Working with panel data covering only two years

[Bert George]

Hi All,

I have a question concerning panel data analysis. I have a dataset covering about 750 municipalities and each variable was measured twice (2008, 2017). It seems that panel data analysis is preferred over pooled because N >> T. So I used xtreg. I also used fixed effects and vce(robust). I had some questions that I really hoped you could help me with:

- I am seeing that stata offers a range of tools for longitudinal analysis. Is there an overview of which tool is best under which conditions? What do you all think of my specific choice?
- In typical linear regression we report a range of assumptions that need to be met (multicollinearity, etc.). Which ones should be reported in the analysis I did? The stata output shows stuff like corr(u_i, Xb), sigma_u, sigma_e, rho... - is there any source available concerning how to interpret this output?
- When doing these types of analyses, which control variables would you add? My IVs focus on characteristics of local leaders and my DVs on local financial performance.

Thank in advance for your help!

[Carlo Lazzaro]

Bert:
welcome to this forum.
Some comments about your post:
- you actually have a small panel (N>T); hence, if your regerssand is continuous, -xtreg- is the way to go. The only concern (which is not methodological but has more to do with your research field) relates to the gap between the two waves of data;
- usually, if you're using default standard errors, -hausman- outcome can give you some clues about going -fe- or -re-;
- see -xtreg- entry in Stata .pdf manual and get yourself familiar with -xtreg- options and outcome. In addition, reading any decent textbook on panel data econometrics is highly recommended (the same entry shows a wide reference section);
- predictors should give a fair and true view of the data generating process you're interested in.