摘要：In this paper we investigate the possible presence of nonlinear dynamics for stock index returns and trading volume in the Chilean stock market. To capture any nonlinear behavior of the series we estimate smooth transition autoregressive (STAR) models and then test them against the linear alternatives. As a complement to this univariate approach, we use Markov-Switching Vector Autoregressive (MS-VAR) models to investigate the empirical relationship between both variables. The results clearly show that the Chilean Stock Market is characterized by the presence of nonlinear patterns in both series (trading volume and stock returns) as well as in their joint relationship. The presence of nonlinearities is a key issue in testing the Efficient Market Hypothesis (EMH), according to which stock returns and trading volume should not be related. Previous research on the efficiency using data from the Chilean stock market, based on linear models, support the hypothesis. However, the nonlinear patters found in the data are a clear signal of misspecification problems in a testing procedure based on a linear approach
