From the abstract of "How Large Are The Classification Errors In The Social Security Disability Award Process" by Hugo Benitez-Silva, Moshe Buchinsky and John Rust (emphasis added):
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 from the Disability Insurance (DI) and Supplemental Security Income (SSI) programs. ... We find that approximately 20\% of SSI/DI applicants who are ultimately awarded benefits are not disabled, and that 60\% of applicants who were denied benefits are disabled.And from the body of the report:
... our results suggest ...that the high reversal rates by the ALJs actually serves to reduce the classication error rates. We find that the low initial award rate at the DDS level produces a high rate of rejection errors at this stage. The DDS centers appear to behave according to a philosophy of when in doubt, reject. However, self-selection is operative: we find that applicants who appeal an initial rejection by the DDS are more likely to be truly disabled than the initial pools of applicants that the DDS evaluated. Therefore, the relatively high acceptance rate by the ALJs, combined with the self-selection in the decision to appeal an initial rejection, signicantly reduces the rate of rejection errors without increasing the rate of award errors.