What is it about?
We illustrate some approaches to view interaction effects in binary choice (e.g., logit and probit) models graphically. Commonly reported marginal effects, computed either as an average or at mean values of other variables, may mask important insights that can be gained by employing these graphs to consider how the interaction effect may vary across the sample or for feasible values of other variables.
Featured Image
Why is it important?
The marginal effects of interaction terms may vary in sign and significance across observations in the sample (or at different combinations values for other variables). Hence, relying on conventionally reported marginal effects without consulting these graphing methods may lead to erroneous conclusions.
Perspectives
As an example, this paper considers transaction cost economics' hypothesized positive effect of the interaction of uncertainty and asset specificity on the use of contractual marketing instead of cash markets. Using the graphing methods, we find support for the hypothesis that is otherwise unapparent. Common empirical practices may be one factor contributing to the relatively lower level of support for this interaction than for its underlying variables when the interaction is omitted, documented by reviews of empirical transaction cost literature.
Dr Jason RV Franken
Western Illinois University
Read the Original
This page is a summary of: Graphical Illustration of Interaction Effects in Binary Choice Models: A Note, Journal of Agricultural Economics, December 2017, Wiley,
DOI: 10.1111/1477-9552.12257.
You can read the full text:
Contributors
The following have contributed to this page
