Title: Adaptive estimation of semi-parametric partially linear predictive regression under heteroskedasticity
Authors: Jiti Gao - Monash University (Australia)
Hsein Kew - Monash University (Australia) [presenting]
Yundong Tu - Peking University (China)
Abstract: Adaptive estimation is considered in semiparametric partially linear predictive regression models with unconditional heteroscedasticity of an unknown form. We develop an adaptive semiparametric estimator weighted by a non-parametric variance estimator. The adaptive estimator is shown to deliver potentially large asymptotic efficiency gains over the conventional unweighted estimator. Monte Carlo simulations confirm this theoretical result. We implement the proposed estimation method by studying the in-sample predictability of US future stock returns using the commonly used financial variables as regressors.