Daily Data is Bad for Beta: Opacity and Frequency-Dependent Betas

Abstract

A stock's market exposure, beta, is not the same when measured across different return frequencies. Sorting stocks on the difference between low and high frequency betas (dBeta) yields large systematic mispricings relative to the CAPM at high frequencies, but significantly smaller mispricings at low frequencies. This result occurs even in large and liquid stocks and remains after applying standard beta measurement corrections. We provide a risk-based explanation for the frequency dependence of betas and alphas by introducing uncertainty about the effect of systematic news on firm value (opacity) into an otherwise frictionless rational expectations equilibrium model with risk-averse investors. Empirically, we document a robust relationship between the frequency dependence of betas and proxies for opacity. Our results suggest that opacity poses significant challenges to using betas estimated from high-frequency returns. Contrary to the conventional wisdom that the CAPM can hold at high frequencies and more factors are needed to price assets at low frequencies, we show that the CAPM may be an appropriate asset pricing model at low frequencies but that additional factors, such as one based on opacity, are necessary at high frequencies.

Keywords: Asset pricing models, systematic risk, factor structure, return frequency, information, price discovery, beta measurement

JEL Classification: G10, G12, G14