How reliable are microsimulation results?: An analysis of the role of sampling error in a U.K. tax-benefit model

Abstract We assess the statistical reliability of microsimulation models in two ways: by comparing simulated outcomes with survey ‘actuals’, and by calculating asymptotic confidence intervals for a variety of summary measures. The confidence intervals we derive take account of re-weighting for differential survey response, and also the effects of imposing revenue-neutrality. They are calculated for a version of one of the most widely-used U.K. tax-benefit simulation models. The results suggest that baseline simulations are reasonably accurate, but that some widely-used measures of the effects of policy changes may be very imprecise estimates of population effects.