estimate the effect of the birth of the first child on wages and hours worked for parents. We focus on the joint response of household members to the appearance
of the child and allow responses to vary for households in which the wife is a continuous participant in the labor force and those in which her participation is
intermittent.
3. Data
The data used for the analysis are from the PSID. Individual level data on husbands and wives are matched so that each of the AunitsB in our panel is a
marriage that lasted at least 5 years within our sample window of 1980 to 1992. The unbalanced panel data set includes data on each couple for all years in which
the marriage was intact, or up to 13 years. We limit our sample to couples in which the husband and wife are between ages 22 and 45, inclusive, and exclude
couples for whom the wife’s first child was born outside the marriage. The final data set consists of 17 744 observations on 2072 couples.
8
The dependent variables in the analysis are the log of the real hourly wage rate Ž
.
9
on the main job in 1985 US dollars and total hours worked during the year. We have used the pay rate available in the PSID for hourly and salary workers only,
rather than the wage rate calculated as labor earnings divided by hours worked. This restricts the sample size in the wage equations, but avoids the measurement
error inherent in the latter measure. Dummy variables for years of age, education, region, and race are included in all equations. Table 1 shows that, on average,
married women without children work more in the market and earn higher wages than women without children, while married men with children work more hours
than men without children.
10
The key independent variables in the base specification are dummy variables indicating years before and after the birth of the couple’s first child. In a
non-linear specification, these dummy variables are broken down into a set of 41 variables, which indicate the number of years before and after the birth. These
measures are based on fertility histories undertaken by the PSID in 1985 and updated in each subsequent year.
8
After excluding observations with missing values, 15 106 observations for husbands and 11 134 observations on wives are used in estimating the wage equations, and 17 403 observations for husbands
and 17 334 observations for wives are used in the hours equations.
9
Wage data are for the current job and are available through 1992. Annual hours data are retrospective and are available for up to 1991.
10
In the PSID, a single Aprimary adultB who is usually the male Ahead,B but is sometimes the Ž
. Awife,B PSID terminology answers all questions about the family. Therefore, the husband is, in
general, reporting the wages of both husband and wife.
Table 1 Ž
. PSID married couples 1980–1992 means, standard deviations, and sample statistics
Variable Full sample
Sample with Sample without
Sample wives Sample wives
w Ž
.x mean s.d.
children children
not continous continous
w Ž
.x w
Ž .x
mean s.d. mean s.d.
participants participants
w Ž
.x w
Ž .x
mean s.d. mean s.d.
Log real wage 2.35
2.34 2.38
2.39 2.30
Ž .
Ž .
Ž .
Ž .
Ž .
rate — husband 0.50
0.50 0.51
0.54 0.46
w x
w x
w x
w x
w x
w x
no. obs. missing 3597
3233 364
1839 1758
Log real wage 1.93
1.91 2.07
1.74 2.00
Ž .
Ž .
Ž .
Ž .
Ž .
rate — wife 0.51
0.51 0.50
0.56 0.47
w x
w x
w x
w x
w x
w x
no. obs. missing 7242
6764 478
5766 1476
Hours worked — 2198
2203 2155
2217 2179
Ž .
Ž .
Ž .
Ž .
Ž .
husband 705
708 672
750 657
Hours worked — 1152
1100 1613
633 1655
Ž .
Ž .
Ž .
Ž .
Ž .
wife 888
880 828
815 630
Age — husband 34
34 34
34 34
Ž .
Ž .
Ž .
Ž .
Ž .
5.7 5.7
5.8 5.7
5.7 Age — wife
32 32
31 31
32 Ž
. Ž
. Ž
. Ž
. Ž
. 5.6
5.6 5.5
5.6 5.5
Years of education 13
13 14
13 13
Ž .
Ž .
Ž .
Ž .
Ž .
— husband 2.3
2.3 2.3
2.4 2.2
Years of education 13
13 14
13 14
Ž .
Ž .
Ž .
Ž .
Ž .
— wife 2.1
2.1 2.1
2.1 2.1
Race — husband 78
78 79
81 76
Ž .
Ž .
Ž .
Ž .
Ž .
Ž .
White 0.41
0.41 0.41
0.39 0.43
Race — wife 78
78 78
81 75
Ž .
Ž .
Ž .
Ž .
Ž .
Ž .
White 0.41
0.41 0.41
0.39 0.43
No. observations — 17 403
15 610 1793
8563 8840
husband No. observations —
17 334 15 556
1778 8520
8814 wife
No. husbands 2066
1810 256
977 1089
No. wives 2065
1809 256
976 1089
Other independent variables include years of education and age, which are entered as a series of dummy variables in order to capture non-linearities in the
Ž effects. Dummy variables for the year of the first birth, region Northeast, North
. Ž
. Central, and South vs. West and race white vs. non-white are also included.
Observations are divided into two subsamples based on the wife’s labor supply behavior. The Acontinuous participantB sample consists of husbands and wives in
households in which the wife participates in each year of the window, other than a year in which she gave birth. The husbands and wives in households in which the
wife does not participate continuously are in the Anon-continuousB subsample. In some analyses, we pool the two samples, and interact a dummy variable for
AcontinuousB with key explanatory variables. Table 1 reports the mean character- istics of these subsamples, which are approximately equal in size.
4. Estimation and simulation of age–wage and age–hours profiles