«Chinhui Juhn University of Houston and NBER cjuhn Kristin McCue U.S. Census Bureau kristin.mccue September 2011 Abstract: Using ...»
Marriage, Employment and Inequality of Women’s Lifetime Earned Income
University of Houston and NBER
U.S. Census Bureau
Abstract: Using Current Population Surveys and Survey of Income and Program Participation
data matched to Social Security earnings records, we summarize changes in marital histories for
different birth cohorts of women and project the associated changes in women’s own lifetime
earnings and in spousal earnings. We find that the gap in lifetime earnings between married and single women appears to have essentially closed for less educated women while a small differential still exists for more educated women. The earnings gap across education categories has increased rapidly in terms of women’s own lifetime earnings. The level of earnings inequality across education categories is higher when marriage and husbands’ earnings is taken into account although the increase is less pronounced. Lifetime earnings of married collegeeducated women diverged most dramatically from those of less educated single women.
JEL Classification: J12, J16, J22, J31 Any opinions and conclusions expressed herein are those of the authors and do not necessarily represent the views of the U.S. Census Bureau. All results have been reviewed to ensure that no confidential information is disclosed. This research was supported by the U.S. Social Security Administration through grant #5 RRC08098400-03-00 to the National Bureau of Economic Research as part of the SSA Retirement Research Consortium. The findings and conclusions expressed are solely those of the authors and do not represent the views of SSA, any agency of the Federal Government, or the NBER.
I. Introduction Over the past four decades, changes in both marriage and women’s employment have reduced the economic importance of traditional families in which the husband works predominantly outside of the home and the wife works and provides child care in the home. For example, among women who are 18-60 years old, the fraction who are currently living with a spouse fell from 74 percent in 1968 to 54 percent in 2010. At age 35, 6 percent of women reported never having married in 1968 while up to 19 percent had never married in 2010.1 Over their lifetimes, both men and women in recent birth cohorts will spend significantly fewer years of their prime aged adult life in marriage, both because they are less likely to ever marry and because they marry at a later age (Stevenson and Wolfers (2007)). Moreover, the patterns have not been uniform across education and skill groups with marriage rates falling more among less educated groups (Isen and Stevenson (2010), Juhn and McCue (2010)).
While marriage rates have fallen, it is well known that employment rates have increased for married women and the fraction of households with dual-earners have increased. In addition, wives of high wage husbands have increased their employment relative to wives of low wage husbands (Juhn and Murphy (1997), Blau and Kahn (2007)). In this paper, we propose to examine the impact of changing marital and employment histories on the distribution of lifetime earnings (both individual and family) for women. We are particularly interested in inequality and therefore compare lifetime earnings of less and more educated women who have had different experiences in both marriage and employment. We also want to know how this inequality across education groups has progressed over time. We therefore compare across birth cohorts beginning with those born 1936-40 and ending with those born 1966-70.
Based on authors’ calculations using the March Current Population Surveys.
The novelty of our exercise largely comes from using the Survey of Income and Program Participation (SIPP) panels matched to Social Security Administration earnings records from 1951-2006. The data allow us to follow an individual woman’s earnings over much of her lifetime and also to observe her spouse’s earnings for the subset of women who are married during their SIPP panel.2 Related work in this area (Gustman and Steinmeier (2001), (2011)) use the Health and Retirement Survey (HRS) matched to Social Security earnings records. In comparison to these studies, we have a wider range of cohorts to compare.
The question we propose to examine is particularly relevant for considering postretirement income security of women since under current rules with spousal benefits, social security benefits depend on women’s marital histories as well as their individual earnings histories. Since employment and earnings increased more among married women than single women, it may be the case that current benefit rules provide limited income security for women who never marry and have low lifetime earnings. At the same time, current rules may be providing redundant income protection for women who have high lifetime earnings of their own.
As pointed out by Liebman (2002), the current system redistributes income based on criteria other than lifetime earnings and has the unintended consequence of redistributing income from two-earner married households to one-earner married households even though they may be equal in terms of potential lifetime earnings.
The paper is organized as follows. Section 2 briefly reviews related papers. Section 3 describes the changes in marriage and employment patterns for women in different education categories. Section 4 reports the main results on lifetime earnings (individual and including We know the entire marital history of a women up to the year of the SIPP survey and characteristics and earnings history of the current spouse although not the characteristics or earnings of previous spouses. For divorced women, we impute earnings of previous husbands using earnings of current husbands (for those women who remarried) and other characteristics as described in the appendix.
spouses’ earnings). Section 5 concludes. Description of sample selection and details regarding various imputation procedures are in the appendix.
II. Related Literature Our paper is closely related to the literature on family earnings inequality (Cancian, Danziger, and Gottschalk (1993), Juhn and Murphy (1997), Blundell, Pistaferri, and Preston (2008)). These papers generally find that the inclusion of wives’ earnings in the consideration of family earnings, as opposed to individual male earnings, reduce the rise in earnings inequality observed since the 1980s. In contrast to these studies we use education levels to group women, as opposed to husbands’ or couples’ earnings, as a better measure of earnings potential. In addition, many of the previous studies on family earnings inequality focused exclusively on continuously married couples whereas we explicitly take into account differences in marital status across women to calculate inequality of lifetime earnings (both their own and including spouses’ earnings).
A related literature has examined the impact of changing labor force participation of women on redistribution under the current Social Security tax and benefit schedules. Gustman and Steinmeier (2001) use the Health and Retirement Survey matched to Social Security earnings data to compare lifetime earnings, social security taxes paid, and benefits received and accrued for individuals and for couples. They find that there is a significant amount of redistribution across individuals in that the lowest earning individuals have the highest benefits to tax ratio and this ratio declines with lifetime earnings. Much of this redistribution occurs within households, however, from high earning husbands to their non-working wives in the form of spousal benefits. When households are the units of observation, they find significantly less redistribution from high earnings to low-earnings households.
Gustman and Steinmeier (2011) update their earlier work by including a younger cohort from the HRS. Somewhat surprisingly, they find that despite significantly higher levels of labor force attachment of wives, not much changed across cohorts and redistribution across households is limited. Liebman (2002) reaches a similar conclusion using a different dataset, the 1991 Survey of Income and Program Participation matched to Social Security earnings data. He finds income-related transfers within cohorts account for only a small percent of total benefits paid by Social Security. Due to the availability of spousal benefits, a large fraction of this redistribution occurs across demographic categories from single households and two-earner married households to single-earner married households.
Coronado, Fullerton and Glass (2000) simulate taxes and benefits paid under various counter-factual scenarios using the Panel Survey of Income Dynamics (PSID). Assigning potential earnings as opposed to actual earnings to labor market non-participants reduces the amount of redistribution. In a similar vein, if wives are assigned one-half of combined spousal earnings, they appear to be significantly less “poor” and the amount of redistribution in the system is reduced. While we do not directly address the redistributive effects of spousal benefits, the patterns we document are important in considering the effects of existing or alternative benefit structures. Relative to these studies, we bring new data to bear that allow us to compare across multiple cohorts.
III. Trends in Marriage Rates and Earnings We first present an overview of changes in marriage rates among women in different education categories. In a previous paper, Juhn and McCue (2010), we documented a reversal in the pattern of marriage across women in different education groups. While the most educated women were the least likely to ever marry in the earlier birth cohorts, the least educated women are now the least likely to ever marry among more recent birth cohorts. This pattern is confirmed in Tables 1 and 2. Table 1 uses 1968-2010 March Current Population Surveys (CPS).
The sample consists of women who were born in 1936-1975.
Table 1 reports the share of women who have ever married by age 35. The table shows that marriage rates fell for women in all education categories. However, the decline was particularly large among the least educated, those with a high school education or less. The share of women who were ever married by age 35 fell from.953 among the 1936-1940 cohort to.795 among the 1971-1975 cohort. Among college graduates, the share fell by a smaller amount from.917 to.836.
Table 2 reports the share who are currently married at age 35. This table examines not only the entry into marriage but also the likelihood of divorce by focusing on current marital status. This alternative marriage variable tells a similar story with marriage rates falling most rapidly for the least educated women. While divorce rates have increased among all women, the increase has been no larger among educated women.
Table 3 presents the cumulative years since age 18 spent in marriage for different cohorts of women up to the age specified. The table also reports marriage statistics separately for our three education categories. In the top panel, which refers to women with a high school degree or less, we find that at age 50 the fraction of years spent in marriage fell from.712 for the 1946cohort to.560 for the 1961-1965 cohort. The bottom panel examines college-educated women. At age 50, the share of potential years spent in marriage fell from.714 for the 1946birth cohort to.665 for the 1961-1965. The table demonstrates that the decline in time spent in marriage during prime-aged years declined for all women but much more significantly for less educated women.
We next turn to the changes in women’s earnings by marital status. Our sample again comes from the March CPS. We focus on women aged 30 to 60 years old to minimize the age difference between married and never married women. We include both wage and salary earnings and self-employment earnings and also include non-participants with zero earnings.
Table 4 reports average earnings by marital status for select years starting with 1970.3 Table 4 shows that in 1970 never-married women earned significantly more than women in all other categories, especially married women. In 2010, however, married women actually had similar or even slightly higher earnings than never-married women.
Table 5 shows earnings by marital status for women in different education categories.