ARMA-GARCH models and trade rules for stock returns

This paper deals with the analysis and modeling of daily returns time series from the Lisbon Stock Exchange. We conclude for the common presence of conditional heteroscedasticity and for the choice of an ARMA(1,1)-GARCH(1,1) t-Student model to represent the behavior of these long time series. Forecasts from this model are then chosen to implement a trade rule, which is compared to other filters, proving to outperform them in several cases.