The DI (Disability Insurance) replacement rate is the ratio of an applicant's expected DI cash benefit (and, when relevant, the Medicare benefit that accompanies DI enrollment after a 24-month waiting period) to their pre-disability earnings. Because the DI benefit formula is progressive — it replaces a higher fraction of earnings for lower-wage workers — the replacement rate is mechanically higher for workers at the bottom of the earnings distribution than for those at the top.
The 2006 paper distinguishes two separate channels through which the progressive benefit formula generates rising replacement rates for low-skill workers, even without any legislative change:
Channel 1 — Bracket effect: The Primary Insurance Amount (PIA) formula's bend points are indexed to the SSA national average wage index (AWI). As real wages at the bottom stagnate below the AWI trend, a larger share of a low earner's AIME falls into the 90% and 32% replacement brackets. In effect, wage divergence between the national average and low-skill workers mechanically pushes more of their AIME into the most-generous bracket.
Channel 2 — Earnings history effect: AIME itself is computed from historical earnings indexed to the AWI. A worker whose real earnings have not grown still sees their AWI-indexed AIME rise in real terms — creating a growing gap between their indexed benefit base and their current earnings, further inflating the replacement ratio.
These two channels compound: the bracket effect raises the rate applied to AIME; the earnings history effect raises the AIME base relative to current earnings. Together they explain why replacement rates rose sharply for low-wage workers without any statutory change to the PIA formula.
Between 1979 and 1999:
Because the denominator (actual wages) fell while the numerator (benefit, indexed to the rising SSA wage series) rose, replacement rates for low-wage workers increased sharply — without any legislative change to the benefit formula:
Autor and Duggan (2003) — 1979 vs. 1999:
| Earnings percentile, male aged 55–61 | Cash replacement rate 1979 | Cash replacement rate 1999 | Total (incl. Medicare) 1999 |
|---|---|---|---|
| 10th percentile | 52% | 74% | |
| 75th–90th percentile | modest rise | modest rise | +4–8 percentage points (pp) only |
Autor and Duggan (2006) — 1984 vs. 2002 (Table 2), including Medicare and net fringe benefits:
| Earnings percentile, male aged 50–61 | Replacement rate 1984 | Replacement rate 2002 | Change |
|---|---|---|---|
| 10th percentile | 68% | 86% | +18 pp |
| 90th percentile | negligible |
The 2006 figures are higher than the 2003 figures because they include net fringe benefits foregone — employer-provided fringe benefits that DI enrollment replaces. Medicare per recipient doubled in real terms from $3,259 (1979) to $7,700 (2005), independently adding pp to the effective replacement rate for 10th-percentile workers by the mid-2000s.
A total replacement rate exceeding 100% means that a low-wage worker at the 10th earnings percentile receives more in DI cash + Medicare than they earned in their pre-disability job — making DI financially dominant over continued low-wage employment.
Replacement rate comparisons across studies require care because two different denominators are in common use:
For the medium earner in 2004: policy model rate = 42% (career-average denominator); AIME-based rate = 48%. The 6-percentage-point gap is purely a measurement artifact.
Three additional factors further complicate comparisons: (1) actual workers have earnings gaps (men: 6 zero-earning years on average; women: 13), which lower their AIMEs and raise their AIME-based rates; (2) most workers claim before Normal Retirement Age (NRA) (74% of men in 2003), incurring actuarial reductions that lower actual benefits. These three effects happen to cancel for the median earner, making the policy model's 42% and actual beneficiaries' 42% identical — by coincidence. Low earners and women have systematically higher actual replacement rates than the policy model implies.
Implication for DI research: The same measurement distinction applies to DI. Studies reporting DI replacement rates as a fraction of AIME (e.g., Autor and Duggan 2003, 2006) will show higher rates than those using career-average earnings as denominator. The steep range reported for low-earning 55–61 year old men uses AIME; the same benefits expressed against career-average earnings would produce lower stated rates.
Butler, Johnson, and Baldwin (1995) provide a complementary finding using WC rather than DI data. They find that the WC replacement rate discourages initial return to work — consistent with the standard moral hazard story — but has no significant effect on post-return employment stability. Among Ontario workers with permanent impairments, higher replacement rates reduced the probability of initial return, but among those who did return, the replacement rate had no statistically significant effect on whether workers achieved single-absence success vs. multiple-absence cycling.
This entry/stability distinction implies that the moral hazard concern about benefit generosity is real at the application and initial return margin (which Autor and Duggan document for DI) but irrelevant to the post-return stability margin (which is dominated by employer accommodation behavior). Policy focused exclusively on cutting benefits to encourage return misidentifies the mechanism of post-return failure.
The replacement rate mechanism is a real, mechanically verifiable force. However, Pattison and Waldron (2013) and Liebman (2015) argue it explains far less of the aggregate DI enrollment growth than Autor and Duggan claim. Their critiques:
The replacement rate is a compelling explanation for educational and earnings gradients in DI receipt; it is less compelling as an explanation for aggregate program growth.