Health insurance and obesity

The insurance market is bedevilled with numerous classic market failures arising from adverse selection and moral hazard. Now, in the latest moral hazard problem, Freakonomics draws attention to an NBER working paper that claims to establish a positive causal link between obesity and health insurance.

Jay Bhattacharya, Kate Bundorf, Noemi Pace, and Neeraj Sood tested the proposition that body weight is influenced by insurance coverage and found strong evidence that being insured increases body mass index and obesity. They trace this moral hazard due to the fact that the obese do not pay for their higher medical expenditures through differential payments for health care and health insurance.

In a pointer to the utility of incentives for healthy behaviours, they write, "These effects are larger in public insurance programs where premiums are not risk adjusted and smaller in private insurance markets where obese might pay for incremental medical care costs in the form of lower wages."

I am inclined to believe that the aforementioned study offers interesting lessons highlighting the pitfalls inherent in randomized program evaluations. While intutively persuasive, this hypothesis appears to be not borne out by macro-level evidence - Europe with higher health insurance coverage, and that too publicly funded, has much less obesity than the US with its limited and mostly private insurance; and within the US too, it has been well established that the under-privileged, blacks and those more likely to be outside the insurance net suffer disproportionately more from the obesity problem than those insured.

At a theoretical level, the two instrumental variables (IV) used to study the causal relationship between health insurance and obesity (the dependent variable) - percentage of workers employed in private firms (for the causal effect of private health insurance) and the generosity of state Medicaid coverage (for the causal effect of public health insurance) - may not offer accurate reflection of the causal relationship. They would translate into the following

1. States with larger share of workers employed by private firms have more of its residents who are insured. States with larger share of workers employed by private firms have less obesity.
2. States with more generous Medicaid coverage had more people insured. States with generous Medicaid coverage had more obesity.

Now both these "instrument leads to treatment (insurance)" and "instrument leads to outcome (obesity)" statements do not reveal the full truth. First, states with a larger share of workers employed by private firms may be a more progressive state (in terms of its ability ot attract private investments) and therefore may be populated by people who may have less propensity to become obese. Second, the states with more genrous Medicaid coverage are likely to be the poorer and less developed states, which are therefore more likely to be populated by alarger share of people who are more vulnerable to becoming obese. It therefore appears that the IV's are closely corelated with the other variables in the relationship, thereby concealing the true nature of the relationship. The IV estimates may have over-estimated the effects of insurance on bodyweight and obesity.