The Efficacy of Neural Networks in Predicting Returns on Stock and Bond Indices

摘要: This paper uses two recently developed tests to identify neglected nonlinearity in the relationship between excess returns on four asset classes and several economic financial variables. Having found some evidence of possible nonlinearity, it was then investigated whether predictive power these variables could be enhanced by using neural network models instead linear regression or GARCH models. Some relationships explanatory large stocks corporate bonds found. It also that are conditionally efficient with respect models, but outperform if performance measures used. In resonance results reported for forecasts bonds, whereas is not statistically significant small intermediate-term government bonds. difference persists even when individual used comparison.