A neural network model to exploit the econometric properties of Austrian IPOs

Applies cointegration and Granger (1969) causality analyses to specify linear and neural network error-correction models for IPOX_ATX (Initial Public Offerings indeX for the Austrian Traded indeX). We use the significant relationship between IPOX_ATX and the Austrian stock market index ATX to forecast IPOX_ATX . For prediction purposes, we apply augmented feedforward neural networks whose architecture is determined by sequential network construction with the Schwartz (1978) information criterion as an estimator for the prediction risk. The results suggest that trading schemes based on the forecasts significantly increase an investor´s return as compared to buy-and-hold or simple moving-average trading strategies