RETIREMENT RESEARCH January 2012, Number 12-2 WHY DO STATE DISABILITY APPLICATION RATES VARY OVER TIME? By Norma B. Coe, Kelly Haverstick, Alicia H. Munnell, and Anthony Webb* Introduction Social Security Disability Insurance (SSDI) applica- come (SSI). The key finding is that when states limit tions and benefit receipts vary greatly by state, which the ability of insurance companies to price coverage has led to concerns about potential inconsistencies based on an individual’s characteristics (“community in the way that states apply disability standards.1 An rating”) and to deny coverage (“guaranteed issue”), earlier brief concluded that more than 70 percent of SSDI application rates decline. This provocative re- the variation across states in SSDI application rates is sult merits further exploration because it implies that explained by state health, demographic, and employ- health care reform, such as the Affordable Care Act, ment characteristics; state policies and politics explain could have spill-over effects to the SSDI program. very little.2 Another concern has been the growth in the SSDI program over time. This brief uses the same data as the earlier analysis to answer a related ques- Trends in Disability Rates tion: How much of the trends in SSDI application rates within states can be explained by the different Within States factors? The earlier brief found substantial variation in SSDI The discussion proceeds as follows. The first application rates across states, with average rates section provides background on the variation in SSDI ranging from 0.5 percent in Utah to 1.4 percent in application rates within states over time. The second Mississippi over the 1993-2009 period.3 This brief recaps the model and the variables. The third section concentrates on changes in application rates over reports the results for a state and year fixed-effects time within states. These variations tend to be less model that identifies the changes within states that dramatic than the differences across states, which is affect the SSDI application rates over time. To better not surprising as many of the health, demographic, understand the results, it also estimates equations and employment factors found to be influential in ex- for two types of SSDI applicants – those who apply to plaining variations across states do not change rapidly SSDI alone and lower-income individuals who apply concurrently to SSDI and Supplemental Security In- * Norma B. Coe is associate director for research at the Center for Retirement Research at Boston College (CRR). Kelly Haverstick is a former CRR research economist. Alicia H. Munnell is director of the CRR and the Peter F. Drucker Profes- sor in Management Sciences at Boston College’s Carroll School of Management. Anthony Webb is a research economist at the CRR. This brief is the second of two adapted from a longer paper (Coe et al. 2011). 2 Center for Retirement Research over time. Figure 1 shows trends in SSDI applica- Health, Demographic, and Employment tion rates for a select group of states and the national Characteristics average. Ohio had the highest growth rate during the period; after dipping initially, its application rate rose Three state-level health variables would all be expect- from 0.6 percent in 1998 to 1.1 percent in 2009. New ed to increase SSDI application rates. These variables York exhibited the lowest growth rate (one of seven come from the Center for Disease Control’s Behav- states with a negative growth rate), with its applica- ioral Risk Factor Surveillance Survey (BRFSS).6 The tion rate dropping from 0.9 percent to 0.8 percent variables are: over the period. The other two states are included to • self-reported fair/poor health status; show the overall boundaries in application rate levels, • smoking (ever smoked more than 100 ciga- as Mississippi tends to have the highest rate and Utah rettes); and the lowest rate. • self-reported body mass index (BMI). Figure 1. Total SSDI Application Rates of Other important factors that could influence SSDI Selected States, 1993-2009 applications are the socioeconomic composition and employability of potential applicants. Variables in the 2.0% analysis include: MS NY OH UT Average • Age of the population. Younger populations are less likely to be insured by SSDI and less 1.5% likely to have a disability that warrants an ap- plication.7 Individuals age 50 plus face a dif- 1.0% ferent screening process, in which it is easier to be accepted, so a state with a relatively older 0.5% population would be expected to have a higher SSDI application rate.8 0.0% • Education. States with a higher proportion of 1993 1995 1997 1999 2001 2003 2005 2007 2009 their population with higher education would Source: Authors’ calculations. be expected to have lower SSDI application rates. The effect of low education is ambigu- ous. Individuals with less than a high school degree may be the most vulnerable, but also The Variables may not have enough steady work history to be insured under SSDI. The data are grouped in three main categories: 1) health, demographic, and employment characteris- • White, non-Hispanic. The impact of race is tics; 2) state policies; and 3) state politics. All data ambiguous. States with a higher proportion cover the period 1993-2009 for the 50 states and the of non-Hispanic whites could be expected to District of Columbia.4 have lower rates of SSDI applications, because The dependent variable is the annual SSDI appli- non-whites are more vulnerable. Or whites cation rate by state, expressed as a percentage of the could have higher application rates because state’s working-age population (age 18-64) not receiv- they are more likely to have steady job histo- ing SSDI benefits.5 The explanatory variables were ries that enable them to qualify for SSDI. originally used to explain variation in SSDI applica- tion rates across states and would be expected to have • Male. States with a higher proportion of a similar impact on variation within states over time. males would be expected to have higher SSDI application rates due to their higher rates of labor force participation. Issue in Brief 3 • Married. States with a higher proportion of • Medicaid buy-in. States with a Medicaid buy- married residents would be expected to have in program have less strict earnings qualifi- lower SSDI application rates since married cations for Medicaid eligibility for disabled people tend to be healthier. individuals who work, allowing better access to health insurance outside of the SSDI • Poor. States with a higher proportion of their program.13 Medicaid buy-in programs are population under the federal poverty line expected to have lower SSDI application rates. would be expected to have higher SSDI ap- plication rates. State Politics Variations over time in employment character- Due to concerns that state politics could influence the istics and unemployment are expected to be associ- administration of this federal program, three variables ated with differences in SSDI application rates. The are included to test whether the governor’s party affili- variables include: ation or tenure in the job appear to have any influence • Occupation and industry. The greater the on application rates. The variables are: proportion of a state’s workforce employed • governor’s party affiliation; in a blue-collar occupation or an agricultural • an indicator for reaching the term limit; and industry, the higher the expected SSDI ap- • an indicator for an incumbent governor.14 plication rate. • Unemployment rate. Because greater unem- ployment lowers the opportunity cost of apply- The Results ing for SSDI, higher unemployment should lead to more applications. In order to explain the variation over time within a • Labor force participation rate. Discouraged state, the analysis estimates a regression equation workers may drop out of the labor force. So relating state SSDI application rates over the period the lower the labor force participation rate, the 1993-2009 to health/demographic/employment higher the expected application rate. variables, state policies, and political factors. The equation includes a variable for each state in order to State Policy isolate the factors affecting within-state changes. The results include the impact on the total SSDI applica- State policies with respect to unemployment insur- tion rate as well as its two components – SSDI-only ance and health programs could also affect applica- and SSDI-SSI concurrent applications – to see if these tion rates.9 The variables include: two sub-populations respond in the same way to dif- • Maximum weeks of unemployment insurance ferent state characteristics or policies. (UI). The longer the duration of UI, the lower Much variation is lost in a state fixed-effects the expected SSDI application rate. model, so significantly fewer of the variables have • UI benefits/average wage. The higher the a statistically significant effect on applications than ratio, the lower the expected SSDI application in the prior analysis of variation across states.15 For rate. example, the percent of the state’s population in fair/ • Strict regulation of private insurance mar- poor health does not change noticeably over a 17- ket.10 States are defined as strictly regulated if year period, so health does not explain any change they have both community rating and guaran- in the application rate within a state even though it teed issue.11 The effect of health care access was an important explanatory variable for variation on the SSDI application rate is theoretically across states. However, despite the loss in variation ambiguous. On the one hand, individuals on health and demographic variables, the variation with access to health insurance might be within a state on some of the policy variables remains. more likely to apply for SSDI because they For example, a number of states made changes to would be less likely to go uninsured during their health insurance regulations during the time the two-year waiting period for Medicare period studied, which could influence application coverage.12 On the other hand, individuals rates.16 might be less likely to apply for SSDI benefits because Medicare coverage is relatively less attractive when they can obtain health insur- ance elsewhere. 4 Center for Retirement Research Effect of Health, Demographic, and proportion of non-Hispanic whites and non-married Employment Variables individuals is positively associated with SSDI-only ap- plications. This effect probably indicates that race is For the health, demographic, and employment vari- important in acquiring the steady job history required ables, the story that emerges is primarily a cyclical for SSDI eligibility, as is not having other sources of one. That is, the variables that change materially over economic support within the household. the period 1993-2009 are those related to the strength of the economy. For total SSDI applications, both the Effect of State Policy Variables unemployment rate and the labor force participation rate have a statistically significant effect (see Figure In terms of state policies, the variable that matters for 2).17 Interestingly, the effect of these two variables within-state changes in SSDI applications is “strict differs between the two sub-populations of applicants. regulation” of the health insurance industry. The The labor force participation rate has a statistically coefficient is negative, suggesting that regulation significant effect on SSDI-only applications, whereas reduces application rates. The specific reason for the unemployment rate affects poor individuals who this effect is unclear, but perhaps the ability to obtain apply concurrently for SSDI and SSI. private health coverage makes the Medicare coverage A few demographic variables are also statistically associated with SSDI less attractive. This effect holds significant. Both low- and high-educational attain- for both the SSDI-only and the SSDI-SSI applications. ment are negatively related to total SSDI applications, driven by the concurrent SSDI-SSI applicants. Those with less than a high school degree tend to be less Effect of State Politics Variables consistently attached to the labor market, so increas- As was the case in variation across states, a Repub- ing the percent of the population with less than a high lican governor is associated with lower within-state school degree means fewer people are insured for SSDI applications. In this case, the effect is driven by SSDI. On the other side of the spectrum, as expected, concurrent SSDI-SSI applications. If Republican gov- increasing the proportion of the population with a ernors indicate a clamping-down on all needs-based post-graduate education also decreases the applica- assistance, individuals may be less likely to apply for tion rate. The results also show that increasing the SSI as well. Figure 2. Impact of Selected Factors on Within-State Variation in Total SSDI Application Rates, 1993-2009 Mean Change in Total SSDI Application Rate = 0.10 Unemployment rate 0.02 Labor force participation rate -0.02 Less than high school -0.01 Post-graduate -0.01 Strict health regulation -0.05 Republican governor -0.02 -0.10 -0.05 0.00 0.05 0.10 Notes: All results are statistically significant at least at the 10-percent level. The results shown for continuous variables are for a one-standard-deviation change; in the case of dummy variables, the results show a change from zero to one. Source: Authors’ calculations. Issue in Brief 5 Conclusion Endnotes This brief uses data collected to examine why SSDI 1 See McVicar (2006); Bound and Burkhauser (1999); application rates vary so much across states to answer and Rupp and Stapleton (1998). questions about what drives within-state variation. The results show that the strength of the economy is 2 Coe et al. (2012). a key driver of SSDI application rates. When un- employment rises or labor force participation falls, 3 Coe et al. (2012). SSDI application rates increase. Several other health, demographic and employment variables also change 4 Data are missing for: Wyoming in 1993, Rhode enough over time to have a significant effect. Island in 1994, Washington, DC in 1995, and Hawaii The most interesting finding is that strict health in 2004 because of lack of coverage in the Behavioral insurance market regulation is correlated with lower Risk Factor Surveillance Survey. Data are also missing SSDI application rates within a state. This is the first for Nevada in 1994 due to lack of detailed data from evidence of which we are aware that health insur- Social Security on SSDI-only applications; we have ance regulation influences SSDI applications, and it 683 observations for concurrent applications for both implies that health care reform, such as the Afford- SSDI and Supplemental Security Income. able Care Act, could have spill-over effects to the SSDI program. This provocative finding merits further 5 The denominator is the number of residents aged exploration in order to understand the precise mecha- 18-64 in a state as of July 1 from the U.S. Census nism through which the effect occurs. Bureau. From this figure we subtract the number of Finally, politics appear to matter. Having a beneficiaries of each program, obtained from the So- Republican governor leads to lower levels of SSDI ap- cial Security Administration Statistical Bulletins (SSA plications. This effect is concentrated among lower- 1994-2009), since current beneficiaries are not at risk income applicants who file concurrently for SSDI and of applying. SSI, which may indicate an overall decline in applica- tions for needs-based assistance. 6 The BRFSS has been administered since 1984 and is the largest ongoing telephone survey in the United States. BRFSS provides detailed data on self-rated health; health-related behaviors such as smoking and drinking; and factors correlated with health condi- tions such as obesity, along with state-of-residence indicators. While the BRFSS data include other health-related variables that may be related to the SSDI application rate (such as alcohol consumption, doctor visits, exercise habits, and mental health mea- sures), these variables were not consistently available for all states over the entire 1993-2009 period. 7 To be insured for SSDI, one must have worked the required number of quarters based on age, and 20 quarters within the last 10 years. 8 Age is specifically in the SSDI determination process because the assessment of the ability to be retrained changes depending on whether an appli- cant is age 50-54 (Approaching Advanced Age), 55-59 (Advanced Age), or 60-64 (Retirement Age). 6 Center for Retirement Research 9 State-mandated employer temporary disability References insurance (TDI) was an important explanatory vari- able in terms of variation of SSDI applications among References for the data sources used in this brief are states. However, since the TDI programs were mostly available in the full paper (Coe et al. 2011). enacted after the Great Depression, they do not change during the period 1993-2009 and therefore are Bound, John and Richard V. Burkhauser. 1999. “Eco- not included in the current analysis. nomic Analysis of Transfer Programs Targeted on People with Disabilities.” In Handbook of Labor 10 Data on state regulations of health insurance were Economics, edited by Orley C. Ashenfelter and Da- compiled from The Henry J. Kaiser Family Founda- vid E. Card, 3417-3528. Amsterdam: Elsevier. tion (2010a; 2010b), and Georgetown University Health Policy Institute (2004). Buchmueller, Thomas and John DiNardo. 2002. “Did Community Rating Induce an Adverse Selection 11 Herring and Pauly (2006). Several studies have Death Spiral? Evidence from New York, Pennsyl- shown that these regulations have a significant im- vania, and Connecticut.” The American Economic pact on coverage and presumably also on subsequent Review 92(1): 280-294. health care access (e.g., Buchmueller and DiNardo 2002; and Long and Stockley 2009). Coe, Norma B., Kelly Haverstick, Alicia H. Munnell, and Anthony Webb. 2011. “What Explains State 12 This hypothesis is explored in Gruber and Kubik Variation in SSDI Application Rates?” Working (2002), who find that individuals with access to health Paper 2011-23. Chestnut Hill, MA: Center for insurance from a spouse are 26-74 percent more Retirement Research at Boston College. likely to apply for SSDI benefits than those without external access to health insurance. Coe, Norma B., Kelly Haverstick, Alicia H. Munnell, and Anthony Webb. 2012. “What Explains Varia- 13 These data were compiled from Kehn, Croake, and tion in Disability Application Rates Across States?” Schimmel (2010); Croake and Liu (2009); Gruman Issue in Brief 12-1. Chestnut Hill, MA: Center for et. al (2008); Jensen (2004, 2006); Georgia Depart- Retirement Research at Boston College. ment of Community Health (https://www.gmwd.org/ WebForms/StaticContent1.aspx); Delaware Health Gruber, Jonathan and Jeffrey Kubik. 2002. “Health and Social Services (http://dhss.delaware.gov/dhss/ Insurance Coverage and the Disability Insur- dmma/); and Commonwealth of Kentucky (http:// ance Application Decision.” Working Paper 9148. manuals.chfs.ky.gov/dcbs_manuals/DFS/VOLIVA/ Cambridge, MA: National Bureau of Economic OMVOLIVA.pdf). Research. 14 The political variables come from National Gover- Herring, Bradley and Mark Pauly. 2006. “The Effect of nors Association (2011) and Council of State Govern- State Community Rating Regulations on Premi- ments (2007). ums and Coverage in the Individual Health Insur- ance Market.” Working Paper 12504. Cambridge, 15 While much variation is lost by relying on within- MA: National Bureau of Economic Research. state changes for identification, the state fixed effects are significant and the Hausman test suggests that Long, Sharon K. and Karen Stockley. 2009. “Health a fixed-effects model is more appropriate than a Insurance in Massachusetts: An Update on Insur- random-effects model. ance Coverage and Support for Reform as of Fall 2008.” Washington, DC: Urban Institute. 16 The descriptive statistics for the variables in the three regressions and the full results are shown in the McVicar, Duncan. 2006. “Why Do Disability Benefit Appendix. Rolls Vary Between Regions? A Review of the Evi- dence from the USA and the UK.” Regional Studies 17 For more details on the results of the SSDI-only 40(5): 519-533. and concurrent SSDI-SSI applicants, see Appendix Table 2 and Coe et al. (2011). Rupp, Kalman and David Stapleton., eds. 1998. Growth in Disability Benefits. Kalamazoo, MI: W.E. Upjohn Institute of Employment Research. APPENDIX 8 Center for Retirement Research Table A1. Descriptive Statistics Across states over time Within states over time Mean Standard deviation Standard deviation Dependent Variables (Percent of Working-Age Population) Total SSDI application rate 0.83 0.24 0.11 SSDI-only application rate 0.43 0.11 0.05 Concurrent SSDI-SSI application rate 0.40 0.15 0.07 Health, Demographic, and Employment Variables Health Fair/poor health 0.15 0.03 0.01 Ever smoke 100+ cigarettes 0.47 0.05 0.02 Overweight or obese (BMI) 0.59 0.06 0.05 Age Profile Age under 18 0.26 0.03 0.02 Age 18-25 0.11 0.01 0.01 Age 25-50 (omitted) 0.35 0.02 0.02 Age 50+ 0.28 0.04 0.03 Education Profile Less than high school 0.15 0.05 0.03 High school degree (omitted) 0.34 0.05 0.02 Some college 0.42 0.06 0.03 Post-graduate 0.09 0.03 0.01 Other Demographics White, non-Hispanic 0.76 0.16 0.03 Male 0.49 0.01 0.01 Married 0.55 0.05 0.02 Poor 0.12 0.04 0.02 Occupation Service occupation 0.43 0.03 0.02 Blue-collar occupation 0.25 0.04 0.02 Other occupations 0.32 0.05 0.03 Industry Agriculture and physical industries 0.29 0.05 0.02 Professional industries (omitted) 0.71 0.05 0.02 Labor Force Unemployment rate 0.05 0.02 0.01 Labor force participation rate 0.67 0.04 0.01 Issue in Brief 9 Table A1. Descriptive Statistics (Continued) Across states over time Within states over time Mean Standard deviation Standard deviation State Policy Variables Length of UI benefits (weeks) 31.66 9.27 9.20 UI benefits/average wage 0.37 0.06 0.02 Strict health regulation 0.13 0.33 0.13 Medicaid buy-in 0.37 0.48 0.43 State-mandated employer TDI 0.10 0.30 N/A State Politics Variables Republican governor 0.54 0.50 0.41 Governor at term limit 0.29 0.45 0.40 Incumbent governor 0.39 0.49 0.47 Source: Authors’ calculations. 10 Center for Retirement Research Table A2. Regression Results with State Fixed Effects for Total SSDI, SSDI-Only, and SSDI-SSI, 1993-2009 Health, Demographic, and Employment Variables (1) Total SSDI (2) SSDI-only (3) SSDI-SSI Fair/poor health -0.069 -0.073 0.007 (0.310) (0.230) (0.190) Ever smoke 100+ cigarettes 0.066 0.068 -0.001 (0.270) (0.140) (0.170) Overweight or obese (BMI) 0.027 0.131 -0.102 (0.250) (0.170) (0.140) Age under 18 0.186 0.025 0.151 (0.430) (0.230) (0.250) Age 18-25 -0.086 -0.189 0.103 (0.400) (0.260) (0.250) Age 50+ 0.314 0.189 0.116 (0.350) (0.190) (0.210) Less than high school -0.525 ** -0.187 -0.341 * (0.260) (0.140) (0.170) Some college 0.062 0.056 0.002 (0.180) (0.110) (0.110) Post-graduate -0.775 * -0.307 -0.468 * (0.460) (0.260) (0.260) White, non-Hispanic 0.248 0.249 ** -0.003 (0.210) (0.120) (0.160) Male 0.651 0.364 0.267 (0.400) (0.290) (0.230) Married -0.199 -0.213 * 0.019 (0.190) (0.120) (0.130) Poor -0.085 -0.147 0.048 (0.210) (0.120) (0.130) Service occupation 0.155 0.079 0.073 (0.230) (0.150) (0.120) Blue-collar occupation -0.238 -0.147 -0.089 (0.290) (0.210) (0.180) Agriculture and physical industries 0.021 0.094 -0.078 (0.270) (0.180) (0.150) Unemployment rate 2.034 *** 0.412 1.608 *** (0.730) (0.430) (0.480) Labor force participation rate -1.544 *** -1.184 *** -0.343 (0.460) (0.300) (0.340) Issue in Brief 11 Table A2. Regression Results with State Fixed Effects for Total SSDI, SSDI-Only, and SSDI-SSI, 1993-2009 (Continued) State Policy Variables (1) Total SSDI (2) SSDI-only (3) SSDI-SSI Length of UI benefits -0.002 0.000 -0.001 (0.000) (0.000) (0.000) UI benefits/average wage 0.084 -0.042 0.130 (0.230) (0.140) (0.140) Strict health regulation -0.054 *** -0.019 * -0.036 *** (0.020) (0.010) (0.010) Medicaid buy-in 0.024 0.006 0.017 (0.020) (0.010) (0.010) State Politics Variables Republican governor -0.021 ** -0.003 -0.018 ** (0.010) (0.010) (0.010) Governor at term limit -0.006 -0.001 -0.005 (0.010) (0.010) (0.010) Incumbent governor 0.005 -0.002 0.007 (0.010) (0.010) (0.010) Constant 1.386 ** 0.879 *** 0.505 (0.530) (0.280) (0.360) Observations 862 862 863 R-squared 0.923 0.853 0.932 Note: * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent. Robust standard errors clustered by state are in parentheses. Also included are a set of year dummies (excluding 1993) and state dummy variables. Source: Authors’ calculations. RETIREMENT RESEARCH About the Center Affiliated Institutions The Center for Retirement Research at Boston The Brookings Institution College was established in 1998 through a grant Massachusetts Institute of Technology from the Social Security Administration. The Syracuse University Center’s mission is to produce first-class research Urban Institute and educational tools and forge a strong link between the academic community and decision-makers in the public and private sectors around an issue of Contact Information Center for Retirement Research critical importance to the nation’s future. To achieve Boston College this mission, the Center sponsors a wide variety of Hovey House research projects, transmits new findings to a broad 140 Commonwealth Avenue audience, trains new scholars, and broadens access to Chestnut Hill, MA 02467-3808 valuable data sources. Since its inception, the Center Phone: (617) 552-1762 has established a reputation as an authoritative source Fax: (617) 552-0191 of information on all major aspects of the retirement E-mail: crr@bc.edu income debate. Website: http://crr.bc.edu © 2012, by Trustees of Boston College, Center for Retire- The research reported herein was performed pursuant to ment Research. All rights reserved. Short sections of text, a grant from the U.S. Social Security Administration (SSA) not to exceed two paragraphs, may be quoted without ex- funded as part of the Retirement Research Consortium. The plicit permission provided that the authors are identified and opinions and conclusions expressed are solely those of the full credit, including copyright notice, is given to Trustees of authors and do not represent the opinions or policy of SSA, Boston College, Center for Retirement Research. any agency of the federal government, or the Center for Retirement Research at Boston College.