Gender Pay Gap and Quantile Regression in European Families

Gender Pay Gap and Quantile Regression in European Families Catia Nicodemo Universitat Autonòma de Barcelona 13th of December

EUROPEAN MARRIED WOMEN: WHY DO(N T) THEY WORK? "To the woman he said, Great will be your pain in childbirth... And to Adam he said: the earth is cursed on your account, in pain you will get your food from it all your life..." Genesis Family roles: A greater participation of women in the labour force has caused changes in: Family forms and responsibilities Labour market opportunities

WOMEN VS MEN Wives suffer two types of DISADVANTAGE OR DISCRIMINATION respect to their husbands: Lower wages for the same characteristics and work Primary responsibility for children, when these are present in the family

Motivation To investigate the gender gap in European families, between husband and wife, and explain why countries with poor policies for childcare, flexibly work, etc, are more unequal. Most of these countries are in the Mediterranean area with a strongly traditional families. To investigate the existence of glass ceiling (the gender gap is bigger at the top of the wage distribution ) or sticky floor (the gender pay gap is bigger at the bottom of the wage distribution) between husband and wife in Europe. To do a cross-country analysis comparing two data-set in two different years, ECHP in 2001 and EU-SILC 2006. To confront two methods of calculating the gender wage gap: Quantile regression and Oaxaca-Blinder

Structure of the paper: DATA AND LITERATURE SAMPLE AND EXPLANATORY VARIABLES OAXACA-BLINDER METHOD UNADJUSTED GENDER PAY GAP QUANTILE REGRESSION CONCLUSION

Data ECHP European Community Household Panel (ECHP), waves 1994-2001 for 15 European Union Member States under Eurostat coordination. The data set covers approximately 130,000 individuals from 60,000 households. The panel data cover a wide range of subjects such as demographics, labor force behavior, income, health, education and training, housing, poverty and social exclusion, etc. EU-SILC The European Union Statistics on Income and Living Conditions (EU-SILC) is an instrument aiming at collecting timely and comparable cross-sectional and longitudinal multidimensional microdata on income, poverty, social exclusion and living conditions. The project was formally launched in 2004 and EU25 coverage was achieved from 2005.

Sample, Countries and Explanatory variable Sample: Married women of working age with or without children matched with their respective husbands. Mediterranean countries: France, Italy, Portugal, Spain and Greece. Age, experience, education, children, household income, log of wage, occupation, type of contract.

Sample, Countries and Explanatory variable Sample: Married women of working age with or without children matched with their respective husbands. Countries: Mediterranean countries: France, Italy, Portugal, Spain and Greece. Age, experience, education, children, household income, log of wage, occupation, type of contract.

Sample, Countries and Explanatory variable Sample: Married women of working age with or without children matched with their respective husbands. Countries: Mediterranean countries: France, Italy, Portugal, Spain and Greece. Explanatory Variables: Age, experience, education, children, household income, log of wage, occupation, type of contract.

Lower wages for the same characteristics and work: Gender pay gap The gender pay gap refers to the difference between wages earned by women and men for the same work. Sometimes the gap in earnings creates conflict in the family and psychological problems for women. Reducing gender pay gap is an important topic in the European political agenda. In 2003 the member countries formulated the plan "to achieve by 2010 a substantial reduction in the gender pay gap in each Member State" (Council Decision 2003 L197/20).

Literature Literature: B. Melly 2006, JEE: Estimation of counterfactual distributions using quantile regression. J. Machado and J.Mata 2005, JAE, Counterfactual decomposition of changes in wage distributions using quantile regression. R.Oaxaca and S.Neuman 2003, IZA, Estimating Market discrimination with Selectivity-Corrected wage equations. J.Albrecth, A.van VUUREN and S.Vroman, 2007, Counterfactual distribution with sample selection adjustments: econometric theory and an application to the Netherlands.

Oaxaca-Blinder method Oaxaca (1973) and Blinder (1973) W H W w = X H βh X w βw (1) W h W w = (X H X w) β H + ( }{{} β h β w)x w }{{} Characteristics Coefficients (2) The selection correction (Heckman two step estimates) lnw }{{} Raw wage gap = β H (X H X W ) + ( β H }{{} β W )X W }{{} Characteristics Coefficients + (λ H θh λ W θw ) }{{} Selectivity (3)

Unadjusted gender pay gap Figure: Raw relative wage gap in European countries: ECHP 2001 and 1994: Relative wage gap is defined = (w h w w )/w h

Unadjusted gender pay gap Figure: Raw relative wage gap in European countries: EU-SILC 2006: Relative wage gap is defined = (w h w w )/w h

Adjusted gender pay gap: Oaxaca decomposition in 2001 with Heckman corrected estimates: ECHP

Adjusted gender pay gap: Oaxaca decomposition in 2006 with Heckman corrected estimates: EU-SILC

Quantile Regression The τ th conditional quantile of Y is given by Q y(τ X) = Xβ(τ) τ (0, 1) Koenker-Bassett (1978) solve by minimizing in β(τ): [ n ] β(τ) = min n 1 ρ τ (Y i X i β), (i = 1,...n), i with the check function ρ τ weighting the residuals µ i asymmetrically: { τµ i if µ i 0, ρ τ (µ i ) = (τ 1)µ i if µ i < 0.

Quantile Regression: Machado and Mata (2005) estimator Machado and Mata propose an estimator of counterfactual unconditional wage distributions based on quantile regression, decomposing the difference of the θ th unconditional quantile between two groups, accord to Blinder-Oaxaca: + 1 F Y 1 F 1 Y 1 1 1 1 (θ T = 1) F Y 0 (θ T = 0) = F Y 1 (θ T = 1) F Y 1 (θ T = 0) }{{} characteristics 1 (θ T = 0) F Y 0 (θ T = 0) }{{} coefficients

Quantile Regression: Melly Estimators(2006) Melly present an alternative estimator of counterfactual distributions Computes first the conditional distribution function: 1 1 F yt (q X i ) = 1(F 1 yt (τ X i ) q) = 1(X i βt (τ) q)dτ = 0 0 = n j=1 (τ j τ j 1 )1(X i βt (τ) q) Then, the unconditional distribution function can be calculated: F yt (q T = t) = 1 n t Fyt (q X i ) The same approach leads to the unconditional counterfactual distribution function.

Quantile Regression Melly Estimators(2006) The θ th quantile of the unconditional distribution is 1 q t (θ) = inf {q : F yt (q X i ) θ} n t t while the θ th quantile of the unconditional counterfactual quantile is: 1 q c1 (θ) = inf {q : F yt (q X i ) θ} n t 0 The θ th unconditional quantile of both groups can be decomposed in analogy with Blinder-Oaxaca as: q 1 (θ) q 0 (θ) = q 1 (θ) q c1 (θ) + q c1 (θ) q 0 (θ) }{{}}{{} characteristics coefficients

Sample Selection Adjustment To remove the selection bias from the wage equation of women we use a semiparametric probit estimate: Q y(y w X) = Xβ w(τ) + h τ (z wγ) τ (0, 1) (4) The term h τ (z wγ) correct the selection at θ th quantile can estimate as: ĥ τ (z wγ) = δ 0 (τ) + δ 1 (τ)λ(z wγ) + δ 2 (τ)λ(z w) 2 (5)

Quantile Regression: Gender Wage Gap

Quantile Regression:Gender Wage Gap

Quantile Regression: Wage Gender Gap The OECD work-family reconciliation index is the sum of indicators as childcare, maternity leave, part-time job (see OECD, 2001)

Conclusions The raw gender gap differs across years and across countries. The distribution of characteristics are constant between husband and wife but women are still paid less than men. This inequality increased in 2006 and can be explained by changes in distribution of characteristics. The sticky floor effect has increased while the glass ceiling effect has decreased Countries with more generous work-family policies have a lower wage gap at the bottom of the wage distribution and a larger pay gap at the top.

Thanks for your attention.