Modelling Non-Linear Cointegration in International Equity Index Futures

摘要

This paper describes a study of cointgration amongest international equity futures indices. Cointegration is an economic and finncial phenomenon which has been attracting increasing attention amongest both academics and practitioners over recent years. We intoriduce the concept of "conditional cointegration" and show how this provides both a framework and a practical motivation for the use of neural networks in financial modelling and forecasting. We apply a non-linear model selection methodology to the problem of modelling relative returens on the FTSE 100 as a non-linear function of a cointegrating residual, conditioned upon other economic and finanical indicators. In the final section we address the issue orf robustness to outliers and influential observations and show that combining a median estimator with the less robust mean-squared-error estimator. using a smple trading rule, almost doubles the out-of-sample risk/return performance of the system.