Do Rising Gas Prices Decrease Obesity?

Acccording to Charles Courtemanche, the answer is yes:

A causal relationship between gasoline prices and obesity is possible through mechanisms of increased exercise and decreased eating in restaurants. I use a fixed effects model to explore whether this theory has empirical support, finding that an additional $1 in real gasoline prices would reduce obesity in the U.S. by 15% after five years, and that 13% of the rise in obesity between 1979 and 2004 can be attributed to falling real gas prices during this period. I also provide evidence that the effect occurs both by increasing exercise and by lowering the frequency with which people eat at restaurants.

I am skeptical with respect to the first pathway, exercise. Seth Roberts summarizes the research:

Garrow and Summerbell (1995, p. 1), after a review of the literature on exercise and weight loss, concluded that “aerobic exercise causes modest weight loss without dieting.” Fogelholm and Kukkonen-Harjula (2000, p. 106), after a similar review, concluded that “the effects of a prescribed exercise programme remain very limited.” A panel report about obesity treatment sponsored by the National Institutes of Health (National Heart,Lung, and Blood Institute, 1998) summarized clinical trials of exercise (mainly aerobic exercise) by stating that “10 of the 12 RCT [randomized clinical trial] articles reported a mean weight loss of 2.4 kg (5.3 lb) . . . or a mean reduction in BMI of 0.7 kg/m2” (p. 45) and even this small difference was an over-estimate. In the two omitted RCTs, the exercise group weighed slightly more than the control group. A recent experiment on the effects of exercise on weight loss, involving 201 overweight sedentary women, failed to find significant differences between the effects of different amounts of exercise (Jakicic, Marcus, Gallagher, Napolitano & Lang, 2003). The range of the effects of exercise was about 2% of the women’s average body weight.

On the other hand, the same paper suggests that the "restaurant" pathway is credible, albeit for other reasons than Courtemanche thinks.