Upload
duongtuyen
View
215
Download
1
Embed Size (px)
Citation preview
Slowing Women's Labor Force
Participation: The Role of Rising Income
Inequality
Stefania Albanesi, FRBNY and CEPRMará José Prados, USC
USC - October 2, 2014
The views expressed here are those of the authors and do not necessarily re�ect
the views of the Federal Reserve Bank of New York or the Federal Reserve System.
1 / 63
Trends in LFP and Wages
I Slowing female participation and convergence in wages sincethe mid-1990s in the US.
.4.4
5.5
.55
.6F
em
ale
labor
forc
e p
art
icip
ation r
ate
1.3
1.4
1.5
1.6
Gender
wage g
ap
1975 1980 1985 1990 1995 2000 2005 2010Year
Gender wage gap Female labor force participation rate
Gender wage gap is computed as mean wage of male workers over mean wageof female workers for full time, full year workersThe values are three−year moving averages.
Gender Wage Gap and Female LFP, married workers
Source: March Supplement of CPS
2 / 63
Trends in LFP and WagesBy education
I Phenomenon is more pronounced for college women.
.3.4
.5.6
.7F
em
ale
la
bo
r fo
rce
pa
rtic
ipa
tio
n r
ate
1.2
1.3
1.4
1.5
1.6
1.7
Ge
nd
er
wa
ge
ga
p
1975 1980 1985 1990 1995 2000 2005 2010Year
Gender wage gap Female labor force participation rate
Gender wage gap is computed as mean wage of male workers over mean wageof female workers for full time, full year workersThree−year moving averages
College
Gender Wage Gap and Female LFP, married workers
.3.4
.5.6
.7F
em
ale
la
bo
r fo
rce
pa
rtic
ipa
tio
n r
ate
1.2
1.3
1.4
1.5
1.6
1.7
Ge
nd
er
wa
ge
ga
p
1975 1980 1985 1990 1995 2000 2005 2010Year
Gender wage gap Female labor force participation rate
Gender wage gap is computed as mean wage of male workers over mean wageof female workers for full time, full year workersThree−year moving averages
High school or less
Gender Wage Gap and Female LFP, married workers
Source: March Supplement of CPS
3 / 63
Trends in LFP and WagesSkill premium by sex
I The growth in the skill premium has accelerated for men andslowed for women since the mid-1990s, resulting in a growingdivergence.
1.4
1.6
1.8
2
1975 1980 1985 1990 1995 2000 2005 2010Year
College Premium, Males College Premium, Females
Premium is computed as mean wage of college workers over mean wageof non−college workers for full time, full year workers
Skill Premium
11
.05
1.1
1.1
51
.2skill
pre
miu
m m
en
/skill
pre
miu
m w
om
en
1975 1980 1985 1990 1995 2000 2005 2010Year
Rolling three years average
all fulltime workers
Ratio of Male to Female Skill Premium
Ratio of mean hourly wages between workers with college degree and those without a college degree, forFT-FY workers, regardless of marital status. Three year centered moving averages.
Source: March Supplement of CPS
4 / 63
Hypothesis
I Rise in skill premium can explain the lack of convergence inparticipation and wages across genders since the early 1990s.
I Mechanism:
I Rise in male earnings generates negative wealth e�ect onfemale participation and market hours.
I Reduced attachment to the labor market, reduces experienceand earnings for skilled women relative to skilled men.
I Positive assortative matching ampli�es e�ect on skilledwomen.
I Objective: Document empirical evidence on participation byeducation and household type, explore mechanismquantitatively.
5 / 63
Hypothesis
I Rise in skill premium can explain both the lack of convergencein participation and wages across genders since the early 1990s.
I Preview of �ndings:
I Mechanism can account for large fraction of lack ofconvergence in female participation to male in 1995-2005 forcollege women and women married to college husbands.
I Mechanism can partially reproduce the lack of convergence infemale wages to males in 1995-2005 and the lower growth inthe skill premium for women.
6 / 63
Related Literature
I Theory: Fernandez (2013), Fogli and Veldkamp (2013)
I Rise in female participation due to learning about costs for thehousehold.
I S-shape as learning slows with most women in the labor force.
I Empirical: Blau and Kahn (2013)
I International comparison, lack of part time and access todaycare.
I Empirical: Fortin (2013)
I Evolution of gender attitudes.
7 / 63
Outline
I Evidence
I Model
I Quantitative Analysis
I Ongoing work
8 / 63
Evidence
Data and Sample
I CPS March Supplement, 1964-2009, and PSID, 1985-2011.
I Adults ages 25 to 64.
I Married (for most of the analysis)
I Labor force participation de�ned as: working or looking forwork at least 40 weeks a year.
10 / 63
Skill PremiumMeasuring the Phenomenon
1.3
1.4
1.5
1.6
1.7
1.8
1.9
2
1976 1980 1984 1988 1992 1996 2000 2004
Skillpremium,full‐9mefull‐yearworkers
Skillpremium,malesSkillpremium,females
Source: March Supplement of CPS
Dashed line corresponds to linear trend for 1976-1991, projected out of sample.
11 / 63
Skill PremiumGender Wage Gap by Education
I The acceleration in the male skill premium post-1995 is alsoassociated with a rise in the gender wage gap for collegeworkers.
Skill Premium Male/Female Wages
Male Female HS Coll
Average 1995-2005
Actual 1.86 1.64 1.29 1.46
Projected 1.75 1.60 1.19 1.32
Actual-Projected 0.11 0.04 0.10 0.14
Actual-Projected% 6.29 2.5 8.4 10.61Source: Authors' calculations from CPS.
12 / 63
Labor supply of womenI The slowing of female participation is driven by prime age
married women. The participation of married men hasremained stable. Female LFP by Age
.4.5
.6.7
.8.9
LFP
1976 1981 1986 1991 1996 2001 2006 2011Year
Married women Single womenMarried men
Labor force participation rates
Source: March Supplement of CPS13 / 63
Measuring the PhenomenonLabor force participation of married women
I Procedure:
I Estimate probit model of labor force participation of marriedwomen for 1964-1994.
I Use estimated parameters to project female labor forceparticipation from 1995-2009.
I Compare projected and actual series.
14 / 63
Labor supply of married womenParticipation by education
.3.4
.5.6
.7.8
1964 1969 1974 1979 1984 1989 1994 1999 2004 2009Year
High school College
Labor force participation of married womenby education level, age−adjusted
Source: March Supplement of CPS
Dashed line corresponds to 1965-1994 participation probit, in and out of sample.
15 / 63
Measuring the PhenomenonAge-adjusted participation by education and husband's education
I The slowdown in participation is most pronounced for marriedwomen with college and with college husbands.
Married Women's Participation
HS Coll HS Husb Coll Husb
Average 1995-2005
Actual 0.57 0.65 0.61 0.60
Projected 0.63 0.74 0.67 0.70
Actual-Projected -0.05 -0.09 -0.06 -0.10
Actual-Projected% -9.5 -12.1 -8.9 -14.0
Source: Authors' calculations from CPS.
16 / 63
Labor supply of womenParticipation by household type, age-adjusted
I The slowdown in participation has been greatest for highschool and college women with college husbands.
Married Women's Participation
Household Types (M-F) HS-HS C-HS HS-C C-C
Average 1995-2005
Actual 0.6 0.56 0.73 0.63
Projected 0.66 0.67 0.79 0.75
Actual-Projected -0.06 -0.11 -0.06 -0.12
Actual-Projected% -9.8 -17 -8 -17Source: Authors' calculations from CPS.
17 / 63
Distribution of Household Types
I The fraction of college-college households is above 20% sincethe mid-1990s.
Figure : Fraction of households by education pairs
���������������������������
���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ���� ����
� �������� ������ ������������������ ���������������� ������������������ ���������������� �������� �����Household type corresponds to husband-wife educational attainment.
Source: March Supplement of CPS
18 / 63
Labor supply of married womenThe role of husband's earnings
05
1015
20%
diff
eren
ce
1 2 3 4 5Quintiles of husband’s earnings
Married Women LFP: Actual vs Projected
Source: March Supplement of CPS
Dashed line corresponds to 1976-1991 participation probit, in and out of sample.
The di�erence between 1995-2009 average and the projected valuesover that period is positively related to husband's earnings.Alt: Participation by husband's deciles
19 / 63
Labor supply of married womenParticipation by husband's earnings
I The slowdown in participation is more pronounced for womenwith husband in the top decile of the male labor incomedistribution, suggesting role of negative income e�ect.
Married Women's Participation
Aggregate Husb <10% Husb 50% Husb> 90%
Average 1995-2005
Actual 0.65 0.75 0.68 0.50
Projected 0.75 0.80 0.76 0.67
Actual-Projected -0.10 -0.05 -0.09 -0.17
Actual-Projected% -13 -6 -12 -25
Source: Authors' calculations from CPS.
Fert.
20 / 63
Negative Income E�ectOne and Two Earner Households
I Average labor earnings of husbands are signi�cantly higher inone-earner households.
Table 2: Sensitivity of female labor participation to husband’s earnings bracket
Coef. Std. Err. z P>z [95% Conf. Interval]
age 0.0060 0.0002 32.11 0.00 0.006 0.006education years 0.0900 0.0011 81.97 0.00 0.088 0.092
degree 0.1930 0.0112 17.30 0.00 0.171 0.215degreeXperc25 -0.2111 0.0169 -12.46 0.00 -0.244 -0.178degreeXperc50 -0.2820 0.0127 -22.12 0.00 -0.307 -0.257degreeXperc75 -0.4821 0.0135 -35.68 0.00 -0.509 -0.456degreeXperc90 -0.8764 0.0120 -72.78 0.00 -0.900 -0.853
constant -1.1311 0.0157 -72.19 0.00 -1.162 -1.100
Table 3: Average wage of husband, by household types
One-earner Two-earner One-earner hh/Period households households Two-earner hh
1976-1993
No College - No College 18.8 16.1 117%No Coll Husb - Coll Wife 22.1 17.1 129%
College Husb - No Coll Wife 29.7 23.0 129%College - College 31.2 24.1 130%
1994-2008
No College - No College 19.4 18.0 108%No Coll Husb - Coll Wife 25.2 20.0 126%
College Husb - No Coll Wife 38.5 29.0 133%College - College 44.7 33.1 135%
in one and two-earner households have become less di↵erent in this last period, whereas college de-
gree men in two earner households are further apart, in terms of earnings, from those in one-earner
households now than what they were before 1994.
Figure 11 shows the labor force participation rates of married women, by household education
types. The graph shows the rates relative to 1993. There is a steep increase in participation until
1993, and after that there is a change in trend for all married women, regardless of education level.
However, the slow down is more pronounced for married women with college degree, since the slope
during the previous decades was steeper than for women without college. The participation of
women married to a college degree husband stopped increasing completely after 1993.
The role of fertility Another potential explanation to the observed behavior of married women’s
labor force participation could be connected to changes in fertility. Women with children are less
likely to participate in the labor force, and hence increases in fertility could a↵ect labor force
participation of married women. However, we can rule out this explanation as the only cause
of the observed behavior of female labor force participation. As shown in figure 12, the decline
in the growth rate of labor force participation was similar for women with and without children.
11
Source: Authors' calculations from CPS.
Regressions21 / 63
Negative Income E�ectHousehold Net Worth
I The corresponding rise in the dispersion of household networth since the mid-1990s likely also contributed to thenegative income e�ect.
0"
200"
400"
600"
800"
1,000"
1,200"
1,400"
1,600"
1989" 1992" 1995" 1998" 2001" 2004" 2007" 2010"
Thou
sand
)dollars)
Net"Worth"by"Household"Educa=on"Type"
C,"C"
HS"wife,"C"husb"
C"wife,"HS"husb"
Hs,"HS"
Mean"
Source: Survey of Consumer Finances
Figure 16: Household Net Worth
3 Model
The model is a non-unitary model of household labor supply, where individuals are “egoistic” and
have independent utility functions, but it is assumed that the household decision-making process
is cooperative, so that household decisions are Pareto e�cient.4, 5
In this economy there is a continuum of measure one of individuals, and each male individual
is assumed to be married to a female individual. Each of them draw a (permanent) productivity
value ✓i from their corresponding distributions. The distributions of ✓is depend on the individual’s
gender and education level. We consider two possible - exogenously given - education levels: less
than college (we call this category high school), and college degree or more (we call this category
college). Therefore, there are four kinds of households defined by the skill level of the partners:
j 2 J = {hs � hs, hs � coll, coll � hs, coll � coll} (measure µe over the set J of household types,
so thatP1
0
P10 µe (em, ef ) = 1). We take the evolution of µe over time as exogenous.
Individual i is indexed by gender s = m, f and age j = 1, .., J .
The model allows for human capital accumulation in the form of learning by doing (cumulative
working experience, non specific, subject to depreciation). This implies that workers lose skills
when they are unemployed or out of the labor force.
Individual labor earnings yij,t are a function of hours worked (li), worker’s stock of human capital
4This result can be achieved by an alternative interpretation of this process, namely assuming that there is a twoperiod decision process, and there is some exogenous sharing rule, given by characteristic of the marriage contract,for instance.
5The model is a particular case of the collective labor supply model in Chiappori (1992), where we assume afixed household welfare function and bargaining weights for all households, whereas in Chiappori’s general case, thehousehold welfare function is allowed to depend on prices, and endowments if any.
15
Source: Survey of Consumer Finances.
22 / 63
Labor Flows
I PN (exit: participation to non-participation) �ow ratesincreased for all, while NP (entry) �ow rates declined forwomen married to college husbands after1995. Patterns byhousehold type:
All Household Type Husband's Education
HS-HS HS-C C-HS C-C HS Coll
PN 1984-1994 0.060 0.056 0.057 0.038 0.048 0.061 0.055
1995-2005 0.073 0.063 0.075 0.045 0.074 0.072 0.065
NP 1984-1994 0.212 0.243 0.190 0.304 0.249 0.209 0.231
1995-2005 0.209 0.220 0.195 0.345 0.220 0.215 0.207Yearly transition rates. Source: Authors' calculations from the PSID.
23 / 63
Summary
I Skill premium for men rises by 6.3% in 1995-2005
I Male/female wage ratio rises by 11% for college workers
I Labor force participation of married women in 1995-2005relative to pre-1995 trend
I 12% lower for college womenI 14% lower for women with college husbandsI 25% lower for women with husbands in the to 10% of the
earning distributionI Similar behavior for hours
I PN �ow rates rise, NP rates fall post 1995, especially forwomen with college and with college husbands.
24 / 63
Quantitative Analysis
ModelHousehold Labor Supply
I Non-unitary households:
I Households comprise two married partners of di�erent gender.I Partners have independent utility functions.
I Marriage and educational attainment exogenous. No divorce.
I Heterogeneity in individual productivity.I Assortative matching by education based on empirical
distribution of household types.
I Intensive and extensive labor supply decisions.
I E�cient market and home hours allocation.
I Wages endogenous due to on the job acquisition of humancapital.
26 / 63
ModelHousehold Problem
maxc isj ,hisj ≥0,l
isj ∈{0,[l ,l ]},bj+1,ks
j+1≥0∑J
j=1 βj∑
s=f ,m λsu(c isj , l
isj + hisj
)
s.t.k isj+1 = F (l isj , k
isj )
Hj = G (hfj , hmj )
∑s=f ,m c isj + qbij+1 ≤
∑s=f .m W is
j l isj + bij for j = 1, 2, ...J
∑
s=f ,m
c isJ ≤ bJ
ks0 = ks
Hj given for j = 1, .., J
27 / 63
ModelHousehold Problem
I Indivisible labor:
l isj =
{0 if pisj = 0
≥ l if pisj = 1
I Home hours requirement:
Hj = G (hfj , hmj )
I Hj exogenously given for all j = 1, 2, ...J
28 / 63
ModelWages and Human Capital
I Wages:W is
j = θsw s(1 + ξsj eis)k isj
for s = f ,m, i = c, hs, for all j = 1, 2, ...J − 1
I �xed individual productivity θs , with gender speci�cdistribution
I gender speci�c baseline wage w f ≤ wm
I skill premium, ξsj , indicator for college e is
I human capital evolution (Imai and Keane, 2004)
F (l isj , kisj ) =
[(1− δ) + A
(l isj)α − Bl isj
]k isj
for δ, α ∈ [0, 1] , A,B > 0
I Endogenous gender wage gap and skill premium due to on thejob acquisition of human capital, even with w s and ξsjexogenous.
29 / 63
Optimal Household Allocation
I Pareto e�cient consumption and home hours allocation.
I Income e�ects:I Participation is zero if own productivity is su�ciently low and
non-labor income is su�ciently high.I Rise in non-labor income causes market hours to drop and
eventually participation to go to zero.I Non-labor income is the sum of partner's labor income and
savings.
I Gender wage gap causes wives' participation and market hoursto be lower than husbands on average, triggering a decline inboth, if the skill premium rises.
30 / 63
Calibration
I Household types (wife-husband):hs − hs, hs − coll , coll − hs, coll − coll
I Four periods: 25-39, 40-54, 55-69, 70+ yo
I Productivity distribution:
I θs ∼ logN(θ̄, σs) for s = f ,m,
I Strategy:
I Set most parameters based on independent evidence.I Set remaining parameters to match 1980 aggregate
participation, labor earnings dispersion by gender,home/market hours ratio.
31 / 63
CalibrationFunctional Forms
I Utility:
u(c isj , l
isj + hisj
)=
(c isj
)1−σ
1− σ − φs(l isj + hisj
)1+ 1γs
1 + 1γs
I σ, φs > 0, γs ∈ (0, 1) for s = f ,m
I Home production:
G (hf , hm) =[ψf(hf)ρ
+ ψm (hm)ρ]1/ρ
I ρ, ψs ∈ (0, 1) for s = f ,m,∑
s ψs = 1
32 / 63
Calibration
Table : Parameters calibrated based on independent evidence
Utility Human capital Home production
σ 1.1 α 0.22 ψf 0.5
γf , γm 3.82, 3.82 A 0.14 ρ 0.65
λf , λm 0.5, 0.5 δ 1− 0.35 {H2,H3} {1.018, 1.031}H1
β 0.97815 B 2.2160e − 04
Time evolution of ξ and w : estimated from data.
33 / 63
Calibration
Table : Parameters calibrated for 1980 to match population moments
Utility Home Production θ distribution
φft φmt H1 σm σf
0.169 0.142 2.74 0.48 0.35
Table : Data/Model Comparison
1980 Data Model
Women Men Women Men
Participation 0.44 0.86 0.44 0.86
Coe�. var. earnings distribution 1.34 0.84 1.34 0.84
Market/home hours 0.922 1.23
34 / 63
Experiment 1: Steady state rise in the skill premium
I Exercise:
I Increase ξsj to match actual 1995-2005 average skill premiumby gender, all other equal.
I Compare 1980 to 1995, as if steady states.
I Results:I large drop in labor supply of women with college husbands,
similar magnitude as in data;I large drop in labor supply of women in college-college
households, approx. 2/3 of drop in data;I rise in skill premium for men larger than for women, and rise in
male/female wage ratio higher for college than high school, asin data.
35 / 63
Experiment 1Response of women's labor supply
I Decline in participation in the model greatest for high schoolwomen with college husbands, as in data. Participation ofcollege women with high school husbands rises, contrary todata.
Married Women's Participation
Household Types HS-HS HS-C C-HS C-C
Model 1980 calibration: 1995-2005 skill premium-actual
percent change 0 −29.5 12.6 −11.3Data Average 1995-2005: actual- pre1995 projection
percent change −10 −17 −8 −17
36 / 63
Experiment 1Response of women's labor supply
I Higher skill premium reduces participation of women withcollege husbands. Participation lower both for high school andcollege women.
Married Women's Participation
HS Coll HS Husb Coll Husb
Model 1980 calibration: at 1995-2005 skill premium-actual
percent change −2.1 −1.7 1.3 −17.2
Data Average 1995-2005: actual-projected using pre-1995 growth
percent change −9.5 −12 −9 −14
37 / 63
Experiment 1Response of wages by gender
I Skill premium rises more for men in the model as in the data.Male/female wage ratio rises more for college than high schoolworkers, as in the data.
Skill Premium Male/Female Wages
Male Female HS Coll
Model 1980 calibration: 1995-2005 skill premium-actual
percent change 25 23.3 1.5 2.9
Data Average 1995-2005: actual- pre1995 projection
percent change 11.4 9.2 8.4 10.61
38 / 63
Experiment 2: Cohort simulations
I Simulate model every 5 years between 1965 and 2005, examinebehavior of cohorts who are 25-39 in each year throughouttheir lifetime.
I Stable decline in φf , capturing labor supply forces (matchesparticipation growth between 1965-1995, projected forward)
I w f /wm and wm as in data, cohort by cohort
I Skill premium:I 1965-1990: Stable process estimated from data.I 1995-2005: New process estimated from data, with higher
level and higher growth rate.I Unanticipated change in the skill premium =⇒ active
households reoptimize in 1995 based on new process.
39 / 63
Experiment 2Inputs
Skill Premium
1980 Level ξf1 0.572 ξm1 0.662
pre − 1995 Trend (5 year growth factor) F 1.036 M 1.054
1995 Level ξf1 0.74 ξm1 0.80
post − 1995 Trend (5 year growth factor) F 1.037 M 1.06
40 / 63
Experiment 2Baseline results
I Participation drops most for women with college husbands,especially those with high school. Participation rises for collegewomen with high school husbands, contrary to data.
Married Women's Participation, 1995-2005 Average
Household Types HS-HS HS-C C-HS C-C
Model actual − pre-1995 projected
percent change 0 −12 7 −4
Data actual − projected using pre-1995 growth
percent change −10 −17 −8 −17
Constant φf results
41 / 63
Experiment 2Baseline results
I Participation drops most for women with college husbands,and for high school women.
Married Women's Participation, 1995-2005 Average
Husband's education Wife's education
HS C HS C
Model actual - projected pre-1995
percent change 1 −14 −1.1 5
Data actual-projected pre-1995 growth
percent change −8.9 −14 −9.5 −12.1
Constant φf results
42 / 63
Experiment 2Baseline results
I Simulation captures small fraction of the gender divergence inthe skill premium and of the slowing convergence in wagesacross gender for skilled workers.
Skill Premium Male/Female Wages
1995-2005 Average Male Female HSch Coll
Model actual - projected pre-1995
percent change 14.6 16 1.0 2.6
Data actual - projected pre-1995 growth
percent change 11.4 9.2 8.4 10.61
Constant φf results
43 / 63
Summary
I Flattening participation of married women accounted by adecline in growth and level of participation of women marriedto college and high income husbands, suggesting role forincome e�ects.
I Quantitative analysis:
I Income e�ects can account for approx. 2/3 of the decline inparticipation of women with college husbands relative topre-1995 trend for 1995-2005.
I Model over predicts fall in participation of women in HS-Chouseholds, does not predict decline in participation of womenin HS-HS and HS-C households.
I Model partially captures relative rise in skill premium for menand in male/female wage ratio for college workers.
44 / 63
International Evidence
I Most OECD countries display a negative correlation betweenthe growth in inequality and women's participation.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0
10
20
30
40
50
60
70
80
90
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
Canada
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
0.5
1
1.5
2
2.5
3
3.5
0
10
20
30
40
50
60
70
80
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
Australia
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
0.5
1
1.5
2
2.5
3
3.5
4
0
10
20
30
40
50
60
70
80
90
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
Germany
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
0.5
1
1.5
2
2.5
3
3.5
4
0
10
20
30
40
50
60
70
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
Spain
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
0.5
1
1.5
2
2.5
3
3.5
4
0
10
20
30
40
50
60
70
80
90
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
France
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
0.5
1
1.5
2
2.5
3
3.5
0
20
40
60
80
100
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
Sweden
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
1
2
3
4
5
0
10
20
30
40
50
60
70
80
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
USA
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0
10
20
30
40
50
60
70
80
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
2008
2011
UK
Female -‐ Ages 25-‐54 (LFPR) Top 10%/Average
Figure : Participation and top 10%/average income ratio. Source: OECD
45 / 63
Ongoing work
I Income tax: incorporate 1993 tax reform.
I Household production: allow for adjustment in householdproduction margin (substitute with market goods).
I Explore trends in φ versus trends in household productiontechnology.
I Fertility choice.
46 / 63
Ongoing Work: Additional ChannelsFertility choice with quality/quantity trade-o�
I Low opportunity cost of additional children in high income one
earner households =⇒ quantity of children may rise inaddition to quality.
I Hypothesis is consistent with rise in fertility for college-collegehouseholds and high income husband households in themid-1990s.
47 / 63
Fertility
48 / 63
FertilityParticipation by husband's earnings & number of children
I Flattening of participation occurs for women with and withoutchildren, suggesting decline in participation not driven byhigher demand for children.
.2.4
.6.8
1975 1985 1995 2005 1975 1985 1995 2005
no children at least one child
10 percentile 50 percentile
90 percentile
fem
ale
LF
part
icip
ation r
ate
s
Year
Graphs by Wife’s own children in household
by husband’s earnings percentiles
Female participation rate
Source: March Supplement of CPS
49 / 63
Back Up Slides
Labor supply of married womenParticipation by household type
.3.4
.5.6
.7.8
1976 1984 1992 2000 2008Year
less than college, both h: college, w: less than college
never married, high school
Female Labor Force Participation Rate
.3.4
.5.6
.7.8
1976 1984 1992 2000 2008Year
h: less than college, w: college college, both
never married, college
Female Labor Force Participation Rate
Household type corresponds to husband-wife educational attainment.
Source: March Supplement of CPS
51 / 63
Labor supply of married womenHours by household/husband type
600
800
1000
1200
1400
1600
1976 1984 1992 2000 2008Year
less than college, both h: less than college, w: college
h: college, w: less than college college, both
Female Labor Market Hours
500
1000
1500
1976 1984 1992 2000 2008Year
husb. earn. in 10th perc husb. earn. in 50th perc
husb. earn. in 90th perc
by husband’s earnings percentile
Female Labor Market Hours
Household type corresponds to husband-wife educational attainment.
Source: March Supplement of CPS
52 / 63
Labor supply of married womenHours by education
800
1000
1200
1400
1600
1976 1984 1992 2000 2008Year
high school college
Married women
Female Labor Market Hours
Source: March Supplement of CPS
53 / 63
Intra-household inequality
.5.7
.91.1
1.3
avera
ge Y
_f/Y
_m
1976 1980 1984 1988 1992 1996 2000 2004 2008Year
Less than college wife College wife
Two−earner householdswhere husband has high school deg.
Wife to husband earnings ratio
.5.7
.91.1
1.3
avera
ge Y
_f/Y
_m
1976 1980 1984 1988 1992 1996 2000 2004 2008Year
Less than college wife College wife
Two−earner householdswhere husband has college deg.
Wife to husband earnings ratio
Source: March CPS
Note: the ratio of wife to husband's earnings was computed for household where both
partners worked for at least 45 weeks that year, and each earned at least $5,000
(constant 2005 USD)
54 / 63
Earnings dispersion by gender1.6
1.8
22.2
2.4
1975 1980 1985 1990 1995 2000 2005 2010Year
Males Females
full−time, full−year workers
Earnings dispersion, 90perc/50perc
1.6
1.8
22.2
2.4
1975 1980 1985 1990 1995 2000 2005 2010Year
Males Females
full−time, full−year workers
Earnings dispersion, 50perc/10perc
Source: March Supplement of CPS
55 / 63
Labor supply of men
I Participation of married men has remained stable. Female LFP
.55
.65
.75
.85
.95
1976 1984 1992 2000 2008Year
all men single (never married) men
married men
Male Labor Force Participation Rate
Source: March Supplement of CPS
56 / 63
Labor Supply of Married Women
I Participation has �attened only for prime age married women.Female LFP
.3.4
.5.6
.7Labor
Forc
e P
art
icip
ation R
ate
1976 1984 1992 2000 2008Year
21−31 y−o. 32−42 y−o
43−53 y−o 54−64 y−o
LFP of married women by age
Source: March Supplement of CPS
57 / 63
Labor supply of married women
Participation by husband's earnings Actual vs Projected
.2.3
.4.5
.6.7
.8.2
.3.4
.5.6
.7.8
1976 1984 1992 2000 2008
1976 1984 1992 2000 2008
10th percentile 50th percentile
90th percentile
LFP, actual LFP, predicted
Year
Graphs by Husband’s earn. percentile
Source: March Supplement of CPS
Dashed line corresponds to 1976-1991 participation probit, in and out of sample.
58 / 63
Labor supply of married women
Participation by husband's earnings and number of children LFP byhusband
.2.4
.6.8
1975 1985 1995 2005 1975 1985 1995 2005
no children at least one child
10 percentile 50 percentile
90 percentile
fem
ale
LF
pa
rtic
ipa
tio
n r
ate
s
Year
Graphs by Wife’s own children in household
by husband’s earnings percentiles
Female participation rate
Source: March Supplement of CPS
59 / 63
Negative Income E�ect
I Strong negative e�ect of husband's income on participation.Two earners
Coef. Robust Std. Err. [95% Conf. Interval]
age 0.005 0.0002 0.0047 0.0053
educ_years 0.108 0.0009 0.1059 0.1096
college -0.171 0.0108 -0.1927 -0.1501
perc25 -0.208 0.0048 -0.2174 -0.1982
perc50 -0.408 0.0086 -0.4252 -0.3913
perc75 -0.427 0.0063 -0.4398 -0.4148
perc90 -0.773 0.0056 -0.7844 -0.762
collegeXper25 -0.024 0.0117 -0.0466 -0.0006
collegeXper~50 0.0147 0.0211 -0.0267 0.0562
collegeXper75 -0.085 0.0126 -0.1099 -0.0602
collegeXper90 -0.107 0.0124 -0.1320 -0.0834
constant -1.028 0.0136 -1.0552 -1.001Source: Authors' calculations from CPS.
60 / 63
Experiment 2Constant φf results
Baseline Results
I Participation drops most for women married to collegehusbands, especially those with high school. But participationrises for college women with high school husbands, contrary todata.
Married Women's Participation, 1995-2005 Average
Household Types HS-HS HS-C C-HS C-C
Model actual skill premium − pre-1995 projected skill premium
percent change 0 −23.3 7.3 −4
Data actual − projected using pre-1995 growth
percent change −9.8 −17 −8 −17
61 / 63
Experiment 2Constant φf results
I Participation drops most for women with college husbands andfor women with college.
Married Women's Participation, 1995-2005 Average
Husband's education Wife's education
HS C HS C
Model actual skill premium − pre-1995 projected skill premium
percent change 1 −8 −1.6 0.5
Data actual-projected using pre-1995 growth
percent change −8.9 −14.0 −9.5 −12.1
62 / 63
Experiment 2Constant φf results
I Simulation captures the gender divergence in the skillpremium. It does not capture the slowing convergence inwages across gender for skilled workers, due to the strong risein participation and hours of college women with high schoolhusbands.
Skill Premium Male/Female Wages
1995-2005 Average Male Female HS Coll
Model actual skill premium − pre-1995 projected skill premium
percent change 14.5 15.9 −0.4 1
Data actual-projected using pre-1995 growth
percent change 11.4 9.2 8.4 10.6
Baseline Results
63 / 63