Autori
Stengos, Thanasis
Prescott, David M.

Titolo
Testing for forecastable nonlinear dependence in weekly gold rates of return
Periodico
European University Institute of Badia Fiesolana (Fi). Department of Economics - Working papers
Anno: 1991 - Fascicolo: 31 - Pagina iniziale: 3 - Pagina finale: 24

We apply nonparametric kernel methods to estimate the conditional mean of weekly gold rates of return in order to examine whether the market overlooks usable nonlinear structure. Linear and ARCH-M models are fitted as parametric benchmarks. The BDS statistic of Brock, Dechert and Scheinkman (1987) is used to test for nonlinear dependence. We find evidence of nonlinearities in the residuals of the linear and ARCH-M models. However, kernel estimation of the conditional mean based on only four lags filters out all detectable nonlinear structure. Nevertheless, the kernel estimator has no forecasting ability whatsoever in 84 one-step-ahead out-of-sample forecasts. We conclude that the structure that the BDS is capturing in the parametric benchmark models pertains to higher moments and it does not help provide better forecasts of the conditional mean once captured. Hence, the Weak Market Efficiency Hypothesis is upheld.



Testo completo: http://hdl.handle.net/1814/370

Esportazione dati in Refworks (solo per utenti abilitati)