Adding community-level variables in household level regression

[Nikos Korompos]

Hi,

I am interested in a model for household consumption, based on repeated cross-sections. I was thinking to incorporate some community-level variables, to control for some supply characteristics. Could I just include the variables (the variation comes from region and year). Given that supply variables are not the variables of interest, should I run a multi-level model, or could I just include them in the household-level regression. If a multi-level regression is requires, could you propose me a good reading?

Thank you.

Best,

Nikos

[Carlo Lazzaro]

Nikos:
assuming that with multi-level regression you refer to -mixed- Stata command, a good introductory textbook (with a quite comprehensive reference for more advanced readings, too) is: https://uk.sagepub.com/en-gb/eur/mul...age/book235963

[Nikos Korompos]

Thank you Carlo. Do you think I can avoid using multilevel modelling in case community variables are not of main interest?

Best,

Nikos

[Carlo Lazzaro]

Nikos:
you can go OLS, but if nesting is relevant, you may have heteroskedasticity problems (see pages 24-26 of the linked reference in my previous post).

[Nikos Korompos]

Thank you Carlo! really useful advice!

[Stephen Jenkins]

Nikos, you might find the first 2 articles at https://academic.oup.com/esr/issue/32/1 of interest and relevant to your question, including the references therein. Your #3 question seems closely related to the issue addressed by Moulton. His paper is cited by the first article (which is Open Access by the way).