Dear STATA experts,

I run an OLS regression including an interaction term.

Code:
 
 reg DepVar IndepVar1 IndepVar2 IndepVar1*IndepVar2 Controls, vce(cluster GGISO)
As there is one control variable (Tax), which has way less observations than all the others and in order to increase sample size, I run the regression twice. First, I run the regression with all controls excluding Tax. Second, I include Tax as control.

In the first regression, the interaction term is highly significant. In the second, Tax is highly significant, but the interaction term isn't. I've checked for high correlation between Tax and any of the other variables, but didn't find high correlation.

Is there any statistical reason, why this might occur?

Thank you very much and best regards! :