Can prior voluntary redistribution improve coordination? We theoretically show that distributive preferences, forward induction and signalling all imply that it can. We then experimentally test our predictions by allowing subjects to redistribute part of their endowment before playing a battle of the sexes game. To identify whether the redistribution option increases coordination, and why, we also run experiments with no redistribution and forced redistribution. Our results show that the redis- tribution option does indeed significantly increase coordination. Disentangling the reasons why, we find that behaviour is most consistent with distributive preferences and one-step of forward induction (rather than signalling or two-steps of forward induction).