OLS Estimation Results for Absolute Poverty, Relative Poverty, and Mortality
| Dependent Variable: | |||
|---|---|---|---|
| Absolute Poverty: Fraction of Families below Federal Poverty Line (1) | Relative Poverty: Fraction of Families with Total Income below 40% of Nonelderly Median Income (2) | Mortality Rate (3) | |
| Parameter estimates | |||
| Early entitlement age | −0.0071 (0.0035) | −0.0072 (0.0045) | 0.0010 (0.0003) |
| Estimated impacts | |||
| Dependent variable mean prior to availability of early claiming | 0.278 | 0.329 | 0.138 |
| Estimated impact of one-year reduction in the early entitlement age as a % of dependent variable mean prior to availability of early claiming | 2.5 | 2.2 | −0.7 |
| Estimated impact of weighted-average actual reduction in claiming age for 1916 cohort (1.35 years earlier) as a % of dependent variable prior to availability of early claiming | 3.4 | 2.9 | −1.0 |
| p-value for placebo test | 0.00001 | 0.00001 | 0.00001 |
Notes:N = 658 in Columns 1 and 2 (ages 65–90); N = 724 in Column 3 (ages 65–98).The first row of table shows the parameter estimate of φ in Equation 4, which is the impact of a one-year increase in the early entitlement age on the poverty or mortality measure. Each column represents a different regression based on a different measure of poverty or mortality. All regressions include the full set of age and year dummies, a quartic in simulated Social Security income for age-65 claiming to control for trends in benefit generosity across cohorts that were not due to early claiming, and additional controls for the percentage in age/year cell that are: veteran status, white;, high school graduate, some college;, college graduate, advanced degree, plus eligibility for Medicare and DI and quarters of coverage for age-72 benefits, respectively. Standard errors clustered by year of birth are shown in parentheses. The second row shows the dependent variable mean for cohorts not eligible for early claiming. The third row is (minus) the quotient of Rows 1 and 2 and represents the estimated impact of a one-year reduction in the claiming age as a percent of the dependent variable for the cohorts not exposed to early claiming. The fourth row is (minus) the quotient of Rows 1 and 2, multiplied by 1.4, and represents the estimated impact of the long-run actual reduction in the claiming age (as measured by the 1916 cohort) as a percent of the dependent variable for the cohorts not exposed to early claiming. The final row shows the p-value associated with the placebo test whereby the early entitlement age is randomly assigned. The null hypothesis is that φ in Equation 4 is zero. The p-value is based on 1,000 replications and is estimated using the estimator of Heβ (2017).