تاثیر طولانی مدت طلاق بر درآمد زنان در سراسر پنجره های متعدد طلاق: چشم انداز دوره زندگی

Abstract We investigate divorce as a turning point for women's earnings trajectories and whether those patterns have changed over recent decades. We use a unique data set that links retrospective family history data from the Survey of Income and Program Participation to longitudinal earnings records from the Social Security Administration. Using a panel fixed-effects model, we retrospectively compare women experiencing a marital dissolution across three divorce windows within a 25-year period (1970–1994) to continuously married women, from 3 years before to 10 years following separation. We find that women who divorced experienced long-lasting earnings increases, particularly among those who did not remarry. The increases in earnings levels are comparatively similar for those divorced at different times; however, percentage increases are significantly smaller for women experiencing divorce in more recent (1990–1994) relative to earlier (1970–1974) years. We also document lasting employment gains and higher earnings among employed women arising from divorce. In combination, the results extend our understanding of divorce as a salient turning point in women's work lives and call attention to possible changes in this process over recent decades.

.

Discussion and conclusions This analysis extends themes in the study of the life course, family, and labor market by assessing the long-term impact of divorce on U.S. women's earnings and employment paths, while also exploring whether those patterns have changed over the past decades. To date, most research asks how women's labor market behavior affects the risk of divorce (Özcan & Breen, 2012). Our study broadens the examination of this relationship to include the attendant effects of divorce itself on women's labor market outcomes. Using longitudinal earnings data matched to a national survey, we followed women whose separations from marriage led to divorce for 3 years before and as many as 10 years after the event, contrasting them with women who remain continuously married across three divorce windows. Our results account for possible differences in women who divorce and who remain continuously married by controlling for constant unobserved heterogeneity as well as observed characteristics. Our findings firmly establish that marital dissolution signals a turning point for women's earnings trajectories. In all three divorce windows, marital dissolution was associated with upward adjustments in real earnings, by $3400 to $1750 annually over 10 post-dissolution years, controlling for time-varying covariates that include earnings prior to the dissolution, work experience, additional education, and children. Such increases correspond to about an 11% to 32% increase relative to divorced women's mean earnings when they were married one year prior to the start of the divorce window. These gains are longer lasting than what most previous research shows. The findings also show that the long-run earnings impact of divorces occurring within a much more limited time period (the early 1970s) as reported by Couch et al. (2013) extend well beyond marital dissolutions occurring at that time. Conceptually, an increase in average earnings among all women can be decomposed into changes in employment and changes in earnings among the employed. Our analysis shows that both contribute to the long-run earnings gains associated with divorce in our sample. The results show that divorce increases the extent of women's employment, particularly in the earliest divorce window. The implication is that marital transitions meaningfully affect women's decisions to work. We also find long-lasting increases in the log earnings of employed women after divorce, from 20 to 28 percent annually, controlling for years of work experience among other characteristics. This is consistent with the notion that women receive a divorce earnings premium relative to those who remained married. This premium could be driven by several factors. First, divorce may promote greater labor supply (e.g., part-time to full-time work), resulting in higher log earnings among employed divorced women. It is also possible that higher earnings are a consequence of increased productivity or different occupational choices made by divorced women relative to continuously married women. While not the focus of our study, it is notable that remarriage emerged as an additional turning point redirecting women's earnings back toward their level prior to the dissolution of their first marriage. To the extent that economic necessity underlies increases in women's earnings after divorce, a modifying effect of remarriage on the divorce-earnings relationship may be expected. This is consistent with previous studies showing that remarriage reduces women's financial needs (Dewilde and Uunk, 2008 and Jansen et al., 2009). An additional goal of our analysis was to examine changes in women's long-run earnings response to divorce over time. With respect to all women in our sample (i.e., Table 3), the positive effects in absolute real dollar terms declined across the windows, but did not differ in statistically significant ways. However, in relative terms (i.e., Table 4), we find a significant reduction in the proportional earnings gains associated with divorce from the first to last window (from 32 to 11 percentage points per annum). Thus, while divorce still represents a turning point in women's earnings trajectories, its relative effect appears to have waned in more recent years. This temporal pattern reflects, in large part, the increase in married women's labor force participation over the study period. By the last divorce window, fewer married women would be labor market entrants or part-time employees who could dramatically ramp up their work effort following divorce. Reinforcing this idea, we find significant declines in the long-run positive effect of divorce on women's probability to be employed over the windows (Table 5). These results are in line with recent work by Tach and Eads (2015), which find smaller drops in women's household income following divorce since the 1980s, due partially to the rise in married women's earnings as a share of pre-dissolution household income. More broadly, such findings are consistent with the life-course tenet that historical context shapes the trajectories of lives following important life events. Further, the supplementary analysis examining educational differences reveals significant long-run increases in women's real earnings following divorce for both higher- and less-educated women across the divorce windows. Higher-educated divorced women experience larger absolute increases, particularly in the most recent divorce window, but the proportional changes are relatively evenly spread across the groups within each window (accounting for earnings growth over the 10-year follow-up period). We also observe a decline in the long-term proportional increase in earnings following divorces among higher- and less-educated women. The decline appears sharper among less-educated women in the latest divorce window (1990–94), in which relative earnings gains are mainly observed in terms of earnings growth over the 10-year follow-up period. While speculative, these findings suggest that among women with relatively low personal earnings, the possibility of expanding work effort to meet the loss of resources that typically accompanies divorce may have diminished over time. This likely is driven by the growth of married women's earnings over past decades, particularly among less-educated married mothers. In combination, our results are salient for understanding the role of the family in shaping labor market disparities among women over the life course. In prior investigations concerned with the intersection of family and gender on women's labor market outcomes, much attention has focused on the wage effects of motherhood and marriage. While our study does not directly speak to that issue, in large part because we do not examine hourly wages, it does add to this literature by offering evidence that the socially common transition of divorce is an additional family-based source of divergence in women's earnings that is long-lasting. The results also are relevant to women's economic well-being. Pronounced increases in earnings following marital dissolution may help buffer women and their dependents from the economic hardship that accompanies divorce. In addition, persistent higher earnings can lead to higher retirement savings and Social Security benefit levels. With that said, earnings gains are not synonymous with offsetting declines in economic well-being following a divorce transition. Divorced women remain a materially disadvantaged group. For example, recent cross-sectional estimates for 2013 show that divorced U.S. women aged 25–59 had a poverty rate of 22 percent compared to 7 percent among married women (authors’ calculations from the 2014 March Current Population Survey). This suggests that there are many women who cannot offset the income declines following divorce through labor market mechanisms. The research presented here suggests this problem is likely to be particularly severe among women with relatively low levels of education. The study is suggestive of additional analyses to be done. Due to data constraints, we could not examine the long-term effects of divorce on work hours, wage rates, or per-capita family income, the latter being a key measure of well-being. Because we use annual earnings as our main measure, our results should not be interpreted as directly comparable to literature on the wage effects of other family status changes, which use hourly wages (e.g., Killewald & Gough, 2013). Furthermore, more systematic analysis of educational differences in the divorce-earnings relationship among women, and how this interacts with subsequent family well-being is clearly warranted. Additionally, life events such as divorce would be expected to influence health outcomes, which in turn, may mediate its labor market impact. Examination of the association between marital history and health is already a topic of much interest in the sociological literature (Hughes and Waite, 2009 and Waite and Gallagher, 2000). Our perspective in bringing methodologies that control for sample selection to those literatures may provide further advancement. The study has several noteworthy limitations. While our models control for constant unobserved heterogeneity, some omitted variables may not be time invariant. For example, data constraints did not allow us to account for cohabitation. The availability of potential partners in marriage markets may also be important. For example, to the extent that women in weak marriage markets partner with men with lower earnings prospects (Harknett & McLanahan, 2004), it could affect both the likelihood of divorce as well as women's work patterns during marriage. We also could not account for non-employment income, such as alimony, child support, or public transfers, although a declining share of divorced women report such income (McKeever & Wolfinger, 2005, Table 6-2). In conclusion, this study provides strong evidence that divorce is a turning point for women's earnings trajectories, particularly for those who do not remarry. The relative effect of divorce on women's long-run earnings, however, has attenuated relative to past periods. Examining how other discrete family events affect the trajectories of individuals and families and how they are shaped by historical context is a promising area for further research. Research examining differences in the association between divorce and women's long-term earnings by age of dissolution would be valuable. Also, additional efforts to develop better longitudinal assessments of the long-run impact of family change events should be undertaken. Linking national survey data with longitudinal administrative records, as done here, represents one promising approach, among others, to advance understanding of these issues.