The existing literature on sub-game perfect risk-sharing suffers from a basic inconsistency. While a group of size "n" is able to coordinate on a risk-sharing outcome, it is assumed that deviating subgroups cannot. I relax this assumption and characterise the optimal contract among all...