The importance of using a test of weak-form market efficiency that does not require investigating the data first

Are financial markets efficient? There are multiple tests for answering this question. Forming a hypothesis and testing should be done before looking at the data, i.e. without data snooping. However, the parameters used in the tests of the efficient market hypothesis are often not decided independent of the data. This paper investigates the consequences of not only this form of data snooping but also the issue of looking at multiple tests. The specific tests compared in this paper are the runs test, the autocorrelation test, and the variance ratio test.