Hi

I am studying how company specific features are impacting reinsurance purchasing behavior in Korea. I used data from year 2001 to 2019, for the insurance companies operated during the years. The panel data is unbalanced.
The dependent variable is reinsurance ratio, calculated as the reinsurance ceded-premium divided by total business premium. Independent variables are loss ratio(UWRISK), risk-based-capital ratio(RBC), financial leverage ratio(FLEV), logarithm of asset size(lnSIZE), herfindahl index for measuring diversification of products(LINE), being conglomerates(CONG), and being a foreign company(FOR). There are seven independent variables and among them, two are categorical variables. I set dummy variable 0 and 1 for being conglomerates or not, and set dummy variable 0 and 1 for being foreign company or not.

I orderly did pooled OLS, RE, FE, and hausman. Hausman result told me FE is more appropriate than RE. However, when I ran xtreg fe, "Note CONG omitted because of collinearity" was noted and CONG was dropped. Maybe FOR should be dropped too but because there are status changes for some companies, FOR was not dropped.
In this case, would there be any way for me to conduct fixed-effect model regression without dropping categorical variables CONG and FOR?
Or, do you think there can be other errors in my model or hausman test?

I will attach the results from Stata. They are in order of how I did in Stata. Please let me know if you need any other information. Thank you in advance.
Array

Array


Array