When wages increase with work experience, estimation of standard labor supply models that assume exogenous wage formation suffers from omitted variable bias and produces downward-biased estimates of the intertemporal elasticity of substitution (IES). We test this theory in a novel way. Using a large data set of the daily labor supply decisions of Florida fishermen, we identify a sample of highly-experienced, near-retirement fishermen, for whom the returns to experience are negligible and the standard model is a close approximation. Using this sample, we estimate an IES of 2.7, more than twice estimates that ignore the role of learning-by-doing.