AAA Spark Meeting of the Regions

Why Subsample-Based Proxies Should Not Be Used as Dependent Variables

• In Event: Statistical Methods

Abstract

Researchers often use subsample-based proxies for constructs such as earnings variability, stock return skewness, and accrual persistence. To study how these distributional properties of a Y variable vary with conditioning variables X, researchers typically regress the corresponding subsample-based proxy (e.g., skewness over a firm-specific rolling-window subsample) on X. However, we show that this standard approach can cause severe biases and produce false findings. We develop alternative methods that address these biases by directly modeling the relevant conditional distributional property of Y for each observation as a function of X. Simulations confirm that our methods perform well even in scenarios in which the standard method is severely biased. Our methods are transparent, robust, and can be implemented in a few lines of code. Use of our methods changes major prior findings.

Authors

• Sudipta Basu, Temple University

• Dmitri Byzalov, Temple University