How to interpret coefficient in log-log regression for percentage increase in x larger than 1%

[Noemi Seng]

Dear community,

In a log-log regression, beta gives the elasticity. Hence, a b is approx. the %-increase in y for a 1%-increase in x. My question is: how can I calculate the % increase in y for a 100% increase in b (I need this for a specific example)? Is it simply 100*b? I have found varying answers to this when looking it up. One is that the exact formula is: (exp(b*ln(2))-1)*100, with ln(2) for a 100% increase, ln(1.10) for a 10% increase and so on. I'm however really unsure whether this is correct. I would appreciate any help!

Best
Noemi

[Noemi Seng]

Maybe Jeff Wooldridge or Clyde Schechter have an idea?

[George Ford]

https://nickchk.substack.com/p/my-unnatural-war-against-the-natural

[Nick Cox]

Cross-posted at https://stats.stackexchange.com/ques...-increase-in-x

Please note our longstanding request to tell us about cross-posting (FAQ Advice #8).

[Noemi Seng]

Dear Nick Cox
oh thank you very much for the advice! I wasn't aware of it. Always happy to learn.

Best,
Noemi

[Nick Cox]

So that means you have posted 89 times and never read the FAQ Advice, as each new message prompt asks you to do?

Reply not expected.

[Noemi Seng]

I did read the FAQ advice but wasn't aware of this one advice. I'm human. Let's be all nice to each other.