Racial disparities in federal disability benefits

Contemporary Economic Policy, Jan, 2007 by Erin M. Godtland, Michele Grgich, Carol Dawn Petersen, Douglas M. Sloane, Ann T. Walker

C. Literature Review

Several previous efforts to model SSA's disability decision-making process shed light on the issue of racial differences in allowance rates. For example, Hu et al. (2001) analyzed SSA administrative data, matched with data from the 1990 Survey of Income and Program Participation and used multistage logit models to examine the determinants of SSA's disability decision at each step of the initial level. They found significant racial differences in denial rates at Step 2 of the determination process (the step at which the severity of the impairment is decided). Specifically, they found that African-Americans were more likely to be denied at Step 2 than whites and that the greater likelihood of being denied was especially pronounced for younger African-American claimants (age less than 35 yrs).

Kreider and Riphahn (2000) analyzed Heath and Retirement Survey (HRS) data using a reduced form multistage logit model of the probability of being awarded benefits at the initial level, the probability of appealing the initial level decision, and the probability of being awarded benefits at the appellate level. One of their objectives was to explore whether men and women have different responses to changes in disability policy, such as benefit amounts. Among other findings, they found that white females and males are no more likely to be awarded benefits than nonwhites at the initial level. However, they found that white males are significantly less likely (at the 10% level) to be awarded benefits than nonwhites at the appellate level.

Benitez-Silva et al. (1999) also analyzed HRS data with a reduced form binary logit model to evaluate individuals' decisions to apply for benefits and SSA's decision to award benefits at the initial and appellate levels. Their principal finding was that self-reported health information is a significant predictor of application and award decisions. They also found that white claimants were no more likely than minority claimants to receive favorable disability decisions at the initial and appellate levels.

Baldwin (1997) analyzed SSA administrative data with a multistage logistic model to estimate factors affecting decisions at the initial level. She also used the Oaxaca decomposition method to test whether the observed difference in SSDI benefit award decisions between men and women was a result of a difference in their characteristics or a difference in how they were treated. She found that women and men were equally likely to satisfy medical criteria, but women over age 55 were more likely to be rejected on the basis of vocational criteria. Baldwin, however, did not include race in her model.

Several factors may account for the differences in findings pertaining to race. First, Hu et al. (2001) used data pertaining to all age groups. Their findings pertaining to race are particularly dramatic for younger African-Americans. Kreider and Riphahn (2000) and Benitez-Silva et al. (1999), on the other hand, used HRS data, which are based on sample of Americans between the ages of 50 and 61. A potential explanation for the difference in Kreider's and Benitez-Silva's findings at the appellate level may be due to the fact that Kreider and Riphahn partition the sample by gender, thereby allowing for greater precision in analyzing the effect of race by gender.