Duggan and Imberman (2009) construct the first explicit five-factor decomposition of the 76% rise in disability insurance (DI) recipiency from 1984 to 2003. Using sequential counterfactual simulations, cohort-tracking, and time-series regressions, they assign percentage shares of the growth to: changes in the age structure, rising DI-insured status (women's labor force participation, LFP), economic conditions (1991 and 2001 recessions), rising replacement rates (income inequality × progressive benefit formula), and liberalized medical eligibility criteria (the 1984 Disability Benefits Reform Act, DBRA). The framework over-explains the growth (total ≈ 106–119%), reflecting that health improvements among the near-elderly actually dampened rolls — absent these improvements, growth would have been even larger. The paper also projects continued DI growth, estimating an equilibrium stock of ~9.8M (from 6.2M in 2004).
"The more liberal medical eligibility criteria represent the most important factor for both men and women... explaining 53 percent and 38 percent, respectively, of the growth in DI receipt during the last two decades."
"Changes in health during the past two decades have slowed rather than added to the growth of the DI rolls. Absent these improvements, the growth in DI enrollment from 1984 to the present would probably have been even greater."
The five-factor decomposition framework is the paper's main contribution — it's more granular than Autor-Duggan (2003, 2006) and a useful predecessor to Ruffing (2014) and Pattison-Waldron (2013). The dominant finding (DBRA explains 38–53%) sits in tension with Ruffing's Social Security Administration (SSA) actuarial estimate that DBRA's long-run cost impact was negligible. The tension resolves partly by definition: Duggan-Imberman identify the DBRA as the correlate of the mental/musculoskeletal award surge (a reasonable but non-causal attribution), while Ruffing's estimate reflects SSA's formal scoring of the legislation's narrowly defined provisions. The methodology is transparent about its limitations (no simultaneous estimation, interaction effects not modeled), and the >100% total is honestly reported and well-explained. The projection section, written for a 2004 conference, was roughly validated: DI rolls peaked near 8.9M (2015) before declining, somewhat below the 9.8M equilibrium projection but in the right direction. The paper predated the recognition that baby-boom aging itself would reverse after ~2015.