Abstract
We investigate the relationship between the unemployment rate and characteristics of applicants for Social Security Disability Insurance using administrative records of the universe of applicants between 1991 and 2008. As the unemployment rate rises, applications shift to those with higher work capacity who are rejected early in the eligibility determination process and have higher pre-application earnings and employment. However, post-application earnings and employment of denied applicants are slightly negatively related to the unemployment rate, suggesting that both compositional changes toward applications with higher work capacity and adverse economic conditions affect their employment and earnings.
Funding statement: Boston Retirement Research Center.
Acknowledgement
We would like to thank Gregory Acs, Pamela Loprest, Nadia Karamcheva and seminar participants at the Urban Institute, the Social Security Administration and Oregon State University for helpful comments. Stephan Lindner was an employee of the Urban Institute while working on this research. This potential conflict of interest has been reviewed and managed by Oregon Health and Science University.
References
Abraham, Katharine G., and Robert Shimer. (2001).National Bureau of Economic Research, . DOI:.10.3386/w8513Search in Google Scholar
Autor, David, and Mark Duggan. 2010. Supporting Work: A Proposal for Modernizing the U.S. Disability Insurance System. Washington, DC: Center for American Progress and the Hamilton Project.Search in Google Scholar
Autor, David, Nicole Maestas, Kathleen Mullen, and Alexander Strand. (2015). “Does Delay Cause Decay? The Effect of Administrative Decision Time on the Labor Force Participation and Earnings of Disability Applicants.”NBER Working Paper No 20840.10.3386/w20840Search in Google Scholar
Autor, David H., and Mark G. Duggan. 2003. “The Rise in the Disability Rolls and the Decline in Unemployment.” Quarterly Journal of Economics 118(1): 157–205.10.1162/00335530360535171Search in Google Scholar
Ben-Shalom, Yonatan, and Arif A. Mamun. 2015. “Return-To-Work Outcomes Among Social Security Disability Insurance Program Beneficiaries.” Journal of Disability Policy Studies 26(2): 100–110.10.1177/1044207315583874Search in Google Scholar
Bound, John 1989. “The Health and Earnings of Rejected Disability Insurance Applicants.” American Economic Review 79(3): 482–503.10.3386/w2816Search in Google Scholar
Bound, John, and Richard V. Burkhauser. 1999. “Economic Analysis of Transfer Programs Targeted on People with Disabilities..” In Handbook of Labor Economics. , edited by O. Ashenfelter, and D. Card, Vol. 33418–3528. Elsevier Chapter 51.10.1016/S1573-4463(99)30042-0Search in Google Scholar
Bound, John, and Timothy Waidmann. 2002. “Accounting for Recent Declines in Employment Rates among Working-Aged Men and Women with Disabilities.” Journal of Human Resources 37(2): 231–250.10.2307/3069646Search in Google Scholar
Brault, Matthew 2012. American with Disabilities: 2010. Washington, DC: U.S. Census Bureau.Search in Google Scholar
Chen, Susan, and Wilbert Van Der Klaauw. 2008. “The Work Disincentive Effects of the Disability Insurance Program in the 1990s.” Journal of Econometrics 142: 757–784.10.1016/j.jeconom.2007.05.016Search in Google Scholar
Coe, Norma B., and Matthew S. Rutledge. (2013). “How Does the Composition of Disability Insurance Applicants Change across Business Cycles?”Center for Retirement Research at Boston College Working Paper No. 2013-5.10.2139/ssrn.2222768Search in Google Scholar
Davis, Steven J., and Till Von Wachter. 2011. “Recessions and the Costs of Job Loss.” Brookings Papers on Economic Activity 1–72.10.3386/w17638Search in Google Scholar
French, Eric, and Jae Song. 2014. “The Effect of Disability Insurance Receipt on Labor Supply.” American Economic Journal: Economic Policy 6(2): 291–337.10.1257/pol.6.2.291Search in Google Scholar
Hellerstein, Judith K, and Melinda Sandler Morrill. 2011. “Booms, Busts, and Divorce.” The B.E. Journal of Economic Analysis & Policy 11(1).10.2202/1935-1682.2914Search in Google Scholar
Hoynes, Hilary, Douglas L. Miller, and Jessamyn Schaller. 2012. “Who Suffers During Recessions?.” The Journal of Economic Perspectives 26(3): 27–47.10.3386/w17951Search in Google Scholar
Lindner, Stephan 2013. “From Working to Applying: Employment Transitions of Applicants for Disability Insurance in the United States.” Journal of Social Policy 42(2): 329–348.10.1017/S0047279412000955Search in Google Scholar
Maestas, Nicole, Kathleen Mullen, and Alexander Strand. 2013. “Does Disability Insurance Receipt Discourage Work? Using Examiner Assignment to Estimate Causal Effects of SSDI Receipt.” The American Economic Review 103(5): 1797–1829.10.1257/aer.103.5.1797Search in Google Scholar
Maestas, Nicole, Kathleen J. Mullen, and Alexander Strand. 2015. “Disability Insurance and the Great Recession.” American Economic Review 105(5): 177–182.10.7249/WR1088Search in Google Scholar
Panis, Constantijn, Roald Euller, Cynthia Grant, Melissa Bradley, Christine E. Peterson, Randall Hirscher, and Paul Steinberg. 2000. SSA Program Data User’s Manual. Baltimore, MD: Social Security Administration.Search in Google Scholar
Rupp, Kalman 2012. “Factors Affecting Initial Disability Allowance Rates for the DI and SSI Programs: the Role of the Demographic and Diagnostic Composition of Applicants and Local Labor Market Conditions.” Social Security Bulletin 72(4): 11–35.Search in Google Scholar
Rupp, Kalman, and David Stapleton. 1995. “Determinants of the Growth in the Social Security Administration’s Disability Programs – An Overview.” Social Security Bulletin 58(4): 43–70.Search in Google Scholar
(Social Security Administration) (2012). Annual Statistical Supplement to the Social Security Bulletin, 2012SSA Publication No. 13-11700.Search in Google Scholar
Solon, Gary, Steven J. Haider, and Jeffrey Wooldridge. (2013)., . DOI:.10.3386/w18859Search in Google Scholar
Von Wachter, Till, Jae Song, and Joyce Manchester. 2011. “Trends in Employment and Earnings of Allowed and Rejected Applicants to the Social Security Disability Insurance Program.” American Economic Review 101(7): 3308–3329.10.1257/aer.101.7.3308Search in Google Scholar
Wixon, Bernard, and Alexander Strand. 2013. “Identifying SSA’s Sequential Disability Determination Steps Using Administrative Data.” In Research and Statistics Note. Office of Retirement and Disability Policy /Office of Research, Evaluation, and Statistics (2013-01).Search in Google Scholar
Appendix: Robustness Checks and Auxiliary Results
Application duration and outcome: Concerning application outcome, we present results for three measures: initial determination, determination including appeals, and determination including appeals and re-applications within 5 years. Table 6 first shows mean values, then results for regressions with state-specific fixed effects, and results for regressions that also include applicants’ characteristics.[28] Both the application duration and the application success rate are negatively related to the unemployment rate. The coefficient for application duration becomes more negative when other covariates are included, whereas the coefficient for application success becomes less negative with such covariates included in the regression. Apparently, applicants who apply more frequently during economic downturns have a longer application processing time and lower application success chance. Still, coefficients for application duration and determination remain negative even after controlling for such compositional changes, suggesting that case workers process claims more quickly and tend to be slightly more strict during economic downturns.[29]
Unweighted regressions: We estimate our regressions using no weights and find very similar results, as shown in Table 7 for applicant characteristics. A few coefficients are somewhat different as compared to Table 1. For instance, coefficients of the log of the share of high-school drop-outs and joint DI/SSI applications are more positive when no weights are used. However, we find no substantial differences for applications determined at step 5 and applicants with a musculoskeletal impairments or mental disorder. Overall, these results suggest that the relationship between short-term fluctuations of the unemployment rate and characteristics of applicants are similar across states.
Variable | Estimate /SE |
---|---|
Applications (b.p.) | 0.57*** |
(0.10) | |
Appl. for DI/SSI (%) | 0.99*** |
(0.14) | |
Appl. for DI/SSI (b.p.) | 0.45*** |
(0.07) | |
Step 2 applicants (%) | 0.50* |
(0.21) | |
Step 2 applicants (b.p.) | 0.15*** |
(0.04) | |
Step 3 applicants (%) | –0.56*** |
(0.15) | |
Step 3 applicants (b.p.) | 0.05 |
(0.03) | |
Step 4 applicants (%) | 0.51** |
(0.19) | |
Step 4 applicants (b.p.) | 0.17*** |
(0.04) | |
Step 5 applicants (%) | –0.22 |
(0.20) | |
Step 5 applicants (b.p.) | 0.20*** |
(0.06) |
p < 0.05
p < 0.01
p < 0.001.
Notes: Each cell of the second column represents a separate regression and displays the coefficient for the unemployment rate in a regression with the dependent variable as given by the first column. All regressions are unweighted and include state fixed effects, year fixed effects, and state fixed effects interacted with linear time trends. All variables are either expressed as percentage of all applicants or as basis points (number of applicants with that characteristic per ten thousand adults for a time period and state). Standard errors are clustered at the state level and displayed in parentheses below corresponding coefficients. The sample size for all regressions is 3,744.
Distributed lag models: Because many applicants for DI do not immediately file an application after losing their job (Lindner 2013), it is plausible that lagged values of the unemployment rate are also associated with applicants’ characteristics and outcomes. We estimate models with two, four and six quarters of lags of the unemployment rate (for regressions with years as frequency we estimate one regression with one year as lag). Results for these distributed lag models are very similar to results from our main specification without lags. Table 8 presents regression results for applicant characteristics with six lags of the unemployment rate (we show summed lagged coefficients in the table). As expected, the correlation between the number of applications and the sum of the lagged unemployment coefficients is larger as compared to Table 1. In terms of coefficients, a few appear to be slightly larger with lags included (e. g., the log of the share of step 2 applications) but our main conclusions are robust to including such lags.
Variable | Estimate /SE |
---|---|
Applications (b.p.) | 0.74*** |
(0.11) | |
Appl. for DI/SSI (%) | 0.91*** |
(0.26) | |
Appl. for DI/SSI (b.p.) | 0.53*** |
(0.09) | |
Step 2 applicants (%) | 1.1*** |
(0.26) | |
Step 2 applicants (b.p.) | 0.26*** |
(0.05) | |
Step 3 applicants (%) | –0.81*** |
(0.19) | |
Step 3 applicants (b.p.) | 0.03 |
(0.04) | |
Step 4 applicants (%) | 0.48 |
(0.30) | |
Step 4 applicants (b.p.) | 0.18** |
(0.07) | |
Step 5 applicants (%) | –0.53 |
(0.37) | |
Step 5 applicants (b.p.) | 0.26* |
(0.11) |
p < 0.05
p < 0.01
p < 0.001.
Notes: Each cell of the second column represents a separate regression and displays the coefficient for the summed unemployment rate coefficients in a distributed lag regression with the dependent variable as given by the first column and six quarters of lags for the unemployment rate. All regressions use the number of applications of a state per period as weights and include state fixed effects, year fixed effects, and state fixed effects interacted with linear time trends. All variables are either expressed as percentage of all applicants or as basis points (number of applicants with that characteristic per 10,000 adults for a time period and state). Standard errors are clustered at the state level and displayed in parentheses below corresponding coefficients. Mean values refer to weighted averages over all states and the time period 1991–2008. The sample size for all regressions is 3,744.
Including years of the Great Recession: We have excluded years of the Great Recession available to us (2009 and 2010) for our main analysis because an increasing fraction of applications filed during these years are still pending. Still, we can estimate regressions for characteristics of applicants with these years included to see whether our main results are robust to including years with much higher unemployment rates and application numbers. Table 9 displays these results. The correlation between the unemployment rate and applications for DI is somewhat smaller as compared to Table 1 and so are most coefficients. Step 2 and 4 applications still rise with the unemployment rate but the coefficient for the relative change in the share of step 4 applicants is now not statistically significant (albeit still substantial). The negative relationship between step 5 applications and the unemployment rate is smaller, however, the coefficient for the change in the relative share of applicants with musculoskeletal impairments is more negative. These differences suggest that changes to applicants’ characteristics during the Great Recession might be different compared to earlier period of economic downturn. Future research could look more carefully at the Great Recession when sufficient post-recession data is available.
Variable | Level |
---|---|
Applications (b.p.) | 0.50*** |
(0.10) | |
Appl. for DI/SSI (%) | 0.67*** |
(0.12) | |
Appl. for DI/SSI (b.p.) | 0.35*** |
(0.07) | |
Step 2 applicants (%) | 0.53*** |
(0.16) | |
Step 2 applicants (b.p.) | 0.14*** |
(0.04) | |
Step 3 applicants (%) | –0.40** |
(0.14) | |
Step 3 applicants (b.p.) | 0.05 |
(0.03) | |
Step 4 applicants (%) | 0.18 |
(0.20) | |
Step 4 applicants (b.p.) | 0.11* |
(0.04) | |
Step 5 applicants (%) | –0.28 |
(0.23) | |
Step 5 applicants (b.p.) | 0.17* |
(0.07) |
p<0.05
p<0.01
p<0.001.
Notes: Each cell of the second column represents a separate regression and displays the coefficient for the unemployment rate in a regression with the dependent variable as given by the first column. All regressions are weighted and include state fixed effects, year fixed effects, and state fixed effects interacted with linear time trends. All variables are either expressed as percentage of all applicants or as basis points (number of applicants with that characteristic per ten thousand adults for a time period and state). Standard errors are clustered at the state level and displayed in parentheses below corresponding coefficients. The sample size for all regressions is 4,160.
© 2017 Walter de Gruyter GmbH, Berlin/Boston