Economic magnitudes within reason

Abstract

A common method of calculating economic magnitudes is to multiply the regression coefficient of the variable of interest by its sample standard deviation. This method is often problematic in finance settings when researchers use granular fixed effects. We show that in many recently published finance papers and for many common finance variables, the sample standard deviation is much larger than the within-group variation that identifies the regression coefficient, and that within-group changes of this magnitude are rare. Without additional assumptions, this common approach can significantly inflate the economic magnitude of the identified effect and impact the comparison of effects among different variables of interest. We recommend using within-group measures of variation to improve the interpretation of economic magnitudes in this setting.