Time-Varying Beta Estimation: A Comparison of DCC-GARCH and Rolling-Window Methods in Turkish Industry Portfolios

This study empirically compares the accuracy of two common methods for estimating time-varying betas in Turkish industry portfolios: rolling-window OLS regression and the DCC model. Using daily returns from 2004 to 2024, the methods are evaluated based on their alignment with CAPM predictions, specifically the insignificance of Jensen’s alpha and the significance of the market risk premium. Findings show that despite its complexity, the DCC model does not outperform the rolling-window approach. The rolling-window approach produces insignificant Jensen’s alpha estimates for more industries and yields slightly higher mean and t-statistics for the market risk premium. These findings challenge the view that rolling-window estimators are inefficient due to assuming beta constancy within short windows and suggest that the DCC model’s reliance on multiple constant parameters imposes a rigid structure that may hinder its adaptability to evolving market conditions. This study contributes to the literature by directly comparing these two widely used methods and highlighting the importance of carefully considering model assumptions when estimating time-varying betas.