Tests of Informational Efficiency in Financial Markets : A Granger Causality Approach

In this study we examine a causal relationship between series of returns and traded volumes in high-frequency data. Our analysis is based on the methodology of Ghysels, Gourieroux and Jasiak (2000), who develop a qualitative framework in which dynamics of financial series are restricted to transitions between a finite number of states, represented by Markov chains with specific transition probabilities. Linear and nonlinear Granger causality tests are used to evaluate statistical causality between series of volumes and returns of various investment vehicles within the parametric and nonparametric frameworks, and for trading and calendar time specifications. Results demonstrate evidence of three causality types, unidirectional from volumes to returns and vice versa, and instantaneous causality. The frequency with which causal relationships are observed increases rapidly as the transaction grid becomes finer, allowing us to infer that while informational efficiency of financial markets holds, opportunities for economic gains may exist at high-frequencies