Butcher and Park 2008 — Obesity, Disability, and the Labor Force

Summary

Butcher and Park investigate how the three-decade rise in obesity is related to the parallel rise in prime-age men's non-employment. Using an Oaxaca-Blinder decomposition on National Health Interview Survey (NHIS) data, they find that changes in obesity alone explain $3$–$13\%$ of the increase in men's non-employment (1984–85 to 2004–05), rising to $34$–$47\%$ when combined with changes in age, race, and ethnicity. The key null finding is that the propensity of the morbidly obese to report worse health, disability, or Social Security Disability Insurance (SSDI) applications has not significantly increased over time — the problem is compositional (more people in a persistently high-risk category), not a worsening obesity-employment relationship.

Key Claims

• Non-employment rise: Prime-age men's (25–54) non-employment rose from $10.3\%$ (1984–85) to $12.5\%$ (2004–05) — a $2.2$ percentage points (pp) increase. NHIS closely tracks Current Population Survey (CPS); National Health and Nutrition Examination Survey (NHANES) overstates growth by more than twofold (excluded from main analysis).
• BMI distribution shift: The median male in 1999–2002 would have been heavier than $\approx 75\%$ of the 1976–80 population. A male just on the cusp of obesity (75th percentile in 1999–2002) would have been heavier than $90\%$ of the earlier population. Variance in body mass index (BMI) has increased, especially at the heavy tail.
• Stable propensity null: From 1984–2005, the morbidly obese are consistently more likely to report musculoskeletal conditions, routine needs disability, and ever having applied for SSDI — but the difference between morbidly obese and normal-weight men has not statistically significantly widened. No supply-side or demand-side shock changed the obese/non-obese outcome gap.
• Decomposition: Obesity alone explains $3.4\%$ (under 1984–85 conditions) to $12.5\%$ (under 2004–05 conditions) of the non-employment rise. Adding age: $14$–$32\%$. Adding age + race/ethnicity with all pairwise interactions: $34$–$47\%$. The amount explained is similar whether evaluated under early or late period conditions — suggesting stable underlying relationships.
• SSDI diagnosis shift: Heart disease fell from $24.9\% \to 11.4\%$ of all SSDI awards (1981–2003); cancer from $16.3\% \to 9.4\%$; musculoskeletal disorders rose from $17.0\% \to 26.3\%$; mental disorders from $10.5\% \to 25.4\%$. The shift toward subjective, harder-to-verify diagnoses is consistent with post-1984 eligibility changes.
• Supply vs. demand indistinguishable: Because health, disability, and employment propensities all show the same stable pattern (no increase in the obesity gap), the paper cannot isolate demand-side factors (employer discrimination, health-cost avoidance) from supply-side factors (health deterioration, SSDI incentives). If health were stable but employment fell, that would have pointed to demand side.
• Contrast with Lakdawalla et al. (2004): B-B-G found obesity explains $50\%$ of disability rise for 18–29 year olds and nearly all for 50–59 year olds (1984–1996). Butcher-Park use a Oaxaca-Blinder approach rather than B-B-G's three-component decomposition, and extend the outcome from disability rates to non-employment.

Data and Method

• Data: NHIS 1984–2005 (main), NHANES 1976–80 and 1999–2002 (BMI distribution), March CPS 1962–2006 (labor force trends). Focus on men aged 25–54.
• Outcomes: (1) Non-employment (not working past 1–2 weeks); (2) self-reported musculoskeletal conditions; (3) routine needs disability; (4) ever applied for SSDI.
• Method: Linear probability models for health/disability/employment outcomes on weight category (underweight, overweight, obese, morbidly obese vs. normal weight), controlling for age, race/ethnicity, and interactions. Oaxaca-Blinder decomposition evaluates how much of the 1984–85 to 2004–05 non-employment rise is attributable to changes in characteristics ($X$ values) vs. returns to characteristics ($\beta$ values).
• BMI categories: Underweight $<18.5$; normal $18.5$–$24.9$; overweight $25$–$29.9$; obese $30$–$39.9$; morbidly obese $\geq 40$.

Concepts Introduced or Extended

• Obesity and Non-Employment

Entities Mentioned

• Kristin F. Butcher
• Kyung H. Park
• Darius Lakdawalla

Quotes

"What we have seen is that there are now more of the category of people—very obese people—who have always reported worse health outcomes, but not much evidence of an increase in the likelihood of reporting worse health outcomes among the very obese."

"Changes in age, race, and ethnicity—which may themselves be markers of changes in underlying health—explain a larger share of the increase in non-employment between 1984–85 and 2004–05 than do changes in obesity measures alone."

My Take

A clear, well-organized descriptive and decomposition paper. The "stable propensity" null finding is the most important result and is consistently replicated across three outcomes (musculoskeletal reporting, routine needs disability, SSDI application), lending it credibility. The Oaxaca-Blinder approach is standard, though as the authors note, unobserved characteristics correlated with obesity and non-employment could bias attribution.

The key limitation is identification: the paper is entirely descriptive/correlational. The obesity-employment relationship likely reflects both selection (less healthy people gain weight) and causation (obesity causes health problems). Without an instrument for BMI, causal claims about how much of the non-employment rise is "caused by" obesity remain imprecise. The $\approx 40\%$ figure for age/race/ethnicity + obesity is probably best interpreted as an upper bound on the compositional contribution to the trend.

The comparison with Lakdawalla et al. (2004) is useful — they found larger effects for disability (especially at extreme ages) — but the extension to non-employment is the novel contribution here. The paper is a good empirical complement to the DI-specific literature (Autor and Duggan 2003) that attributed declining labor force participation to SSDI incentive changes, since it shows underlying health deterioration (a supply-side factor) can explain a substantial share of the trend independent of programmatic changes.