Interaction terms in a Logit Model

[Dillan Patel]

Hi,

I am studying the effect of trade unions on wages, and the corresponding impact on poverty incidence. So far, I have run a regression with 'poverty' as my binary dependent variable (calculated via an income measure), union (a binary variable) being my main independent variable, alongside many controls. The code for that is below:

logit poverty union i.SEX i.GORWKR i.HIQUL22D i.AGEEUL i.ETHUKEUL i. MARSTA i.MPNR02 i.INDC07M i.SC20MMJ i.FTPT i.JOBTYP, r
margins, dydx(*) atmeans post
//

If I want to see whether being a Male unionised individual increases/decreases this probability of being in poverty compared to a female unionised individual, would I interact Male with union? I have attempted in the code below:
logit poverty MALE##union i.GORWKR i.HIQUL22D i.AGEEUL i.ETHUKEUL i. MARSTA i.MPNR02 i.INDC07M i.SC20MMJ i.FTPT i.JOBTYP, r
margins, dydx(*) atmeans post


Hope this makes sense.

Thank you.

[George Ford]

male union male*union

see this for interpreting interaction terms in non-linear models.

https://journals.sagepub.com/doi/pdf/10.1177/1536867X1001000211
https://blog.stata.com/2016/07/12/effects-for-nonlinear-models-with-interactions-of-discrete-and-continuous-variables-estimating-graphing-and-interpreting/

you could run the model using ols rather than logit. with all categorical variables, I think even the predictions will be on the unit interval. marginal effects will be nearly identical, but the interpretation of the interaction term will be straightforward.