Keeping it within bounds: Regression analysis of proportions in international business

Abstract

International business researchers commonly estimate proportions, percentages, rates, or fractions – so-called “proportional dependent variables”. In this paper, we posit that two regression strategies are particularly pertinent to the international business field: Tobit and fractional regression. Reviewing recent international business research, we find that, while fractional regression is rarely used, analyses from Tobit regression are often incomplete or erroneously interpreted with consequences for the validity of the reported results. Accordingly, we clarify how researchers should choose between Tobit and fractional regression and interpret their results. We present insights based on simple simulations and data examples with associated Stata code and a decision tree for choosing between types of models for use with proportional dependent variables.