Omitted variables due to Multicollinearity

[Javid Imranov]

Hi,

I am running two regressions of my banking risk proxies which are Z-score and Non-Performing Loan Ratio (NPLr) on the list of bank-specific and macro-economic variables. Initially, I obtained annual data for variables. I will use panel data with entity and time fixed effect but the problem is that when I run regression with time-fixed effect, my macro-economic variables (GDP growth, Inflation, Unemployment) become omitted due to collinearity. As I understand it happens due to the fact that I use the same value for each company within certain time period (1 year). So, I wanted to ask if there is a way to solve this problem without obtaining quarterly data for all the variables or it's the only way?
I've attached a screenshot from Stata so that you could see that.

Regards,

[Maarten Buis]

My first question here would be why you would expect the period to matter net of GDP growth, inflation, and unemployment? Time can do things by making the subjects older. However, the period on its own does not do anything, it is just a proxy for "other things that happen at the same time point". If you already adjust for the main factors, why would you add period fixed effects, what is it supposed to capture, and can you capture it directly? It is still possible that you want to enter time fixed effects but you will need to justify it.

As an aside, the Statalist FAQ asks everybody to post under their real name. This has a long tradition on this list, and we think that it has helped to make the interactions on this list more professional and friendly. So we would like to continue that tradition. You can change your login name by clicking on “Contact us” located at the bottom right-hand corner of every page.

[Javid Imranov]

I would like to use the time fixed-effect variables because they can highlight any specific years that exhibit substantially higher banking risk levels, particularly those leading up to the recent financial crisis. The firm fixed-effect variables are going to be used to capture the unobserved characteristics, such as management decisions and corporate culture, which are unique to each individual firm and yet, impact the total risk faced by that bank.