Robustness Checks In the Scaling and Modeling of Pay

A number of recent studies relied on nominal pay in estimating the effects of individual differences on pay. We show that this practice may lead to results that are substantially different from the results of logarithmic pay models, the commonly accepted models in the literature, and that the differences between the two types of models are considerable with regard to interaction effects, though less so with regard to main effects. We recommend that researchers should conduct robustness checks by estimating log pay models when estimating nominal pay models. We further suggest that previous research that relied on nominal pay models should be reexamined, and we identify studies involving interaction effects that are particularly likely to be sensitive to the differences between the two types of models.