We study the efficiency of dynamic portfolio choices using the nonparametric methods of Dybvig (1988) and Post (2003). We compare a dynamic portfolio task against an equivalent static Arrow-Debreu problem under two alternative environments: (1) nonpooled with 2^T terminal states and (2) pooled with T+1 unique terminal states. The results suggest that, within each environment, efficiency is lower in a static format and when the number of final states is larger. In the nonpooled dynamic task, which allows for path dependent strategies, we find that a form of stop-loss strategy drives efficiency losses.