Firm fixed effects or industry fixed effects

[Ingo Brooks]

Hello,

I have a large panel dataset containing firm-level data (firm characteristics, accounting data plus some macroeconomic indicators).

I would like to estimate a linear regression where I regress an accounting variable (e.g. ROE) on a set of firm characteristics (e.g. employee turnover) and some macroeconomic indicators (e.g. unemployment rate). Due to significant differences across the firms and, in particular the industries, it seems important to include some sort of fixed effects in the analysis. What is the better choice: firm fixed effects or industry fixed effects? What are the pros and cons of the two types of fixed effects?

Thanks a lot for any help on this.

Kind regards,
Ingo

[Carlo Lazzaro]

Ingo:
welcome to this forum.
The best approach is skimming through the literature of your research field and see what Others did in the past when presented with the same research topic.
As an aside, please note that if you're going to use a fixed effect specification under (I assume) -xtreg-, all time-invariant predictors will be cancelled out.

[Phil Bromiley]

To add to Carlo's comment, you're more likely to get a very helpful answer if you follow the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output and sample data using dataex.

If you say there are "significant differences across firms" then you'll probably want firm effects because they control for all firm factors that do not vary over time. Firm effects automatically control for industry (assuming firms stay in the same industry). As Carlo points out, firm fixed effects do make it impossible to estimate stable firm characteristics, although there are some estimators mid-way between fixed and random effects (xthtaylor and mundalk estimators come to mind) that allow estimation of some stable firm characteristics along with control for some unspecified firm characteristics. Industry fixed effects let you estimate stable firm factors, but do not control for any omitted firm factors.

[Ingo Brooks]

Carlo and Phil,

Many thanks for your very helpful comments and sorry for not having properly followed the FAQ on asking questions on Stata list. I will definitely take a careful look at them.

The point about the omitted firm factors seems very important to me, in particular since accounting variables are often highly correlated. Therefore, I will try to stick with the _xtreg, fe_ command whenever possible. While some of the accounting variables do only have very limited within variation, they nevertheless fluctuate over time a bit such that I should not encounter problems with time-invariant predictors.

Thanks again for your help.

Ingo