The Decomposition of Coefficients in Censored Regression Models: Understanding the Effect of Independent Variables on Taxpayer Behavior

This paper discusses the interpretation of coefficients in censored regression models. Although a substantial amount of empirical tax research employs censored regressions to model taxpayer behavior, the common interpretation of model coefficients overlooks the substantive information that coefficients can provide about the effects of the independent variables. This article presents a decomposition technique for a censored regression model with an upper bound. It shows that a change in an independent variable affects the dependent variable by changing the expected value of the dependent variable for cases not at the limit and by changing the cumulative standard normal distribution function associated with the limit The paper applies this interpretative technique to a recently published study. The paper concludes by noting that the information provided by this interpretative technique contributes to a more precise understanding of the effects of a given tax policy.