Bootstrap confidence intervals for the simultaneous equations model under heavy-tailed contamination

A simulation study illustrates the good accuracy of the bootstrap bias-corrected accelerated confidence interval (BCa) by Efron [1] in the simultaneous equations model when the errors are generated from a normal distribution contaminated by a heavy-tailed distribution. When the sample size is small, the BCa confidence interval is shown to have a substantially smaller coverage error than the normal asymptotic interval, which is, however, often used in practice.