Widows Waiting to Wed? (Re)Marriage and Economic Incentives in Social Security Widow Benefits

by Michael J. Brien, Stacy Dickert-Conlin, and David A. Weaver
ORES Working Paper No. 89 (released January 2001)

Text description for Figure 1.
Widows' Marriage Rates By Age Category

Panel A descriptors: Y-axis = Average marriage rate; X-axis = Year (1968–1995).

The legend consists of two categories: (1) 50–59 year olds, and (2) 60–70 year olds.

The two vertical lines show the year in which the current remarriage rule passed into law (1977) and became effective (1979).

Panel A (descriptive analysis) examines average marriage rates by selected years 1968–1995. For widows aged 60–70, there is a decrease in the remarriage rate after the passage of the law in 1977, suggesting widows aged 60 or older delayed marrying in 1978 in anticipation of the law becoming effective in 1979. There is an increase in the marriage rate of those widows when the law became effective in 1979, which implies that a good number of women would have preferred to marry, but did not pursue matrimony at that time because of the penalty in the Social Security system.

For widows age 50–59, marriage rates decline around the law change, yielding the expected response to that change. Note that other factors, such as the overall decline in marriage rates during this period, are not accounted for.

Panel B descriptors: Y-axis = Percentage difference from 1968; X-axis = Year (1969–1995).

The legend consists of two categories: (1) 50–59 year olds, and (2) 60–70 year olds.

The vertical lines show the year in which the current remarriage rule passed into law (1977) and became effective (1979).

Panel B (regression results) gives the regression results of interest. In all cases, the marriage rates for widows under age 60 is further below the baseline year than the rate for widows who are at least at 60. For widows under age 60, the law did not have significant effects.

For widows over age 60, the data show that in 1977, their marrying rate was 2 percent below the baseline year, and in 1978 (the year the law was passed) the marriage rate was 13 percent lower than the baseline year. Those rates are statistically different from one another at the 6 percent level, supporting the possibility that widows over age 60 delayed marriage until their marriage penalty was eliminated. In 1979 (the year the law became effective) the marriage rate was 20 percent higher at the baseline—the largest positive deviation in the sample period. (Full regression results are available in Appendix Table 1.)

Notes: The numerator is the weighted number of marriages in the age category and year from states that were in the marriage-registration areas for all years between 1968 and 1995 and that reported previous marital status on their marriage certificates. The denominator is the number of widows in the age category and year in all states. (See Appendix Table 1 for details on coefficients and standard errors.)

Sources: Author's calculations from the 1968–1995 Vital Statistics and Current Population Survey.

Text description for Figure 2.
Divorced women's marriage rates, by age category

Panel A descriptors: Y-axis = Average marriage rate; X-axis = Year 1968–1995).

The legend consists of two categories: (1) 50–59 year olds, and (2) 60–70 year olds.

The two vertical lines show the year in which the current remarriage rule passed into law (1977) and became effective (1979).

Panel A (descriptive statistics) shows that the marriage rate patterns for divorced women do not exhibit the trends shown for widows. This panel shows slight increases in the marriage rates of divorced women aged 60–70 in 1979. Again, the marital trends for divorced women aged 60–70 are not as striking as those for widows.

Among divorced women at least age 60, the marriage rates in 1977, 1978, and 1979 are 20, 24, and 10 percent below the marriage rate for the baseline year of 1968. The rates are not statistically different from one another at standard levels.

Panel B descriptors: Y-axis = Percentage difference from 1968; X-axis = Year (1969–1995).

The legend consists of two categories: (1) 50–59 year olds, and (2) 60–70 year olds.

The vertical lines show the year in which the current remarriage rule passed into law (1977) and became effective (1979).

Panel B (regression results) shows that the deviations from the baseline year (1969) are significantly different for divorced women who are at least age 60 relative to women under age 60 in 1970, 1971, 1975, 1976, 1980, and 1981. Unlike the trend for widows, there is no clear pattern surrounding the 1979 law change. (Full regression results are available in Appendix Table 1.)

Among divorced women under age 60, the marriage rates decline significantly between 1979 (29 percent below the baseline) and 1980 (47 percent below the baseline), hinting at the possibility that these divorced women were substitutes for the widows who were now free to marry without penalty.

Notes: The numerator is the weighted number of marriages in the age category and year from states that were in the marriage-registration areas for all years between 1968 and 1995 and that reported previous marital status on their marriage certificates. The denominator is the number of widows in the age category and year in all states. (See Appendix Table 1 for details on coefficients and standard errors.)

Sources: Author's calculations from the 1968–1995 Vital Statistics and Current Population Survey.

Text description for Figure 3.
Widow's marriage rates by age and year category

Panel A descriptors: Y-axis = Average marriage rate; X-axis = Age (50–70).

The legend consists of five categories: (1) 1968–1973, (2) 1974–1978, (3) 1979–1983, (4) 1984–1990, and (5) 1991–1995.

Panel A (descriptive statistics) examines single-age marriage rates for various birth cohorts of women, using the 1979 law change as the point of reference. The chart shows that all of the time periods after the 1979 law change (1979–1995) reveal a large relative decline in marriage rates at age 59 and a relative increase in marriage rates at age 60 for widowed women. This trend did not exist before 1979, suggesting that widows may postpone or avoid marriage if they are very close to being able to marry without a penalty on the Social Security widow benefits.

The chart shows that in the periods before and after the law change, the trends in marriage rates relative to the baseline age of 50 years are statistically the same for all age groups up to and including 58-year-old widows. In the post-1979 law period, 59 year-old widows are less likely to marry (101 percent below 50 year olds) than in the pre-1979 law period (86 percent below 50 year olds). Conditional on the time trend, all age categories at age 60 or older are more likely to marry in the post-1979 law period. These differences are statistically significant and show that the law decreased marriage rates of 59 year olds and increased the marriage rate of women who were at least age 60.

Panel B descriptors: Y-axis = Percentage difference from age 50; X-axis = Age (51–70).

The legend consists of two categories: (1) 1968–1978, and (2) 1979–1995.

Panel B (regression results) show that prior to 1979, there is some incentive to wait until age 60 to marry because of the difference between keeping 50 percent of widow benefits for a marriage after age 60 versus zero widow benefits for marriage before age 60. The marriage rate at age 59 is 86 percent below the baseline age of 50, while the rate at age 58 is only 68 percent below the baseline age. This difference is significant at the 1 percent level, whereas unlike the post-law period, the age 59 effect is not statistically different from the age-60 effect (84 percent below the baseline). (Full regression results are available in Appendix Table 2.)

Notes: The numerator is the weighted number of marriages among widows in the age category and year from states that were in the marriage-registration areas for all years between 1968 and 1995 and that reported previous marital status on their marriage certificates. The denominator is the number of widows in the age category and year in all states. (See Appendix Table 2 for details on coefficients and standard errors.)

Sources: Author's calculations from the 1968–1995 Vital Statistics and Current Population Survey.

Text description for Figure 4.
Divorced women's marriage rates by age and year category

Panel A descriptors: Y-axis = Average marriage rate; X-axis = Age (50–70).

The legend consists of five categories: (1) 1968–1973, (2) 1974–1978, (3) 1979–1983, (4) 1984–1990, and (5) 1991–1995.

Panel A (descriptive statistics) shows that there are no similar patterns for divorced women as those shown for widows (Figure 3). The analysis conducted in Figure 3 (widows) is repeated in Figure 4 (divorced women), showing no statistical or economic differences between the pre- and post-law periods or between the effects of the age 58-, 59-, or 60-year-old dummies. Marriage rates decrease monotonically around age 60 in the post-1979 law change period.

Panel B descriptors: Y-axis = Percentage difference from age 50; X-axis = Age (51–70).

The legend consists of two categories: (1) 1968–1978, and (2) 1979–1995.

Panel B (regression results) gives the regression results of interest; full regression results are available in Appendix Table 2.

Notes: The numerator is the weighted number of marriages among divorced women in the age category and year from states that were in the marriage-registration areas for all years between 1968 and 1995 and that reported previous marital status on their marriage certificates. The denominator is the number of divorced women in the age category and year in all states. (See Appendix Table 2 for details on coefficients and standard errors.)

Sources: Author's calculations from the 1968–1995 Vital Statistics and Current Population Survey.

Text description for Figure 5.
Widow's marriage counts around 60th birthday

Panel A descriptors: Y-axis = Average monthly count; X-axis = Month around 60th birthday (-24 through +24).

The legend consists of two categories: (1) 1968–1978, and (2) 1979–1995.

Panel A (descriptive statistics) shows that during the 1979–1995 period, the peak number of marriages for widows within the 24-month range is the woman's 60th birthday, with especially high counts in the 3 months following a woman's 60th birthday. A sharp decline in the few months prior to the 60th birthday precedes that peak. This pattern is similar to, but not as pronounced in the years preceding the 1979 law change.

Panel B descriptors: Y-axis = Percentage difference from -24 months from 60th birthday; X-axis = Months from 60th birthday (-23 through +23).

The legend consists of two categories: (1) 1968–1978, and (2) 1979–1995.

Panel B (regression results) shows that there are 10 months after age 59 (60th birthday, +1, +2, +3, +8, +12, +18, +22, +23, +24 months) where the marriage counts are statistically and economically higher in the post-1979 law period relative to the pre-1979 law period. The trend around the 60th birthday is dramatic. The counts in all months preceding the 60th birthday are the same or lower than the baseline month (24 months before the 60th birthday). One month before the 60th birthday, the marriage count is 72 percent lower than the baseline month (this is the largest deviation from the baseline). On the other hand, the number of marriages on the 60th birthday is 56 percent higher than the baseline month. These counts are statistically different from one another at the 1 percent level. One, 2, and 3 months following the 60th birthday, the number of marriages is still 52, 36, and 25 percent higher than the baseline month, suggesting that the marriage penalty in Social Security is most influential on women who are very close to age 60. (Full regression results are available in Appendix Table 3.)

Notes: The numerator is the weighted number of marriages among widows in the month and year category. These are only from states that were in the marriage-registration areas for all years between 1968 and 1995 and that reported previous marital status on their marriage certificates. (See Appendix Table 3 for statistical significance of individual coefficients.)

Source: Author's calculations from the 1968–1995 Vital Statistics.

Text description for Figure 6.
Divorced women's marriage counts around 60th birthday

Panel A descriptors: Y-axis = Average monthly count; X-axis = Month around 60th birthday (-24 through +24).

The legend consists of two categories: (1) 1968–1978, and (2) 1979–1995.

Panel A (descriptive statistics) shows that the pattern describes in Figure 5 (widows) is not the same for divorced women. In the 1979–1995 period, there is a flat trend in marriage counts in the months preceding the 60th birthday and a much less pronounced increase at the 60th birthday. There appears to be no trend for divorced women's marriage counts in the period prior to the law change in 1979.

Panel B descriptors: Y axis = Percentage difference from -24 months from 60th birthday; X axis = Months from 60th birthday (-23 through +23).

The legend consists of two categories: (1) 1968–1978, and (2) 1979–1995.

Panel B (regression results) confirm the finding from panel A. Unlike counts for widows, the marriage counts for divorced women on their 60th birthday are below the baseline month before and after 1979 law change. (Full regression results are available in Appendix Table 3.)

Notes: The numerator is the weighted number of marriages among divorced women in the month and year category. These are only from states that were in the marriage-registration areas for all years between 1968 and 1995 and that reported previous marital status on their marriage certificates. (See Appendix Table 3 for statistical significance of individual coefficients.)

Source: Author's calculations from the 1968–1995 Vital Statistics.