Sébastien Laurent

This presentation first introduces the Autoregressive Conditional Beta (ACB) framework of Blasques, Laurent and Christian Francq (2024, JoE), for modeling time-varying regression coefficients in (financial) time series. The ACB is a score-driven model allowing conditional betas to evolve dynamically with a structure similar to GARCH models. We then discuss the penalized QMLE of ACB models for identifying zero, constant, and time-varying betas as recently introduced by Francq, Laurent and Schnaitmann (2025). Finally, we extend the framework to the Realized ACB model as proposed by Artemova, Francq and Laurent (2026). The RACB jointly exploits returns and realized beta measures. Monte Carlo evidence and financial applications illustrate the benefits of dynamic, penalized, and realized-information-based beta modeling.