Twice chosen: Spouse matching and earnings among women in first and second unions

Abstract

This study examines spouse matching for females in second order marriages or cohabitating relationships. It is based on detailed data from longitudinal Swedish population registers. We follow women who marry, divorce, and subsequently remarry compared to women who marry and remain married over the course of the study interval. The earnings of both groups, along with the earnings of their respective husbands, are modeled on the basis of data in the year prior to their marriages. From the regressions we obtain spouse-to-be pairs of earnings residuals and measure the correlation between spouses within each marital regime. Overall, we find significant positive residual correlations for both sample partitions. The correlation is smaller for the first of two marriages for women who divorce than for women who marry and remain married. For the second of successive marriages, however, the correlation is larger than that for women who marry but once. Also for twice-married women, we find evidence of matching between successive husbands. Women who marry men with unmeasured positive earnings capacities, in the event of divorce, tend to select and match in a similar fashion the second time around.

Keywords:

• Marriage
• Assortative mating
• Divorce

Acknowledgments

We are grateful for comments from Magnus Wikström, Jörgen Hellström, and participants at a seminar at Umeå University. Åström acknowledges financial support from The Graduate Program of Population Dynamics, Umeå University. Westerlund acknowledges financial support from The Swedish Council for Working Life and Social Research (FAS)and the Aging and Living Conditions Program at the Center for Population Studies, Umeå University.

Notes

1 Statistical Abstract of the United States, 2012: Table 1336.

2 As noted in Section 4, in estimating the earnings regressions, subsamples based on minimum earnings thresholds at SEK 50,000 and SEK 100,000 are formed on the basis of individual restrictions. Table 4 likewise presents estimates for the entire sample as well as for subsamples based on the earnings thresholds. Since Table 4 necessarily requires residual pairs in order to estimate residual correlations, we base the restrictions on both spouses satisfying the respective minimum thresholds. As the table indicates, these exclusions affect the magnitudes of the estimated correlations but not the essential conclusions of the study.

3 The marriage year indicators in Table 6 extend only through 2001, not 2003 as in Table 2. Recalling that all data are measured prior to marriage, 2001 is the final year for which a woman can provide complete data if she is to marry, divorce and remarry.

4 See, for example, Zimmer (2004) and references therein.