Evaluating public health interventions often requires predicting long-run outcomes from short- run behavioral responses. This paper examines whether commonly used behavioral measures can serve as reliable surrogates for long-term effects. Drawing on a large-scale field experiment on vaccination incentives in Sweden, I assess which short-term responses predict sustained behavioral change and which do not. The findings speak directly to debates about prediction, measurement, and external validity in experimental research, with implications for policy design and evaluation.