Estimating beta: Forecast adjustments and the impact of stock characteristics for a broad cross-section

Abstract

Researchers and practitioners face many choices when estimating an asset's sensitivities toward risk factors, i.e., betas. Using the entire U.S. stock universe and a sample period of more than 50 years, we find that a historical estimator based on daily return data with an exponential weighting scheme as well as simple shrinkage adjustments yield the best predictions for future beta. Adjustments for asynchronous trading, macroeconomic conditions, or regression-based combinations, on the other hand, typically yield very high prediction errors and fail to create market-neutral anomaly portfolios. Finally, we document a robust link between stock characteristics and beta predictability.

Predicting Systematic Risk With Macroeconomic and Financial Variables

“…Some studies augment this approach by modeling betas as an unobserved time-varying parameter following a certain process, such as an AR model or a random walk, and/or as functions of some exogenous state variables (e.g., Shanken 1990;Ferson and Schadt 1996;Adrian and Franzoni 2009). Other studies consider different adjustments, such as adjustments for asynchronous trading (e.g., Scholes and Williams 1977;Dimson 1979), shrinkage (e.g., Vasicek 1973;Karolyi 1992;Cosemans et al 2016), and weighting schemes (Hollstein, Prokopczuk, and Simen 2019), to the Fama-MacBeth approach to obtain better beta estimates, which might in turn provide better beta forecasts. Another popular approach is based on the theoretical developments in modeling and forecasting time-varying conditional variances and covariances (e.g., Bollerslev, Engle, and Wooldridge 1988).…”

“…On the other hand, there is a large literature pointing to the advantages of using high-frequency data in estimating and forecasting betas. For example, studies such as Andersen et al (2006), Ghysels and Jacquier (2006), Hooper, Reeves, and Ng (2008), Papageorgiou, Reeves, and Xie (2016), Hollstein, Prokopczuk, and Simen (2019), and Cenesizoglu et al (2019) show that one can obtain better beta estimates and/or forecasts using daily data. Other studies such as Bollerslev, Patton, and Quaedvlieg (2016) and Hollstein, Prokopczuk, and Simen (2020) show that one can improve on these estimates/ forecasts using even higher frequency intraday data.…”

“…We find that bond market and portfolio-specific variables better hedge the market risk of these portfolios. Hollstein, Prokopczuk, and Simen (2019) analyze the effect of several empirical choices on the performance of beta estimates and forecasts. They find that a historical estimator based on daily return data with an exponential weighting scheme as well as simple shrinkage adjustments provides the best beta forecasts.…”

“…The two papers closest to ours are Ghysels and Jacquier (2006) and Hollstein, Prokopczuk, and Simen (2019). Ghysels and Jacqiuer model quarterly realized betas as an ARX model to analyze the power of these variables in explaining the in-sample variation of realized betas.…”

Can International Market Indices Estimate TASI’s Movements? The ARIMA Model

“…Unsurprisingly, specific movements of market indices may influence investors to buy, sell, or hold, especially if these occurrences take place at the right time. Therefore, an accurate estimation of the coming index movement may help investors to perform better while taking the right investment decision [13][14][15][16][17][18].…”