Proportional treatment effects for count response panel data: effects of binary exercise on health care demand.

We define conditional and marginal treatment effects appropriate for count data, and then conduct an empirical analysis for the effects of exercise on health care demand using panel data from the Health Retirement Study. The response variables are office visits to doctors and hospitalization days, and the treatments of interest are light and vigorous exercises. We found that short-run light exercise increases health care demand by 3-5%, whereas long-run light exercise decreases it by 3-6%. We also found that short-run vigorous exercise decreases health care demand by 1-2%, whereas long-run vigorous exercise decreases it by 1-3%. However, many of these numbers are not statistically significantly different from zero. These findings suggest that it will be difficult to reduce health care cost much by encouraging people to do more exercise--at least in the short-run.