A change in payment mechanism for inpatient care from per diem to per episode creates two incentives – a marginal and an average price effect – to change length of stay. The decrease in marginal price per day to zero should reduce the length of stay, while an increase in average price per inpatient stay should increase the length of stay. This study uses data from a natural experiment to estimate both marginal and average price elasticities, and to test whether the length of stay falls after the introduction of prospective payment in a sample of 8509 severely mentally ill patients. We estimate that the marginal price elasticity is zero, but the average price elasticity is between 0.16 and 0.20. The results were generally robust for short- and long stayers, and for persons admitted early and late after the change in payment mechanism. The model controlled for hospital fixed effects and individual random effects. Copyright © 2002 John Wiley & Sons, Ltd.