NOT JUST EDUCATION: GENDER WAGE GAP IN THE ALBANIAN LABOUR MARKETS THROUGH OCCUPATIONAL SEGREGATION, WORK EXPERIENCE, AND CHILD CARE

Transcription

1 NOT JUST EDUCATION: GENDER WAGE GAP IN THE ALBANIAN LABOUR MARKETS THROUGH OCCUPATIONAL SEGREGATION, WORK EXPERIENCE, AND CHILD CARE JUNA MILUKA*, CAREN GROWN** *Department of Economics, Business and Administrative Sciences, University of New York, Tirana, Albania **American University, Economics Department, Washington, DC, USA Paper prepared for presentation at the World Bank International Conference on Poverty and Social Inclusion in the Western Balkans WBalkans 2010 Brussels, Belgium, December 14-15, 2010

2 Copyright 2010 by author(s). All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. 2

3 NOT JUST EDUCATION: GENDER WAGE GAP IN THE ALBANIAN LABOUR MARKETS THROUGH OCCUPATIONAL SEGREGATION, WORK EXPERIENCE, AND CHILD CARE Juna Miluka University of New York Tirana Caren Grown American University Abstract Understanding the mechanisms of gender discrimination in the labour market is especially important due to the influences that it may have on income inequality, education of future generations, occupational distribution, women s position and opportunities, intergenerational inequality, and ultimately poverty. Specific to the case of Albania, this issue becomes especially important given the prospect of the country in joining the European Union. If existing issues of inequality are not considered, Albania runs the risk of increasing wage inequalities even further if women do not have the needed skills for the European markets and become even less competitive. This paper estimates the gender wage gap, and the different sources that account for such disparity. The paper finds that education is key and should be given special consideration by the policy makers, but other important sources such as occupational segregation, work experience, and child care play a crucial role. Key Words: gender wage gap, labor market, occupational segregation, education, gender differences, Albania J01, J31, J71 Corresponding Author. Dr. Juna Miluka is an Assistant Professor at the University of New York Tirana. Rruga: Komuna e Parisit, Tirana, Albania. Tel: ; Fax: ; junamiluka@unyt.edu.al Dr. Caren Grown is an Economist in Residence at American University, Washington, DC. 3

4 1. Introduction The issue of wage discrimination in the labour markets appears to always be in style in the runways of economics. Despite the many studies conducted on the topic and the well documented procedures, it continues to be very relevant since wage discrimination is still part of the labour markets. In the context of a developing country, this topic becomes even more important since in combination with other aspects of a developing economy, it may have stronger and lasting impacts. One such country, for which empirical studies regarding issues of labour markets in general, and wage discrimination in particular are lacking, is Albania. Albania underwent major transformations, which in part also translated to transformations in the labour market as a result of the change from centralized planning to an open market economy. In the midst of these transformations, women found themselves in a more critical position due to their dual burden inside and outside of the household. In order to understand the factors responsible for the unequal rewards between males and females, it is of interest to analyse the position of women in the labour market as measured by the gender wage gap and wage decomposition. Understanding discrimination in the labour market is especially important due to the influences that it may have on income inequality, education of future generations (stemming from the expectations formed by the parents), occupational distribution, women s position and opportunities, intergenerational inequality, and ultimately poverty. Before 1990, the labour market in Albania was characterized by state controlled individual decision-making and a high degree of centralization. In the early 1990s, the labour market was liberalized; this period was followed by a high degree of unemployment due to the shutting down of major industries and overcrowding in administrative jobs. From the mid 1990s on, there has been a considerable increase in private businesses mainly due to the remittances by the emigrants; however, the rate of long term unemployment still remains high especially among women (Cuka et al., 2003). 4

5 5 The transitional period in Albania was characterized by a large vacuum, which brought back traditional law, increased vulnerability of women in the labor market, and deteriorated economic status (Lawson and Saltmarshe, 2000). In this new environment, women faced a critical situation since state enterprises, which employed the majority of women, collapsed and so did social protection associated with these jobs. In addition, market reforms increased earning inequalities through wage and price liberalizations, and changed the characteristics of employment (The World Bank, 2002). The unavailability of social protection and economic rights reinforced women s homemaking roles (Tarifa, 1994). Consequently, women faced more work within the family, but less mobility and chances to find jobs. With the state stopping child care provision or long paid maternity leaves, women s schedules required more accommodation giving rise to long-term structural discrimination in the labor market (Kligman, 1996). With the additional burdens placed on them, women who already during communism despite their education levels were in lower occupation and were being paid less, became and easily targeted group for discrimination. Statistics from the Albanian Ministry of Labour and Social Affairs (2005) show that in 2004, the employment levels were 38.3 percent for women and 60.1 percent for men; whereas unemployment was 17.5 percent for women and 12.4 percent for men. In terms of participation in the labour force, in 2003, 70.5 percent of men in the working age participated in the labour force compared to 46.7 percent for women (Albanian Ministry of Labour and Social Affairs, 2005). In 2004, men were at 68.6 percent, while women were at 46.4 percent. Women still overwhelmingly remain in the social-state-service sector, where they comprise 80 percent of employees (Albanian Ministry of Labour and Social Affairs, 2005). Men are 2 times more in lawmaking, senior officials, leading executives, while women are mostly found as specialists and regular employees (Albanian Ministry of Labour and Social Affairs, 2005). As a result of occupational segregation, women s wages are also lower than those of men. In addition, in the

6 6 urban areas females frequent universities more than men as a way to deal with the harder employment opportunities that they face in the labour market (The World Bank, 2002). Not only the market, but also curriculum training, promotes occupational segregation. Statistics from the Albanian Ministry of Labour and Social Affairs show that in 2004 the vast majority of vocational training programs for females were concentrated in English, Italian, computer, secretary, sewing, and cosmetics. Only three courses offered skill training in trades, which makes it easier to find a job in the labour market. There is a tendency for the vocational training to be in the traditional female fields (AMoLSA, 2005). Furthermore, the educational curriculums retain the reproductive and gender stereotypes in the selection of the fields of study (AMoLSA, 2005). In this way gender roles and occupational segregation are promoted. There is a lack of career orientation related to the skills required in the labour markets; rather it still focuses on traditional vocations (AMoLSA, 2005). Given that Albania had achieved almost universal literacy rates, high levels of women s education, high participation in the labor force, and extensive child care and maternal health during communism, it has often been trivial for policy makers to acknowledge the true position of women and the well masked burdens of paid and unpaid work. Not recognizing that women might be in a less favorable position than men, and not adequately addressing such issue leads to the gender imbalances. As a result of changes in the labor markets in Albania since the fall of the communist period and the need for empirical investigation on the issue of wage discrimination in the labor markets as a way of unraveling the position of women in them, this paper draws on the literature and methodology of wage decomposition to apply it to labour markets in Albania. More specifically it will estimate the gender wage gap, and the different sources that account for such disparity.

7 7 2. Brief Literature Review During communism in almost all of the transition economies, wages were set according to the industry specific wage grids responding only to worker s education and experience (Munich et al., 1999). There was a policy of full employment, and women enjoyed high education and health care access (Munich et al., 1999). The fall of communism ended the wage regulation, responding to increased returns to education and causing increased wage dispersion (Svejnar, 1999). As a result of skill composition, labor market institutions, and specific country history and culture, gender wage dispersion has affected countries differently. During the early transition, the gender wage gap diminished in Eastern Europe, but it increased in Russia and Ukraine (Brainerd, 2000). During mid transition, Newell and Reilley (2000) report that the gender wage gap has remained relatively stable for most countries in the transition economies. Human capital, work experience, occupational segregation, social norms, and household characteristics are all possible sources of the gender wage gap. Human capital theory suggests that wages are a function of education and work experience (Mincer, 1974). Therefore, changes in human capital are expected to lead to differences in wages. There are increasing returns to education, which implies that highly skilled workers get paid more than low skilled. Returns to education more than doubled in Romania in 2000 compared to the levels under central planning (Andren et al., 2004). Skill related wage differences rose in the transition economies following the system change (Svejnar, 1999). In the case of Bulgaria, Giddings (2002) shows that the high levels of human capital that women had acquired during communism helped them in the transitional period by favoring their earning and improving their economic conditions. Women s higher human capital helped reduce the gender wage gap in Russia (Oglobin, 2005). In the case of Albania, as in many other transitional economies, the policies of the communist regime provided free education for all, full labor force participation, and a fairly small gender pay gap (Mango and Silova, 2004). As a result of the high education levels inherited by women educated

8 8 in the communist period, we would expect human capital to play no role or favor women in the wage gap. Work histories or work experience is also expected to affect the gender wage gap. Women are found to have higher home time than men due to the responsibilities of child bearing and rearing (Kunze, 2000). Whereas men usually have less frequent interruptions in the labor market, women have interrupted work histories due to the family responsibilities (Kunze, 2000). Since work experience is also one of the main component influencing wages, lower or interrupted work experience will be rewarded less by the market, thus influencing the gender wage gap. Since women often have lower or interrupted work experience, we would expect the gender wage gap to be in their disfavor. With the fall of communism, the social support for child care suffered great losses. Day care was no longer provided by the state; rather it became the individual s responsibility. Since women remain the main care provider, child bearing and rearing places additional demands on women, requiring them to take time off from the labor market. Thus, we would expect the interrupted work experience of women to increase the gender wage gap in Albania. Occupation segregation by industry or job status, also contributes to the gender wage gap. Women comprise the majority of workers in the service occupations, while men are largely found in manufacturing jobs and industries (Kunze, 2000). Differences in occupation bring forth differences in wages for two reasons. First, different occupations require different skills, and since skills in certain occupations are paid more than skills in others, occupational segregation may increase the wage gap (Oglobin, 1999). Thus, if women are segregated in the lower paying occupations, they will receive lower wages. Second, from simple supply and demand analysis we know that if demand stays the same, but supply increases, prices will fall. If women are concentrated in certain occupations, then supply for those occupations increases giving employers some degree of monopsony power (Joshi and Paci, 1998), thus pulling wages down. Jurajda

9 9 (2003) finds that in the Czech Republic and Slovakia, occupational segregation explained over 1/3 of the gender wage gap. Oglobin (1999) finds that gender differences in education and work experience are not enough to explain the gender wage gap in Russia; rather it is occupation segregation which is the main determinant of the gender wage gap. Occupation segregation is found to account for percent of the gender wage gap in Russia (Oglobin, 1999; Oglobin, 2005). As a result of the similarities of labor market practices across the region, occupational segregation is expected to play a major role in most of the transition economies (Oglobin, 1999). Indeed, in Albania women are mostly concentrated in the service industry, and largely employed by the public administration (Albanian Ministry of Labor and Social Affairs, 2005). We expect a lager gender wage gap as a result of occupational segregation in Albania. Social norms may also affect the gender wage gap by promoting occupational segregation. Social norms may dictate certain gender roles, which influence concentration of women in certain occupations, thus increasing occupational segregation (Oglobin, 1999). Social norms may also affect employer preferences who may see women better fit for certain occupations than men. For the case of the Czech Republic and Slovakia, employers strongly preferred men to women in many occupations (Svejnar, 1999). In addition to the usual preference for men in maintenance and repair, employers also preferred male to female employers in professional, administrative, and service occupations; 36 to 58 percent preferred males, while under 10 percent preferred women (Svejnar, 1999). The legacy of social norms placing women in certain occupations that are regarded as socially fit for females is also expected to be found in Albania. Thus, social norms are expected to widen the gender wage gap in Albania. Household decision making regarding labor market choices may override the individual choice. Family and housework responsibilities are found to explain a large part of the unadjusted gender wage gap (Andren and Andren, 2007). Women often make a choice regarding their occupation dictated by their household characteristics. Having children increases preferences for

10 10 participation of women in the public sector employment because it provides more flexibility (Gang et. al., 2006). This also relates to the lack of social support for women in terms of day care and to the role of women as the main child care providers. Thus, household characteristics may also influence women s preference for certain types of jobs, often including public sector which might provide more flexibility but lower pay. As a result, gender pay gaps might arise. Albania is no exception, thus we would expect the gender pay gap to increase as a result of having children for women. 2.1 Methodological Review The literature on wage decomposition and the methodologies followed to implement such procedures are very vast. Many studies in labour economics have measured returns to education, and wage discrimination in the labour market. As a response to the need to understand the sources of labour market discrimination, wage decomposition methods were developed as to shed light on the processes of discrimination faced by different groups in the labour markets. Although the procedures of wage decomposition are now well documented and accepted over the field, ranging from parametric to non-parametric estimates, the importance and relevance of these methods continues to be very prominent. Even though many studies have been conducted to measure wage discrimination between men and women, black and white and other groups, the topic of trying and unravel the differences in wages is still very relevant. The milestone of the literature on wage decomposition is by Oaxaca (1973) and Blinder (1973) which measures the sources of discrimination in the labour market. Oaxaca (1973) measured the average extent of discrimination against female workers in the United States and provided a quantitative assessment of the sources of male-female wage differentials using data from the from the 1967 Survey of Economic Opportunity. He estimated the effects of discrimination as the residual left after adjusting for sex differential for differences in various

11 11 characteristics. Like other previous studies, Oaxaca (1973) found that the sex differential in wages was quite large, and that unequal pay for equal work was not the principal factor in explaining the wage differentials. Rather, it was the concentration of women in lower paying jobs that was responsible for the large pay differentials. His results suggested that the bulk proportion of the male-female wage differential was due to the effects of discrimination. At the same time, Blinder (1973) developed a similar methodology, but extended the analysis by making a distinction between reduced form and structural wage equations, supplying estimates of both and examining the differences in results. In addition to male-female wage differentials among whites, he also analyzed white-black differentials among men. Blinder (1973) found that synthesizing both reduced form and structural wage equations, discrimination of various sorts (including occupational status and job seniority) account for 70 percent of the overall race differential, and 100 percent of the overall sex differentials. Subsequent to these earlier studies, other techniques were developed providing further improvements. One such methodology by Brown, Moon, and Zoloth (1980) incorporated the treatment of occupations as endogenous by supplementing the wage equation with an occupation selection equation (Kunze, 2000). Although this methodology improves upon accounting for the share of the gender wage gap due to within and between occupations (Kunze, 2000), it does not address the treatment of residuals. An alternative method that takes into account the residual distribution was developed by Juhn, Murphy, and Pierce (1993). This methodology is best for two or more time periods, and two or more countries (Kunze, 2000). Although Juhn, Murphy, and Pierce (1993) advanced in the residual distribution, they lacked in the modelling of the covariates (Lemieux, 2002). Lemieux (2002) is regarded as the superior method since it accounts for both the residual and covariate distribution. Juhn, Murphy, and Pierce (1993) do not explicitly account for the changes in the covariates; rather they model them as the unexplained partion after accounting for the changes in the residual (Lemieux, 2002). Therefore, the changes

12 12 in the covariate distribution in the Juhn, Murphy, and Pierce (1993) methodology are a mixture of the true effects of changes in the covariate distribution and the changes unexplained by the other factors (Lemieux, 2002). Lemieux (2002) fixes this problem by modelling residuals as in the Juhn, Murphy, and Pierce (1993) methodology, and incorporating the weighting method of changes in the covariate distribution as in DiNardo, Lemieux, and Fortin (1996). The main contribution of Lemieux (2002) stands in the: 1) simplified and unified procedure using standard OLS regressions augmented by a logit or probit model, and 2) in the extension of the approach to model residuals as a function of unmeasured skills and skill prices, thus considering them not just as the unexplained part of the regression, but rather as the pricing of unmeasured skills giving them empirical content. For the above explained reasons, and since the Oaxaca, Blinder (1973) decomposition is the most used method for cross sectional data, and Lemieux (2002) is the latest decomposition technique unifying the decomposition methods, we use both techniques for comparison. 3. Data The data for this paper is obtained from the 2005 ALSMS (Albanian Living Standard Measurement Survey study) conducted by INSTAT (Albanian Institute of Statistics) with technical supervision by the World Bank. The 2005 ALSMS is a standard household survey, which in addition to the usual household roster, community characteristics, food consumption, and the like, also includes a labour module. The labour module provides various information on labour force participation, hours worked, wages, types of jobs, employment grid, etc. The sample is stratified into four regions: Coastal, Central, Mountain, and Tirana. It contains 3,680 households from which 5,540 individuals are of the ages 15 and above and included in the labor module. A total of 1,829 individuals included in the labour module report no wages, thus leaving the final sample with 3,703 individuals included in the labour module and that have reported

13 13 wages. There are 1,111 females and 718 males in the labour module that report no wages. The average education of those that do not report any wages is of 8.13 years of education. Descriptive statistics in Table 1 show a statistically significant wage difference both in log and non-log terms of wages between males and females. Males average monthly wages (28,657 Albanian New Lek) are over 1.7 times higher than those of females (16,907 Albanian New Lek). A first peak at the signs of discrimination in the labour market is seen in the rewards to education between the two sexes. Figure 1 shows that across ages, women have on average more education than men, with the exception of age 65 and above, where there are much lower levels of education for women than men. This age category pertains to the era of education prior to the Communist system in Albania, which is known for very large gender inequalities in education. Females have higher education on average, years as compared to the years of males, however as seen from the wages above, the market does not reward these assets equally. Notably from Table 2, males receive higher monthly wages in all of the education categories, with the exception of having no diploma, having studies abroad, or having conducted graduate studies in Albania, for which no significant differences are found. In order to see the wage differences by educational categories in amounts, the same table, but with monthly wages in New Lek 1 is presented below, Table 3. Table 4 shows that a potential reason through which the market rewards differently is the different occupations held by males and females. The overwhelming majority of the highly paid jobs such as legislators, senior officials and managers are held by the males. Males occupy 6.20 percent of these positions, while females occupy only 1.95 percent. There is an also large difference in another well paid occupation category, such as craft and related trades. Men in this 1 Average December, 2005: 1 USD= Albanian New Lek.

14 14 latter category represent percent, while women represent only percent. Women comprise the majority of professionals, technicians and associate professionals. They are also found in the majority of the service, clerks, and elementary occupations. Table 5shows the monthly log wages by gender and occupation. With the exception of clerks, where we find no significant differences, throughout the occupations, women are rewarded less than men. The concentration of women in occupations that do not fairly reward them for their skills can partly affect the age wage profile. Figure 2, shows that across age levels (with an exception between the ages of 60 and 65) women s wages consistently stay below those of men. Furthermore, while men s wages seem to remain fairly constant from age 40 on to the retirement period, women s wages show greater fluctuations. In addition to the occupational issue, another possible factor influencing the disparity of wages between the two groups is the difference in experience, which is associated with job tenure. As seen in Table 1, and Figure 3, women have on average 4 years less experience than men. Women start out with higher levels of experience, but then, after reaching the peak, their experience is consistently lower than that of men. The lower experience, which can be a result of child bearing and lack of social support in child care, is associated with lower tenure on the job, and thus with lower pay. 4. Econometric Model In this paper we use both the Oaxaca, Blinder (1973) and Lemieux (2002) methodologies to analyse the gender wage gap and its decomposition in the labour market in Albania. As noted earlier, the Lemieux (2002) technique: 1) yields results which are easily interpretable and have economic meaning, 2) go beyond the decomposition of means to decomposing wages

15 15 and wage dispersion over the full distributional case, and 3) models residuals as the pricing of unmeasured skills, rather than as the unexplained part of the regression (Lemieux, 2002). 4.1 Oaxaca Blinder Decomposition Following Oaxaca-Blinder (1973), the wage differential between two groups, males vs. females in our specific case, may be decomposed in : 1) the proportion of the differential attributed to the shift of the coefficients b f 0 - b m 0, which is typically regarded as pure discrimination of the rent of being of a specific sex; 2) the explained part attributed to the differences in the coefficients b f i and b m i, and the differences in the average characteristics or endowments and ; and 3) the unexplained or interaction between the coefficients and the average characteristics. Thus, stemming from the 2 basic equation used in this analysis the human capital earnings function from Mincer (1974) (1) ln w = c + rs + b 1 E + b 2 E 2 + e, where w is hourly wage, c is a constant, S is years of schooling, E is years of experience in the labor market, and e is the error term we can write the raw wage differential as: (2) R= b f 0 + i b f i (b m 0 + i b m i ) = E + C + U where E= portion of differential attributed to differences in endowments (3) E = i b f i ( - ), C = portion of differential attributed to changes in coefficients (4) C= i (b f i - b m i ), U= the unexplained portion of the differential due to the shifts in the coefficients b f 0 - b m 0, and D= portion of the differential attributed to discrimination = C + U 2 The notation used in this section derives mainly from Lemieux (2002).

16 Lemieux Decomposition Following Lemieux (2002), and using standard OLS regressions augmented by a probit model, the gender wag gap is decomposed into 1) changes in the regression coefficients, 2) changes in the distribution of the covariates, and 3) changes in the residuals, which are modelled as a function of unmeasured skills and skill prices. More specifically, in this approach, we create counterfactual wages controlling for 1) changes in prices, b, 2) changes in endowments, x, and 3) changes in unobservable, u. The first step is to run separate OLS regressions for males and females. Keeping the same endowments and error terms from the female regression, we create female counterfactual wage regressions, using the b s from the male regression. This way we can see what the female wage equation would look like if females were paid according to male wages. After controlling for changes in the price of skills, we can control for changes in endowments by creating a female counterfactual wage that keeps the b s from the female wage equation, but that gives females the endowments, x, from the male wage equation. Thus, we can see how the average wages for females would change, were they to be paid according to the female wage equation, but having the endowments of men. In order to give females the average endowments of males, we run a probit equation on the entire sample of being female (using as many controls as possible), and use the propensity score to weight the female wage equation. Below we formally present the above summery of the methodology used Decomposition of Wages through Changes in the Regression Coefficients Referring to the previously mentioned wage equation from Mincer (1974) (1) ln w = c + rs + b 1 E + b 2 E 2 + e, where w is hourly wage, c is a constant, S is years of schooling, E is years of experience in the labour market, and e is the error term, let's consider a more general form of the above equation, (5) y if = x if β f + e if,

17 17 where i is an indicator for each individual and f stands for female (a regression equation for females), x if is a 1 X k vector of covariates (including a constant), β f is a k X 1 vector of parameters, and e if assumed to have E(e if x if ) = 0. In terms of our wage equation, y if is the log hourly wages for females, x if is a vector of human capital and other control variables. The OLS estimated regression equation is (6) y if = x if b f + u if, where u if is the regression residual, which by construction is uncorrelated with the covariates and has a mean of zero. The sample average of y for females is (7) = b f, where = i ; = i. Consequently, we can apply the same equation to the earnings of males, in which case we would have a sample average of (8) = b m where m stand for male. Stemming from Oaxaca (1973) and Blinder (1973), we can decompose these changes in means as (9) = (b f b m ) + ( ) b m, where the first term on the right is the difference in the estimated parameters, and the second term is the difference in the mean values of the covariates between females and males. Another way of interpretation is that b m represents a counterfactual value of y that would be obtained if the parameters for females were replaced by the parameters of males. Going back to the wage equation, it represents the average wage for females if the returns to human capital were the same as those of males. This counterfactual can be written as

18 18 (10) = b m and it can be used to rewrite the decomposition of the difference between the average value of y for females and males, such as (11) = ( b f ) + ( b m ) = ( ) + ( ) The individual-specific counterfactual wage (9) y a if = x if b m + u if = y if + x if (b m b f ) can be computed either by obtaining the sample average of x if and applying = b m, or by computing directly the sample average of (12) = ω if y a if Decomposition of Wages Through the Distribution of Covariates The decomposition of wages through the distribution of covariates may be achieved by constructing a counterfactual weight ψ i, which yields the distribution statistic that would have existed if the distribution of x for the females had the same distribution as males. The main idea behind this type of decomposition rests in the estimation of a probit model in order to compute the re-weighting factor ψ i. The re-weighting factor ψ i is constructed by pooling together the male and female samples and estimating a porbit model for the probability of being male. Conditional on x the estimated probit model estimates the predicted probability of being a male. We can denote the predicted probability as (13) P if = Prob(sex=male x if ) and the re-weighting factor as (14) ψ i = [(1 - P if )/ P if ] X [P t /(1 P t )], where P t is the unconditional probability that an observation is male. This procedure has the advantage of not suffering from the dimensionality problem of a cell by cell procedure and it can

19 19 incorporate several controls by including various independent variables in the probit model. In this context the distribution of females with the distribution of covariates of males can be obtained by weighting y if by ψ i. There is no agreement in the literature on the inclusion of control variables in the wage regression (Kunze, 2000), leaving them to the discretion of the researcher and to the question that needs to be answered. In addition to the standard education and experience variables in our analysis we also include additional control variables. The number of children, and the person s marital status are included because they may serve as a measure of the implications that women s double burden may have on their wages. The lack of social support and state provided child care makes women the primary care givers of their children. Thus, the number of children reflects the cost of lost experience for women (Grimshaw and Rubery, 2002). A married women with children might be viewed from the employer as less productive, since she might need more time off work and be considered less dedicated to work due to her family engagements. As a result, the employer might offer women lower wages. On the other hand, a married man might be regarded as more stable and dedicated to work since it is the wife that is expected to take care of the household. Married men may also just receive preferential treatment (Weichselbaumer and Winter-Ebmer, 2005). The distance index and social capital index are included to control for the costs or benefits of social support. Women who live in areas with adequate transportation, and have social capital that facilitates child rearing might be more productive and mobile. We control for the percentage of females in each occupation as to control for occupational segregation. This variable has been widely used in the literature to capture female occupational segregation (Jurajda and Harmgart, 2007; Andren and Andren, 2007). Lastly, we control for regional differences, which can play a role in terms of market segmentation and supply side, as well as social, economic and cultural aspects.

20 20 5. Results The Oaxaca decomposition results over the entire group of workers show that the principal sources of the gender wage gap are education, work experience, occupational segregation, and number of children. In table 6, the wage differential between males and females, is decomposed into three parts accounting for: 1) differences in endowments, 2) differences in coefficients, and 3) the interaction between endowments and coefficients. The total difference in endowments is insignificant. This result is in accordance with the fact that women have on average more education than men presently in the labour market. As it has been found in other transitional economies with high levels of education for women, the differences in endowments do not contribute to the gender wage gap. The total difference in the regression coefficients between females and males, which account for the largest part of the decomposition favours males (-0.510). The difference in coefficients is interpreted as a form of discrimination applied by the market in offering different rewards for the same skills. It means that given women s endowments, the difference between what they are actually paid and what they would get paid if given the male wage structure is negative, indicating a superior wage structure for males. If women were paid men s wages for their endowments they would get paid more. Lastly, the interaction between endowments and coefficients, which is referred to as the unknown part of the regression favours women. The positive values of education in the detailed decomposition for endowments, coefficients, and the interaction indicate that the higher levels of education for women give them an advantage. However, education is not enough to make up for the other sources which negatively affect their wage structure. An important variable that accounts for a large part of the differential in wages is experience. Women have on average less experience than men, which is associated with the fact that women take time off for child bearing and rearing. This is negatively rewarded by the labour market putting women at a disadvantage in the economic ladder. The

21 21 impact of having children is negative and is another major factor negatively affecting coefficients and putting women at a disadvantage. If there is lack of social support and child care possibilities, then, having children is associated with a discontinuity of participation in the labour market, a decrease of the stock of human capital, and therefore lower rewards in the labour market. As elsewhere in the literature, occupation segregation for women is also found to have a large impact in widening the gender wage gap in Albania, possibly through crowding of women in certain occupations and lower wages. Lastly, the shift in the constant term (-1.042), which is usually attributed to pure discrimination in the labour market (Blinder, 1973) or else as pure premium of being a specific sex, largely favours males. As would be expected, the decomposition for the highly skilled workers (holding university degrees and above) shows a smaller gender wage gap (-0.309), table 4.7. Differences in the coefficients explain the gender wage gap mainly through the different rewards of education, while differences in endowments and interactions are insignificant. Work experience and occupational segregation remain important sources of the gender wage gap in the endowment differences, but they lose significance in the coefficient differences. Unlike the overall decomposition, the number of children is not significant, and being married has a positive effect. In the case of highly educated workers, it might be easier to overcome the lack of state support in child care. Highly educated women are less vulnerable to taking time off from the labour markets since they might have larger means of support. In the case of highly skilled workers, markets seem to be less discriminatory as also supported by the loss in significance of the constant term. The lower educated workers display the largest gender wage gap (-0.539). Table 8, shows the largest constant term shift (-1.054), and unlike the previous results differences in endowments favor men. The males in the low skilled workers continue to obtain a superior wage structure. The negative impact of occupational segregation and number of children is the largest for this group. This means that low skilled women suffer larger discrimination in the labour

22 22 market. They are more likely to suffer occupational segregation and more vulnerable to staying out of the labour market for longer periods due to child care responsibilities. The increased magnitudes with which occupational segregation and child care affect women of lower education reduces their wages and further increases the gender wage gap. Lastly, for the low education group as for the overall sample, the distance index matters for the difference in the coefficients, suggesting the importance of mobility. Turning to Lemieux (2002) decomposition, from table 9, column 1, the mean log wage difference between males and females in Albania is , which means that females earn approximately 36% 3 less than males. From column 2 we see that females have lower wage variance due to lower residual wage variance. The predicted variance is higher for females, suggesting higher between group inequalities, however the residual variance of wages is higher for males, suggesting larger within group inequalities. Unlike other cases where women have both lower returns to their skills and lower human capital, in the case of Albania women receive lower prices for their human capital, but they are relatively more educated than men, therefore they have a more compact distribution of covariates than men. Women in Albania are thus in the low wage/low dispersion, while males are the high wage/high dispersion. The results from the counterfactual analysis are given in rows 6, 7, and 8 of Table 9. As in the Oaxaca-Blinder (1973) decomposition, most of the wage gap between the groups is explained by the changes in the regression coefficients (row 6). It is interesting to see that the variance in row 7 column 2 has a negative sign, driven by the larger negative difference in the residual variance of covariates. This is consistent with human capital theory, which states that to the extent that residual wage dispersion is due to unmeasured differences in human capital skills, residual wage dispersion should increase when the price to human capital increases. Therefore, if 3 This number is calculated by taking the exponential of the mean log wage difference between females and males, subtracting 1, and multiplying by 100 to get the percentage value.

23 23 males receive higher returns to their measured human capital, the dispersion of their unmeasured human capital is expected to also be higher. Tables 10 and 11 show the Lemieux (2002) decomposition for the high and low educated workers, respectively. As we saw in the Oaxaca (1973) decomposition, the differences in the coefficients account for the majority of the gender wage gap. With the reduction of the gender wage gap, the wage dispersion also decreases. Women in the highly educated group have lower wage variance than in the case of all workers. Unlike the results in Table 9., the predicted and residual variance for highly educated females is lower than that of highly educated males, suggesting lower between and within group inequalities. For the lower educated group, as in the case of the Oaxaca (1973) decomposition, the majority of the gap is explained by differences in the coefficients in addition, the differences in endowments also explain some of the gender wage gap. The increase in the gender wage gap for this group is associated with larger wage dispersion. The total and predicted wage variance for females is larger in the low education group than in the high education group; the same is true for males. As in the overall case, women have higher predicted variance indicating higher between group inequality, and lower residual variance indicating lower within group inequality. From Figure 4 we can also see that the two wage distribution have quite different shapes. Visually, the gap between the two densities is the gender wage gap, which is much larger on the left hand side and middle of the distribution. The gender wage gap starts to shrink on the right hand side of the distribution, and it greatly vanishes for the top skilled individuals, suggesting that wages for the highest skilled women are similar to the wages for the highest skilled men. As Figure 5 shows, when females are given the regression coefficients of the males, the two distributions look almost identical, and the gender wage gap gets significantly reduced. Thus, it suggests that the b s account for most of the gender wage gap.

24 24 When females are given the covariates of the males as in Figure 6, their wage distribution becomes tri-modal, which suggests that if women were given the covariates of males and were paid according to their wage structure, they would score even lower. In this case females would have lower b s and lower covariates. When we look at figure 7, where females get both the b s and the covariates of the males, the figure looks closer to Figure 5, where only the b s are of the males. This finding suggests that the differences in the distribution of the covariates have a small impact on the wage distribution. This is in line with the earlier findings from the Oaxaca, Blinder (1973) decomposition, which showed that women s covariates are not enough to make up for the differences in the wage distribution. From figures 8-10 we see that in the highest education groups, the gender wage gap decreases. This is seen by the lower gap in the two distributions. In the case of the highly educated workers, since they share very similar levels of education, the two distributions look very alike in the case when females are given the b s of males, and in the case when females are given both the returns and covariates of males. Lastly, in figures 11-13, for the low education group we see that the b s account for most of the differences in distribution. When females are given the b s of males, the wage distributions are very similar; whereas when females have only the covariates of males, their wage distribution becomes bimodal. In this case they would get even lower mean wages. Giving low educated females both the b s and covariates of males reduces their mean wages. Lower endowments put them at a further disadvantage, for which the increased b s are not enough to make up for the difference. 6. Conclusions This paper provides a detailed account of the decomposition of the wage gap between men and women in the Albanian labour markets using two different estimation methodologies.

25 25 Using 2005 Albanian Living Standard Measurement Survey (2005 ALSMS) data, both the Oaxaca, Blinder (1973) and Lemieux (2002) wage decomposition techniques show the existence of pure labour market discrimination through a pure rent of being male. The majority of the gender wage gap is accounted for by the different rewards provided by the labour market. The different rewards provided by the labour market, the pure rent of being male, experience loss, occupational segregation, and child care, all reduce women s wages and put them at a disadvantageous position. The results of wage decomposition in Albania share similarities with other countries in the region. Like the case of Russia and Ukraine, the gender wage gap in Albania favors men, and occupational segregation plays an important role in increasing the gender wage gap. Education is not enough to give women in Albania an overall wage advantage as it did in Bulgaria; however, high levels of women s education help reduce the gender wage gap in Albania. The main implication that comes out of the decomposition results and that is robust to the different methodologies applied is that education although key, it is not enough. Rather, it is other factors such as occupational segregation, lower work experience as a result of discontinued experiences in the labour market, and child care that account for the majority of the gender wage gap. Women who are currently in the labour markets, the majority of whom have been educated during the Communist period, have on average more education that men. However, education is not enough to make up for the labour market discrimination in terms of wages. If women kept their current endowments, where education is the main player, and were paid according to the wage structure of men, their average wages would score higher than that of males. Instead, if their education levels decreased and were the same as those of males, they would earn even less than they do now. Consequently, there are three main messages that come out of this paper: 1) Education is key and should be given special consideration by the policy makers, but other important sources

26 26 such as occupational segregation, work experience, and child care play a crucial role 2) although education is not enough to make up for the gender wage gap, if education levels of the females were to decrease, the gender wage gap would be increased even further, 3) the problem is much bigger for the low educated group where education levels are lower and pure discrimination in the labour markets is deeper. Therefore, policy makers should concentrate on designing policies that fight gender segregation and that offer equal pay for equal work. In order to prevent occupational segregation it is important that policies are designed not only for the labour market, but also for the educational system as to stop promoting curriculums that influence choice of women into mainly dominated female occupation. In addition, equal pay for equal work policies should be designed in conjunction with policies similar to those of affirmative actions promoting the hiring of women in fields which are predominantly male. As is the case with many transition economies, there is often a mismatch between skills and occupations. Policies should be designed such that they match women s skills and education with the appropriate occupation. Lastly, to alleviate the loss of experience and discontinuity in the labour market, as a result of child bearing and rearing, policies should be designed to either share the child care responsibilities between both males and females, or have better provision for child care.

27 REFERENCES Alderman, H., J. Behrman, S. Khan, and R. Sabot (1996). Decomposing the Regional Gap in Cognitive Skills in Rural Pakistan. Journal of Asian Economics 7: Albania: Urban Growth, Rural Stagnation and Migration, A Poverty Assessment. World Bank, (2007). Albania: Poverty Reduction Strategy. (2006). Annual Progress Report, IMF. Altonji, Joseph G and Thomas A. Dunn (1996). The Effects of Family Characteristics on the Return to Education. The Review of Economics and Statistics 78.4: Andren, Daniela, and Thomas Andren (2007). Occupational Gender Composition and Wages in Romania: From Planned Equality to Market Inequality? IZA Discussion Paper No Andren, Daniela, John S. Earle, and Dana Sapatoru (2004). The Wage Effects of Schooling under Socialism and in Transition: Evidence from Romania, 1950:2000. Upjohn Institute Staff Working Paper No Anker, Richard (1985). Comparative Survey in Working Women in Socialist Countries: The Fertility Connection. Geneva: International Labour Office. Boserup, E (1970). Womens Role in Economic Development. London: Earthscan Press. Brown, Randall S., Marilyn Moon, and Barbara S. Zoloth (1980). Incorporating Occupational Attainment in Studies of Male-Female Earnings Differentials. Journal of Human Resources 15.1: Browning M. and R. Subramaniam (1995). Gender-Differences in India: Parental Preferences or Marriage Costs.. Blinder, Alan S (1972). Wage Discrimination: Reduced Form and Structural Estimates. The Journal of Human Resources 8.4: Brainerd, Elizabeth (2000). Women in Transition: Changes in Wage Differentials in Eastern Europe and the Former Soviet Union. Industrial and Labor Relations Review 54.1: Cuka, Elida, Harry Papapanagos, Natasha Polo, and Peter Sanfey (2003). Labor Market Developments in Albania: An Analytical Overview. Review of Development Economics 7.2: Das Gupta, Monica (1987). Selective Discrimination against Female Children in Rural Punjab,India. Population and Development Review 13. 1:

28 Deaton, Angus (1997). The Analysis of Household Survey. Baltimore: The Johns Hopkins University Press. 28 DiNardo, John, Nicole Fortin, and Thomas Lemieux (1996). Labor Market Institutions and the Distribution of Wages, : A Semiparametric Approach. Econometrica 64.5: Dreze, J. and Geeta Kingdon (2001). School Participation in Rural India. Review of Development Economics 5: Echevarria, Cristina, and Antonio Merlo (1999). Gender Differences in Education in Dynamic Household Bargaining Model. International Economic Review 40.2: Falkingham, Jane and Arjan Gjonca (2001). Fertility Transition in Communist Albania, Population Studies 55.3: Gang, Ira N., John Landon-Lane, and Ralitza Dimova (2006). Where to Work? The Role of the Household in Explaining Gender Differences in Labour Market Outcomes. Departmental Working Papers No , Rutgers University, Department of Economics. Gender and Employment in Albania. (2005). Albanian Ministry of Labor and Social Affairs, Committee for Gender Equality, and UNDP, Albania. Gender in Transition. (2002). The World Bank. Gender Inequalities in Education. (2002 b). The World Bank. Giddings, Lisa A. (2002). Changes in Gender Earnings Differentials in Bulgaria s Transition to A Mixed-Market Economy. Eastern Economic Journal 28.4: Greene, William (2002). Econometric Modeling Guide. Vol. 1. New York: Econometric Software, Inc. Grimshaw, Damian, and Jill Rubery (2002). The Adjusted Gender Pay Gap: A Critical Appraisal of Standard Decomposition Techniques. Prepared as part of the work by the co-ordinating team of the Group of Experts on Gender and Employment Commissioned by the Equal Opportunities Unit in the European Commission. Haderi, Sulo (1999). Inflation and Stabilization in Albania. Post-Communist Economies 11.1: Heinen, Jacqueline (1997). Public/Private: Gender - Social and Political Citizenship in EasternEurope. Theory and Society, Special Issue on Recasting Citizenship 26.4: Holmes, Jessica (2003). Measuring the Determinants of School Completion in Pakistan: Analysis of Censoring and Selection Bias. Economics of Education Review 22:

29 Joshi, H. and Paci, P. (1998). Unequal Pay for Women and Men: Evidence from the British Birth Cohort Studies. Cambridge, MA: The MIT Press. Juhn, Chinhui, Kevin M. Murphy, and Brooks Pierce (1993). Wage Inequality and the Rise in Returns to Skill. Journal of Political Economy 101.3: Jurajda, Stepan (2003). Gender Wage Gap and Segregation in Enterprises and the Public Sector in Late Transition Countries. Journal of Comparative Economics 31: Jurajda, Stepan, and Heike Harmgart (2007). When are Female Occupations Paying More? Journal of Comparative Economics 35: Kanbur, Ravi (2002). Education, Empowerment and Gender Inequalities. Cornell University. Kennedy, Peter (2003). A Guide to Econometrics (Fifth Edition). Cambridge, MA: The MIT Press. Kingdon, Geeta Gandhi (1998). Does the Labour Market Explain Lower Female Schooling in India? The Journal of Development Studies 5.1: Kligman, Gail (1996). Women and the Negotiation of Identity in Post-Communist Eastern Europe. Identities in Transition. repositories.cdlib.org. Korovilas, James P. (1999). The Albanian Economy in Transition: The Role of Remittances and Pyramid Investment Schemes. Post-Communist Economies 11.3: Kunze, Astrid (2000). The Determination of Wages and the Gender Wage Gap: A Survey. IZA Discussion Paper No Lemieux, Thomas (2002). Decomposing Changes in Wage Distributions: A Unified Approach. Canadian Journal of Economics 35.4: Mincer, Jacob (1974). Schooling, Experience, and Earnings. NBER. Changes in Wage Inequality, , Research in Labor Economics 16 (1997): Munich, Daniel, Jan Svejnar, and Katherine Terrell (1999). Returns to Human Capital under the Communist Wage Grid and During the Transition to a Market Economy. CERGE-EI Discussion Paper No. 29. Neal, Derek A., and William R. Johnson (1996). The Role of Premarket Factors in Black-White Wage Differences. The Journal of Political Economy 104.5: Netz, Janet and Jon D. Haveman (1999). All in the Family: Family, Income, and Labor Force Attachment. Feminist Economics 5.3:

30 Newell, Andrew, and Barry Reilly (2000). The Gender Wage Gap in the Transition from Communism: Some Empirical Evidence. William Davidson Institute Working Paper No Oaxaca, Ronald (1973). Male-Female Wage Differentials in Urban Labor Markets. International Economic Review 14.3: Oglobin, Constantin G. (1999). The Gender Earnings Differential in the Russian Transition Economy. Industrial and Labor Relations Review 52.4: Oglobin, Constantin G. (2005). The Gender Earning Differential in Russia After a Decade of Economic Transition. Applied Econometrics and International Development 5.3: O Neill, Donal, Olive Sweetman, and Dirk Van de gaer (2006). The Impact of Cognitive Skills on the Distribution of the Balck-White Wage Gap. Labour Economics 13: Radcliffe, S. (1991). The Role of Gender in Peasant Migration: Conceptual Issues from the Peruvian Andes. Review of Radical Political Economics 23: Rose, Elaina (2000). Gender Bias, Credit Constraints and Time Allocation in Rural India. The Economic Journal 110: Silova, Iveta and Cathryn Magno (2004). Gender Equity Unmasked: Democracy, Gender, and Education in Central/Southeastern Europe and the Former Soviet Union. Comparative Education Review 48.4: Svejnar, Jan (1999). Labor Markets in the Transitional Central and East European Economies. Handbook of Labor Economics, Vols: 3-4. North Holland: Amsterdam. Tarifa, Fatos. Disappearing from Politics: Social Change and Women in Albania. Women in the Politics of Postcommunist Eastern Europe. New York: M.E. Sharpe, Inc. Weichselbaumer, Doris, and Rudolf Winter-Ebmer (2005). A Meta-Analysis of the International Gender Wage Gap. Journal of Economic Survey 19.3: Table 1. Descriptive Statistics Variables Males Females Total P-Value Ln Wage Monthly Wage

31 Age Years of Schooling Experience No Diploma 0.17% 0.48% 0.26% 0.32 Primary % 2.70% 4.69% 0.00 Primary % 34.25% 39.29% 0.00 Secondary 20.64% 20.16% 20.50% 0.78 Vocational 22.79% 22.82% 22.79% 0.99 University 9.39% 19.40% 12.26% 0.00 Postuniversity Albania 0.44% 0.34% 0.41% 0.61 Postuniversity Abroad 0.05% 0.24% 0.10% 0.18 Total Observations Table 2. Log Wage by Education Categories Variables Males Females Total P-Value No Diploma Ln Wage Primary 4 Ln Wage Primary 8 Ln Wage Secondary Ln Wage Vocational Ln Wage University Ln Wage University in Albania Ln Wage University Abroad Ln Wage Post Grad in Albania Ln Wage Post Grad Abroad Ln Wage Total Observations Table 3. Monthly Wages by Education Categories Variables Males Females Total P-Value No Diploma Wage Primary 4 Wage Primary 8

32 Wage Secondary Wage Vocational Wage University Wage University in Albania Wage University Abroad Wage Post Grad in Albania Wage Post Grad Abroad Wage Total Observations Note: Unit in Albanian New Lek. Table 4. Main Occupations by Gender Occupations % Males % Females Total P-Value Legislators, Senior Officials and Managers 6.20% 1.95% 1.11% 0.00 Professionals 6.36% 20.15% 10.32% 0.00 Technicians and Associate Professionals 5.23% 12.60% 7.35% 0.00 Clerks 1.67% 2.84% 2.00% 0.07 Service, Shop and Market Sales Workers 12.67% 16.40% 13.74% 0.02 Skilled Agricultural and Fishery Workers 18.79% 21.31% 19.52% 0.16 Craft and Related Trade Workers 28.69% 10.51% 23.47% 0.00 Plant and Machine Operators and Assemblers 12.98% 6.03% 10.98% 0.00 Elementary Occupations 7.43% 8.20% 7.65% 0.49 Table 5. Monthly Wages by Occupation Variables Males Females Total P-Value Legislators and Managers Wage Professionals Wage Technicians

33 Wage Clerks Wage Service Workers Wage Skilled Agricultural and Fishery Wage Craft and Trade Wage Plant and Machine Operators Wage Elementary Occupations Wage Total Observations Note: Unit in Albanian New Lek. Table 6. Oaxaca Decomposition Variables Endowments Coefficients Interaction Education (0.007)*** (0.088)*** (0.008)*** Experience (0.021)*** (0.198)* (0.030) Experience (0.019)*** (0.117)*** (0.030)*** Occupation (0.007)*** (0.052) (0.010) Distance Index (0.002) (0.004)*** (0.004)** Social capital (0.001) (0.003) (0.001) Number of Children (0.002) (0.039)*** (0.004)** Married (0.005) (0.068) (0.007) Divorced (0.004) (0.001) (0.004) Living together (0.001) (0.001) Widow (0.006) (0.001) (0.007) Coastal (0.003) (0.018) (0.001) Central (0.004)** (0.017) (0.003) Mountain

34 (0.006) (0.017) (0.002) Urban (0.005)*** (0.034)** (0.007)** Constant (0.148)*** Total (0.017) (0.026)*** (0.020)*** Observations Standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% (+) indicates an advantage for females, (-) indicates an advantage for males Table 7. Oaxaca Decomposition for Higher Education Variables Endowments Coefficients Interaction Education (0.008) (0.913)* (0.009) Experience (0.120)** (0.445) (0.142) Experience (0.102)** (0.286) (0.131) Occupation (0.015)** (0.106) (0.021) Distance Index (0.007) (0.034) (0.010) Social Capital (0.008) (0.014) (0.007) Number of Children (0.003) (0.074) (0.007) Married (0.029)** (0.145)** (0.033)* Divorced (0.002) Living together (0.001) (0.004) (0.003) Widow (0.002) (0.005) (0.002) Coastal (0.011) (0.023) (0.005) Central (0.013) (0.024) (0.004) Mountain (0.016) (0.026) (0.006) Urban (0.007) (0.126) (0.005) Constant (0.934)

35 Total (0.040) (0.054)*** (0.046) Observations Standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% (+) indicates an advantage for females, (-) indicates an advantage for males Table 8. Oaxaca Decomposition for Lower Education Variables Endowments Coefficients Interaction Education (0.003)** (0.118)*** (0.005)** Experience (0.016)** (0.244) (0.023) Experience (0.017)*** (0.142)** (0.027)* Occupation (0.008)*** (0.064) (0.010) Distance Index (0.001) (0.003)** (0.004)* Social Capital (0.003) (0.001) Number of Children (0.002) (0.045)*** (0.004)* Married (0.004) (0.082) (0.005) Divorced (0.005) (0.001) (0.005) Living together (0.001) (0.001) Widow (0.009) (0.001) (0.010) Coastal (0.002) (0.023) (0.001) Central (0.003)** (0.021) (0.004) Mountain (0.006) (0.021) (0.002) Urban (0.005)*** (0.036)** (0.008)* Constant (0.186)*** Total (0.018)** (0.029)*** (0.022)* Observations 3114 Standard errors in parentheses. * significant at 10%; ** significant at 5%; *** significant at 1% (+) indicates an advantage for females, (-) indicates an advantage for males

36 36 Table 9. Log Wage Distribution of Females and Males in Albania Mean Variance Total Predicted xb Residual (1) (2) (3) (4) 1.Female Female with Male b's Female with Male b's and X's Male Female - Male Difference Effect of 6.b (Row 1- Row 2) x (Row 2 - Row 3) Residual (Row 3 - Row 4) Table 10. Log Wage Distribution for the Highly Educated Mean Variance Total Predicted xb Residual (1) (2) (3) (4) 1.Female Female with Male b's Female with Male b's and X's Male Female - Male Diference Effect of 6.b (Row 1- Row 2) x (Row 2 - Row 3) Residual (Row 3 - Row 4) Table 11. Log Wage Distribution for the Low Educated Mean Variance Total Predicted xb Residual (1) (2) (3) (4) 1.Female Female with Male b's Female with Male b's and X's Male Female - Male Diference

37 37 Effect of 6.b (Row 1- Row 2) x (Row 2 - Row 3) Residual (Row 3 - Row 4) Figure 1. Age Education Gender Profile

38 38 Figure 2. Age Wage Gender Profile Figure 3. Kernel Density Estimates of Experience by Gender

39 39 Figure 4. Kernel Density Estimates of Predicted Ln Monthly Wage by Gender Figure 5. Females with Males Regression Coefficients

40 40 Figure 6. Females with Males Covariates Figure 7. Females with Males b s and Covariate Distribution Figure 8. High Education Females with Males Regression Coefficients

41 41 Figure 9. High Education Females with Males Covariates Figure 10. High Education Females with Males b s and Covariate Distribution Figure 11. Low Education Females with Males Regression Coefficients