Interpreting the outcomes of two-part models

Abstract

Interaction effects capture the impact of one explanatory variable x1 on the marginal effect of another explanatory variable x2. To explore interaction effects, so called interaction terms x1x2 are typically included in estimation specifications. While in linear models the effect of a marginal change in the interaction term is equal to the interaction effect, this equality generally does not hold in non-linear specifications (AI, NORTON, 2003). This paper provides for a general derivation of marginal and interaction effects in both linear and non-linear models and calculates the formulae of the marginal and interaction effects resulting from the Two-Part Model, a commonly employed censored regression model. Drawing on a survey of automobile use from Germany, we illustrate several subtleties inherent to the substantive interpretation of interaction effects gleaned from non-linear models.