In a novel experimental design we study dynamic public good games in which wealth is allowed to accumulate. More precisely each agent's income at the end of a period serves as her endowment in the following period. In this setting growth and inequality arise endogenously allowing us to address new questions regarding their interplay and effect on cooperation levels. We find that average cooperation levels in this setting are high (between 20-60% of endowments) and that amounts contributed do not decline over time. Introducing the possibility of punishment leads to lower group income, but less inequality within groups. In both treatments (with and w/o punishment) inequality and group income are positively correlated for poor groups (with below median income), but negatively correlated for rich groups (with above median income). There is very strong path dependence: inequality in early periods is strongly negatively correlated with group income in later periods. These results give new insights into why people cooperate and should make us rethink previous results from the literature on repeated public good games regarding the decay of cooperation in the absence of punishment.