilities as their primary occupation. These are, almost exclusively, women on official maternity or childcare leave.
Table 1. Labor Force Participation and Employment by Gender, 1994-1996 (percent).(a) Description Women Men Labor Force Participation Rate(b) 93.2 93.8 Unemployment Rate 19.8 16.7 Wage-Employment Rate(c) 78.3 79.8 Self-Employment Rate(d) 1.8 3.4 Consider Wage Employment Their Main Occupation(e) 85.4 91.3 On Maternity and Child-Care Leave(e) 7.5 .0 Consider Maternity, Childcare, and Household Responsibilities Their Main Occupation(e) 7.8 .0 Actually Worked(f) 83.0 85.6 Worked Part-Time(g) 26.7 12.7 Average Weekly Hours Worked(h) 38.8 43.6 a Computed from the RLMS data, Rounds 5-7, for women aged 18-55 and men aged 18-60. b Labor force is defined as those who, at the time of the interview, either were employed (including self-employed) or were not employed but wanted to find a job. c Percentage of labor force that worked for a firm, enterprise, organization, or institution. d Percentage of labor force that worked, but not for a firm, enterprise, organization, or institution. e Percentage of the wage-employed. f Worked in the last 30 days before the interview at their primary place of employment at least 20 hours per week, percentage of those considering wage employment their main occupation. g Percentage of those who considered wage employment their main occupation and actually worked; part-time is defined as employed for less than 35 hours per week. h Computed over those who considered wage employment their main occupation and actually worked.
The data indicate a relatively small (2.6-point) gender differential in percentages of the employed who actually worked (Table 1). At the same time, the gender differences in hours worked and in worker status (part-time or full-time) appear to be significant. As may be seen from Table 1, women work about five hours less per week than men do, and the percentage of women working part-time is about twice as high as that of men.
Table 2. Earnings by Gender, 1994-96.(a) Description Women Men Received Full Payment Last Month (percent) 52.5 44.6 Mean Monthly Earnings Paid in Full(b) 441 660 Female-Male Earnings Ratio .668 a Computed from the RLMS data, Rounds 5-7, for women aged 18-55 and men aged 18-60 who considered wage employment their main occupation and worked in the 30 days before the interview at their primary place of employment at least 20 hours per week. b Wages, bonuses, and benefits after taxes, in thousands of December 1995 rubles, received from the primary place of employment in the 30 days before the interview; the geometric mean computed as the antilog of the mean natural logarithm of earnings, calculated over those who were not owed back wages.
The earnings figures calculated from the RLMS data are shown in Table 2. The percentages of those workers who received their monthly earnings in full - only 44.6% for men and 52.5% for women - reflect the depth of the wage arrears crisis. The gender differential in the rate of non-payments - 7.9 percentage points in favor of women - is worthy of note.
The high incidence of non-payment complicates calculations of mean monthly earnings and of the gender differential. Unfortunately, the RLMS does not contain information on whether the reported wages are earned during the reference period and paid in full or back wages paid in the reference period. Data on back wages owed, however, are available. Thus, a meaningful gender earnings differential may be calculated by estimating mean earnings only over those workers who were not owed back wages and hence received their reference period earnings in full. The female-male real earnings ratio calculated using this approach equals 0.67, which is close to the estimates of the gender pay gap in the Soviet economy.(14)
Table 3. Human Capital Endowments by Gender, 1994-1996 (percent).(a) Variable Women Men Education General Secondary 74.0 64.9 Ordinary Vocational 6.6 12.3 Secondary Vocational 16.0 20.7 Specialized Secondary 35.4 18.8 Higher 21.3 18.8 Mean Years of Schooling 12.0 11.6 Experience(b) 0-4 Years 13.7 10.0 5-9 Years 14.4 14.6 10-19 Years 31.5 29.8 20-29 Years 29.9 26.1 [greater than or equal to]30 Years 10.5 19.5 Mean Years of Experience 16.7 19.1 a Computed from the RLMS data, Rounds 5-7, for women aged 18-55 and men aged 18-60 in the labor force. b Potential labor market experience computed by subtracting total years of schooling after the 8th grade plus 15 from a respondent's age.
The gender differences in human capital endowments are reflected in Table 3, where the Russian labor force is characterized using traditional human capital variables - education and experience. Showing high educational attainment in general, the data suggest that women are better educated than men in terms of both general and specialized schooling. Men are more inclined toward vocational education, which corresponds to the traditional gender job stereotyping and occupational patterns.(15)
Since the RLMS lacks data on actual labor market experience, potential experience is estimated applying a commonly used method: years of potential experience are calculated by subtracting total years of schooling after the eighth grade plus fifteen from a respondent's age. Given the strong labor force commitment of Russian women, gender-specific deviations of these estimates from the actual labor market experience are expected to be trivial.
The figures in Table 3 suggest that women's mean labor market experience is 2.4 years shorter than men's. Given the negligible gender differential in years of schooling, this may be explained mainly by a younger official retirement age for women (55) than for men (60). While the percentage of women with 30+ years' experience is 9 points lower than that of men, women are more concentrated in the groups of workers with 10-29 years of experience, that is, in the middle part of the experience-earnings profile, which is typically higher than the sides of the curve. Thus, in general, Russian women appear to have an advantage over Russian men in terms of measured human capital endowments.
The RLMS data also allow us to examine the patterns of occupational and sectoral segregation in the Russian transition economy. Table 4 reflects the industrial distribution of employment by gender. The industry classification in the table is based on the standard one-digit schema adjusted for idiosyncrasies of the Russian economy. The segregation index of 0.324 indicates an overall level of industrial gender segregation that is relatively modest by international standards.(16) The patterns of industrial segregation formed during the Soviet era remain distinct. More than 40% of the employed men and less than 17% of the employed women are occupied in extractive industries, agriculture, construction, and transportation, while the respective percentages in trade, personal services, health, and education are 12% and 43%. Interestingly enough, women continue to dominate finance, insurance, and real estate, constituting 76% of the employed in this industry that used to be an underdeveloped, low-paid branch of the Soviet public administration but has changed so profoundly - becoming a lucrative sector of the emerging market economy - that it no longer fits the traditional stereotype of a "female" industry.
Table 4. Industrial Distribution of Employment by Gender, 1994-1996 (percent).(a) Industry(b) Women Men Manufacturing 21.2 24.4 Extractive Industries 2.2 5.3 Agriculture 7.2 12.3 Construction 3.1 10.7 Transportation 4.2 11.7 Trade and Personal Services 14.3 5.6 Business and Repairs Services .8 3.5 Utilities 3.8 5.8 Health 10.3 2.3 Education 18.3 4.2 Public Administration 5.7 6.3 Finance, Insurance, Real Estate 2.3 .7 Others 6.7 7.1 Segregation Index(c) .324 a Computed from the RLMS data, Rounds 5-7, for wage-employed women aged 18-55 and men aged 18-60. b The classification of industries is based on the standard one-digit schema adjusted for idiosyncrasies of the Russian economy. c Computed as follows: [S.sub.i] = .5 [[Sigma].sub.i] [absolute value of [p.sub.im] - [p.sub.if]] where [p.sub.im] is the proportion of men employed in industry i and [p.sub.if] is the proportion of women employed in industry i.
An aspect of the sectoral distribution of workers that is new for Russia and may affect the gender earnings differential is the structure of employment by enterprise ownership type. The mass privatization in Russia in 1992-94 put an end to the state monopoly in ownership of productive assets and created a sizable private sector, where employment policies and wage-setting practices may differ from those in the state sector. As follows from the RLMS data, some gender differences are present in the employment distribution by ownership type. The private sector gives jobs to 13.3% of the employed women and 16.7% of the employed men, while the respective figures for the state sector are 62.0% and 53.3%.(17)
A detailed examination of the male and female occupational distributions clarifies further the patterns of gender job segregation in the Russian transition economy.(18) Unlike the index of industrial segregation, the occupational segregation index calculated at the one-digit level of aggregation, 0.514, is exceptionally high by international standards.(19)
As suggested in the literature (Gunderson 1989; Kidd and Shannon 1996), further gender segregation is likely to occur within the one-digit categories. To capture this segregation, the four-digit occupations within each one-digit category were grouped according to a common rule: designate an occupation as "male-dominated" or "female-dominated" if more than 70% of those in the occupation are men or women, respectively (Gunderson 1994:23-24). The results are presented in Table 5. The index of occupational segregation by gender calculated using this classification of occupations, 0.673, shows a very high degree of gender job segregation. Overall, 71.3% of working women are employed in the female-dominated occupations and almost 73.8% of men have "male"jobs.
The patterns of occupational segregation by gender reflected in Table 5 may be explained, in general, by the stereotypes - common in many countries - of women and their supposed abilities mirrored by the characteristics of "female" and "male" occupations (Anker 1997). For example, "caring nature," "skill and experience in household-related work," "greater manual dexterity," "greater tolerance of monotonous work," and "attractive physical appearance" help "qualify" women for such occupations as nurse, teacher, cleaner, cook, weaver, sewer, bookkeeper, clerk, and shop assistant. At the same time, "disinclination to supervise others," "lesser ability in science and mathematics," "lesser physical strength," and "lesser willingness to face physical danger" negatively affect women's acceptability in the "male" occupations, such as manager, physical scientist, engineer, mechanic, transport equipment driver, freight handler, and police officer. The high degree of gender job segregation in Russia indicates that these common stereotypes are more pronounced in Russia than in the Western societies.
Some extremes and peculiarities of Russian occupational segregation by gender are worthy of note. First, 79% of economists, 94% of accountants, and 98% of bookkeepers in Russia are women. Second, men are practically absent from such occupations as pre-school and primary-school teacher and nurse, and all clerical jobs are [TABULAR DATA FOR TABLE 5 OMITTED] "female." Finally, mechanics and drivers are almost exclusively men. Apparently, these idiosyncrasies are inherited from the Soviet labor market and reflect the influence of its specific institutions and policies. At the same time, the content and image of such "female" occupations as economist and accountant have changed notably during the transition. In the Soviet past, these occupations were characterized by humdrum, low-paid clerical work, whereas during the transition they became more of what they are in the developed market economies. As in the case of the finance industry, it is interesting that women have kept their high prevalence in these occupations. This may be explained, however, by the fact that these occupations require specialized training, which takes time to acquire.
The Gender Earnings Differential Decomposition
The standard approach to examining the gender wage differential is based on estimating separate wage or earnings equations for women and men and then decomposing the aggregate (gross) pay differential into the components explained by workers' productivity-related characteristics (endowments) and the unexplained (residual) component, which is often attributed to discrimination. The method is suggested by Oaxaca (1973) and Blinder (1973) and has been followed by many authors.
Oaxaca calculated the gender wage differential as
(1) ln [W.sub.m] - ln [W.sub.f] = [X[prime].sub.m][B.sub.m] - [X[prime].sub.f][B.sub.f],
where [W.sub.m] and [W.sub.f] are log wages of men and women, respectively, [X.sub.m] and [X.sub.f] are vectors of mean productivity-related characteristics of men and women, and [B.sub.m] and [B.sub.f] are coefficient estimates in the OLS regression equations for men and women.(20) The differential may then be decomposed as
(2) ln [W.sub.m] - ln [W.sub.f] = [X[prime].sub.m] - [X[prime].sub.f]) [B.sub.m] + [X[prime].sub.f] ([B.sub.m] - [B.sub.f]),
(3) ln [W.sub.m] - ln [W.sub.f] = ([X[prime].sub.m] - [X[prime].sub.f]) [B.sub.f] + [X[prime].sub.m] (B.sub.m] - [B.sub.f]).
On the right-hand sides of (2) and (3), the first term is the component of the gender earnings differential explained by different characteristics of men and women, and the second term is usually interpreted as the component reflecting discrimination, but, in fact, also accounts for possible difference in unobserved productivity-related characteristics. The coefficients of the explanatory variables in the earnings equations may be interpreted as prices of skills associated with worker characteristics, X.
A problem with this method is that the two equations do not yield the same estimates for the wage differential components, since equation (2) implies the male wage structure in the absence of discrimination, while equation (3) assumes that in the absence of discrimination the female wage structure would prevail. Cotton (1988) and Neumark (1988) show that in an actual nondiscriminatory setting neither the male nor the female wage structure would prevail, and that the true nondiscriminatory structure should lie somewhere between the two. The wage differential decomposition, then, is given by
(4) ln [W.sub.m] - ln [W.sub.f] = ([X[prime].sub.m] - [X[prime].sub.f]) [B.sub.p] + [X[prime].sub.m] ([B.sub.m] - [B.sub.p]) + [X[prime].sub.f]([B.sub.p] - [B.sub.f]),
where [B.sub.p] is the estimated nondiscriminatory wage structure, and, on the right-hand side of the equation, the first term is an estimate of the productivity differential, the second term is an estimate of the male wage advantage, and the third term is an estimate of the female wage disadvantage. Drawing on the theory of discrimination, Oaxaca and Ransom (1994) showed that the nondiscriminatory wage structure is given by the least squares coefficient estimates obtained from the pooled sample of males and females.(21) The decomposition of the gender wage differentials implemented in the present study is based on the latter approach.
As previously noted, due to the high incidence of wage arrears and lack of information in the RLMS on whether or not reported wages are earned in the reference period and paid in full, the regression sample should include only those workers who were not owed back wages[middle dot] However, the estimates obtained this way are subject to so-called selectivity bias. The essence of this problem is explained by Manski (1989). Since the probability that an individual's earnings are present in the regression model may be influenced by some of the factors that also influence observed earnings, analyses of the latter may yield biased estimates of mean wages earned by individuals with given measured characteristics and biased coefficients of the earning functions. If inclusion in the wage regression sample is selective of those with higher earned wages, the mean of observed wages will be higher than the mean of wages earned (a positive bias). If inclusion in the sample is selective of those with lower earned wages, the mean of observed wages will be lower than that of wages earned (a negative bias). If men and women differ in the direction and magnitude of this selectivity bias, the estimate of the gender earnings differential based on observed wages will be biased.
A technique suggested by Heckman (1979) is used to obtain consistent estimates of the coefficients in the earnings equations and of the gender earnings differential. First, probit equations predicting inclusion in the earnings regression sample are estimated separately for each gender. Then, a selection correction variable, [Lambda], is calculated as the inverse of the Mills' ratio:
(5) [[Lambda].sub.i] = [Phi]([Z[prime].sub.i][Gamma])/[Phi]([Z[prime].sub.i][Gamma]),
where [Mathematical Expression Omitted] and [Mathematical Expression Omitted] are, respectively, the probability density function and the cumulative probability function for the standard normal variable, [Gamma] is the vector of probit coefficient estimates, and [Z.sub.i] the vector of individual i's characteristics that are instrumental in determining the probability of receiving earnings in full.(22) Finally, the earnings equations with [Lambda] as an explanatory variable are estimated for each gender and for the pooled sample as follows:
(5) ln [W.sub.i] = [X[prime].sub.i]B + [[Lambda].sub.i]b + [[Epsilon].sub.i].
The gender earnings differential then may be decomposed as
(6) ln [W.sub.m] - ln [W.sub.f] = ([X[prime].sub.m] - [X[prime].sub.f]) [B.sub.p] + [X[prime].sub.m] ([B.sub.m] - [B.sub.p]) + [X[prime].sub.f]([B.sub.p] - [B.sub.f]) + ([[Lambda].sub.m][b.sub.m] - [[Lambda].sub.f][b.sub.f]).
In this equation [X.sub.m] and [X.sub.f] are vectors of mean characteristics of men and women, respectively, [B.sub.m] and [B.sub.f] are coefficient estimates in the earnings regression for men and women corrected for selectivity, and [b.sub.m] [[Lambda].sub.m] and [b.sub.f][L.sub.f] are selection correction terms for men and women. The last term in the equation is an estimate of the gender earnings differential bias due to the sample selection; that is, the unbiased mean log earnings and the gender differential may be expressed as
(7) [D.sub.c] = (ln[W.sub.m] - [[Lambda].sub.m][b.sub.m]) - (ln[W.sub.f] - [[Lambda].sub.f][b.sub.f]).(23)
The Regression Model
The earnings regression sample includes working-age individuals who considered wage employment their main occupation, worked during the 30 days before the interview at their primary place of employment at least 86 hours (20 hours per week), and received their earnings in full. The earnings measure used in the earnings equations is the amount of after-tax earnings (including wages, bonuses, and benefits) received by an individual in the reference month from her or his primary place of employment. These nominal earnings are adjusted for inflation by estimating the regression equations with a set of dummy variables that relate to the reference months,
(8) ln ([W.sub.i]) = [X[prime].sub.i]B + [T[prime].sub.i]A + [[Epsilon].sub.i],
where [T.sub.i] is the vector of time dummy variables for individual i and A is the vector of estimated coefficients of the time dummies. Then the time-adjusted log earnings are calculated as follows:
(9) ln ([Y.sub.i]) = ln ([W.sub.i]) - [T[prime].sub.i]A.
To examine the effect of gender differences in hours worked and worker status, log hours worked in the reference month, a dummy variable for part-time work, and their interaction term are included in the gender-specific earnings functions, allowing for both a part-time shift and different slopes for hours worked by part-time and full-time workers. The inclusion of the status variables is justified by the fact that the earnings of part-time workers may differ from those of full-time workers due not only to reduced hours, but also to different wage systems and wage-setting practices for part-time and full-time workers. The regression equation then may be expressed as follows:
(10) ln ([Y.sub.i]) = [b.sub.p]P + [b.sub.ph]Pln (H) + [b.sub.fh]Fln(H) + [X[prime].sub.i]B + [[Epsilon].sub.i].
In this equation P is a dummy for part-time work, defined as less than 35 hours per week, and Pln(H) and Fln(H) are interactions of log weekly work hours with part-and full-time status. Then, the coefficient estimates, [b.sub.p], [b.sub.ph], and [b.sub.fh] are used to adjust each person s earnings for work hours by assuming a standard 41-hour workweek, as follows:
(11) [Mathematical Expression Omitted].(24)
The difference (In Y- In [Y.sup.s]) shows the effect of hours worked and worker status on earnings, which, in terms of gender differential, may be expressed as
(12) [Mathematical Expression Omitted],
where [D.sub.h] is the gender earnings differential due to the effect of hours and status, [D.sub.g] is the gross gender differential in monthly earnings, and [D.sub.s] is the differential in earnings corrected for hours and status. Log earnings adjusted for hours and status, In [Y.sup.s], are used as the dependent variable in the regression equations for the further gender differential analysis.
Descriptions of the independent variables in the earnings equations are given in Table 6. A subset of independent variables accounts for conventional human capital characteristics - formal education and labor market experience. Education is characterized by dummy variables for different types of education (see footnote 15). All education variables in the model equal zero if an individual has completed no education beyond the incomplete secondary school? The coefficients of the dummies reflect marginal values of each type of education. Given that the character and quality of education depend on the type of school and given the gravitation of different genders to different types of school, variables controlling for the type of education appear to be more meaningful than the commonly used control for uniform years of schooling or highest degree obtained. Labor market experience is accounted for by years of potential experience calculated as explained above, and this variable's square is included in the model to fit the typical experience-earnings profile (Mincer 1974).
Another subset of independent variables in the earnings equations accounts for occupational and sectoral segregation. According to Reskin et al. (1986) and Blau and Ferber (1992), occupational segregation may reflect three different factors: women's and men's relative training levels, their different labor force commitment, and the impact of employers and of governmental policy.
Several studies show that the results of the pay differential decomposition vary substantially with the inclusion or exclusion [TABULAR DATA FOR TABLE 6 OMITTED] [TABULAR DATA FOR TABLE 7 OMITTED] of controls for occupation in the earnings equation (Blau and Ferber 1987; Neumark 1988; Gunderson 1989). Inclusion of occupation controls increases the proportion of the explained gender gap (Brown et al. 1980; Miller 1987; Groshen 1991; Dolton and Kidd 1994). Bloom and Killingsworth (1982) argued, however, that controlling for occupation in the earnings equations is appropriate only if the differences in occupational outcomes are presumed given and the purpose is to measure only the extent of unequal pay for equal work. But this measure may understate the "discrimination" part of the gender differential if women are denied access to more highly paid jobs. Gunderson (1989) suggested that to obtain a measure of the earnings differential that reflects differences in the occupational distribution of women, one should exclude occupation from the earnings regressions. However, the differential decomposition in this case may overstate the unexplained portion of the gender differential ascribed to discrimination, since a significant part of gender differences in the occupational distribution may be explained by differences in worker choices rather than by employer job discrimination.(26)
In the regression model applied in this study, sectoral segregation is reflected by twelve industry dummies and three dummies for ownership type (see Table 6). To capture occupational segregation by gender both across the one-digit occupational categories and within each category, fifteen dummy variables defined according to the classification in Table 5 are included in the earnings equations. One more dummy variable controls for occupations with supervisory responsibilities across all occupational groups. Finally, the model is augmented with region-specific dummy variables to allow the intercept to vary across regions.
Decomposition of the Gender Earnings Differential: Empirical Results
The estimated effect of hours worked and worker status on the gender earnings differential is shown in Table 7. The coefficient estimates for equation (10) are reported in Table 8. The equalization of hours and status decreases the gender earnings differential only by 3 log percentage points. Thus, the higher incidence of part-time work among women and fewer hours worked by them compared to men can explain only a minuscule part of the gender pay differential.
Table 8. Estimated Coefficients of Hours Worked and Worker Status Variables. (Standard Errors in Parentheses) Variable Women Men P -1.130 .610 (.759) (1.128) PlnH .487(***) .120 (.184) (.316) FlnH .167 .311(***) (.120) (.095) *** Statistically significant at the .01 level.
[TABULAR DATA FOR TABLE 9 OMITTED]
To examine the gender differential in hours-adjusted earnings, initially, the basic human capital model (Mincer 1974) is adopted with the adjusted monthly earnings as a function of education and experience and with the regional controls. The analysis is then extended by inclusion of variables that reflect occupational and sectoral distribution of employment. Finally, the regression model with selectivity bias correction is estimated. The coefficient estimates are presented in Table 9. The results of the gender earnings differential decomposition based on each of these models are shown in Table 10, where worker characteristics are categorized into several groups representing education, experience, industry, type of ownership, occupation, and region.
Overall, the three models are consistent in revealing an important fact: most of the gender earnings differential in the Russian transition economy is due to gender job segregation. The basic human capital model shows that the differential cannot be explained by gender differences in education and experience. On the contrary, it indicates that the gross gender differential would have been even higher (by 1.0 log percentage points) had women's human capital endowments been the same as men's. As shown by the extended models, occupational, industrial, and sectoral segregation by gender accounts for 82.3% of the uncorrected gross gender differential and 92.6% of the gross differential corrected for selectivity. Overall, given the offsetting effect of human capital endowments, the two models explain, respectively, 74.9% and 84.3% of the gross differential, leaving for the unexplained residual only 9.5 and 5.2 log percentage points, respectively.(27)
As shown in Table 9, correction for selectivity bias does not, essentially, change the regression results. For women the selectivity bias (b) is slightly negative, and for men it is slightly positive. Due to the gender differences in the direction and magnitude of selectivity bias, the corrected gross gender earnings differential is 4.7 log percentage points smaller than that based exclusively on observed wages. The selectivity-corrected estimate of the gender log-earnings differential corresponds to a female/male earnings ratio of 71.7%, a medium gender earnings ratio by European standards.(28) Since the selectivity bias is not very pronounced, the models with and without its correction yield virtually the same estimates of the regression coefficients. The further analysis focuses on the selectivity-bias-corrected results.
The coefficient estimates in all three models indicate significantly positive returns to college education for both genders. However, the marginal effect of college education for women is notably greater than for men. As may be calculated from the regression model, ceteris paribus, the gender pay gap for individuals holding a college degree is 21 log percentage points narrower than that for individuals with only incomplete secondary education. A similar, although smaller, gender differential in marginal effects on earnings characterizes the returns to specialized secondary education, which are statistically significant for women and insignificant for men. Thus, receiving higher or specialized secondary education appears to be more important for women than for men. The effects of other forms of education on earnings are generally insignificant for both genders. The effect of vocational education tends to be negative, which may be explained by negative selection by ability into this category of schools (see footnote 15).
The gender differences in educational attainment weighted by the nondiscriminatory earnings structure result in a 1.4 log percentage points differential in favor of women. The experience-earnings profiles for both men and women are rather typical, with a positive coefficient of the experience variable and a negative coefficient of its square. Like education, experience is more important for women than for men. The female experience-earnings profile is steeper and reaches its maximum at 19.0 years, where the earnings are 23.3 log percentage points higher than those for a woman without labor market experience. The respective figures for the male experience-earnings profile are 18.0 years and 9.3 log percentage points. The effect of the gender differences in labor market experience weighted by the nondiscriminatory earnings structure, 1.4 log percentage points, is also in favor of women.
These women's advantages in human capital endowments, however, are overwhelmed by their disadvantages caused by the gender-specific industrial distribution of employment. The most advantageous industries in terms of nondiscriminatory earnings structure, extractive industries, construction, and transportation, are male-dominated, while the most disadvantageous industries, education and healthcare, are "female." Women's prevalence in the most highly paid finance industry does not influence the gross gender earnings differential noticeably, since the share of total employment in this industry remains minuscule.
The effect of occupational segregation by gender on the gender pay differential is even stronger than that of industrial segregation.(29) As shown in Table 10, significant women's disadvantages in terms of nondiscriminatory earnings structure are (a) women's prevalence in lower-valued one-digit occupational categories, such as "clerks" and "service and market workers"; (b) lower-paid "female" occupations within such one-digit categories as "technicians and associate professionals" and "unskilled workers"; and (c) small percentages of women in the "male" occupations within such categories as "craft" and "plant and machine operators and assemblers." At the same time, the nondiscriminatory earnings structure values the "female" professional occupations, especially accountants and economists, higher than the "male" professional occupations; this women's advantage reduces the gender earnings differential, but only by 1.2 log percentage points. Men's dominance in the managerial occupational category and in occupations with supervisory responsibilities across all one-digit groups does not add much to the gross pay differential, since managers account for only a small fraction of total employment, and the gender difference in percentages of employees with supervisory responsibilities is fairly small.
The estimated nondiscriminatory marginal effect of working in the state sector, where the percentage of women is higher than that of men, is negative, while the marginal effects of working in the private sector, where men slightly prevail, is positive. However, since the differences in patterns of employment by ownership type between women and men are rather small, this factor is relatively unimportant in terms of its influence on the gender earnings differential. Also unimportant in explaining the gender earnings differential is worker ownership of the firm, due to the very small percentage of workers who may be considered true firm owners or co-owners.(30)
Since job segregation by gender appears to be the main determinant of the gender earnings differential, an important question is why women would gravitate to jobs that pay substantially less. One of the possible explanations is discrimination by employers. [TABULAR DATA FOR TABLE 10 OMITTED] Some evidence of this discrimination in Russia at the outset of the transition was found by Standing (1994; 1996:28183). In the Russian Labor Flexibility Survey conducted in mid-1994, 28.4% of enterprise managers said they had a preference for men when recruiting production workers, 12% preferred women, and 59.9% had no preference. On the recruitment of white-collar employees, nearly 80% of firms expressed no gender preference, 11.7% reported a preference for men, and 9.1% said they preferred women. A strong preference for men was recorded in traditionally male industries such as engineering, basic metals, and
construction materials, and a net preference for women existed in such "female" industries as textiles, garments, and food processing.
The reasons given for preferring men were the conventional ones, found in other countries, and primarily related to a perception that men had better or more relevant skills. An interesting fact is that 11% of those who expressed a preference for men cited labor regulations. Among those who preferred women, the main reason given was women's greater reliability related, in particular, to the perception that women were less likely to drink or slack on the job.
In general, Standing concludes, while there is some employer discrimination, it does not seem to be substantial. The RLMS respondents' answers to the question "Do you agree that men and women have equal abilities for any work?" asked in Round 7 provide further support for this conclusion: 87.1% of female and 72.6% of male employees with supervisory responsibility - those who make or influence employer decisions - answered this question affirmatively.(31)
Another possible reason why women could choose lower-paid jobs is that factors other than earnings are likely to play a more important role in women's work preferences than in men's. Some evidence in support of this hypothesis may be found in the RLMS data from Round 4. Two survey questions are relevant. One question, addressed to the unemployed respondents, asked, "What kind of job would you like to find?" The other question, addressed to individuals 28 years of age or younger, asked respondents to evaluate the importance of different characteristics of their work using a five-grade scale, where "1" meant "not at all important," and "5" meant "very important."
The answers of both groups of respondents, summarized in Table 11, clearly show gender differences in job preferences. On the one hand, high pay is substantially more important for men than for women; on the other hand, women show relatively higher preferences for working part-time and having a flexible work schedule, for work that is not physically hard or harmful, for work that is not very responsible and allows more family time, and for workplaces that are close to home and provide childcare facilities.
It is easy to see that these patterns of employee preferences correspond well to the Russian version of the traditional gender stereotyping discussed earlier. However, as Anker (1997) noted, the word "preferences" should be put within quotation marks, because even when an individual chooses to accept work in a particular occupation, this decision is influenced by learned cultural and social values that stereotype occupations as "male" or "female" and often discriminate against women. By influencing workers' choices, as well as those of employers, these socially defined stereotypes form "female" and "male" occupations and sectors of the economy. Furthermore, once a job is perceived as "female," its "female" characteristics are reinforced further through legislation and labor market policies, such as reduced work hours and flexible working conditions.
Consequently, the lower pay in "female" industries and occupations in the Russian transition economy is determined by the interaction of the institutional factors inherited from the Soviet past with the forces of the emerging market. The former define preferences of both employers and employees, while the latter form labor demand curves explained by the competitive theory of discrimination (Becker 1971) and labor supply curves explained by the compensating differentials model (Anker 1997).
The purpose of this study has been to examine the gender earnings differential in Russia after the first three years of transition from a centrally planned to a market economy. The study has used data drawn from the Russia Longitudinal Monitoring Survey (RLMS), the only nationally representative household survey conducted so far in the Russian Federation.
The gender pay gap appears to be the key determinant of gender inequality in the Russian transition economy. Adjusted for time trend, the gender differential in log monthly earnings has been calculated at 41.0 log percentage points (in favor of men), which corresponds to a female/male earnings ratio of 66.4%. The higher incidence of part-time work among women and fewer hours worked by them compared to men account only for a small part of the gender pay gap. Equalization of each person's hours and status by assuming a standard 41-hour workweek has yielded a [TABULAR DATA FOR TABLE 11 OMITTED] gender earnings differential of 38.0 log percentage points (gender earnings ratio of 68.4%). After the correction of the observed earnings for selectivity bias which existed due to a high incidence of wage arrears - the estimated gender earnings gap falls to 33.3 log percentage points, which corresponds to a female/male earnings ratio of 71.7%, a medium gender earnings ratio by the standards of the developed market economies.
The three regression models employed in the study - the basic human capital model, the extended model with variables that reflect the occupational and sectoral distribution of employment, and the extended model with a correction for selectivity bias - are consistent in revealing that most of the gender earnings differential in the Russian transition economy is accounted for by occupational and industrial segregation by gender. The basic human capital model has shown that the gender pay gap cannot be explained by women's disadvantages in measured human capital endowments. As shown by the extended models, the gender differences in educational attainment weighted by the nondiscriminatory earnings structure result in a pay differential of 1.4 log percentage points in favor of women, and the effect of the gender differences in labor market experience is also 1.4 log percentage points in favor of women; that is, the gross gender earnings differential would have been 2.8 log percentage points higher if women's human capital endowments had been the same as men's.
The correction for selectivity bias has not changed the regression results significantly. Occupational, industrial, and sectoral segregation by gender account for 82.3% of the uncorrected gross gender differential and for 92.6% of the gross differential corrected for selectivity bias. Overall, given the offsetting effect of human capital endowments, the extended models without and with correction for selectivity explain 74.9% and 84.3% of the gross gender pay gap, respectively. As a rule, the most advantageous industries, in terms of the nondiscriminatory earnings structure, are male-dominated, while the most disadvantageous industries are "female." The effect of occupational segregation by gender on the gender pay differential is even more pronounced than that of industrial segregation. Women prevail in lower-valued occupations both across and within the one-digit categories and are under-represented in more highly paid blue-collar occupations.
Women's prevalence in industries and occupations that pay substantially less may be explained by employee preferences rather than employer discrimination. The latter, while present to some extent, does not seem to be substantial. Evidence found in the RLMS data suggests that high pay is substantially more important for men than for women and, on the other hand, women show relatively higher preferences for better labor conditions, flexible or reduced hours, and workplaces that are close to home and provide childcare facilities. It is important to note, however, that even when an individual "prefers" a particular occupation, this preference is influenced by learned cultural and social values.
It may be argued, then, that the new economic conditions have not essentially changed the gender job stereotypes and attitudes socially learned from the Soviet past. Despite the changes in the labor market induced by radical economic reform, this institutional legacy of the Soviet era is sustained in the new employment and wage-setting practices. Consequently, the lower pay in the "female" industries and occupations and the gender pay gap in general are determined by interaction of the institutional factors inherited from the Soviet past with the forces of the emerging market. The former influence preferences of the labor market participants, while the latter form labor demand curves explained by the competitive theory of discrimination and labor supply curves explained by the compensating differentials model.
These results imply that policies aimed at narrowing the gender pay gap in the Russian transition economy should be focused on decreasing the degree of occupational and industrial segregation by gender rather than on equalizing women's and men's human capital endowments. The results of the study also suggest that measures seeking to reduce the incidence and degree of domination of men or women in particular industries and occupations through improving and leveling labor conditions in all industries and occupations for both men and women and through changing the images of the occupations that have been traditionally "male" or "female" are more likely to improve the relative earnings position of Russian women than attempts to introduce special labor legislation protecting women and new laws against discrimination by employers.
An interesting finding of the study is that women have kept their high prevalence in such industries as finance and real estate and in such occupations as economist and accountant. Relatively low-paid in the past, these industries and occupations have become prestigious and highly paid during the transition and no longer fit the traditional stereotype of "female" jobs. This, however, does not significantly influence the gender earnings differential yet, since the shares of total employment in these industries and occupations remain small.
The data used in this study may be obtained from the RLMS Web site (http://www.cpc.unc.edu/projects/rims/data.html). The computer programs (SPSS syntax files) used to process the data and generate econometric results are available from the author on request.
1 The tariff system specified wage rates for each industrial branch and each job within a branch. The actual wage rates were defined by multiplying these base rates by special uniform coefficients that reflected workers' skills, labor conditions, and regional settings.
2 Nearly 60% of the men and 50% of the women who participated in an immigrants survey reported that Soviet women had fewer opportunities than men to hold responsible positions in the economic bureaucracy (Linz 1996).
3 In the context of a worsening demographic situation and decreasing birth rate, the legislative acts concerning women passed in this period contained mainly provisions for women with children, including an increased leave payment and the right for part-time work at the current place of employment. A rather progressive change in the legislation, the provision that allowed the father, grandparent, or other person to take care of the child, had only an ideological value, since no special measures to motivate men to take the leave were introduced (Posadskaya 1993:167-69).
4 The Taganrog data may be unrepresentative nationally, since they are rather specific to a medium-sized southern industrial city and were collected in a limited number of state-owned enterprises and organizations, about 50% of which belonged to heavy industry.
5 According to the Russian Labor Flexibility Survey (RLFS), although 56% of industrial firms still used the tariff wage system in 1994, it was not used consistently. based on the statutory minimum wage, which was held far below the subsistence level, the tariff system lost touch with actual wages and could no longer perform its centralizing function (Standing 1996:115, 116).
6 In a World Bank survey of Russian industrial firms, 93% of the respondents referred to available resources as a wage benchmark; 78% of the respondents named firm revenues as a wage constraint, while only 3% mentioned regulation (Commander and Coricelli 1995:179). In the RLFS, 62% of the respondents stated that price was the only factor that influenced their enterprise's wages, 18% said that productivity was taken into account, and 10% referred to the expansion of production (Standing 1996:115, 116).
7 According to the World Bank survey, in over 83% of cases wages were set by the enterprise administration, and only in 17% of cases was there any form of explicit bargaining (Commander and Coricelli 1995:179). About two-thirds of respondents said worker demands through trade unions or collectives were of no importance in the wage setting (Commander et al. 1996:30-31).
8 Starting wages of the grading schedule in budget organizations corresponded to the minimum wage. The amount paid to other grades of employee was determined by multiplying the minimum rate by the ratio provided for their respective category. For example, a teacher or a doctor in the 10th category received 4 times the minimum wage.
9 The World Bank Survey showed that between 1990 and 1994 the Gini coefficient on wages for all employees rose from 0.22 to 0.32 and the decile ratio went from 2.1 to 4.9 (Commander et al. 1997:11).
10 For example, in October 1994, the statutory minimum monthly wage was 20.5 thousand rubles, while the average monthly wage was 265 thousand rubles (Russian Economic Trends).
11 The survey has been coordinated by the Carolina Population Center (CPC) at the University of North Carolina at Chapel Hill.
12 The following outline of the RLMS sampling techniques draws on the detailed project description provided by the CPC team (http://www.cpc.unc.edu/rims/).
13 To account for a significant percentage of "discouraged workers," the participation rates and the unemployment rates are calculated defining the labor force as those who either work or want to find a job.
14 The estimates of earnings and, consequently, of the gender pay differential presented in Table 2 may be biased, since the workers who happened to receive their earnings in full may not adequately represent all those who worked. This problem is addressed in the following section.
15 General secondary education in Russia (10-11 years of schooling) is, approximately, equivalent to high school in the United States. Ordinary vocational schools (PTU) offer vocational training based on incomplete secondary education (7-8 years of schooling). Secondary vocational schools provide general secondary education along with vocational training. Students of specialized secondary schools acquire some special (technical, medical, pedagogical, art) knowledge, along with or in addition to general secondary education. Specialized secondary education is considered superior to vocational schools. The latter, especially ordinary vocational schools, have a reputation as a place for general school drop-outs with low academic qualifications.
16 The segregation index represents the proportion of workers who would have to change jobs for the occupational distribution of men and women to be the same (Duncan and Duncan 1955). The industrial segregation indexes at the one-digit level of aggregation computed by Blau and Kahn (1996) are 0.320 for Germany, 0.247 for Hungary, 0.426 for Sweden, 0.291 for Switzerland, 0.349 for the United Kingdom, and 0.343 for the United States.
17 The rest are employed at firms or organizations with mixed or ambiguously defined ownership type.
18 Occupations were coded by the RLMS team according to the four-digit International Standard Classification of Occupations: ISCO-88 (Geneva: International Labour Office, 1990). Considerable care was devoted to taking into account the idiosyncrasies of the Russian labor market.
19 The gender segregation indexes at the one-digit level of aggregation computed by Blau and Kahn (1996) are 0.422 for Germany, 0.408 for Hungary, 0.461 for Sweden, 0.322 for Switzerland, 0.440 for the United Kingdom, and 0.357 for the United States.
20 Much of the methodology outlined here is drawn from Ashraf (1996).
21 This method of nondiscriminatory wage structure estimation was originally proposed by Neumark (1988).
22 The specification of Z is based on the understanding of the wage arrears problem in the Russian economy outlined above. Vector Z includes time dummy variables that relate to the reference months, a dummy for part-time status, and dummies for industry, ownership, occupation, and region. The detailed specification of Z and discussion of the sample selection results are available from the author upon request.
23 Although decomposing the last term in equation (6) in the same manner as the rest of the equation is a common practice, the results it yields are not very meaningful. The technique of the gender earnings differential decomposition in the presence of selectivity bias suggested in this study has not yet been proposed in the literature.
24 This is a modification of the technique suggested by Blau and Kahn (1996), who used straight hours (not log hours) in the regression equation. This implies an exponential relationship between hours and earnings, which is not theoretically justified.
25 Variation in educational attainment within the incomplete secondary school cannot significantly influence the regression results, since only 0.4% of women and 2.1% of men have not finished this school.
26 Gill (1994) suggested a method of separating the contribution of occupational differences to the pay differential (racial in his study) into a part not attributable to discrimination (characteristics and choice) and a part due to differential access to occupations. The method is based on analyzing individuals' occupational aspirations using multinomial logit techniques to identify the determinants of the probability that an individual desires employment in an occupation. This analysis, however, could not be implemented in the present study, since the RLMS data sets do not contain the requisite data on individuals' occupational aspirations.
27 Jones (1983) has shown that while the "discrimination" term as a whole can be given consistent and meaningful interpretation, this term "cannot be further decomposed in any intelligible fashion," since the structure of the coefficient estimates critically depend on the model specification - in particular, on the reference points for the dummy variables structures.
28 The gender earnings ratios calculated by Blau and Kahn (1996) with adjustment for hours are 77.3% for Sweden, 72.7% for Austria, 70.2% for Germany, 65.4% for the United States, 64.9% for Hungary, 64.6% for Switzerland, and 61.4% for the United Kingdom.
29 Experimenting with the regression model has shown that occupational segregation remains a key contributor to the gender earnings differential even if the occupation dummy variables reflect only standard one-digit categories. In this case, however, occupational segregation explains only 29.0% of the gender pay gap, versus 49.9% explained by the model suggested in the paper. This confirms that "female" and "male" occupations within the one-digit categories are an important determinant of the gender earnings differential. The detailed results of the decomposition of the gender earnings differential with one-digit occupations in the regression model are available from the author upon request.
30 In Newell and Reilly's (1996) study of the gender pay gap in Russia in the immediate aftermath of the Soviet era, gender differences in worker characteristics and, in particular, in occupational distribution explain only a small proportion of the overall gender earnings differential. Comparisons of their results with the results in the present paper should take into account the following. First, since in Newell and Reilly's model the only education variable is years of high school completed, the model fails to account for more important human capital factors such as college education. Second, in order to facilitate the estimation of occupation-specific wage equations, the authors aggregated occupational groups even beyond the one-digit level, leaving out occupational segregation by gender that occurs within these highly aggregated groups. Further, Newell and Reilly's model does not include industry variables. Finally, as the authors acknowledged, their occupation-specific wage equation estimates exhibit poorly determined skill price coefficients and generally provide inadequate fits to the data.
31 While it may be argued that the results of both surveys are likely to understate the degree of employer discrimination, since people may be reluctant to admit prejudices based on stereotypes, this bias is not likely to be large. As Standing (1996:280) correctly noted, there is nothing illegal or improper in admitting a preference, nor is it in the Russian character to be reticent about voicing it.
Anker, Richard. 1997. "Theories of Occupational Segregation by Sex: An Overview." International Labour Review, Vol. 136, No. 3 (Autumn).
Ashraf, Javed. 1996. "Is Gender Pay Discrimination on the Wane? Evidence from Panel Data, 19681989." Industrial and Labor Relations Review, Vol. 49, No. 3 (April), pp. 537-46.
Becker, Gary S. 1971. The Economics of Discrimination, 2nd ed. Chicago: University of Chicago Press.
Blau, Francine D., and Marianne Ferber. 1987. "Discrimination: Empirical Evidence from the United States." American Economic Review: Papers and Proceedings; Vol. 77, No. 2, pp. 316-20.
-----. 1992. The Economics of Women, Men, and Work, 2nd ed. Englewood Cliffs, N.J.: Prentice Hall.
Blau, Francine D., and Lawrence M. Kahn. 1996. "Wage Structure and Gender Earnings Differentials: An International Comparison." Economica, Vol. 63, pp. S29-S62.
Blinder, Alan S. 1973. "Wage Discrimination: Reduced Form and Structural Estimates." Journal of Human Resources, Vol. 8, No. 4, pp. 436-55.
Bloom, David E., and Mark R. Killingsworth. 1982. "Pay Discrimination Research and Litigation: The Use of Regression." Industrial Relations, Vol. 21, No. 3, pp. 318-39.
Brown, Randall S., Marilyn Moon, and Barbara Zoloth. 1980. "Incorporating Occupational Attainment in Studies of Male-Female Earnings Differentials. "Journal of Human Resources, Vol. 15, No. 1, pp. 3-28.
Clarke, Simon. 1997. "The Financial System and Demonetarisation of the Economy." Paper prepared in support of the ILO/ICFTU campaign on the non-payment of wages in Russia. Centre for Comparative Labour Studies and the Institute for Comparative Labour Relations Research (ISITO), Moscow.
Commander, Simon, et al. 1996. "How Russian Firms Make Their Wage and Employment Decisions." In Simon Commander, Qimiao Fan, and Mark E. Schaffer, eds., Enterprise Restructuring and Economic policy in Russia. Washington, D.C.: World Bank, pp. 15-51.
Commander, Simon, John McHale, and Ruslan Yemtsov. 1995. "Russia." In Simon Commander and Fabrizio Coricelli, eds., Unemployment, Restructuring, and the Labor Market in Eastern Europe and Russia. Washington, D.C.: World Bank, pp. 147-92.
Cotton, Jeremiah. 1988. "On the Decomposition of Wage Differentials." Review of Economics and Statistics, Vol. 70, No. 2, pp. 236-43.
Dolton, Peter S., and Michael Kidd. 1994. "Occupational Access and Wage Discrimination." Oxford Bulletin of Economics and Statistics, Vol. 56, No. 4, pp. 457-74.
Duncan O. D., and B. Duncan. 1955. "A Methodological Analysis of Segregation Indexes." American Sociological Review, Vol. 20, pp. 210-17.
Gill, Andrew M. 1994. "Incorporating the Causes of Occupational Differences in Studies of Racial Wage Differentials." Journal of Human Resources, Vol. 19, No. 1, pp. 21-41.
Goskomstat SSSR. 1990. Narodnoye Khozyaystvo SSSR v 1989 g. Moscow: Finansy i Statistika.
Groshen, Erica. 1991. "The Structure of the Female/Male Wage Differential." Journal of Human Resources, Vol. 26, No. 3, pp. 457-72.
Gunderson, Morley. 1989. "Male-Female Wage Differentials and Policy Responses." Journal of Economic Literature, Vol. 27, No. 1, pp. 46-72.
-----. 1994. Comparable Worth and Gender Discrimination: An International Perspective. Geneva: International Labour Office.
Heckman, James. 1979. "Sample Selection Bias as Specification Error." Econometrica, Vol. 47, No. 1, pp. 153-61.
Jones, F.L. 1983. "On Decomposing the Wage Gap: A Critical Comment on Blinder's Method." Journal of Human Resources, Vol. 18, No. 1, pp. 127-30.
Katz, Katarina. 1994. "Gender Differentiation and Discrimination: A Study of Soviet Wages." Diss., Goteborgs Universitet (Sweden).
-----. 1997. "Gender, Wages, and Discrimination in the USSR: A Study of a Russian Industrial Town." Cambridge Journal of Economics, Vol. 21, No. 4 (July), pp. 431-52.
Kidd, Michael P., and Michael Shannon. 1996. "Does the Level of Occupational Aggregation Affect Estimates of the Gender Wage Gap?" Industrial and Labor Relations Review, Vol. 49, No. 2, pp. 317-28.
Lapidus, Gail Warshosky. 1993. "Gender and Restructuring: The Impact of Perestroika and the Aftermath on Soviet Women." In Valentine M. Moghadam, ed., Democratic Reform and the Position of Women in Transitional Economies. Oxford: Oxford University Press.
Linz, Susan. 1996. "Gender Differences in the Russian Labor Market." Journal of Economic Issues, Vol. 30, No. 1, pp. 693-716.
Manski, Charles F. 1989. "Anatomy of the Selection Problem." Journal of Human Resources, Vol. 24, No. 3, pp. 344-60.
McAuley, Alastair. 1981. Women's Work and Wages in the Soviet Union. London: George Allen & Unwin.
Miller, Paul. 1987. "The Wage Effect of Occupational Segregation of Women in Britain." Economic Journal, Vol. 97, No. 388, pp. 885-96.
Neumark, David. 1988. "Employers' Discriminatory Behavior and the Estimation of Wage Discrimination." Journal of Human Resources, Vol. 23, No. 3, pp. 279-95.
Newell, Andrew, and Barry Reilly. 1996. "The Gender Wage Gap in Russia: Some Empirical Evidence." Labour Economics, Vol. 3, No. 3 (October), pp. 33756.
Oaxaca, Ronald L. 1973. "Male-Female Wage Differentials in Urban Labor Markets." International Economic Review, Vol. 14, No. 3, pp. 693-709.
Oaxaca, Ronald L., and Michael R. Ransom. 1994. "On Discrimination and the Decomposition of Wage Differentials." Journal of Econometrics, Vol. 61, pp. 521.
Ofer, G., and A. Vinokur. 1992. The Soviet Household under the Old Regime: Economic Conditions and Behavior in the 1970s. Cambridge: Cambridge University Press.
Posadskaya, Anastasia. 1993. "Changes in Gender Discourses and Policies in the Former Soviet Union." In Valentine M. Moghadam, ed., Democratic Reform and the Position of Women in Transitional Economies. Oxford: Oxford University Press.
-----, ed. 1994. Women in Russia. A New Era in Russian Feminism. London: Verso.
Reskin, B., et al. 1986. Women's Work, Men's Work. Washington, D.C.: National Academy of Sciences.
Russian Economic Trends. Working Centre for Economic Reform, Government of the Russian Federation, Russian European Centre for Economic Reform. (On line version: http://cep.lse.ac.uk/datalib/ret/)
Standing, Guy. 1994. "The Changing Position of Women in Russian Industry: Prospects of Marginalization." World Development, Vol. 22, No. 2, pp. 271-83.
-----. 1996. Russian Unemployment and Enterprise Restructuring: Reviving Dead Souls. New York: St. Martin's Press.
Constantin G. Ogloblin is a Ph.D. candidate at Kent State University and Visiting Instructor of Economics at John Carroll University. This research was supported by funds from the U.S. Agency for International Development. The author thanks Gregory Brock, whose initiative made this study happen, and Donald Williams for helpful comments.
|Printer friendly Cite/link Email Feedback|
|Author:||Ogloblin, Constantin G.|
|Date:||Jul 1, 1999|
|Previous Article:||What is a promotion?|
|Next Article:||Disability retirement among German men in the 1980s.|