Aggregation of Test Statistics: An Evaluation of the Evidence on Contracting and Size Hypotheses; Statistics vs. Economics

An exact chi-square test, an asymptotic normal test, and a Bayesian procedure are used to evaluate the ability of contracting and size theories to explain the data. It is concluded that 6 variables - managerial compensation, leverage, size, risk, interest coverage constraints, and dividend constraints - have significant explanatory power. These tests are more powerful than tests emanating from any one study because they aggregate results across studies. Leftwich comments that Christie's tests improve the statistical precision of statements about the body of empirical tests of economic consequence theories. However, the basic problems in that literature are not primarily statistical. The contribution of Christie's tests is limited because the fundamental question of the relation between the empirical regularities and the theory remains unexamined.