This paper provides an answer to the question of why agents make self-serving decisions under moral hazard and how their self-serving decisions can be kept in check through institutional arrangements. Our theoretical model predicts that the agents' power and the manner in which they are held accountable jointly determine their propensity to make self-serving decisions. We test our theory in the context of financial investment decisions made under moral hazard using others' funds. Across three studies, using different decision-making tasks, different manipulations of power and accountability, and different samples, we show that agents' power makes them more likely to behave in a self-serving manner under moral hazard, but only when the appropriate accountability mechanisms are not in place. Specifically, we distinguish between outcome and procedural accountability and show that holding agents accountable for their decision-making procedure reduces the level of self-serving decisions under moral hazard and also curbs the negative consequences of power. Implications for decisions under moral hazard, the psychology of power, and the accountability literature are discussed.