# Reverse causality between predictors, interactions and results' interpretation

In my model I have clear reverse causality between two predictors. I am mainly interested in esitmating the interaction between the two predictors mentioned above.

I am wondering how this form of reverse causality could affect the interpretation of the relative coefficients, especially the interaction.

    Y = B0 + B1 X1 + B2 X2 + B3 X1*X2 + e


directions in causations:

   X1 ---> Y
X2 ---> Y
X1 <--> X2

• If you know exactly the causality direction (from one variable to another one), it is sufficient to use standard regression tools. – Karel Macek Mar 29 '17 at 9:39
• If you are asking for the ways how to determine reverse causality, common sense may help, as cited here: statisticshowto.com/reverse-causality – Karel Macek Mar 29 '17 at 9:40
• As stated in the question the reverse causality is known and between the predictors, the dependent variable is related to the predictors in a single direction of causation. Thus, I am not worried about common issues of endogeneity due to reverse causality causing correlation among one or more predictors and the error term (I edited the question with clearer information on the issue). – Caserio Mar 29 '17 at 9:44