Fixed effect model - unobserved heterogeneity

[Louisa Krekel]

Hi everyone,


I investigate the effect of advertising bans on tobacco consumption, thus my dependent variable is tobacco consumption (logcons) and my explanatory variables are advertising ban dummies (weak, limited and comprehensive). My control variables are price (logprice), income (loggdp) and unemployment rate (logunemp).

I'm using a fixed effect model to control for unobserved heterogeneity (characeteristics that do not change over time) and include time-dummies

xtreg logcons logprice loggdp logunemp lim compr i.year , fe robust cluster (Country)



But now I have a question: If i have unobserved variables that vary across countries AND time - how can I estimate this model? In this case something like a change in country-specific regulations affecting smoking behaviour (smoking bans in restaruants, schools etc.).

Thnaks a lot: Highly appreciate any comment or help.

Louisa

[Jorge Eduardo Perez Perez]

Duplicate post: http://www.statalist.org/forums/foru...nd-individuals

If these unobservables are not correlated with your regressors then that wouldn't be a problem. In the case of country specific regulations, they're most likely correlated with advertising bans, so ommiting them from the model would lead to ommited variable bias. You should get information on when these regulations were enacted and include dummy variables, 0 before the regulation was enacted, 1 afterwards, for each country.