Nonlinear Behaviour of Returns in an Emerging Stock Market

"It is generally recognized that volatility in the prices of securities is due, in part, to traders continuously revising their preference sets in response to the arrival of unanticipated information. In particular, traders may update beliefs about the value of an asset in response to information on both market microstructure and the macro-economy. We argue that the microstructure characteristics of the Trinidad and Tobago Stock Exchange (TTSE) are consistent with serial correlation, volatility clustering, and nonlinearity in stock returns. The underlying behavioural patterns arise because informed traders may posses ""long-lived"" information sets arising from poor dissemination and disclosure of market information. The paper has two objectives: (1) to investigate a price-volume relationship for stock returns, accounting for conditional heteroscedasticity and nonlinearity; and (2) to determine whether volume is a sufficient proxy for information flow. Our results show that simple linear autoregressive models of stock returns display significant nonlinearities, which can partly be explained by functional and variable misspecification in the functions describing the mean and variance of returns. In particular, we find that prices and volume are not sufficient statistics for the conditioning information set of traders because both volume and the real effective exchange rate are significant predictors for the distribution of stock returns."