The Minimum Approval Mechanism (MAM) was introduced by Masuda et al.

(2014) as a mechanism aimed at mitigating free riding in the social dilemma context of a public good game. The MAM is a two-stage mechanism which theoretically achieves the socially optimum level of public good provision, according to various equilibrium concepts (e.g., backward elimination of weakly dominated strategies, level-k, or minimax regret). We study the robustness of this mechanism to the introduction of endowment heterogeneity. We explore, theoretically and experimentally, how endowment inequalities affect the effectiveness of the MAM at improving the level of provision. We find that the mechanism is still Pareto-improving under endowment heterogeneity, but that its efficiency diminishes as inequality is increased. Our experimental findings indicate a significant weakening of the mechanism under endowment inequalities, surpassing our theoretical predictions. A close examination of individual behaviors reveals a significant drop in contributions compared to the uniform case, prompted by even minor inequalities. Intriguingly, our findings challenge conventional assumptions by showing that inequality aversion drives contributions in a public good game with endowment disparities only under certain assumptions. We explore the impact of beliefs about the contributions of advantaged player as potential motivations through guilt aversion and Kantian preferences.