1
$\begingroup$

I've been working through some econometrics worksheets recently. I came to a problem that I was hoping to get some help with. Assume we have a model,

(1)     yi = βo + β1chemi + β2disti + εi
and we know that
E[ εi | chemi, disti ] = 0.
Further, assume that we only have data to estimate the regression using the following model:
(2)     yi = βo + β1chemi + γi.

Then, on average, what will be the estimate of β̂1 if we exclude disti from the regression, in terms of the parameters from model (1), and why?

Many thanks.

$\endgroup$
  • 2
    $\begingroup$ Google "omitted variable bias". $\endgroup$ – Dimitriy V. Masterov Sep 29 '17 at 3:03
0
$\begingroup$

Since you only have one explanatory variable in the model you estimate, you have $\hat{\beta}_1=\hat{c}ov(y,chem)/\hat{v}ar(chem)$

According to the correct model this gives $\hat{\beta}_1=\hat{c}ov(\beta_1 chem,chem)/\hat{v}ar(chem)+\hat{c}ov(\beta_2 dist,chem)/\hat{v}ar(chem)$

which provides the bias

$\hat{\beta}_1-\beta_1=\beta_2\hat{c}ov(dist,chem)/\hat{v}ar(chem)$

It is zero only if dist and chem are uncorrelated or $\beta_2=0.$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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