Robust Generalized Empirical Likelihood for Heavy Tailed Autoregressions with Conditionally Heteroscedastic Errors

Abstract

We present a robust Generalized Empirical Likelihood estimator and confidence region for the parameters of an autoregression that may have a heavy tailed error, and the error may be conditionally heteroscedastic of unknown form. The estimator exploits two transformations for heavy tail robustness: a redescending transformation of the errors that robustifies against innovation outliers, and weighted least squares instruments that ensure robustness against heavy tailed regressors. Our estimator is consistent for the true parameter and asymptotically normal irrespective of heavy tails.

Keywords: Empirical Likelihood, autoregression, tail trimming, tail estimation

JEL Classification: C13, C22