Abstract
When comparing assets, differences in sensitivity to economic variables are highly relevant, yet tests for such differences are absent from traditional studies. We therefore examine a range of tests for parameter differences across regression equations of asset returns. Simplistic approaches, which ignore conditional heteroskedasticity and/or serial correlation, suffer test distortions in simulations calibrated to asset returns, and more ‘robust’ methods also distort, for different reasons, but among these are the best methods. The tests suggest dramatic and time-varying differences between small and large firm sensitivity to market risk, the default premium and the term structure.
COinS