Indirect Inference and Small Sample Bias — Some Recent Results

Abstract Macroeconomic researchers use a variety of estimators to parameterise their models empirically. One such is FIML; another indirect inference (II). form ‘informal’ whereby data features are ‘targeted’ by the model — i.e. parameters chosen so that model-simulated replicate closely. Monte Carlo experiments show in small samples prevalent macro data, both FIML informal II produce high bias, while formal II, which joint probability data- generated auxiliary maximised under simulated distribution, produces low bias. They also FII gets this bias from its power rejecting misspecified models, comes turn fact distribution restricted model-specified parameters, sharply distinguishing it rival models.