4
$\begingroup$

Similar questions have been asked before, but all of them focus on the dummy or interaction term.

Say run an OLS regression on the model:

$\ln( housePrice )= \beta_1 \times pollutionLevel + \beta_2 \times D_N + u$

where $D_N$ is a dummy that indicates whether there is a school nearby the house.

The interpretation for $\beta_1$ and $\beta_2$ are simple enough, but in the model:

$ln( housePrice ) = \beta_1 \times pollutionLevel + \beta_2 \times D_N + \beta_3 \times pollutionLevel \times D_N + u$

it's not so clear.

I understand the interpretation of $\beta_3$, but how does the interpretation of $\beta_1$ change? Is $\beta_1$ now just the effect of pollutionLevel when there isn't a nearby school, or is that totally wrong?

Thanks in advance for any help!

$\endgroup$
5
$\begingroup$

Yes, that is correct in your case. A good way to convince yourself of that statement follows.

Say you want to find the impact of the pollution level on the log of house prices.

$$ \dfrac{\partial \ ln(housePrice)} {\partial \ pollutionLevel} = \beta_1 + \beta_3 \times D_N $$

where the impact of the pollution level on the percentage change in house prices when there is no school nearby $(D_N=0)$ is simply $\beta_1$.

$\endgroup$
  • 2
    $\begingroup$ (+1) Isn't it the symbol $\partial$ more common notation than $\delta$ to denote partial derivatives? $\endgroup$ – javlacalle Apr 12 '15 at 20:43
  • $\begingroup$ Changed it now! I had a moment of weakness and forgot how to type the partial derivative symbol in LaTex format $\endgroup$ – Giaco.Metrics Apr 12 '15 at 21:32
2
$\begingroup$

One way to generally look at this is via marginal effects as in @Giaco.Metrics' response. Another general technique is a distinction of cases.

For $D_N = 0$ (no school nearby, reference group), your equation simplifies to:

$\ln(housePrice) = \beta_1 \times pollutionLevel + u$,

i.e., you have intercept 0 and slope $\beta_1$ in the reference group.

For $D_N = 1$ (school nearby), you get

$\ln(housePrice) = \beta_1 \times pollutionLevel + \beta_2 + \beta_3 \times pollutionLevel + u\\ = \beta_2 + (\beta_1 + \beta_3) \times pollutionLevel + u$,

i.e., you have intercept $\beta_2$ and slope $\beta_1 + \beta_3$ in the school group. So $\beta_2$ is the difference in intercepts and $\beta_3$ the difference in slopes.

$\endgroup$
  • $\begingroup$ +1 for the "graphical" explanation of OLS using slopes and intercepts. $\endgroup$ – Giaco.Metrics Apr 12 '15 at 21:38

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.