Autore: Skouras, Spyros
Titolo: Risk Neutral Forecasting
Periodico: European University Institute of Badia Fiesolana (Fi). Department of Economics - Working papers
Anno: 1998 - Fascicolo: 40 - Pagina iniziale: 1 - Pagina finale: 61

This paper develops statistical and computational tools for modelling returns forecasts to be used by a risk neutral investor. Any forecast with the same sign as the conditional mean optimises the loss function derived from this agents' decision problem, so the class of optimal predictors is rather broad. We exploit the fact that optimal forecasting in this context can be seen as an extension of binary quantile regression in order to provide consistent estimators for optimal predictors. Further properties of these estimators are explored using simulations and favourable comparisons with least squares procedures are made. Unfortunately, our estimators are difficult to compute but an optimisation algorithm tailor-made for this purpose is provided. Our results provide a statistically valid method for selecting certain types of 'investment rules' according to popular optimality metrics.




Testo completo: http://www.iue.it/ECO/WP-Texts/ECO98-40.pdf

Esportazione dati in Refworks (solo per utenti abilitati)