Hi all,

I want to interact a dummy variable (CSR) with the cash flow. However, the interaction term seems to have a multicollinearity issue with the individual effect of the cash flow. When I remove the individual effect of cash flow, the coefficient, direction, and significance level become accurate. Nonetheless, the inclusion of the cash flow in the model leads to changing the direction of the interaction term.

I tried to utilize the mean-centered values of the variables to have the interaction term, but it didn't resolve the issue. As the interaction deals with only two variables, i think PCA isn't a viable option either. Above all, I can't remove any of the variables.

The only way that resolves the issue of multicollinearity is by using the inverting value of the cash flow in the interaction while maintaining the absolute value of the cash flow as an individual effect in the model.

Is this an acceptable way of dealing with the problem? If yes, how would i interpret the coefficient? Please.