Accuracy versus Incentives A Trade-Off for Performance Measurement

Health-care providers are increasingly subject to measurement of noisy performance signals. I show that shrinkage estimation, commonly used to improve the accuracy of performance measures, blunts performance incentives. I study this phenomenon in the setting of hospital performance measurement for heart attack mortality. Analysis of Medicare claims shows that shrinkage estimation substantially dilutes incentives, particularly for smaller hospitals, whose measured performance increases by only 20-40 percent of true performance improvements. Alternative methods like increasing the time-span of measurement can improve accuracy without reducing incentives.