Upload
grape
View
27
Download
5
Embed Size (px)
DESCRIPTION
We use several models to estimate the gender wage gap and compare the results. Our comparisons suggest that all methods produce similar results, though dispersion within each of them varies.
Citation preview
Gender wage gap in Poland
Gender wage gap in Poland
A comparative analysis of available methods
Lucas Augusto van der VeldePhD Candidate
Research Assistant in GRApE
Faculty of economic sciences
University of Warsaw
May 31, 2014
Gender wage gap in Poland
Table of contents
1 Introduction
2 Methods & Speci�cations
3 Data
4 Results
5 Conclusions
Gender wage gap in Poland
Introduction
Introduction
Motivation
Our work
Goal: Provide a guide for the practitioner
How: Compare the gender wage gap in di�erent methods and speci�cations
Data: Polish LFS 2012
Gender wage gap in Poland
Introduction
Introduction
Motivation
Our workGoal: Provide a guide for the practitioner
How: Compare the gender wage gap in di�erent methods and speci�cations
Data: Polish LFS 2012
Gender wage gap in Poland
Introduction
Introduction
Motivation
Our workGoal: Provide a guide for the practitioner
How: Compare the gender wage gap in di�erent methods and speci�cations
Data: Polish LFS 2012
Gender wage gap in Poland
Introduction
Introduction
Motivation
Our workGoal: Provide a guide for the practitioner
How: Compare the gender wage gap in di�erent methods and speci�cations
Data: Polish LFS 2012
Gender wage gap in Poland
Methods & Speci�cations
Methods under analysis
Linear regressions
Oaxaca Blinder decompositions (5)
John, Murphy and Pierce
Di Nardo, Fortin and Lemieux
Machado Mata
Nopo
Firpo, Fortin and Lemieux
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFL
Selection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
Summary
How should the perfect decomposition method look like?
Characteristics OB JMP DFL MM Nopo FFLSelection Bias OK OK OK
Dimensionality Curse OK OK OK OK OK
Detailed decomposition OK OK OK
Quantile decomposition OK OK OK OK
Common Support OK
Functional Form OK OK
Compare across time OK OK OK
Gender wage gap in Poland
Methods & Speci�cations
What we expect with respect to...
The selection bias: the adjusted gap increases when women experience
more selection than men
The addition of new variables increases the adj. gap when variation withinis larger than between
The common support: the adj. gap increases when the non-matchedwomen are better endowed than men.
Gender wage gap in Poland
Methods & Speci�cations
What we expect with respect to...
The selection bias: the adjusted gap increases when women experiencemore selection than men
The addition of new variables increases the adj. gap when variation
within is larger than between
The common support: the adj. gap increases when the non-matchedwomen are better endowed than men.
Gender wage gap in Poland
Methods & Speci�cations
What we expect with respect to...
The selection bias: the adjusted gap increases when women experiencemore selection than men
The addition of new variables increases the adj. gap when variation withinis larger than between
The common support: the adj. gap increases when the non-matched
women are better endowed than men.
Gender wage gap in Poland
Data
The Sample
Polish Labour Force Survey (2012)
Male Female C-Support
Hourly wage 11.91 11.00 0.12
Age 40.64 41.29 0.04
Experience 19.15 17.89 0.07
Agriculture 0.41 0.21 0.32
Construction 0.15 0.01 0.39
Industry 0.35 0.43 0.12
Services 0.08 0.35 0.49
Secondary 0.75 0.62 0.20
Tertiary 0.16 0.33 0.28
Social sciences 0.07 0.28 0.40
Medicine 0.01 0.08 0.26
Engeneering 0.59 0.18 0.65
Teaching 0.01 0.03 0.11
Gender wage gap in Poland
Data
Di�erent speci�cations
Basic: Age, experience, education levels, married, kids, rural, cities,Mazowieckie
Industry: Industry dummies for agriculture (reference), manufacture,construction and services.
Industry plus: "Industry" + �rm size and ownership type
Occupations: 9 occupational dummies (ISCO-1 codes)
Tenure: "Basic" + tenure
Education: 9 educational �eld dummies
Gender wage gap in Poland
Results
Comparison across methods
Gender wage gap in Poland
Results
Comparison across speci�cations
Gender wage gap in Poland
Results
Comparison of the methods
Gender wage gap in Poland
Conclusions
Conclusions
With respect to the gap
The adjusted gap is 20% of female gap - two times the size of the raw gap.
There is evidence of a glass ceiling in Poland
We did not �nd evidence of segregation on industries nor on �eld of study
Comparison of methods
On average similar results, but great variations between methods.
After correcting for the selection bias and the common support thedi�erences were greater
Gender wage gap in Poland
Conclusions
Questions or suggestions?
Thank you for your attention
Gender wage gap in Poland
Conclusions
Questions or suggestions?
Thank you for your attention
Gender wage gap in Poland
Conclusions
References
Blinder, A. (1973): "Wage Discrimination: Reduced Form and StructuralEstimates", Journal of Human Resources, 8, 436-455.
DiNardo, J. , N. Fortín, and T. Lemieux, 1996 �Labor market institutions and thedistribution of wages, 1973-1992: a Semi-parametric approach� Econometrica,Vol. 64, No.5, 1001 -1044.
Firpo, S., Fortín, N., and Lemieux, T. 2009 Unconditional Quantile regressions,Econometrica, Vol. 77, No. 3, 953-973
Fortín, N., T. Lemieux and S. Firpo, 2010 �Decomposition methods inEconomics� NBER Working paper 16045
Juhn, C., K. M. Murphy, and B. Pierce (1993): "Wage Inequality and the Rise inReturns to Skill", Journal of Political Economy", 101, 410-442.
Machado, J. A. F., and J. Mata (2005): "Counterfactual Decomposition ofChanges in Wage Distributions using Quantile Regression", Journal of AppliedEconometrics, 20, 445-465.
Nopo, H(2008) Matching as a Tool to Decompose Wage Gaps, The review ofEconomics and Statistics, May 2008, vol. 90, No. 2, Pages 290 � 299.