The empirical anomalies related to the Capital Asset Pricing Model (CAPM), typified by the size, value, and momentum effects, lie at the center of multi-factor asset pricing models. Especially, the model by Fama and French (1992, 1993) incorporates the excess returns on two portfolios capturing the size and value premiums as the additional factors. They estimated and tested this three-factor model using twenty-five equity returns from portfolios sorted by stocks’ sizes and book-to market ratios. The research in empirical finance has been focusing on multi-factor asset pricing models since then, and mainly geared toward identifying asset pricing anomalies, thereby new pricing factors.
Finding a new factor typically begins with grouping stocks by a characteristic, such as size, book to market ratio, or past return performances. Then, econometric analyses follow, verifying if there exist significant, abnormal returns not explained by the incumbent pricing factors, and testing if a new model embedding an additional factor made from the anomaly variable is rejected. That is, empirical asset pricing involves the construction of panel data sets of returns, andthe ensuing statistical investigation of those data series with some economic restrictions.
In the paper, we develop a new framework and a new set of statistical tools for high frequency panels and use them to reexamine Fama-French regressions. Our approach utilizes some recent econometric research on models with high frequency observations. Fama-French regressions have still been analyzed largely within the classical regression framework. There are at least two dimensions that we may look into for a new opportunity using our approach. First, asset return data sets are available at several different frequencies, e.g., daily, monthly and yearly. However, very few attempts have been made to address the issue of how to use these data sets provided at multiple frequencies.
In modern financial markets, information flows almost in real time and assets are traded at high frequencies. Thus, a valid asset pricing model under the premise of well-functioning markets must delineate relationships between asset returns and pricing factors at the (high) frequency of market clearing. This implies that a proper integration of higher frequency models is needed to accurately estimate and test asset pricing models at lower frequencies.
Download
Evaluating Factor Pricing Models Using High Frequency Panels
