Effects of receipt of Social Security retirement benefits on older women’s employment

ABSTRACT

Labor force participation of women has declined since 1999; however, labor force participation of women 62+ has increased. The 2000–2006 waves of Health and Retirement Study (HRS) data, the initial years of the continuing upward trajectory, were used to test the effects of receipt of Social Security retirement benefits on older women’s employment. The models tested: (a) the effect of receipt of Social Security retirement benefits on whether employed; and (b) for women receiving Social Security retirement benefits, the effect of age elected receipt of benefits on whether employed. Both models included the effects of human capital characteristics and income sources. Receipt of Social Security benefits, pension income, and current age reduced the likelihood of employment; while educational level, good to excellent health, and nonmarried marital status increased the likelihood of employment. The older the woman was when she elected Social Security benefits, the more likely she was to be employed.

KEYWORDS:

• Employment
• retirement
• Social Security
• women

Notes

¹ Gross Domestic Product rose from $9,817 billion (B) in 2000 to over $13,194 B in 2006; personal consumption was $6,739 B in 2000 and rose to over $9,224 B in 2006; and residential construction rose from almost $447 B to approximately $765 B over the same period. Unemployment went from 4% in January 2000 to 4.4% by December 2006—with the peak at the mid-point of the period at 6.3% June 2003 (U.S. Bureau of Labor Statistics, 2011) as the result of a minor recessionary period during 2001 (Toossi, 2013).

² Two preliminary tests were conducted in Model 2: (a) interaction effects between current age and age of elected receipt of Social Security benefits, and (b) an endogeneity test of age of elected receipt of Social Security benefits. First, as a woman’s age increases, particularly for the age group 62–70, the likelihood of employment is expected to decrease. However, her employment may differ across these ages depending on when she initially elected receipt of Social Security benefits. To test this effect, an interaction term between those two variables was included (employed x age), but it was not statistically significant (p value = .593). Therefore, the interaction term was not included in Model 2. Second, there is potential endogeneity of age of elected receipt of Social Security benefits. One can argue a simultaneous decision between working or not and timing of receipt of Social Security benefits. To test for this effect, an augmented regression test (Durbin–Wu–Hausman test) for endogeneity was conducted. The first step was to use ordinary least squares (OLS) regression with the dependent variable age of elected receipt of Social Security benefits. Current age, age expected to elect receipt of Social Security benefits, and marital status were included as independent variables. Then a residual variable of age of elected receipt of Social Security retirement benefits was calculated. Next, logistic regression was used with employment (whether employed or not employed) as the dependent variable and the augmented residual variable of age of elected receipt of Social Security benefits, original variable of age of elected receipt of Social Security benefits, current age, education, health status, and marital status as independent variables. The coefficient of the augmented residual variable of age of elected receipt of Social Security benefits was not statistically significant (p value = .308), which indicates that the variable of age of elected receipt of Social Security benefits is not an endogenously determined variable.