Country Fixed Effect - Effect of Independent on Dependent Variable

[Paulo Alexandri]

Hello everyone,

for my Master Thesis, I am analyzing the effect of financial crises on earnings management. Herefore, I want to use an OLS regression including the earnings management proxies as the dependent variable (Modified Jones Model etc.) and a crisis indicator (0 in non-crisis and 1 in crisis periods) as the independent variable. In addition, I included other independent variables such as leverage and ROA to control for firm characteristics. My panel data includes 18 countries across 28 crisis periods from 1990 to 2018 (unbalanced). I am only interested in the overall relationship between the crisis indicator and the earnings management proxies. Now I am wondering if it is appropriate to use country fixed effects to control for the effect of different regulatory settings etc. on earnings management. OR if I should use interaction terms or separate regressions for each country. In the end, the only thing that matters for me is if in general earnings management decreases or increases during crises.

Thank you in advance.

Paulo

[Clyde Schechter]

First a pedantic digression: you have observational data, so you cannot control anything. You can adjust for regulatory effects, etc. I realize that it is common to refer to this as "controlling for..." but it is an abuse of language.

If you want to model a single overall effect of crises on earnings management, then there is no role for interaction terms. Just add country-level fixed effects and you are done. Interaction terms would give you separate estimate of the earnings management:crisis relationship for each country--which is what you say you don't want.

[Paulo Alexandri]

Hello Mr Schechter, thank you very much for your answer that was very helpful! And please excuse the misuse here of the phrase "controlling for".

Best,

Paulo