Say I have an augmented growth regression, where my Y is GDP growth, and my Xs are the classical MRW variables, + international aid + corruption + the interaction term between the two.
Basically, I'm interested in seeing if lower corruption increases the effectiveness of international aid, which means I'd like to see a positive sign on my interaction term.
However, I also think I should see a negative sign on corruption, as the higher corruption, the lower should growth be.
Now, my questions:

  1. Is this possible as I describe it? Can I see a negative component but a positive interaction term? In which case, how should I interpret it?
  2. Or is there a reason the coefficient on corruption should be positive, and still indicate that higher "institutional quality" has a positive impact on growth by itself and also in interaction with aid?

1 Answer 1


Ignore the other variables. Let $g$ be growth, $C$ corruption and $A$ aid. You look at

$$g = \alpha C + \beta A + \gamma A\cdot C$$

You ask, "does lower corruption increases the effectiveness of foreign aid?"

For this you have indeed to examine the cross-partial derivative

$$\frac {\partial g}{\partial A\partial C} = \gamma$$

This is "how growth is affected at the margin by Aid, if corruption changes"... So if lower corruption increases the effectiveness of Aid, $\gamma$ should be negative, not positive...

...if your corruption index is mapped as "the higher the value of $C$ the higher the corruption".

If your "corruption index" is mapped in reverse (and so in reality it measures "non-corruption"), the anticipated sign reverses also.


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.