An Empirical Test of the Theory of Efficient Markets of Stock Prices

Abstract

A structure of the statistical tests motivated by Cromwell, Labys&Terraza[23]has been used to build linear and nonlinear predictability models.Most importantly, the variance ratio test and that of AR-GARCH model is used to test the dual hypotheses of the random walk and efficiencyin stock markets. Whilein all or nothing condition of market efficiency, the variance ratio tests show weak signs of predictability and in contrast to the AR-GARCH model that shows strong signs of predictability. Testing efficiency over time shows that price-fluctuations between periods of predictability and unpredictability and theseare not correlated through indices. This study then contributes to the empirical evidence that the efficient market hypothesis should not be an all or nothing condition but be stated as a time varying condition where prices fluctuate between periods of efficiency and inefficiency. It is found that market microstructure can cause problems for certain measuring frequencies and a sufficiently risk averse investor may be happy to pay a premium to avoid any unforecastable asset price volatilities as inLeroy[11]andLucas[12]. Three random walk models also do not prevent questioning the validity of predictability of stock prices.