Quantifying the Beta Estimation Bias and its Implications for Empirical Asset Pricing

45 Pages Posted: 4 Aug 2022 Last revised: 23 Mar 2023

See all articles by Timo Wiedemann

Timo Wiedemann

University of Münster - Finance Center Muenster

Date Written: August 2, 2022


Trading is not continuous, leading to different trading times for different assets. While this microstructure effect is known to bias covariance estimates between single stocks and the broad market, it has been hard to quantify the bias without restrictive assumptions in real empirical settings due to the lack of broadly available transaction-level data. This paper uses transaction data from the TAQ database, which provides information about the time of the last trade (TTC) for each stock and day. I show that the relative bias in a stock’s beta linearly increases in this measure of asynchronicity and that betas can be underestimated by up to 50%. Testing the widely used method of Dimson (1979), I show that while the approach decreases the bias in the beta estimate, the number of lags typically used in empirical studies only accounts for a small fraction of it. More precisely, many studies restrict themselves to only include three lags of the independent variable leaving a bias which still amounts to 30% for less frequently traded stocks. These findings translate into important implications for the empirical asset pricing literature and I highlight three cases in which the bias affects results.

Keywords: Measurement error, Asynchronous trading, Microstructure noise, Bias correction, Liquidity, Idiosyncratic volatility

JEL Classification: C1, C18, G12

Suggested Citation

Wiedemann, Timo, Quantifying the Beta Estimation Bias and its Implications for Empirical Asset Pricing (August 2, 2022). Available at SSRN: https://ssrn.com/abstract=4179365 or http://dx.doi.org/10.2139/ssrn.4179365

Timo Wiedemann (Contact Author)

University of Münster - Finance Center Muenster ( email )

Schlossplatz 2

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
PlumX Metrics