We build a dataset based on 23 experiments that introduce heterogeneous endowments into linear public good games. We use it to measure the effect of inequality on cooperation. This method allows an investigation of a large panel of inequality scenarios, with maximum representativeness in terms of the strength of inequality and design features. It offers the possibility to study the effect of the strength of inequality, a distinctive feature of our paper compared to the past experimental literature which has focused mainly on the existence of inequality. We also explore the contribution gaps between the relatively rich and relatively poor in heterogeneous groups. We discuss the interaction of time (dynamics) and punishment with inequality. We find that not only the presence, but also the strength of inequality has a negative impact on cooperation, but that the marginal effect becomes less negative as the level of inequality increases. We find that the rich contribute more than the poor in absolute amounts, while the poor contribute more as a proportion of their endowment. Both these gaps increase with the strength of inequality. Finally, punishment strongly attenuates the effect of inequality on aggregate cooperation, but has contrasted effects on the contribution gaps between the rich and the poor. There is no significant effect of inequality on the dynamics of contributions.