There are widespread differences in total factor productivity across producers in the U.S. and around the world. To help explain these variations, we devise a general test for misallocation in input choices – the underuse of effective inputs and overuse of ineffective ones. Misallocation implies that conditional on total input use, the return to using a particular input is not zero (a positive return implies underuse, and a negative return implies overuse). We measure misallocation across hospitals, where inputs and outputs are better measured than in other industries. Applying our test to a sample of 1.6 million Medicare beneficiaries with heart attacks (of which 436 thousand were admitted by ambulance), we reject the hypothesis of productive efficiency; moving a patient from a 10th percentile to a 90th percentile hospital with respect to misallocation, holding spending constant, is predicted to increase survival by 3.1 percentage points. With misallocation accounting for as much as 25 percent of the variation in hospital productivity, our results suggest that how the money is spent, rather than how much money is spent, is central to understanding productivity differences both in health care, and in the rest of the economy. (Joint with Amitabh Chandra -Harvard University and NBER, Carrie Colla -Dartmouth .)