Health Mismatch Index

Definition

The Health Mismatch Index (HMI) is an occupation-level measure of the share of workers who report at least one health-related difficulty with a task that is required in their occupation. It was constructed by Rutledge, Zulkarnain, and King (2019) by linking Survey of Income and Program Participation (SIPP) health topical module questions (1996–2008 panels, four cross-sections: 1997, 2002, 2005, 2010) to the Occupational Requirements Survey (ORS), a Bureau of Labor Statistics (BLS)/Social Security Administration (SSA) survey of required physical and cognitive abilities across $\approx 20{,}000$ occupations. The Index operationalizes — at the population level — the same health-occupation fit concept that the SSA Sequential Determination Process formally assesses at Steps 4 and 5: can a worker perform the requirements of their own occupation (Step 4) or any occupation (Step 5) given their health limitations?

Key Ideas

Matched ability pairs: 12 SIPP health difficulty variables are matched to 12 ORS occupational requirement variables — lifting 10 lbs and 25 lbs, standing/sitting for one hour, stooping/crouching/kneeling, reaching overhead, pulling/pushing, fine manipulation, climbing stairs, near visual acuity, verbal communication, hearing in one-on-one conversation, hearing on the telephone.
Threshold rule: An occupation is coded as requiring a given ability if its ORS share-of-workers-requiring-that-ability exceeds the median across all occupations. Three universal requirements (near visual acuity, verbal communication, hearing in conversation) are assigned to all occupations.
Three severity thresholds: HMI(1+), HMI(2+), HMI(3+) — share with at least 1, 2, or 3 requirement mismatches. Results are reported for all three; the 1+ version is the main index.
Occupation-level construction: The analysis collapses SIPP data to $178$ unique occupations with at least $20$ workers in at least one panel; small occupations are merged with similar ones.

Key Findings (Rutledge, Zulkarnain, and King 2019)

$\approx 7\%$ of U.S. workers have at least one health mismatch with their current occupation's requirements (mean $6.65\%$ , 1997–2010). $\approx 2.6\%$ have two or more; $\approx 1.2\%$ have three or more.
Most common mismatches: lifting 25 lbs. ( $2.3\%$ ), standing for one hour ( $1.7\%$ ), hearing in conversational settings ( $1.5\%$ ).
High-HMI occupations: licensed practical nurses/licensed vocational nurses (LPNs/LVNs) ( $14.6\%$ ), teacher assistants ( $13.8\%$ ), personal care aides ( $13.8\%$ ), maids/housekeeping cleaners ( $11.6\%$ ), nursing/home health aides ( $10.8\%$ ), cooks ( $10.5\%$ ). Health care and education dominate — not traditional blue-collar work.
Earnings and hazard gradients: Lower-wage occupations and more hazardous occupations have higher HMI. High-performance/problem-solving occupations have lower HMI — mismatched workers are forced out faster.
The HMI declined from $7.4\%$ (1997) $\to$ $6.1\%$ (2010) — despite workforce aging. The decline is concentrated in the 1-mismatch group; the 2+ and 3+ shares were roughly stable.
Predicts Social Security Disability Insurance (SSDI) receipt: In occupation-level Tobits, $+10$ pp HMI $\to$ $+0.45$ pp share receiving SSDI in next 16 months (marginal effect $0.045$ , $p < 0.01$ ). Nearly doubles the mean ( $0.49\%$ ). Effect is concentrated in workers $50$ – $64$ ( $0.099^{***}$ ) vs. $18$ – $49$ ( $0.022$ , not statistically significant [n.s.]).
Predicts health-related job exit: Significant only at 2+ and 3+ severity thresholds — not at 1+. Single-requirement mismatches may be accommodatable; multi-requirement mismatches are not.
Age concentration: $\approx 70\%$ of mismatched workers are $50+$ , consistently across panels.

How It Works

The ORS reports, for each occupation, the share of sampled employers whose workers are required to perform each ability. The SIPP health topical module asks respondents whether they have difficulty with a matched set of tasks (not whether those difficulties affect their work specifically — a limitation). After collapsing to occupation-year cells, the paper regresses SSDI receipt (or health-related job exit) shares on the HMI plus controls: O*NET (Occupational Information Network)-derived indices (hazardous environment, high performance/problem-solving, social skills/teamwork), automation share, occupational demographics (age, sex, education, race/ethnicity, marital status), employer-sponsored insurance coverage, median earnings, and median family income. Tobits with $[0,1]$ bounds are used because the outcome means are low.

Why It Matters

Operationalizes Steps 4–5 at scale. The HMI is the only population-level measure that maps directly to the vocational assessment logic SSA applies at the individual claim level. Prior research on Disability Insurance (DI) eligibility either focuses on Step 3 (Listings-based) or uses coarser health proxies.
Supply-side constraint on DI enrollment growth. The HMI declined 1997–2010 while the DI rolls were growing. This implies that worsening health-job fit was not a driver of roll growth in this period; the growth occurred despite improving occupational health fitness. See DI Growth Decomposition.
Identifies target occupations for intervention. High-HMI occupations (health care workers, teacher assistants, custodial/food service workers) are observable and targetable for retraining, accommodation mandates, or DI application outreach — the jobs where the pipeline from mismatch to SSDI is most active.
Validates the ORS as a research tool. The ORS (fielded by BLS in partnership with SSA beginning in the early 2010s) is better suited to DI eligibility research than the Dictionary of Occupational Titles (DOT, last updated 1991) or O*NET, because it directly measures employer-reported required abilities using SSA-aligned criteria.

Open Questions

The 16-month outcome window likely understates cumulative mismatch-to-SSDI conversion, since the onset-to-application lag often exceeds two years. What is the longer-run elasticity?
The 2013 HMI appears to have spiked to $\approx 11\%$ in the non-comparable 2014 SIPP redesign. If real, this would suggest the Great Recession reversed the earlier improvement. Future comparable data would be needed to adjudicate.
Can the HMI be constructed using administrative records (SSA earnings data + ORS) rather than SIPP, enabling longitudinal individual-level analysis rather than cross-section occupation-level analysis?
Does the ORS vary sufficiently across similarly titled occupations at different firms to capture within-occupation heterogeneity? The current design averages across all employers in an occupation code.

Sources

Rutledge Zulkarnain and King 2019 — The Relationship Between Occupational Requirements and SSDI Activity