0

I'm new to econometrics and was wondering when including an interaction term do I have to put each variable separately as control variables? I know this is the case for a difference in differences.

Y = a + B1 X1 + B2 X2 + B3 X1*X2 + e

However, I recently read a paper that had this regression equation

ln(imports) = a + B1 StringencyIndex + B2 
  CovidDeaths + B3 StringencyIndex * 
  HealthIndex + B4 CovidDeaths * HealthIndex 
  + e

Now, I'm curious if B3 and B4 are correct. HealthIndex is not included separately as a control variable and can we run the regression in this way as well?

kjetil b halvorsen
  • 77,844
chunguc1004
  • 513
  • 1
    You don’t have to do anything, but it’s typically advised to include the linear terms in order to capture any effects that the variable has on its own. It might help to include a citation for and a link to the paper so we can see what they did. For instance, if they did some kind of stepwise selection or screening based on p-values, I could believe that they wound up omitting such a feature. (That both techniques tend to be frowned upon by statisticians is really immaterial. People still do both.) – Dave Apr 10 '22 at 18:10
  • Please look at the linked question and answers. There's a good discussion of why omitting "main effects" isn't usually a good idea, although there are some circumstances when it can be OK. If that thread doesn't answer your question, please pose a question that gets directly to the issues that remain unaddressed. – EdM Apr 12 '22 at 14:17

0 Answers0