@wwwojtekk Studying log(y) can go the wrong way. The most prominent example is the case of gravity models: https://t.co/HcedDfZtE0 Another case is fitting the Feldstein formula for net-of-tax income: https://t.co/EzueZgqWhB Different from what Bella…
The answer to that question is heteroskedasticity. See the seminal paper by Santos Silva and Tenreyro (2006). https://t.co/nf81Mhcsgq If the error term implied by this approximation does not follow a very specific (quadratic) form of heteroskedasticity,…
@jondr44 there's actually a well-known paper that is kinda about this question, https://t.co/X866X83zOl
"Log of gravity" OLS is biased when estimating log-linear models in the presence of heteroskedasticity #heteroscedasticity #elasticity https://t.co/93Wz9Y2JNz