Using Monte Carlo experiments, we examine the performance of indirect inference tests of DSGE models in small samples, using various models in widespread use. We compare these with tests based on direct inference (using the Likelihood Ratio). We find that both tests have power so that a...