I am estimating a model of the following form:

log(y) =  b1 x + b2 x^2 + b3 log(z1) + b4 z2

This is an econometric model with a focus on the impact of x x^2 (the inverted U relationship), where x ranges from 0 to 1. I included the other variables z1 and z2 just to be clear that there are other controls (some which are log-transformed and some not).

I get a clear inverted U relationship - slope check at end points and a marginal plot confirms that.

My question is the following: How can I say something about the change in x impacting y (not log(y))? This article (https://library.virginia.edu/data/articles/interpreting-log-transformations-in-a-linear-model) provides a nice way to interpret in case I don't have the x^2 term. How to do this correctly when I have the x^2 term?

For example, if I am able to do one of the following that will be wonderful:

  • If x increases by 0.1 from its current sample mean what will be impact on y?
  • If x increases by 10% from its current sample mean, what will be the impact on y?



1 Answer 1


I think the following approach will work. Comments and suggestions would be greatly appreciated.

Now, if I evaluate log(y1) = log(y) evaluated at x=x1 and log(y2) = log(y) evaluated at x=x2, then

log(y2) - log(y1) = -b1 x1 - b2 x1^2 + b1 x2 + b2 x2^2

Taking exponents on both sides and simplifying:

y2/y1 = Exp(-b1 x1 - b2 x1^2 + b1 x2 + b2 x2^2)

So, if I now substitute b1, b2 coefficients and select two points x1 and x2 (x2>x1, and both are smaller than the inflection point) in the RHS, can I then interpret that as the ratio between y2/y1?

Say, that the ratio is evaluated as 1.4, then can I state that moving x from x1 to x2, yields a 40% increase in y?

From my experiments with predicted values, this does work out. However, I am concerned that I am making any errors by not considering standard errors.


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.