3
$\begingroup$

Obviously, error terms can be correlated with explanatory variables. E.g. see this.

However, I came up with an argument why they cannot be, so I am wondering where the mistake in my argument is.

Assume we have a "true" linear regression model $$y=\beta_0+\beta_1x_1+\beta_2x_2+\epsilon$$

Where $E(\epsilon|x_1,x_2)=0$

But assume that instead we run the regression $$y=\beta_0+\beta_1x_1+\eta$$ Where $$\eta = \beta_2x_2 +\epsilon$$

Then $E(\eta|x_1)=E(\beta_2x_2 +\epsilon|x_1)=\beta_2\cdot E(x_2|x_1)$ Now, if we assume that the relation between $x_1$ and $x_2$ is purely linear, then we know that $E(x_2|x_1)=\gamma x_1$, so that $$E(\eta|x_1)=\beta_2\gamma x_1$$

But if this is true, then why can't we simply state the model

$$y=\beta_0+(\beta_1+\beta_2\cdot \gamma)x_1+\zeta$$ so that $E(y|x_1)=\beta_0+(\beta_1+\beta_2\cdot \gamma)x_1$, and $$E(\zeta|x_1)=0$$

In this adjusted model, the error term is uncorrelated with $x_1$, even though we have an omitted variable. So if we do the regression, we will have an unbiased estimator of $\beta_1+\beta_2\cdot \gamma$.

What is wrong with this argument? Is the only reason that this doesn't work simply because we cannot measure $\gamma$ and $\beta_2$?

$\endgroup$
2
  • $\begingroup$ Why do you have to assume any (or a linear one specifically) association between $x_1$ and $x_2$? $\endgroup$
    – IWS
    Nov 17, 2017 at 9:28
  • $\begingroup$ Because I think that if there is no correlation between the two, then the ommission of $x_2$ is not problematic in the first place, so then the whole question would be irrelevant. $\endgroup$
    – user56834
    Nov 17, 2017 at 9:37

1 Answer 1

4
$\begingroup$

You seem to have rediscovered linear projection coefficients.

You are right that $\zeta$ has zero conditional mean given $x_1$, and hence OLS consistently estimates the linear projection coefficient of $y$ onto $x_1$. In terms of your true model, you consistently estimate $\beta_1+\beta_2\gamma$. In terms of your true model, you are however interested in estimating $\beta_1$. So unless $x_2$ is irrelevant for explaining $y$ (i.e., $\beta_2=0$) or $x_1$ and $x_2$ are unrelated ($\gamma=0$), OLS won't estimate the parameter you seek to estimate.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.