Financial Market Efficiency Tests

This paper provides a selective survey of the voluminous literature on tests for market efficiency. The ideas discussed include standard autocorrelation tests, multi-period regression tests and volatility tests. The formulation and estimation of models for time-varying volatility are also considered. Dependence in second-order moments plays an important role in implementing and understanding tests for market efficiency. All of the reported test statistics and model estimates are calculated with monthly data on value-weighted NYSE stock prices and dividends. The distributions of the test statistics under various alternatives, including fads and bubbles, are illustrated through the use of Monte Carlo methods. In addition to the standard constant discount rate present value model, we postulate and simulate a new fundamental price relationship that accounts for the time-varying uncertainty in the monthly dividend growth rates. Allowing the discount rate to be a function of the time-varying uncertainty in the dividend process results in a simulated fundamental price series that is broadly consistent with most of the sample statistics of the actual data.