Microsoft Word - 00_tresc.docx © 2013 Nicolaus Copernicus University. All rights reserved. http://www.dem.umk.pl/dem D Y N A M I C E C O N O M E T R I C M O D E L S DOI: http://dx.doi.org/10.12775/DEM.2013.009 Vol. 13 (2013) 163−174 Submitted September 30, 2013 ISSN Accepted December 30, 2013 1234-3862 Joanna Małgorzata Landmesser* Decomposing the Gender Gap in Average Exit Rate from Unemployment A b s t r a c t. In the paper, we analyse the exit rates from unemployment, taking into account gender differences. The process of leaving the unemployment state was examined for each sex separately using the parametric hazard models. The objective was to present a decomposi- tion of inequalities between men and women when leaving unemployment. The application of the modified Oaxaca-Blinder decomposition technique allowed us to isolate the factors ex- plaining the observed inequalities. We found, that the gender gap is explained almost exclu- sively by differences in the effects of men’s and women’s characteristics. K e y w o r d s: duration of unemployment, parametric hazard models, gender gap, Oaxaca- Blinder decomposition. J E L Classification: J16, J64. Introduction Different levels of the economic activity between women and men are often analyzed in the economic literature. It is noted that men are more fre- quently associated with the labor market, while women, to a lesser extent (due to their involvement in the family career). The review of the various aspects related to the activity of women and men in the labor market is pro- vided, for example, by Altonji, Blank (1999). Numerous empirical studies tend to focus on the gender wage gaps (Blinder, 1973; Oaxaca, 1973; Beblo et al., 2003). The findings of these * Correspondence to: Joanna Landmesser, Department of Econometrics and Statistics, Warsaw University of Life Sciences, Nowoursynowska 166, 02-787 Warszawa, Poland, e-mail: joanna_landmesser@sggw.pl. Joanna Małgorzata Landmesser DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 164 studies show that males earn substantially higher wages than females. The part of each wage differential is due to differences in „objective” characteris- tics such as education and work experience. However, a partial differential remains even when male-female differences in these traits are controlled for. Also the occupational segregation by gender is discussed in the literature (Petrongolo, 2004). The studies show that women have a weaker position in the labor market. Women are often discriminated against even if their quali- fications are higher than for men. Different attitudes of men and women are also reflected in the opportuni- ties to take employment. Gender is a significant factor affecting the move- ments between labor market states. It is an important variable in determining the chances of finding a job. Many studies have shown that women have a lower probability of finding a new job (Katz, Meyer, 1990b), especially on a permanent basis (Edin 1989), and are exposed to more frequent periods without work (Steiner, 1989; Jensen, Westergard-Nielsen, 1990). Women are at a disadvantage even if their job-search activity is higher than for men. Analyses focused on the description of leaving the unemployment state, which were carried out in Poland so far, disregard the issue of gender behav- iour. They generally do not consider that the effect of different factors may depend on gender. In empirical models, gender only shifts the dependent variable and there are not estimated separate equations for men and women (see for example Malarska (2007)). The opposite approach propagate Gonza- lo, Saarela (2000), Ollikainen (2003), Tansel, Taşçi (2010) or Landmesser (2008), who estimated separate hazard models for both sexes. The present study is an analysis of exit rates from unemployment, taking into account gender differences. In the paper, the process of leaving the un- employment state was examined for each sex separately using the tools of duration analysis – the parametric hazard models. The results confirmed that the influence of explanatory variables on the chance to exit the unemploy- ment depends on the sex of the individual. However, the main objective of the study was to perform a decomposi- tion of inequalities between men and women when leaving unemployment. The quantitative dimensions of the various causes of unequal chances by taking a job are not well known, but the Oaxaca-Blinder microeconometric decomposition technique applied allowed us to isolate the factors explaining the inequalities. Decomposing the Gender Gap in Average Exit Rate from Unemployment DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 165 1. The Analysis Method The study comprises the duration of time a person spent in the unem- ployment state (T). Modeling of such a variable requires the use of survival analysis tools. Constructed hazard models are suitable for the analysis of censored observations and for studying the influence of individual character- istics on the chances of leaving the unemployment state (cf. Lancas- ter, 1979). The basic function describing the distribution of the duration times is the survival function [ ] )(1Pr)( tFtTtS −=>= , which expresses the probability of survival beyond a certain point in time t (cf. Kalbfleisch, Prentice, 2002). The hazard function (risk, intensity function) is the ratio of the density func- tion and the survival function. It is the limit of probability that the episode is completed during the time interval [t, t+dt] given that it has not been com- pleted before the moment t, for dt→0: [ ] . Pr lim )( )( )( 0→ dt tTdttTt tS tf th dt ≥+<≤ == (1) The hazard rates describe the intensity of the transition from one state to another. Using the parametric proportional hazards models, it is possible to speci- fy hazard as a function of time and some explanatory variables: ).exp()()( 0 βxx jj thth ′= (2) Models in this class differ in their assumptions about the baseline hazard )(0 th . The courses of hazard functions can take many forms, from monoton- ic to non-monotonic. To model the monotonic hazard, the Weibull distribution with two pa- rameters ),( pW λ is often used. The Weibull distribution described original- ly the dispersion of the fatigue life of materials (Weibull, 1939). It is also used by researchers modeling the duration of unemployment time (due to the decreasing intensity while leaving the state). For a variable ),(~ pWT λ survival and hazard functions are defined as )exp()( pttS λ−= and λ1)( −= pptth , for 0>λ , 0>p and 0≥t . The estimated parameter p is called a shape parameter. If p = 1, then the hazard function is constant; if p > 1, it monotonically increases, while for p < 1 the hazard monotonically decreases. The form of the Weibull hazard model after taking the parameter- ization )exp( βx′=λ is as follows: Joanna Małgorzata Landmesser DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 166 ).exp()( 1 βxx j p j ptth ′= − (3) In order to describe the process of leaving the state of unemployment the Weibull proportional hazard models were estimated for both men and wom- en separately. This was followed by the analysis of inequalities observed using a modification of the Oaxaca-Blinder decomposition technique. The idea of Oaxaca-Blinder decomposition can be applied whenever we need to explain the differences between two comparison groups. Let MW xx , be the characteristics of men and women, respectively, and let MW ββ , be the returns to these characteristics and Φ - the average level of dependent variable in the econometric model (e.g. the average hazard rate). The popular Oaxaca-Blinder decomposition for the gender gap at the aggregate level is as follows (Oaxaca, 1973; Blinder, 1973): ].)()([])()([)()( MWWWMMMWMMWW βxβxβxβxβxβx '''''' Φ−Φ+Φ−Φ=Φ−Φ (4) The equation (4) is based on one group’s characteristics and the estimat- ed coefficients of another group’s equation. The first term on the right hand side of the equation expresses the difference of the potentials of both groups (women and men). If the characteristics x used to estimate an econometric model exhausted all the factors affecting the chance to leave the unemploy- ment state, it is possible to assume that the second term on the right hand side of this equation represents the amount of discrimination. This expres- sion is the result of differences in the estimated parameters, and so in the „prices” of individual characteristics of men and women. Blinder argued that „the latter sum […] exists only because the market evaluates differently the identical bundle of traits if possessed by different demographic groups” (Blinder, 1973, pp. 438–439). The contribution of the differences in characteristics and coefficients of individual variables (the detailed decomposition) can be easily found when linear equations are used. The methodology proposed by Yun (2004) pro- vides a way to apply the Oaxaca-Blinder decomposition to a non-linear func- tion for both aggregate and detailed decompositions. A modification made by Yun allowed the use of the above concepts to determine the contribution of each explanatory variable in Weibull regression to the total difference. The formula given by Yun has the following form (as given in Ortega Masagué, 2008): Decomposing the Gender Gap in Average Exit Rate from Unemployment DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 167 ],)()([])()([ )()( 11 MWWW k i MMMW k i x MMWW ii WW βxβxβxβx βxβx '''' '' Φ−Φ+Φ−Φ= =Φ−Φ ∑∑ = Δ = Δ β (5) where: MMW M i M i W i x xx W i βxx ')( )( − − =Δ β and ,1 1 =∑ = Δ k i xi W )( )( ' MWW M i W i W ixW i ββx − − =Δ ββ β and ,1 1 =∑ = Δ k i i W β k is the number of explanatory variables in the model, ,W Mx x - the mean levels of characteristics for men and women, respec- tively. The detailed decomposition methodology proposed by Yun is the de- composition of differences in the first moment, i.e. differences in the mean value of the variable of interest. The method „does not depend on the func- tional form as long as the dependent variable is a function of a linear combi- nation of independent variables and the function is once differentiable” (Yun, 2004, p. 275). In order to obtain a proper weight Yun evaluated the value of the function using mean characteristics and used a first order Taylor expansion. 2. The Empirical Data Unemployment is a problem frequently connected with specific regions. In our research work we try to analyze in detail the situation in district Słupsk in north Poland (voivodeship Pomorskie). In this region, many inhab- itants of rural areas were previously employed in the state-owned agricultur- al farms. However, a large number of these farms went bankrupt after the political system change in 1989. The former employees, women and men, turned out to be the most helpless social group in Poland. Therefore, it seems desirable to analyze the situation of the inhabitants of this region on the labor market over the past 20 years. The study was conducted using individual data of people registered as unemployed in the District Labor Office in Słupsk. The selected sample consisted of 4 372 people registered in the office from January 1990 to Au- gust 2007. There were 2 203 women and 2 169 men randomly selected (women constituted 50.4% and men 49.6% of the selected sample). The data Joanna Małgorzata Landmesser DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 168 about each person took the multiepisode form and contained a detailed histo- ry of the office customers. At the data basis it was established how long un- employment episodes lasted (in days) or how long they are still going on (in the case of censored episodes). Each person could register many times in the labor office in his or her history. Therefore, for 4 372 examined people a total of 10 118 episodes of being unemployed was noted (10% of these episodes were censored) (see Table 1). In the case of women, censored episodes constituted larger share of total episodes than it was for men. This was due to the fact that women had on average longer episodes of unemployment than men. For example, uncen- sored episodes for women lasted on average 413.1 days and for men 271.2 days. The information on the average duration in the unemployment state during a single episode is also presented in the Table 1. Table 1. The number and the average duration of unemployment episodes Number of episodes All persons Women Men Total 10118 100% 4786 100% 5332 100% Censored 1007 10% 664 13.9% 343 6.4% Uncensored 9111 90% 4122 86.1% 4989 93.6% Average duration of episode (in days) All persons Women Men Total 406.5 523.2 301.7 Censored 1049 1207 745.5 Uncensored 335.4 413.1 271.2 On average, a woman registered in the labor office 2.55 times and a man 3.02 times (more often, but for shorter periods). The average age of women at the beginning of the unemployment episode was 31.99 years old, while for men it was 33.66 years old. Detailed information on the age of registrants, their level of education and place of residence are shown in the Table 2. Women registered in the labor office were usually younger than men, more frequently they had higher or secondary education level than in the case of men, they rarely had vocational education level. They are character- ized by a higher proportion of residence in the city. Episodes of unemploy- ment were also examined for the payment of unemployment benefits, train- ing benefits and social security benefits. It was found that men, more often than women received unemployment benefits, while women were often as- signed training allowances and social security benefits. Decomposing the Gender Gap in Average Exit Rate from Unemployment DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 169 Table 2. The structure of unemployment episodes by selected characteristics Characteristics All persons Women Men Age-group 17–24 years old 3068 30.3% 1512 31.6% 1556 29.2% 25–34 years old 2755 27.2% 1353 28.3% 1402 26.3% 35– 44 years old 2408 23.8% 1200 25.1% 1208 22.7% 45–54 years old 1742 17.2% 699 14.6% 1043 19.6% over 55 years old 145 1.4% 22 0.5% 123 2.3% Education level tertiary 767 7.6% 517 10.8% 250 4.7% vocational secondary 1957 19.3% 1170 24.4% 787 14.8% general secondary 710 7.0% 516 10.8% 194 3.6% basic vocational 3100 30.6% 1240 25.9% 1860 34.9% lower second. or primary 3584 35.4% 1343 28.1% 2241 42.0% Place of residence town 5079 50.2% 2551 53.3% 2528 47.4% village 5039 49.8% 2235 46.7% 2804 52.6% The information obtained from the database of the labor office allowed to establish a set of potential explanatory variables in the models describing the intensity of leaving the unemployment state. Most of them are dichoto- mous variables, such as: − „gender” (number 1 coded the male sex), − set of variables for the five age categories („age1724”, „age2534”, „age3544”, „age4554”, „age55over”), − set of variables for the education level („tertiary”, „vocational second- ary”, „general secondary”, „basic vocational”, „lower secondary or pri- mary”), − „married” (1, if a person is not of free marital status), − „town” (1, if the person lives in the city), − „disabled” (1, if a person is disabled), − „unemployment benefits” (1 for those who receive unemployment bene- fits), − „training benefits” (1 for those who receive training benefits), − „social security” (1 for social security beneficiaries from polish Zakład Usług Społecznych). In addition, the models used information about the consecutive unem- ployment episode number for the person (the variable called „episode_nr”). 3. The Results of Empirical Analysis In order to analyse the distribution of unemployment durations, first the nonparametric Kaplan-Meier survival function S(t) - separately for men and women - was plotted (Figure 1 (A)). Joanna Małgorzata Landmesser DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 170 (A) (B) Figure 1. Nonparametric Kaplan-Meier estimator for the survival function (A) and plots of the hazard functions in Weibull model (B) The survival curve for men indicates a lower likelihood of continuation in the state of unemployment than for women (see Figure 1(A)). There was a statistical significance of differences in the courses of the survival func- tions for both sexes. Then, based on the total sample, the parametric Weibull proportional hazard model for the chance to leave the unemployment state was estimated (the parameter estimates of this model are listed in Table 3, part (A)). For example, the interpretation of the parameter by the variable "gender" is as follows: an opportunity to leave the unemployment state in the case of men is about 58.2% higher than in the case of women (exp (0.459) = 1.582). A positive value of the parameter βk means that a one unit change in the k-th explanatory variable results in the relative increase of hazard. A negative value is associated with the relative decrease of hazard, respectively. We conclude, that the chances of leaving unemployment rise among men, people at a younger age, better educated, married, living in the city, with the next in turn episode without a job, and they decrease with a disability and the fact of receiving unemployment and training benefits or social security payments. Figure 1 (B) shows two graphs of Weibull hazard functions derived from the estimated model for "gender" = 0 and "gender" = 1. These functions de- crease (as p < 0) and at any time women are characterized by the lower in- tensity of leaving the unemployment than men. The next step was to estimate two Weibull hazard models for time spent in unemployment state for men and women separately (the estimation results are presented in Table 3, part (B) and (C)). These models differ in parameter estimates standing by the relevant variables. Compared to the reference group (women and men aged 55 years and over, respectively) younger wom- 0. 00 0. 25 0. 50 0. 75 1. 00 S ur vi va l f un ct io n 0 2000 4000 6000 analysis time gender = 0 gender = 1 Kaplan-Meier survival estimates 0 .0 02 .0 04 .0 06 .0 08 H az ar d fu nc tio n 0 2000 4000 6000 analysis time gender=1 gender=0 Weibull regression Decomposing the Gender Gap in Average Exit Rate from Unemployment DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 171 en are more likely to exit from unemployment than young men. For women the positive effect of higher education is stronger than for men. Marriage intensifies the exit from unemployment among men. The level of the corre- sponding parameter for women shows a decrease in employment opportuni- ties. Among the remaining parameters, attention should be paid to the coeffi- cients which appear by the variables associated with financial benefits. The negative effect of unemployment benefits and training allowances on the opportunities to leave unemployment for women is smaller than for men, but the impact of social security payments are stronger for women. Apart from that, the estimated values of the parameter p indicate a steeper decline in employment opportunity for women over time. Table 3. The results of the Weibull model estimation for the whole sample (A), for women only (B) and for men only (C) Variable All persons (A) Women (B) Men (C) βiA exp(βiA) βiW exp(βiW) βiM exp(βiM) gender 0.459 *** 1.582 – – – – – – age1724 0.633 *** 1.884 1.213 *** 3.362 0.713 *** 2.039 age2534 0.526 *** 1.692 1.208 *** 3.348 0.502 *** 1.653 age3544 0.410 *** 1.508 1.185 *** 3.269 0.293 *** 1.340 age4554 0.324 *** 1.382 1.096 *** 2.993 0.217 ** 1.243 tertiary 0.571 *** 1.769 0.769 *** 2.157 0.355 *** 1.427 vocational secondary 0.316 *** 1.372 0.425 *** 1.529 0.211 *** 1.235 general secondary 0.348 *** 1.416 0.428 *** 1.534 0.275 *** 1.317 basic vocational 0.168 *** 1.183 0.263 *** 1.300 0.113 *** 1.120 married 0.115 *** 1.121 -0.099 *** 0.906 0.287 *** 1.332 town 0.095 *** 1.100 0.057 * 1.058 0.128 *** 1.136 disabled -0.275 *** 0.759 -0.153 * 0.858 -0.321 *** 0.726 episode_nr 0.069 *** 1.072 0.098 *** 1.104 0.051 *** 1.052 unemploy. benefits -0.290 *** 0.748 -0.231 *** 0.793 -0.372 *** 0.690 training benefits -0.613 *** 0.542 -0.554 *** 0.575 -0.670 *** 0.512 social security -0.698 *** 0.497 -0.787 *** 0.455 -0.425 *** 0.654 cons -5.503 *** 0.004 -6.148 *** 0.002 -5.104 *** 0.006 p 0.777 *** – 0.770 *** – 0.797 *** – Number of episodes 10118 4786 5332 lnL -17804.816 -8356.835 -9357.641 Note: *** significant at 1%; ** significant at 5%; * significant at 10%. However, estimates of the parameters obtained in the models (B) and (C) are difficult to compare to each other, due to different empirical subsamples. Thus, subsequently observed inequalities between women and men in leav- ing the unemployment were decomposed using the modified Oaxaca-Blinder technique. The results on an aggregated basis are presented in Table 4 (the aggregation consisted in an accumulation of effects for related variables). Joanna Małgorzata Landmesser DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 172 Table 4. Decomposition of the gender gap in hazard rates from unemployment (re- sults of modified Oaxaca-Blinder decomposition technique) Value % Observed differential (total) -0.001507 100 Value % Value % Characteristics –0.000100 6.63 Returns –0.001407 93.37 age-group –0.000113 7.50 age-group –0.000953 63.28 education level –0.000388 25.76 education level –0.000450 29.89 married 0.000124 –8.24 married 0.000134 –8.91 town –0.000036 2.41 town 0.000051 –3.39 disabled –0.000016 1.09 disabled –0.000008 0.54 episode_nr 0.000115 –7.63 episode_nr –0.000158 10.48 benefits 0.000215 –14.26 benefits –0.000022 1.48 Note: The factors, that cause diversity of chances on the labor market to the greatest extent, are in bold. There is a negative difference between the mean values of the hazard function for men and women (–0.001507), meaning that women typically have lower chances of leaving the unemployment state than men. Decompo- sition, which was carried out, made it possible to isolate the factors explain- ing the inequality observed to a different extent. It turns out that the differences in exit rates are only in the 6.63 percent explained by the individual characteristics of women and men (vectors xW and xM). The gender gap in the chances of exit from unemployment recog- nized in this way comes from the fact that women are different from men due to certain characteristics relevant in the labor market. The effect of the different education levels of women and men can be noticed. Even though women on average are better educated than men, they have rarely technical education, which results in a smaller probability of employment than in the case of men. The differences in transition rates from unemployment are re- duced by more frequent training allowances and social security benefits for women. However, gender inequalities examined should be assigned in the ma- jority – in 93.37% – to the coefficients βW and βM of estimated hazard mod- els (rather than to the differentiation of individuals characteristics). We find out that people with the same characteristics, if they are of different sexes, have various chances of exiting unemployment. The gender gap in unem- ployment rates is therefore explained not by differences in the characteristics of men and women but by differences in the labor market returns to their characteristics. A different "evaluation" of men’s and women’s characteris- tics is a major cause of the inequality. It can be assumed that employers fa- vor men. Women are discriminated against in the labor market. Different opportunities are mainly due to prejudices associated with the woman's age Decomposing the Gender Gap in Average Exit Rate from Unemployment DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 173 (generally it is believed that younger women are less involved in work for family reasons) and her education level. Conclusions The analysis conducted shows that the differences in the intensity of leaving unemployment are explained by individual characteristics of women and men only in a limited way. To a much greater extent these differences can be attributed to the „valuations” of men’s and women’s characteristics carried out by the labor market. Inequalities observed in opportunities for men and women would probably occur even if the characteristics of both groups were identical. The calculated differences result from the parameter values, and thus it is important that all the parameters in the models have the statistical properties expected. Therefore, for a correct inference – using the method presented in the article – the careful selection of variables for hazard models is very im- portant. References Altonji, J. G., Blank, R. M. (1999), Race and Gender in the Labor Market, in Ashenfelter, O., Card, D. (eds.), Handbook of Labor Economics, Volume III, North-Holland, Amster- dam, DOI: http://dx.doi.org/10.1016/S1573-4463(99)30039-0. Beblo, M., Beninger, D., Heinze, A., Laisney, F. (2003), Methodological Issues Related to the Analysis of Gender Gaps in Employment, Earnings and Career Progression, European Commission. Blinder, A. S. (1973), Wage Discrimination: Reduced Form and Structural Estimates, Journal of Human Resources, 8(4), 436–455, http://dx.doi.org/10.2307/144855. Edin, P. A. (1989), Unemployment Duration and Competing Risks: Evidence from Sweden, Scandinavian Journal of Economics, 91(4), 639–653, DOI: http://dx.doi.org/10.2307/3440211. Gonzalo, M. T., Saarela, J. (2000), Gender Differences in Exit Rates from Unemployment: Evidence from a Local Finnish Labour Market, Finnish Economic Papers, 13(2), 129–139. Jensen, P., Westergård-Nielsen, N. (1990), Temporary Layoffs, in Hartog, J., Ridder, G., Theeuwes, J. (eds.), Panel Data and Labour Market Studies, North-Holland, Amster- dam. Kalbfleisch, J., Prentice, R. (2002), The Statistical Analysis of Failure Time Data, Second Edition, John Wiley and Sons, New York, DOI: http://dx.doi.org/10.1002/9781118032985. Katz, L. F., Meyer, B. D. (1990), Unemployment Insurance, Recall Expectations, and Unem- ployment Outcomes, Quarterly Journal of Economics, 105(4), 973–1002, DOI: http://dx.doi.org/10.2307/2937881. Lancaster, T. (1979), Econometric Methods for the Duration of Unemployment, Econometrica, 47, 939–956, DOI: http://dx.doi.org/10.2307/1914140. Joanna Małgorzata Landmesser DYNAMIC ECONOMETRIC MODELS 13 (2013) 163–174 174 Landmesser, J. (2008), Analiza aktywności ekonomicznej kobiet wiejskich z wykorzystaniem ekonometrycznych modeli hazardu (The Analysis of Economic Activity of Women in Rural Areas Using the Econometric Hazard Models), Roczniki Naukowe Stowarzyszenia Ekonomistów Rolnictwa i Agrobiznesu (Annals of the Polish Associa- tion of Agricultural and Agribusiness Economists), 1/X, 233–239. Malarska, A. (2007), Diagnozowanie determinantów bezrobocia w Polsce nieklasycznymi metodami statystycznymi. Analiza empiryczna na podstawie danych BAEL, (Diagnos- ing the Determinants of Unemployment in Poland with Non-classical Statistical Meth- ods. The Empirical Analysis Based on LFS-data), Wydawnictwo Biblioteka, Łódź. Oaxaca, R. L. (1973), Male-Female Wage Differentials in Urban Labor Markets. Internation- al Economic Review, 14, 693–709, DOI: http://dx.doi.org/10.2307/2525981. Ollikainen, V. (2003), The Determinants of Unemployment Duration by Gender in Finland, VATT Discussion Papers, 316, Helsinki. Ortega Masagué, A. C. (2008), Gender Gaps in Unemployment Rates in Argentina, Económica, La Plata, 54(1–2), 161–202. Petrongolo, B. (2004), Gender Segregation in Employment Contracts, Discussion Paper 4303, CEPR, DOI: http://dx.doi.org/10.1162/154247604323068032. Steiner, V. (1989), Causes of Recurrent Unemployment - An Empirical Analysis, Empirica, 16, 53–65, DOI: http://dx.doi.org/10.1007/BF00924940. Tansel, A., Taşçi, H. M. (2010), Hazard Analysis of Unemployment Duration by Gender in a Developing Country: The Case of Turkey, IZA Discussion Paper, 4844, Bonn, DOI: http://dx.doi.org/10.1111/j.1467-9914.2010.00480.x. Weibull, W. (1939), A Statistical Theory of the Strength of Materials, Ingeniörs Vetenskaps Akademiens, Handlingar, 151, Stockholm. Yun, M. (2004), Decomposing Differences in the First Moment, Economic Letters, 82, 275–280, DOI: http://dx.doi.org/10.1016/j.econlet.2003.09.008. Dekompozycja nierówności płciowych w przeciętnych stopach wyjścia ze stanu bezrobocia Z a r y s t r e ś c i. W pracy przeprowadzono analizę stóp wyjścia ze stanu bezrobocia, uwzględniającą zróżnicowanie płciowe. Proces opuszczania stanu bezrobocia badano dla obu płci osobno wykorzystując parametryczne modele hazardu. Głównym celem było dokonanie rozkładu nierówności między kobietami i mężczyznami podczas opuszczania bezrobocia. Zastosowana zmodyfikowana mikroekonometryczna technika dekompozycji Oaxaca-Blindera pozwoliła na wyodrębnienie czynników wyjaśniających zaobserwowane nierówności. Otrzy- mano, że nierówności płciowe są wyjaśniane w większości przez różniące się „wyceny” cech kobiet i mężczyzn dokonywane przez rynek. S ł o w a k l u c z o w e: czas trwania w bezrobociu, parametryczne modele hazardu, nierów- ności płciowe, dekompozycja Oaxaca-Blindera.