We conduct public goods experiments in which subjects earn real money, are subject to a deduction that is distributed to their group, and can lie about their earnings. Performing better on incentivized real effort tasks results in more cheating. Cheating is only weakly correlated with the gains from cheating. We observe no costly cheating — subjects will not pay to cheat. The correlation between ability and cheating persists when earnings are associated with luck. We see the same performance effect in experimental treatments with more redistributive public goods. This correlation persists when we control for other-regarding preferences. High ability subjects cheat more in a classic die game in which they privately report the results of tossing a die. We conclude that ability is positively correlated with cheating.