Abstract
Numerous studies have modeled learning as a function of cumulative experience as
measured by the number of times an individual or firm has used a given
technology. These models are characterized by the useful—though
limiting—assumption of uniformity across units of experience. More recent work
has departed from this assumption of uniformity and has suggested that learning
may result not simply from repeated use but also from expanding the difficulty
or variety of situations to which a technology is applied.
Using data on the performance of cardiac surgeons in New York State over a thirteen-year period, we consider how a surgeon’s performance in terms of patient mortality is affected by both the mean and variance of the difficulty of his or her recent cases. We find preliminary evidence that surgeons with a higher mean difficulty of recent cases exhibit lower risk-adjusted levels of patient mortality. Further, we find that surgeons with high mean difficulty and low variation in difficulty perform significantly better than those with high mean difficulty and high variation. Our findings suggest a mechanism by which the entrance of angioplasty—a competing technology that limited cardiac surgeons to a narrower range of more difficult cases—may have improved the performance of cardiac surgeons.