Cawley Moran and Simon 2009 — The Impact of Income on the Weight of Elderly Americans

Summary

Uses the Social Security (SS) Benefits Notch — a policy accident that gave cohorts born 1915–1917 permanently $\approx 7\%$ higher SS benefits — as an instrumental variable (IV) for unearned income, to estimate the causal effect of income on body weight among elderly Americans. Drawing on National Health Interview Survey (NHIS) 1990–1996 data ($N \approx 70{,}000$), the IV estimates find no significant effect of income on body mass index (BMI) or on any clinical weight category (underweight, healthy weight, overweight, obese) for either sex. The confidence intervals are tight enough to rule out economically meaningful effects, particularly for men. This is the first published causal estimate of income on weight for the elderly, complementing Schmeiser (2008) which found small positive income-weight effects for young low-income women using Earned Income Tax Credit (EITC) variation.

Key Claims

• IV result: income has no detectable effect on weight. All second-stage coefficients are statistically indistinguishable from zero for both males and females across every weight outcome.
• Tight confidence intervals rule out practically meaningful effects:
  • Men: a permanent $\$1{,}000$/year income increase will not change weight by more than $2/3$ of a pound. Income elasticity of BMI bounded at $|\varepsilon| < 0.06$.
  • Women: bounds slightly wider — weight change no more than $1.3$–$1.4$ pounds. Income elasticity of BMI bounded at $|\varepsilon| < 0.14$.
• The ordinary least squares (OLS) correlation between income and obesity is negative for women (higher income → lower obesity) but is biased by endogeneity (weight affects earnings → SS benefits) and confounding (time preference, education).
• The Notch instrument is strong: partial $F$-statistics of $24.2$ (women) and $28.8$ (men), well above the Stock-Wright-Yogo threshold of $10$. Pre-retirement BMI in 1976 shows no systematic difference between Notch cohorts and controls, supporting instrument validity.
• The permanent nature of the windfall strengthens the test: treatment cohorts received elevated SS benefits for every year of retirement ($>10$ years by the observation window), so estimates capture the long-run steady-state effect of a persistent income change on weight — a more powerful test than a one-year windfall.
• Null result undermines WHO's income claim: the World Health Organization (WHO) attributed part of the global obesity epidemic to rising incomes, but a permanent $6$–$7\%$ income increase for elderly Americans had no measurable effect on obesity.
• Elderly obesity is a major fiscal concern: Medicare spends $35\%$ more on obese individuals from age 70 onward — an additional $\$39{,}000$ in lifetime costs; obesity prevalence among men aged 60–74 rose from $8.4\%$ (1960–62) to $35.8\%$ (1999–2000).

Concepts Introduced or Extended

• Obesity and Non-Employment

Entities Mentioned

• John Cawley
• John Moran
• Kosali Simon

Quotes

"Our IV estimates yield little evidence that income affects weight or the probability of being underweight, healthy weight, overweight, or obese among elderly Americans."

"For a man who is 5 ft 10 in tall, the extra income will not lower or raise weight by even two-thirds of 1 pound."

My Take

A clean, tight null result on a question with clear policy stakes. The SS Notch instrument is well-suited to this question: it generates large, permanent income differences across nearly identical cohorts, the first-stage is strong, and the pre-treatment placebo holds. The main external validity concern — that surviving elderly are a selected, habit-entrenched sample — is acknowledged honestly. The result matters for Disability Insurance (DI)/obesity policy: if income transfers don't reduce obesity, policies designed to reduce obesity-driven disability via income support alone will not work. It also complements the Cawley-Ruhm (2011) synthesis which notes this paper alongside Schmeiser (2008) as the only high-quality causal estimates of income on weight. The contrast — no effect for elderly unearned income, small positive effect for young women's earned EITC income — is consistent with Lakdawalla-Philipson theory predicting that earned income has different weight effects than unearned income (via physical activity).