gender and age wage patterns in germany

Gender and Age wage patterns in Germany

Gender and Age wage patterns in GermanyAn analysis form the GSOEP

Lucas van der Velde(with J. Tyrowicz and I. van Staveren)

Annual Conference of the Verein fur SocialPolitik,

September 2015


Introduction

Introduction

Motivation

Introduce age-factors in the study of the gender wage gap.Ageing process in Europe.How to craft efficient policies to reduce Gender Wage Gap?

What we do

Explore the effects of the life-cycle in women’s earnings.Separate age-cohort effectsUse the DiNardo, Fortin and Lemieux (1996) decomposition.Data: German Socio-Economic Panel for 1984-2008


Introduction

How the gender wage gap changes over age?

Hump-shaped pattern

Differences increasing over the age-years


Introduction


Hump-shaped pattern

Unequal distribution of activities within the household.

Child bearing and child rearing and its expectation.

Gender bias in the measurement of human capital.

Statistical discrimination from the employers.



Introduction


Hump-shaped pattern


“Hysteresis effect”

“Double standard of ageing”


Data and method

Sample:The German Socio-Economic Panel

Yearly surveys covering a broad range of topics.

We restricted the sample to West Germany between1984-2008.

Almost 500 000 observations13 000 individual are observed for a decade or longer.2 300 of them are present in each wave.

We only keep observations from German nationals aged 25-59.


Data and method

A quick look at the sample


Data and method

A first glance at the gender wage gap

Notes: Dependent variables: tenure, experience, small kids in the household, married,education level and year.


Data and method

What are we doing

We pursue three analyses:

1 Decompose the GWG using the DiNardo, Fortin and Lemieux(1996) decomposition for different cohorts across time.

2 Panel analysis of determinants of changes in the AdjustedGWG over time.

3 Double decomposition.


Results

Decomposition at different agesResults for the adjusted GWG

Gender wage gap in different age groups (1984-2006).

Notes: adjusted gap estimated at the mean with the DiNardo, Fortin and Lemieux (1996) decomposition;smoothed (averaged over three years). Each bar represents a year in the sample, bars of similar colors correspond

to the same cohort. Red lines represent women’s participation rate, measured in the right axis.


Results

Panel estimates: age effects on the Adjusted GWG

Mean 1st Quartile 3rd Quartile25-29 Base level30-34 0.122*** 0.067 0.110*35-39 0.134*** 0.068 0.159**40-44 0.192*** 0.159*** 0.226***45-49 0.213*** 0.193*** 0.307***50-54 0.155** 0.125 0.312***55-60 0.195* 0.180 0.365**Year -0.010*** -0.008* -0.014**Observations 175 175 175R-squared 0.649 0.624 0.661

Notes: ***,**,* indicate significance at the 1 %, 5% and 10% level respectively. Thedependent variable is the adjusted gender wage gap calculated at different points of

thedistribution. All estimates include cohort specific effects and participation rates formen and women.


Results

Additional controls: Household characteristics

Mean q(.25) q(.75) Mean q(.25) q(.75)30-34 0.159** 0.063 0.152** 0.122** 0.063 0.06835-39 0.166** 0.062 0.162** 0.119** 0.041 0.09640-44 0.213** 0.143** 0.181** 0.154** 0.099** 0.134**45-49 0.219** 0.182** 0.175** 0.155** 0.127** 0.136**50-54 0.207** 0.175** 0.151* 0.112** 0.083** 0.113**55-59 0.435** 0.442** 0.316** 0.260** 0.268** 0.261**

% female -1.014** -1.129** -0.080main earner

% Tertiary -0.641 ◦ -1.410** 1.325*educated

Observations 175 175 175 175 175 175R-squared 0.461 0.456 0.269 0.434 0.455 0.286

Notes: **,*,◦ indicate significance at the 1 %, 5% and 10% level respectively. Thedependent variable is the adjusted gender wage gap calculated at different points ofthedistribution. All estimates include participation rates for men and women.


Results

Additional controls: Child-related concerns

Mean q(.25) q(.75) Mean q(.25) q(.75)30-34 0.133** 0.107** 0.054 0.160** 0.146** 0.02935-39 0.127** 0.078** 0.075◦ 0.168** 0.123** 0.080◦40-44 0.158** 0.133** 0.115** 0.206** 0.197** 0.129**

Places for kids <3 0.006 0.013** -0.004

Fertility change 0.313* 0.229 -0.230

Observations 76 76 76 80 80 80R-squared 0.360 0.341 0.209 0.633 0.464 0.558

Notes: **,*,◦ indicate significance at the 1 %, 5% and 10% level respectively. Thedependent variable is the adjusted gender wage gap calculated at different points ofthedistribution. All estimates include participation rates for men and women.


Results

Double decomposition

Age Characteristics Residuals Unexplained1984-1989

30-34 -0,05 0,08 0,0535-39 -0,04 0,04 0,1540-44 -0,17 0,2 0,1345-49 0,35 -0,45 0,3650-54 0 0,01 0,22

1990-199930-34 0 -0,1 0,1435-39 0 -0,43 0,5640-44 0,03 -0,02 0,1145-49 0 -0,07 0,250-54 0,01 -0,28 0,4

2000-200830-34 0,05 -0,17 0,1435-39 -0,18 0,03 0,2240-44 -0,11 -1,16 1,4345-49 -0,12 -0,47 0,7450-54 -0,18 -0,53 0,8


Results

Double decomposition: how different

Changes in the adjusted GWG: Panel analysis

Notes: The estimations also include controls for the adjusted GWG in the initial period


Conclusions

Conclusions

1 We separate cohort and age effect to understand changes in theGWG over age-years

2 The gender wage gap increases with age, possibly in anon-monotonic fashion.

Steep increase in early career and later stabilization (mean)Continuous increase in the later stages (q.75)

3 Variables connected to cohort specific trends do not affect our mainresults

4 Policy implication: measures to tackle the GWG should take intoaccount also the pos-productive age.


Conclusions

Questions or suggestions?

Thank you for your attention


Appendix

Institutional context in Germany

Reasons

1 Restrictions on pregnant women employment.

2 Lenght of the maternity leaves (up to three years).

3 Maternity benefits (amount and non-relation to the labormarket history).

4 Only part-time work compatible with maternity benefits.

5 Insuficient childcare facilities.

6 Social constraints: the persistence of the KKK (children,kitchen and church).


Appendix

Fertility patterns


Appendix

Day care facilities


Appendix

Household earnings


Appendix

Double decomposition: one cohort

Age Characteristics Residuals Unexplained

30-34 -0,08 0,11 0,0435-39 -0,01 -0,12 0,1540-44 0,16 -0,19 0,1545-49 0,02 -0,41 0,250-54 -0,26 0,25 0,05


Appendix

Introduction to the DiNardo, Fortin and Lemieuxdecomposition (1996)

Given a joint distribution of wages and characteristics of the form

fj(wi ,j) =

∫fi ,j(w |x) f (x |g = i , t = j)dx (1)

Where i represents the gender, male or female, and j represents the period

We can derive a counterfactual wage structure of the form by reweightingfemale observation to make them more similar to males.

fj(wcf ,j) =

∫ff ,j(w |x) Ψj(x)fj(x |g = f , t = j)dx (2)

where Ψ(x) is the reweighting factor and equals

Ψj(xj) =fj(x |g = m, t = j)dx

fj(x |g = f , t = j)dx(3)


Appendix

Introduction to the DiNardo, Fortin and Lemieuxdecomposition (1996)

Thanks to Bayes rule, we can estimate Ψj(xj) as follows

Ψj(xj) =Pr(g = m|x , j = t)Pr(g = f )

Pr(g = f |x , j = t)Pr(g = m)(4)

We decompose the differences as

fj(wm,j)− fj(wf ,j) = [fj(wm,j)− fj(wcf ,j)] + [fj(w

cf ,j)− fj(wf ,j)] (5)

The first term represents the unexplained component; and thesecond, the explained.


Appendix

Double decomposition

Presented in Simon and Welch(1985) to study convergence inblack workers’ wages

They decompose the change between periods t−1 and t in 4components

1 The relative changes in characteristics from t−1 to t2 Differences in characteristics in t3 Differences in wage structure in t4 The relative changes in wage structures from t−1 to t

The last component is similar to the unexplained component fromthe previous decomposition, though it is ”cleaner” for thecomparison across time. A simple difference between the adjustedgaps in two periods will also reflect the changes in thecharacteristics used as a base (women from each period) while inthis case, we use the same characteristics in the two periods

gender and age wage patterns in germany

Economy & Finance