Asymptotic and bootstrap evaluations of the size and power of test statistics

This thesis is concerned with Monte Carlo evaluations of test statistics in finite samples. Two types of Monte Carlo simulations are performed. One is based on comparing a test statistic with its asymptotic distribution and a second compares a test statistic with its bootstrap distribution.; This thesis first considers the finite sample properties of a test for overidentifying restrictions that is based on the Kullback-Leibler information criteria (KLIC), an alternative to the generalized method of moments (GMM). It is found that the KLIC-based test statistic does not solve the overrejection problem common to GMM but yields superior size-adjusted power.; Secondly, an heteroskedasticity-robust bootstrap test for an unknown structural break is proposed when the regressors can be nonstationary. The robust bootstrap test and its double bootstrap version essentially removed all size distortions.; Finally, the impacts of changing the trimming rule on the size and power of the tests for structural change are analyzed. The choice of the trimming rule has virtually no effect but the bootstrap test statistics have more power than their asymptotic counterparts. In all instances, empirical examples are studied using Canadian and U.S. macroeconomic time series.