How Large Are the Classification Errors in the Social Security Disability Award Process?

TitleHow Large Are the Classification Errors in the Social Security Disability Award Process?
Publication TypeReport
Year of Publication2003
AuthorsBenitez-Silva, H, Buchinsky, M, Rust, J
InstitutionSUNY-Stony Brook
Call Numberwp_2003/B-Silva_etal.pdf
KeywordsDisabilities, Methodology, Social Security
Abstract

This paper presents an audit of the multistage application and appeal process that the U.S. Social Security Administration (SSA) uses to determine eligibility for disability benefits of the Disability Insurance (DI) and Supplemental Security Income (SSI) programs. We study a subset of individuals from the Health and Retirement Survey (HRS) who applied for DI or SSI benefits between 1992 and 1996. We compare the SSA s ultimate award decision a (i.e., after allowing for all possible appeals) to the applicant s self-reported disability status d (recorded at the first HRS survey after their initial application for benefits). We use these data to estimate classification error rates under the hypothesis that applicants self-reported disability status d is the relevant measure of true disability and the SSA s ultimate award decision a is a noisy but unbiased indicator of d This truthful, accurate reporting hypothesis allows us to estimate the magnitude of classification errors in the SSA award process and obtain insights into the patterns of self-selection induced by varying delays and award probabilities at various levels of the application and appeal process. Overall we find that 22 of SSI/DI applicants who are ultimately awarded benefits are not disabled, and that 59 of applicants who were denied benefits are disabled. We construct a computerized disability screening rule using a subset of objective health indicators that the SSA uses in making award decisions that results in significantly lower classification error rates than does SSA s current award process. This suggests that there may be cheaper, faster, and more accurate ways to make disability determinations than the SSA s current disability award process. We also estimate classification errors under the assumption that both a and d are noisy but unbiased indicators of an (unobserved) underlying indicator of true disability, t . However, the estimated classification error rates remain virtually unchanged under this alternative hypothesis.

Notes

NIH Grant AG12985-02

URLhttp://ms.cc.sunysb.edu/ hbenitezsilv/dice.pdf
Endnote Keywords

Social Security/Disability/Disability/Classification errors

Endnote ID

10382

Citation Key5540