Transcript Gender wage differentials in Sweden 1993–2002

Wage differences between women and men in
Sweden – the impact of skill mismatch
by
Mats Johansson and Katarina Katz
Katarina Katz, Department of Economics and statistics,
Karlstad University, Sweden
1
ABSTRACT
We investigate skill mismatch and its impact on gender
differences in wage gap and in returns to education in
Sweden 1993 to 2002.
Women are more likely to have more formal education than
what is normally required for their occupation
(overeducation), while men are more likely to have less
(undereducation).
Over- and undereducation contribute far more to the gender
wage gap than years of schooling and work experience. In
decompositions, adjusting for skill mismatch decreases the
gender wage gap by between one tenth and one sixth.
This is roughly a third to a half as much as is accounted for
by segregation by industry. Thus, taking skill mismatch into
account is essential for the analysis of gender wage
differentiation, even though it does not alter the result that
the estimated returns to education are smaller for women
than for men in Sweden.
•
Published by the Institute for Labour Market Policy
Evaluation, Uppsala, Sweden
•
as IFAU Working Paper 2007:13
•
www.ifau.se
2
Aim of study:Join insights and approaches from
two fields of economics:
 Economics of gender
• Gender and wages
─Decomposition of the gender wage gap
 Economics of education
• Over-, Under- and Required Education
─ORU and wage effects of skill mismatch
3
Standard result from the ORUliterature:
• Women are more often overeducated than
men.
• Men are more often undereducated than
women.
4
Standard results from the ORUliterature:
• An overeducated person earns more than a person
doing the same job who has only the required education.
• An overeducated person earns less than a person with
the same education whose job requires this education
• An undereducated person earns less than a person
doing the same job who has the required education.
• An undereducated person earns more than a person
who has the same education but whose job doesn’t
require more.
5
Why mismatch?
• Differences in ability (unobserved heterogeneity)
• More other human capital compensates for
undereducation/overeducation compensates for
less other human capital.
• Search theory (cost of finding a good match)
• Spatial restrictions (local labour markets).
• Assignment theory (supply and demand on the
labour market do not match).
• BUT WHY THE GENDER DIFFERENCE?
6
Frank (1978): Married women are tied movers/stayers
Weak or no support for Frank’s hypothesis in Büchel (2000),
McGoldrick & Robst (1996), Battu et al. (2000) and Büchel &
Battu (2003). Married women are not more overeducated than
single.
Rubery et al (1989): Women, particularly, part-timers are often
overqualified and underpaid (education & content of work).
They also tend to undervalue the skill level of their jobs.
Many ORU-studies disregard gender and hardly
any place it at the centre of attention.
7
Our research questions:
1. Are women more often overeducated for their
jobs and men more often undereducated in
Sweden too?
2. Does skill mismatch explain the (observed)
gender difference in returns to education?
3. Does skill mismatch explain any part of the
persistent gender wage differential in Sweden?
4. Have changes in skill mismatch had an impact
on the development of the gender wage
differential in the 1990s?
8
Data
•HINK/HEK
–repeated yearly cross-section 1993-2002
(collected by Statistics Sweden).
– 6 000-10 000 employed individuals aged 2064 each year (self-employed excluded).
– register data on level of education
– interview data on occupation
9
Structure of study:
• Estimated probability of being over/undereducated, separately by gender (multinomial
logit).
• Estimated wage equations to measure returns to
education with/without controls for over- and
undereducation, separately by gender (OLS).
• Oaxaca-Blinder decomposition of the gender
wage differential with a ORU-model.
• Juhn-Murphy-Pierce decomposition of the
change in gender wage differential 1993-2002.
10
Explanations of gender wage
differentials
•
•
•
•
human capital
job segregation
wage discrimination
value discrimination
11
To operationalise empirically:
• Wage equation: ln Wi =Xiβj
j = f, m
• Oaxaca Blinder decomposition
Dt  ln Wmt  ln W ft  X t  mt  X mt  t
• ”Explained”
(endowment) term
• Unexplained term
Dt  ln Wmt  ln W ft  X t  ft  X ft  t
12
Wage equation
•
•
•
•
•
•
•
•
•
•
•
Dependent variable: ln (hourly wage)
Independent variables:
Experience + experience squared
Years of education
Years of undereducation (if positive)
Years of overreducation (if positive)
Industry (12)
Country of birth (3 categories)
Region (3 categories)
Marital status
Children under 18 in household
13
ORU-measure
• We use Socio Economic Index (SEI) which defines the
level of education required for occupations.
• From the level we impute years of education.
• We use a non-standard specification of the ORU wage
equation:
• ln W = α + β1jAE + β2j OE + β3jUE + βjX + ε j = f, m
where
• AE = years of actual education
• RE = years of required education
 AE  RE if AE  RE 
OE  

0
if
AE

RE


 RE  AE if RE  AE 
UE  

0
if
RE

AE


14
Robustness of data and measure?
• Swedish studies using the Level of Living Survey
show (le Grand, Szulkin, Thålin & Korpi)
– Levels of male/female OE and UE similar to ours.
– High correlation between SEI-based measure and
respondents self-assessment of required education.
• Study using Living Conditions Survey
(Oscarsson and Grannas)
– Alsp find similar levels when using SEI
– Find substantially less OE when using a more
classification (SSYK)
15
Over-, Under and Adequate
Education Sweden 1993-2002
0,600
0,500
0,400
AE Women
OE Women
UE Women
AE Men
0,300
OE Men
UE Men
0,200
0,100
0,000
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
16
•
* Probabilities were also estimated by a
multinomial logit model which confirmed
the descriptive results.
17
18
Factors associated with the higher probability of
overeducation:
• Level of education:
– Women: Long (≥ 3 yrs) secondary or long (≥ 3 yrs) higher education.
– Men: Short secondary education.
• Field of education:
– Both women and men: Transport & communication, services.
• Industry:
– Women: Transport & communication, trade
– Men: Transport & communication, public administration, culture
& recreation.
• Sector:
– Women: Private sector
– Men: Local government
• Work time:
– Both women and men: Part-time work
19
Factors associated with the lower probability
of overeducation:
• Level of education:
– Women: Short higher education.
– Men: Short and long higher education.
• Field of education:
– Women: Health care
– Men: Health care, science, technology & manufacturing.
• Industry:
– Women: Financial and insurance, health care, recreation and
culture
– Men: Hotels & restaurants, construction
• Sector:
– Women and men: Central government.
20
Family and female overeducation:
• Men are more likely to be undereducated and less likely
to be overeducated if they are married. Women are
slightly more likely to be overeducated if they are single.
• Women with children aged 0-6 are somewhat less likely
to be overeducated than women without children. The
difference between women without children and with
children age 7-16 is not statistically significant.
• Thus, the Swedish data do not support the hypothesis
that female overeducation is due to restricted labour
markets for women with family committments.
21
Education parameters in traditional and ORUmodel
0,09
Trad.
model
Men
0,08
0,07
ORUmodel
Men
0,06
0,05
Trad.
model
Women
0,04
0,03
ORUmodel
Women
0,02
0,01
•
02
20
01
20
00
20
99
19
98
19
97
19
96
19
95
19
94
19
19
93
0
Both models include experience, education, industry, large
city dummy, marital status, children, country of birth.
22
• Standard Mincer model
• Returns to actual education, no control for over- or
undereducation (per year)
– 4.6% for women
– 5.8% for men
• ORU-model:
• Returns to education 2002 (per year):
• For those having the appropriate level:
– 5.6 % for women
– 7.7 % for men
• For a year of overeducation:
– 2.3 % for women (5.6% - 3.4%)
– 3.2 % for men (7.7% - 4.5%)
• Reward per year of undereducation:
– 2.9 % for women
– 3.5 % for men
23
Gender wage differentials in Sweden 1993–2002 (unadjusted and
adjusted), female average wages in percent of the male wages.
Adjusted differential
Unadjusted
differential
Adjusted,
female
Adjusted,
male
1993
85.0
88.6
89.9
1994
85.6
87.2
91.4
1995
85.1
88.5
90.9
1996
86.2
90.5
92.5
1997
86.0
89.1
91.5
1998
84.5
88.7
90.9
1999
85.1
89.4
90.8
2000
86.0
90.1
91.1
2001
85.4
89.3
90.0
2002
86.8
90.0
91.7
24
Decomposition of the gender gap in log wages using the
parameters estimated for women
Total wage
gap
Total
endow
ments
Work
experience
Educa-tion
Skill mismatch
Industry
Other
1993
0.163
0.042
0.008
-0.005
0.014
0.026
-0.001
1994
0.155
0.018
0.006
-0.004
0.015
0.001
-0.001
1995
0.161
0.039
0.005
-0.008
0.014
0.029
-0.002
1996
0.148
0.049
0.005
-0.007
0.009
0.041
0.000
1997
0.150
0.035
0.000
-0.006
0.013
0.030
-0.002
1998
0.169
0.049
0.000
-0.005
0.014
0.042
-0.001
1999
0.162
0.050
0.001
-0.008
0.014
0.042
0.000
2000
0.151
0.046
0.003
-0.016
0.014
0.048
-0.002
2001
0.158
0.045
0.000
-0.019
0.020
0.046
-0.001
2002
0.142
0.037
0.005
-0.014
0.015
0.032
0.000
25
Decomposition of the gender gap in log wages using
the parameters estimated for men
Total wage
gap
Total
endow
ments
Work
experience
Educa-tion
Skill mismatch
Industry
Other
1993
0.163
0.042
0.008
-0.005
0.014
0.026
-0.001
1994
0.155
0.018
0.006
-0.004
0.015
0.001
-0.001
1995
0.161
0.039
0.005
-0.008
0.014
0.029
-0.002
1996
0.148
0.049
0.005
-0.007
0.009
0.041
0.000
1997
0.150
0.035
0.000
-0.006
0.013
0.030
-0.002
1998
0.169
0.049
0.000
-0.005
0.014
0.042
-0.001
1999
0.162
0.050
0.001
-0.008
0.014
0.042
0.000
2000
0.151
0.046
0.003
-0.016
0.014
0.048
-0.002
2001
0.158
0.045
0.000
-0.019
0.020
0.046
-0.001
2002
0.142
0.037
0.005
-0.014
0.015
0.032
0.000
26
Results of Oaxaca decomposition
• A quarter to a third of gap ”explained” if
parameters for women are used, 2/5 if
those for men are.
• Contribution of actual education < 0, of
experience very small
• 1.5 - 3 percentage point differential is
attributable to mismatch
• 3-6 percentage point differentiable
attributable to industry
27
Results of JMP decomposition
• Total change 1993-2002 very small – 2.1
percentage points.
• But according to JMP it is the net of larger
changes in opposite directions
– Gender specific factors tend to decrease gap
• This is mainly the mainly ”unobserved skills” term
which includes discrimination
– Changes in wage structure tend to increase it
• This is mainly the ”unobserved wage structure” i.e.
wage dispersion
28
Summary
• only about half of employed women and men have an occupation
that matches their level of education.
• women are more often overeducated than men
• women are less often undereducated than men
• more undereducation and less overeducation among those with long
experience
• formal educational requirements and the education of the employee
correspond more closely in the public than in the private sector
• independently of skill mismatch women received smaller rewards to
education than men.
• skill mismatch does contribute to the gender wage gap.
• with the same over- and under-education the gender wage gap would
decline by 1.4 (female eq.) or 2.3 (male eq.) percentage points.
• skill mismatch accounts for a considerably larger part of the
endowment term than traditional human capital variables
29
Suggestion for further work 1
• Panel studies – with a gender perspective
• To control for unobserved heterogeneity when
doing a gender comparison.
• To observe OE and UE changes over careers and
see if there is a gender difference.
• Korpi & Thålin (2007) find that the overeducated
do not ”catch up” in terms of wages – but is this
equally the case for women and men?
• To observe effects of changes in family status and
of parental leave on both incidence of and wage
effects of overeducation and see if these vary
according to gender.
30
Suggestion for further work 2
• More detailed study of the occupations in which the
expert- and or self-assessed required education
levels often differ from the actual.
– Scrutiny of assigned required education – do the
classifications have a gender bias?
– Qualitative study – in-depth interviews with respondents and
employers about the occupations that quantitative study
identify as having high levels of overeducation.
– Integrate experience and results from comparable worth
studies – for instance the ”equal opportunity wage revisions”
that are mandatory for Swedish employers.
31
Suggestion for further work 3:
• Integrate over- and undereducation in the
mainstream of studies of gender wage
differentials and of gender differences in
careers (job mobility and wage increases).
32
Suggestion for further work 4
• Similar studies focussing on immigrants –
integrating both gender and ethnic
dimension.
33