Numerical reasoning research points to the value of considering multiple causal mechanisms that give rise to a statistic. However, much evidence suggests that people prefer simple causal explanations, and discount one causal explanation when another is salient. This led us to ask when and how people integrate causal alternatives. We presented causal reasons for changes in statistics (e.g., traffic fatalities) that were either directly relevant or irrelevant to an action, and asked participants how willing they were to take each action. By presenting two causal reasons in counterbalanced order, we were able to consider the extent to which participants integrated causes in their action decision. We found broad support for discounting but also some evidence for integration of explanations, and will discuss circumstances that lead people to integrate multiple explanations rather than discount.