9617925 Horowitz The bootstrap is a method of estimating the distribution of an estimator or test statistic by resampling one's data. It amounts to treating the data as if they were the population for the purpose of evaluating the distribution of interest. Under suitable conditions the bootstrap provides a better finite-sample approximation to the distribution of a test statistic and better finite-sample critical values than does first-order asymptotic theory. As a result, the differences between the true and nominal levels of tests and overage probabilities of confidence intervals are often much smaller using critical values obtained through the bootstrap than critical values based on first-order asymptotic theory. Bootstrap has demonstrated for a number of economic applications spectacular reductions in finite-sample distortions of the levels of tests. The proposed research develops methods for applying the bootstrap to several important problems in applied econometrics where the bootstrap may have substantial advantages over alternative methods but that cannot be handled satisfactorily using standard bootstrap theory. These are: a. Hypothesis tests based on non-smooth estimators; b. Bandwidth selection for semiparametric estimators; c. Hypothesis tests based on instrumental-variables (IV) estimation with weak instruments. ??
