Robust Subgame Perfect Implementation

We revisit robustness of the Moore-Repullo mechanism for subgame perfect implementation to a small amount of incomplete information. Contrary to the findings of Aghion et al. (2012) that the mechanism cannot robustly implement a non-monotonic social choice function in a interdependent-value setting, we show that a slightly modified Moore-Repullo mechanism robustly implements any social choice function in a private-value setup. We also construct a two-stage mechanism that robustly implements any social choice function in a unique pure-strategy truth-telling equilibrium.