Health care demand price elasticies are often estimated from samples conditioned to include only sike people. This paper shows that not only may such estimates be statistically biased, but even when properly estimated they can only be interpreted as short-run price responses.