Linear regression with "panel-like" data and multiple observations per time period

[Jonas Voss]

Dear statalisters,

I have a dataset of loans between banks and companies and want to examine potential effects of loan- and borower-characteristics on banks. The dataset is of the following exemplary structure:

* Example generated by -dataex-. For more info, type help dataex
clear
input long loanid float bankvariable int bankid float companyvariable long companyid float monthyear
145  0 6 37  497  5
181  0 6 30   11  5
101  0 6 37  497  5
183  0 6 30   11  5
177  3 6 21  243  6
196  3 6 20  920  6
162  3 6 18  706  6
324 -1 6 12  474 10
240 -1 6 16 1140 10
372 -1 6 22  400 10
end

As the same banks make loans at different times in the sample period, the data has a cross-sectional as well as a time-dimension. The data on bank-level variables is available monthly, so one bank also sometimes makes multiple loans per month. I want to analyse the data via linear regression using bank- and time-fixed effects. As my level of time in the data is monthly, my concern is if I´ll have problems due to the fact that there are multiple entries per bank per time period. Does this somehow invalidate usual OLS- or Fixed effects-assumptions? I could transform the loan- and borrower-level data to the bank-month-level by taking averages per bank. Would this solve the aforementioned problem (if this actually is a problem)?

Unfortunately i could not find comparable cases in literature, which is why i am posting here. Any help would be really appreciated.

Best regards
Jonas

[Carlo Lazzaro]

Jonas:
welcome to this forum.
If your concern is the so calle -repeated time values- per panel, this is cot an issue.
Provided that you do not plan to use time-series operators (such as lags and leads), you can simply -xtset- your dataset with -panelid- only and go -xtreg-.

[Jonas Voss]

Thank you Carlo!