In this paper we study the relationship between wealth, income distribution and growth in a game-theoretic context in which property rights are not completely enforceable. We consider equilibrium paths of accumulation which yield players utilities that are at least as high as those that they could obtain by appropriating higher consumption at the present and suffering retaliation later on. We focus on those subgame perfect equilibria which are constrained Pareto-efficient (second best). In this set of equilibria we study how the level of wealth affects growth. In particular we consider cases which produce classical traps (with standard concave technologies): growth may not be possible from low levels of wealth because of incentive constraints while policies (sometimes even first-best policies) that lead to growth are sustainable as equilibria from higher levels of wealth. We also study cases which we classify as the "Mancur Olson" type: first best policies are used at low levels of wealth along these constrained Pareto efficient equilibria, but first best policies are not sustainable at higher levels of wealth where growth slows down. We also consider the unequal weighting of players to trace the subgame perfect equilibria on the constrained Pareto frontier. We explore the relation between sustainable growth rates and the level of inequality in the distribution of income.