Time-Varying Coefficient Estimation in SURE Models. Application to Portfolio Management*. (20th March 2019)

• Record Type: Journal Article
• Authors: Casas, Isabel; Ferreira, Eva; Orbe, Susan
• Is Part Of: Journal of financial econometrics. Volume 19:Number 4(2021)
• Page Start: 707
• Page End: 745
• Publication Date: 2019-03-20
• Subjects: asset pricing -- five-factor model -- nonparametric -- SURE -- time-varying; C01 -- C03 -- C14; Capital market -- Law and legislation -- Periodicals; Financial institutions -- Law and legislation -- Periodicals; 338.544205
• DOI: 10.1093/jjfinec/nbz010
• Languages: English
• ISSNs: 1479-8409

Abstract

Abstract: This paper provides a detailed analysis of the asymptotic properties of a kernel estimator for a seemingly unrelated regression equations model with time-varying coefficients (tv-SURE) under general conditions. Theoretical results together with a simulation study differentiate the cases for which the estimation of a tv-SURE outperforms the estimation of a single regression equations model with time-varying coefficients. The study shows that Zellner's results cannot be straightforwardly extended to the time-varying case. The tv-SURE is applied to the Fama and French five-factor model using data from four different international markets. Finally, we provide the estimation under cross-restriction and discuss a testing procedure.