Although incentives can be a powerful motivator of behavior when they are available, an influential body of research has suggested that rewards can persistently reduce engagement after they end. This research has resulted in widespread skepticism among practitioners and academics alike about using incentives to motivate behavior change. However, recent field studies looking at the longer-term effects of temporary incentives have not found such detrimental behavior. We design an experimental framework to study the dynamic behavior under temporary rewards, and find that although there is a robust decrease in engagement immediately after the incentive ends, engagement returns to a post-reward baseline that is equal to or exceeds the initial baseline. As a result, the net effect of temporary incentives on behavior is strongly positive. The decrease in post-reward engagement is not on account of a reduction in intrinsic motivation, but is instead driven by a desire to take a ‘break’, consistent with maintaining a balance between goals with primarily immediate and primarily delayed benefits. Further supporting this interpretation, the decrease in post-reward engagement is reduced by contextual factors (such as less task difficulty and higher magnitude incentives) that reduce the imbalance between effort and leisure. These findings are contrary to the predictions of major established accounts and have important implications for designing effective incentive policies to motivate behavior change.