A Step Towards Demystifying High-Beta Stocks Asset Pricing Puzzles: A New Bootstrap Pricing Error Test for High-Beta Stocks Under Conditional Correlation and Heteroskedasticity

Abstract

High-beta stocks seem to be an asset pricing mystery involving puzzles that have been intensively discussed in the most recent finance literature (Christoffersen and Simutin, 2017; Moreira and Muir, 2017). This papers derives novel implications for pricing high-beta stocks in the presence of dynamic correlations and conditional heteroskedasticity. First, it establishes that the widely-used asymptotic Gibbons, Ross, Shanken (1989) test statistic fails as an asset pricing test when applied to high-beta stocks because it dramatically overrejects the null hypothesis. Second, it proposes a new bootstrap procedure that considerably lowers the size distortions and that under certain scenarios generates the correct size in line with the stringent criterion for robustness (Bradley, 1978; Serlin, 2000).

Keywords: GRS test, Wald test, dynamic correlation, conditional heteroskedasticity

JEL Classification: C12, C15