13.3 Regression assumptions

exogeneity assumption: In a regression context, the assumption that the variables in the model (or perhaps just our treatment variable) is uncorrelated with the error term

• basically the same conditions as for identifying a causal effect
• if something is still in the error term and is correlated with $X$, we haven’t closed this path
• other way to say this: $X$ is correlated with $\epsilon$ and so is endogenous
• endogeneity problem: bias –> estimate gives the wrong answer on average
• here: omitted variable bias