We run an experiment in China, Paraguay and Uganda to determine how preferences for redistribution depend on the cause of the original inequality. Unlike previous studies, we move beyond the rudimentary luck/skill distinction in designing the activities subjects use to generate their income. First, we employ tasks that differentiate between two types of luck: (1) chance, in which the subject's payoff is the realization of a lottery chosen exogenously by the experimenter, and (2) risk, in which the subject herself chooses between lotteries with known payoff distributions. Second, we decompose skill into its constituent categories: (a) aptitude, as measured by an IQ test, and (b) effort, as measured by a computerized real-effort task that controls for baseline ability. We hypothesize that subjects will exhibit a preference for redistributing their group's income in order to achieve a more equitable distribution, and that the magnitude of this preference is decreasing with how much control subjects feel they have over their income level. Specifically, we hypothesize that redistribution levels will consistently be greatest in the chance treatment and least in the effort treatment, and that we can predict the likelihood of subjects redistributing more income from risk than from aptitude based on a priori knowledge about a country's average beliefs about locus of control. Though some of our results are only marginally statistically significant, we find that we can confirm our hypothesized universal attitudes toward chance and effort. Furthermore, Ugandan subjects, with on average an external locus of control, exhibit higher redistribution for aptitude relative to risk, whereas Chinese and Paraguayan subjects, with on average an internal locus of control, exhibit the opposite. Our results are related to the question of designing optimal tax policies when inequality arises, or is perceived to arise, from different sources.