Absence of conflict of interest.
Citation
Haider, S. J., & Loughran, D. S. (2008). The effect of the social security earnings test on male labor supply: New evidence from survey and administrative data. Journal of Human Resources, 43(1), 57-87.
Highlights
- The study’s objective is to examine the impact of 1983 and 2000 policy changes that removed the Social Security earnings test for certain age groups on the earnings and employment outcomes of affected men
- The study is a nonexperimental analysis that examines the correlations between the 1983 and 2000 policy changes and earnings and employment status. The study compares the outcomes of men affected by the policy changes with the outcomes of men of similar ages not affected by the policy changes.
- The study finds that the 1983 policy change is associated with few changes to the observed outcomes and that the 2000 policy change is associated with higher earnings and hours worked per week for affected men.
- The quality of causal evidence presented in this report is low because the study does not demonstrate that men affected by the policy changes are similar to the men not affected by the policy change, nor does it control for possible differences. This means we are not confident that the estimated effects are attributable to the 1983 and 2000 policy changes; other factors are likely to have contributed.
Intervention Examined
Social Security Earnings Test
Features of the Intervention
The Social Security earnings test reduces the benefits of Social Security beneficiaries who have earnings above an earnings test threshold. The 1983 policy change eliminated the earnings test for people ages 70 and 71. The 2000 policy change eliminated the earnings test for people between the full retirement age (around age 65 at the time) and age 70.
Features of the Study
The study is a nonexperimental analysis that uses regression models to examine differences in annual earnings and employment status between men affected by the policy changes and men of similar ages not affected by the policy changes. The study uses data from three sources. The first is the March Current Population Survey from 1976 to 2005, which includes self-reported earnings and other information. The second is the New Beneficiary Data System, which includes information from 1983 to 1989 on demographics, labor supply, health, income, wealth, and earnings for those who first received Social Security benefits between mid-1980 and mid-1981. The third data source is the 2004 Social Security Benefit and Earnings Public Use File, which includes information from 1990 to 2004 on demographics, Social Security eligibility and benefits, and earnings for those who received Social Security benefits in 2004. The authors conducted most study analyses using data from 1978 to 1986 or from 1996 to 2003, and the analyses involve between about 4,000 to 300,000 person-by-year observations.
Findings
Earnings
- The study found that the 1983 policy change was not associated with changes in earnings, but the 2000 policy change was associated with larger growth in earnings for those ages 66 to 68, higher annual earnings for those ages 65 to 69, and declined increases in earnings as people aged from 66 to 69
Employment
- The study found that the 1983 policy change was associated with an increase in weeks worked per year for those ages 70 and 71 and that the 2000 policy change was associated with an increase in hours worked per week for those ages 65 to 69.
Considerations for Interpreting the Findings
The study is a nonexperimental analysis that does not demonstrate that the men affected by the policy changes were similar to the comparison group, and did it not control for pre-intervention measures of outcomes. These differences, rather than the policy changes, could account for the findings. The study cannot restrict the analyses using CPS data to only those eligible for Social Security, so some people might be included who are not eligible for Social Security and thus should not change their behavior in response to the changes to the earnings test, which might bias the results.
Causal Evidence Rating
The quality of causal evidence presented in this report is low because the study does not demonstrate that the men affected by the policy changes were similar to the men not affected by the policy changes or control for possible differences between these groups. This means we are not confident that the estimated effects are attributable to the 1983 and 2000 policy changes; other factors are likely to have contributed.