Fixed effects in cross-country firm-level research

[Sneha Thayyil]

Greetings,
I wanted to request for advice on fixed effects - in the context of a cross-country analysis of firms, in countries surveyed by the same organization using the same sampling methods to do so, but in different years. Each country appears for only one year, however.

An indicative example is - firms surveyed in USA in 2020, firms surveyed in UK in 2021, firms surveyed in Italy in 2022, etc.

In such a model, a number of fixed effects can be considered - country, industry, province, year (as it differs by country), country x year, industry x year, country x industry, country x province, industry x province, province x year, country x industry x province x year..etc. 'x' here refers to the interaction.

However, some fixed effects may be nested in others, at least when more than one is included in the model specification.

So I wanted to request for advice about how to think over which effects may be nested in others, and accordingly which one(s) to control for.

Thank you!

[Carlo Lazzaro]

Sheha:
I would include -i.industry- and -i.year- as predictors and cluster the standard errors at -country-level, if countries are at least 30.

[Sneha Thayyil]

Thank you so much for your helpful reply, Carlo.
If I may ask, what could be the thought process behind this? Would there be a reference to consult?

Gratefully,
Sneha

[Carlo Lazzaro]

Sneha (I beg your pardon for mispelling your given name in my previous reply):
I would be interested in investigating whether, othe things being equal, 1-unit variation in -industry- and -time- (level of categorica variables in your case) play any role in varying the regressand. In addition, I would cluster the standard errors at country level because, in all likelihood, firms located in the same country share unobservable local characteristics (say, fiscal system).
I would consult the literature in your research field and see wht others did when addressing the same research question.