Gender Unemployment Gaps: Evidence from the New EU Member ... · and increases in Poland, Slovenia,...
Transcript of Gender Unemployment Gaps: Evidence from the New EU Member ... · and increases in Poland, Slovenia,...
Gender Unemployment Gaps: Evidence fromthe New EU Member States
Alena Bicáková
CERGE-EI, Prague
European User Conference for EU-LFS and EU-SILC
March 5, 2009, Mannheim
Gender Differences in the Labor Market Outcomes
• gender wage gaps• gender unemployment gaps
Previous research focused predominantly on the the first one.
Only two cross-country papers on the second:
Azmat, Guell, Manning (JOLE 2006)
Stefanova-Lauerova, Terrell (Comparative Econ Studies 2007)
Why studying U-gap is (even more) important?
• affects lifetime / long-term earnings and income volatility• job security, unemployment stigma, skill deterioration• may affect female labor force participation,
discouraged worker• evidence on discrimination in hiring (audit studies)• may force women to accept worse jobs• may affect the observed wage gap
(Olivetti, Petrongolo JLE 2008)• may-be a trade-off between the two gaps depending on LM
institutions (wage flexibility)
Azmat, Guell, Manning JOLE 2006
based on cross-sectional cross-country comparisonusing ECHP data
they conclude that
gender unemployment gap tends to be higher in
• countries with higher overall unemployment rate• countries with lower female labor force participation• countries with lower observed gender wage gap
(effect of wage compressing institutions)
classic South - North divide
Gender U gap and Female LFP
AT
BE
DE
DK
ES
FI
FR
GR
IE
IT
LU
NL
NO
PT
SEUK
CZ
EE
HU LT
LV
PL
SI
SK
−.05
0.0
5.1
.6 .7 .8 .9 .6 .7 .8 .9
Old EU New EU
Une
mpl
oym
en G
ap
Female Labor Force Participation
New EU versus Old EU
High female labor force participation(similar to Nordic countries, Denmark, France ..).
Medium sized unemployment gaps.
No or negative gender unemployment gaps in Baltic countries.
Negative correlation between gender unemployment gapsand female labor force participation - observed for the Old EU
is not present among the New EU member states.
New Member States - Detailed Evidence
Gender unemployment gaps of prime age individuals in 2007
Country Male U Female U ratio difference t-stat
Czech Rep. 0.035 0.066 1.86 0.030 10.36Estonia 0.045 0.048 1.07 0.003 0.35Hungary 0.063 0.069 1.10 0.006 1.92Latvia 0.057 0.053 0.92 -0.004 -0.54Lithuania 0.038 0.041 1.09 0.003 0.70Poland 0.079 0.089 1.13 0.010 2.37Slovakia 0.087 0.124 1.42 0.037 5.90Slovenia 0.032 0.058 1.82 0.026 5.19
Source: EU LFS, own calculations, weighted, t-stat for the difference betweentwo independent variables with binomial distribution
Gender Unemployment Gaps 1996-2007
−5−2.5
02.5
5
−5−2.5
02.5
5
−5−2.5
02.5
5
1997 1999 2001 2003 2005 2007 1997 1999 2001 2003 2005 2007 1997 1999 2001 2003 2005 2007
1997 1999 2001 2003 2005 2007 1997 1999 2001 2003 2005 2007 1997 1999 2001 2003 2005 2007
1997 1999 2001 2003 2005 2007 1997 1999 2001 2003 2005 2007
Czech Republic Estonia Hungary
Latvia Lithuania Poland
Slovakia Slovenia
up95/low95 d_Urate
Year
New Member States - Specific Features
Pre-1989• zero unemployment rate, so zero gap• very high female labor force participation (not PL and HU)• working woman norm, supported by the state• informal child care, two-generational households
Post-1989• reduced child care provision• different social values• changes in anti-discrimination laws• different labor market policies
This paper
Document the gender unemployment gaps in the 8 New EUmember states.
Analyze the observed unconditional gaps via Oaxaca-Blindertype decomposition.
Analyze the observed unconditional gaps in terms of flows fromand to unemployment.
Explain the variation in gender unemployment gaps across theNew Member states.
Shed more light on the relationship between the genderunemployment gaps and female labor force participation.
Data: EU Labor Force Survey (1996-2007)
annual data from Q2, in cross-sectional analysis focus on 2007
prime age individuals (25-54 year old)
Country N in 2007Czech Republic 25,639Estonia 2,405Hungary 31,333Latvia 3,565Lithuania 7,148Poland 21,321Slovakia 11,409Slovenia 7,642
We exclude individuals in compulsory military service (very few)
Education
sample = labor force
Women are on average more educated than men
except in Czech Republic and Slovakia (at older ages).
Age 25-34 Age 35-44 Age 45-54 TotalM F M F M F M F
Czech Rep. 2.10 2.12 2.12 2.07 2.10 1.97 2.10 2.05Estonia 2.17 2.37 2.14 2.43 2.27 2.34 2.19 2.38Hungary 2.06 2.21 2.03 2.08 2.06 2.00 2.05 2.09Latvia 1.84 2.23 2.05 2.29 2.07 2.23 1.98 2.25Lithuania 2.21 2.42 2.17 2.32 2.17 2.24 2.18 2.33Poland 2.15 2.34 2.05 2.15 2.02 2.05 2.08 2.18Slovakia 2.11 2.14 2.10 2.05 2.09 2.00 2.10 2.06Slovenia 2.13 2.28 2.08 2.15 2.03 2.03 2.08 2.15
Weighted, year 2006.
Marital status and number of children less 15
sample = labor force
Marital status No. of ChildrenM F M F
Czech Rep. 0.631 0.668 0.667 0.595Estonia 0.527 0.519 0.600 0.545Hungary 0.601 0.623 1.155 1.019Latvia 0.612 0.552 0.778 0.742Lithuania 0.763 0.717 0.975 0.939Poland 0.755 0.755 2.038 2.026Slovakia 0.689 0.714 0.987 0.89Slovenia 0.563 0.646 0.887 0.934
Weighted, year 2006.
Flexible decomposition I
construct J subgroups based on discrete versions of X -s
the overall gender unemployment gap Ugap defined as thedifference between the female uF and male uM unemploymentrate can be written in terms of the J sub-groups
Ugap = uF − uM =∑
j
wFj uF
j −∑
j
wMj uM
j
where uGj is the unemployment rate in subgroup j for gender G
and wGj is the share of subgroup j among gender G
Flexible decomposition II
adding and subtracting terms for the overall gender-neutralunemployment rates weighted by the gender specific weights∑
j wFj uj and
∑j wM
j uj , we get
Ugap =∑
j
wFj (uF
j − uj)︸ ︷︷ ︸A
+∑
j
wMj (uj − uM
j )
︸ ︷︷ ︸B
+∑
j
(wFj − wM
j ) uj︸ ︷︷ ︸C
A and B is the part of the U-gap due to gender differences inthe respective subgroups
C is the part of the U-gap due to gender differences in thedistribution across the subgroups [i.e. differences in observedcharacteristics]
Flexible Oaxaca-Blinder Results 2007
18 groups based on age(6) and education(3)U gap A B C (A+B)
Czech Rep. 0.030 0.014 0.011 0.005 0.025Estonia 0.003 0.005 0.005 -0.006 0.010Hungary 0.006 0.005 0.004 -0.003 0.009Latvia 0.000 0.004 0.004 -0.009 0.008Lithuania 0.003 0.004 0.004 -0.005 0.008Poland 0.010 0.010 0.009 -0.008 0.018Slovakia 0.037 0.015 0.013 0.009 0.028Slovenia 0.026 0.015 0.014 -0.003 0.029
A =P
j wFj (uF
j − uj), B =P
j wMj (uj − uM
j ), C =P
j(wFj − wM
j ) uj
A + B is the U-gap if women and men equally distributedacross J groups
Flexible Oaxaca-Blinder Results: Summary
The within-group gender unemployment gap turns out to bepositive everywhere, although close to zero for Baltic countriesand Hungary.
Except for Czech Republic and Slovakia, women have morefavorable distribution across age and education categories thanmen, therefore partly reducing the unemployment gap.
17 % of the U-gap in CZ and 24 % of the U-gap in SK causedby unfavorable distribution of women across age and education
Gender Unemployment Gaps 1996-2007
−.05
0
.05
−.05
0
.05
−.05
0
.05
1999 2001 2003 2005 2007 1999 2001 2003 2005 2007 1999 2001 2003 2005 2007
1999 2001 2003 2005 2007 1999 2001 2003 2005 2007 1999 2001 2003 2005 2007
1999 2001 2003 2005 2007 1999 2001 2003 2005 2007
Czech Republic Estonia Hungary
Latvia Lithuania Poland
Slovakia Slovenia
Unexplained U−gap Raw U−gap
Year
Parametric Oaxaca-Blinder
Model the probability of being unemployed
use three specifications
• Pr(U = 1|LFP = 1, X ) = F (α + β ∗ FEM)
• Pr(U = 1|LFP = 1, X ) = F (α + β ∗ FEM + Xγ)
• Pr(U = 1|LFP = 1, X ) = F (α + β ∗FEM + Xγ + FEM ∗Xδ)
where X is a set of human capital or family relatedcharacteristics
F (Xβ) = Xβ (LPM) or F (Xβ) = Φ(Xβ) (probit)
estimation for 2007 data, by country separately
Conditional Unemployment Gap
Coefficient of female dummy in LPM, robust standard errors
human C = six age categories, educM, educH
family = number of children (0 to 4 and above), marital status
no Xs + human C Xs + family Xs + interactionscoeff se coeff se coeff se coeff se
cz 0.031 0.003 0.024 0.003 0.023 0.003 0.012 0.024ee -0.001 0.009 0.006 0.009 0.006 0.009 0.075 0.063hu 0.004 0.003 0.006 0.003 0.007 0.003 0.016 0.015lv -0.010 0.008 -0.001 0.008 -0.003 0.008 0.062 0.063lt -0.001 0.005 0.005 0.005 0.005 0.005 -0.007 0.049
pl 0.012 0.004 0.020 0.004 0.021 0.004 0.037 0.025sk 0.038 0.006 0.027 0.006 0.029 0.006 -0.102 0.040si 0.023 0.005 0.027 0.005 0.031 0.005 0.067 0.025
Parametric Oaxaca-Blinder: Summary
After conditioning on human capital and family variables(comparing the same individuals) gender unemployment gapsremain similar to the unconditional onesbut decreases in Czech Republic and Slovakia,and increases in Poland, Slovenia, (and Lithuania).
After adding interactions with female dummy (allowing theeffect of RHS variables differ by gender) female effect alonecaptures gender unemployment gap of the base categoryyoung, low educated, single, no children
and is no longer significant except in Slovakia (becomes hugeand negative) and Slovenia (doubles).
Cost of Children - Women of Age < 44
children1 children2 children3 children4coeff se coeff se coeff se coeff se
cz 0.088 0.011 0.074 0.010 0.136 0.018 0.211 0.047ee -0.017 0.024 -0.003 0.028 -0.031 0.027 0.050 0.077hu 0.034 0.010 0.035 0.011 0.079 0.017 0.082 0.040lv -0.002 0.020 0.019 0.022 0.002 0.029 -0.054 0.021lt -0.010 0.016 -0.027 0.016 -0.012 0.023 0.031 0.047pl 0.006 0.012 -0.003 0.013 -0.011 0.017 -0.025 0.024sk 0.052 0.020 0.049 0.021 0.081 0.028 0.154 0.046si -0.025 0.019 -0.053 0.017 -0.031 0.025 -0.038 0.049
Coefficients from the linear probability regression with human capital andfamily characteristics, estimated for women less or 44 years old, in 2007.
Labor Force Participation by Age Group in 2007
.6
.7
.8
.9
1
.6
.7
.8
.9
1
.6
.7
.8
.9
1
27 32 37 42 47 52 27 32 37 42 47 52 27 32 37 42 47 52
27 32 37 42 47 52 27 32 37 42 47 52 27 32 37 42 47 52
27 32 37 42 47 52 27 32 37 42 47 52
Czech Republic Estonia Hungary
Latvia Lithuania Poland
Slovakia Slovenia
males females
Age
Gender U gap and Female LFP
CZ
EE
HULT
LV
PL
SI
SK
CZ
EE
HULT
LV
PL
SI
SK
−.05
−.04
−.03
−.02
−.01
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8
.6 .7 .8 .9 .6 .7 .8 .9
Prime−age Female LFP 25−29 Age Female LFP
Une
mpl
oym
en G
ap
Female Labor Force Participation
Determinants of gender unemployment gap
Steady state unemployment rate
u =δ
δ + λ
where δ firing rate and λ is the rate of leaving unemployment
Condition that inflows equal outflows: (1 − u) δ = λ u
Azmat et al. (2006): gender differences in both δ and λ
Stefanova-Lauerova and Terrel (2007): gender differences in λ
δ typically assumed exogenous in job search models
Flow analysis - U to E transition
no Xs + human C Xs + family Xs + interactionscoeff se coeff se coeff se coeff se
cz -0.038 0.009 -0.033 0.008 -0.042 0.009 -0.061 0.024ee 0.010 0.019 -0.004 0.019 -0.024 0.020 -0.050 0.065hu -0.028 0.005 -0.023 0.005 -0.034 0.005 -0.033 0.014lv 0.004 0.017 0.004 0.017 -0.001 0.017 -0.039 0.050lt -0.061 0.017 -0.057 0.017 -0.061 0.018 -0.056 0.061
pl -0.083 0.005 -0.090 0.005 -0.103 0.005 -0.070 0.016sk -0.038 0.008 -0.026 0.008 -0.038 0.008 0.019 0.023si -0.021 0.010 -0.022 0.010 -0.030 0.010 -0.047 0.028
Coefficients of female dummy from different LPM specifications. Robuststandard errors. All available years, year fixed effects included.
Flow analysis - E to U transition
no Xs + human C Xs + family Xs + interactionscoeff se coeff se coeff se coeff se
cz -0.040 0.001 -0.038 0.001 -0.037 0.001 -0.084 0.006ee -0.010 0.003 -0.017 0.003 -0.017 0.003 -0.081 0.017hu -0.015 0.001 -0.015 0.001 -0.015 0.001 -0.045 0.004lv -0.015 0.005 -0.021 0.005 -0.023 0.005 -0.044 0.022lt 0.005 0.002 0.000 0.002 0.001 0.003 0.013 0.016
pl -0.007 0.001 -0.011 0.001 -0.011 0.001 -0.021 0.006sk -0.035 0.002 -0.036 0.002 -0.035 0.002 -0.061 0.013si -0.012 0.001 -0.014 0.001 -0.015 0.001 -0.022 0.005
Coefficients of female dummy from different LPM specifications. Robuststandard errors. All available years, year fixed effects included.
Cost of Family - U to E transition - Women Age <44
children1 children2 children3 children4coeff se coeff se coeff se coeff se
cz -0.124 0.025 -0.084 0.025 -0.153 0.032 -0.219 0.044ee 0.066 0.050 0.008 0.050 0.057 0.066 0.041 0.087hu 0.006 0.015 0.026 0.016 -0.011 0.021 -0.022 0.035lv -0.084 0.066 -0.059 0.070 -0.104 0.100 -0.111 0.144lt 0.017 0.055 0.007 0.056 -0.139 0.069 -0.092 0.091pl -0.008 0.017 -0.003 0.017 -0.018 0.020 -0.011 0.025sk -0.014 0.027 -0.027 0.025 -0.039 0.030 -0.053 0.033si -0.025 0.032 -0.003 0.032 -0.063 0.049 -0.124 0.077
Coefficients from the linear probability regression of the women’s transitionrate from unemployment to employment, human capital and familycharacteristics and year fixed effects included.
Conclusion I
Unexplained part (within education and age categories) ofunemployment gap is present in all countries but is close tozero for some.
Baltic countries have much lower unemployment gap, mostlyunrelated to marital status and children.
In Slovenia - U-gap only among younger women irrespective ofwhether married or with children.
Marital status and children increase U-gap in other countries.
Conclusion II
The largest and most persistent U-gap is in the Slovakia, CzechRepublic, and Poland.
Variation in LFP of women after childbirth seem to account forthe respective labor market costs of children, which explainmuch of the observed gender unemployment gaps.
The observed changes in LFP are in line with the generosity ofthe country-specific maternity and parental leave policies.