Dear Statalist,

I have a fairly simple doubt, forgive me if it's rather basic, but I was wondering if you could clear something for me. When we have reverse causality (as a result of Endogeneity), how does this affect the standard errors? Are they overestimated? Does it depend on the sign of the bias?

More concretely, I have a simple regression where 2 variables influence one another in an upwards bias way, and I was wondering up to what extent can I still use my results (since I'm not really using the coefficient values, only if they are significant or not). Since my coefficients will be overestimated, will the t-stat also be overestimated, and thus my variables be "overly significant", so to speak?

I've tried to search for some references on this matter but I didn't find much to go from except that "endogeneity affects and biases estimates".

Any help you could give me, I would be beyond grateful.


Merci,
Anthoine