So-called "gravity choice models" estimate demand using revealed preference. For example, a vacationer who passes a nearby polluted lake to travel a more distant one tells us that she values the pollution reduction associated with the more distant lake by more the increased “travel time” that she “pays” for it. Similarly, a consumer who passes a lower-quality grocery store to visit a more distant one tells us that he values the higher quality shopping experience by more the increased travel time that he pays for it. In both of these applications, the opportunity cost of travel is estimated or assumed to be about $25/hour, close to the median wage.
For health care gravity choice models, however, consumers behave as if their travel costs are much, much higher. For example, a typical patient will not travel five minutes for elective surgery at a more distant hospital with much lower mortality risk. The implied opportunity cost of travel is several orders of magnitude higher than in these other contexts, e.g., as much as $100/minute. This means that competition is extremely localized, or that a merger between two nearby hospitals, located even a small distance away from a non-merging rival, can have enormous effects.
However, the real reason for the observed reluctance to travel may not be the patient's aversion to travel, but rather the preferences of the patient's physician.
DeveshRaval and Ted Rosenbaum at the FTC show this by estimating travel costs by comparing the change in hospital choice between the first and second births for women who move and switch hospitals. If the travel costs are all that is determining hospital choice then the woman should switch to a nearby hospital for her second child's birth. But the data show that the woman is often likely to go to hospitals that are not as close at second birth. The authors conclude that there must be that something other than distance, like physician preference, determining her choices.
Their corrected estimates of travel cost imply that demand falls by 5.4% for an extra minute of travel, about half of what it would be without the correction.
Bottom line: hospital competition is localized, but about 50% less localized than it would appear from looking at simple choice models.