[jlarsen]

You really ought to read the paper.

1. They do control for demographics.

2. They do account for temperature, by comparing geographically close cities with different activity inequality levels.

3. They do account for "cost". They repeatedly compare cities and countries with roughly the same income distributions.

Also, as far as I can tell, the paper draws _exactly_ the opposite conclusions as what you state. In no way do they claim that rich people exercise more, nor do they claim that demographics have a strong correlation with exercise.

From what I read, they come to two conclusions:

-There is a correlation between activity inequality and obesity.

-There is a negative correlation between activity inequality and walkability.