We analyze the impact of heterogeneous beliefs in economies where agents have recursive preferences of the Kreps-Porteus type and differ in their subjective beliefs about the distributions of future quantities. We show that the set of Pareto optimal allocations is given as a solution to a planner's problem with Pareto weights evolving according to an Ito process. In a Markov environment, belief heterogeneity can be incorporated without the need for additional state variables. Using a parsimonious economy, we show that agents with distorted beliefs are not driven out of the market for an empirically relevant range of parameters when preferences are nonseparable across time. Return decomposition reveals nonlinearities in contributions of belief heterogeneity across payoff horizons.