Eric Ghysels

Overview

Eric Ghysels is an economist (University of North Carolina at Chapel Hill, Kenan-Flagler Business School and Economics Dept.) specializing in financial econometrics, high-frequency data analysis, and mixed-frequency methods. He is a co-inventor of the MIDAS regression framework and has made foundational contributions to realized volatility measurement, intraday seasonality, and stochastic volatility theory.

Key Contributions / Features

• MIDAS Regression (Ghysels-Santa-Clara-Valkanov 2004, 2006) — Co-developed the Mixed Data Sampling regression framework, which parameterizes distributed lags using Beta or Almon polynomial weights (2 parameters) to forecast low-frequency outcomes from high-frequency regressors. Applied to volatility forecasting in Journal of Econometrics 131 (2006): 59–95 with Santa-Clara and Valkanov; showed that realized power $\tilde{P}^{(m)} = \sum|r_{t-j/m}|$ dominates realized variance, squared/absolute returns, and daily range as a volatility predictor; MSE ratios vs. ABDL ARFI(5,d) benchmark: 0.606–0.912 in-sample, 0.714–0.897 out-of-sample. See MIDAS Regression.
• Risk-Return Tradeoff (Ghysels-Santa-Clara-Valkanov 2005 JFE) — "There Is a Risk-Return Tradeoff After All," Journal of Financial Economics 76(3): 509–548. Co-demonstrated a positive conditional risk-return tradeoff using MIDAS-estimated conditional variance, reversing earlier negative findings that used parametric GARCH variance proxies.
• Stochastic Volatility Survey (Ghysels-Harvey-Renault 1996) — Co-authored the comprehensive handbook chapter on stochastic volatility in Handbook of Statistics Vol. 14; standard reference for SV model taxonomy and estimation methods.
• Intraday Seasonality and Volatility Persistence (Andersen-Bollerslev 1997a) — Co-authored with Andreou (2002) a study of rolling-sample volatility estimators; also cited in foundational intraday periodicity work; Forsberg-Ghysels (2004) WP explains theoretically why absolute returns outpredict squared returns as volatility proxies (Bernstein inequality for power variation).
• Realized Power Variation Motivation — The empirical finding that $\tilde{P}^{(m)}$ beats $\tilde{Q}^{(m)}$ is grounded in Barndorff-Nielsen–Shephard (2003/2004) theory: realized power excludes jump variation, yielding a higher-persistence predictor of the continuous integrated variance component.

Related

• Pedro Santa-Clara
• Rossen Valkanov
• MIDAS Regression
• Realized Volatility
• Ghysels-Santa-Clara-Valkanov (2006)
• Neil Shephard