Collinearity issue (master's thesis)

[Angelo De Santis]

Hello everyone, I am currently using Stata for my master’s thesis in Economics and Business, and I’ve been facing some difficulties lately. My objective is to verify whether the introduction of the EU-ETS system had an effect on Italian trade flows through a difference-in-differences analysis, from 1995 to 2022, using the gravity model.
The treatment group consists of trade flows between Italy and countries that adopt the EU-ETS, while the control group consists of trade flows between Italy and countries outside the EU-ETS system (as shown in the sample).
The issue is that when running the command, Stata reports collinearity problems, and I am unable to visualize the coefficients of the independent variables of interest. I am attaching the necessary files below.

If you also notice any other issues, I would be grateful if you could point them out. Thank you in advance for your help!

[Nick Cox]

Cross-posted at https://www.reddit.com/r/stata/comme...asters_thesis/

Please see https://www.statalist.org/forums/help #12 for our policy on attachments and #8 for our policy on cross-posting.

Sorry, but I am not an economist and never do your kind of project, so I stop there myself.

[Carlo Lazzaro]

Angelo:
have you addressed these issues to your supervisor?
What's her/his take about them?

[Angelo De Santis]

Hello, Carlo. I talked about it with her yesterday. We tried to solve the problem for more than a hour. She suggested me to include more countries in the sample, because the distance between countries is fixed every year. So this might be why stata takes my code down. Before adding my countries, she gave me an advice telling me about this forum, in order to see if there are any other ways to solve the problem.

[Nick Cox]

I doubt that active members here will download your dataset and do file to try to see what the issue is.

There are two general reasons already explained or implied in our FAQ: downloading attachments doesn't carry zero risk, and that's more work than people are generally willing to do.

Your dataset is likely too big to be posted otherwise, but showing the command you used and the error you got is more likely to elicit specific advice. It could be that your model is ill advised, but no-one can tell if you don't show specific details.

[Carlo Lazzaro]

Angelo:
as an aside to Nick's helpful recommendation, see if this today's thread can help you out somehow: Using xtdidregress command with 2 cohorts from different birth years - Statalist

[Angelo De Santis]

Thank you both for your help!