Panel Fixed Effects & Interactions

[Tomas Jankelevich]

Hello,
Im working with a panel of several banks of different states and i want to mesure if the death of a CEO (president) affects the banks performance. My regression are the following:

xtreg ln_profits post_death  l.ln_assets i.year, fe
xtreg ln_deposits post_deat  l.ln_assets i.year, fe

My doubts are the following:
1. I think i need to make also fixed effects of the states because the banks may differ between states, how do i add another fixed effect to my regression? Can i do that or i will have an omited variable?
2. Is my dummy post_death capturing the effect i expect or should i create different dummys to specify if we are in the first year after a death(+1), second(+2), third (+3), etc.
3. If i want to add another dummy that specify rural vs urban places. Do i need to create an interaction with post_death? How do i do that?

This is my data:

state id year president turnover death ln_profits ln_deposits post_death
"ny" 701 1928 "c s nisbet"     0 .       13.5912  15.41431 
"ny" 701 1929 "c s nisbet"     0 .       13.688297 15.469843
"ny" 701 1930 "c s nisbet"     0 .       13.754994   15.5269
"ny" 701 1931 "c s nisbet"     0 .        13.66891 15.430566
"ny" 701 1932 "c s nisbet"     0 .       13.286278  15.36186
"ny" 701 1933 "c s nisbet"     1   1    13.181445  15.35412        1
"ny" 701 1934 "a v morris"     0 .      12.750996 15.370708       1
"ky" 702 1928 "l e harrower"   0 .     13.321004 14.139345
"ky" 702 1929 "l e harrower"   0 .     13.372365 13.958246
"ky" 702 1930 "l e harrower"   0 .      13.39825 13.852264
"ky" 702 1931 "l e harrower"   0 .     13.362202 13.526426
"ky" 702 1932 "l e harrower"   0 .    13.299565  13.50915
"ky" 702 1933 "l e harrower"   0 .     13.27321 13.568895
"ky" 702 1934 "l e harrower"   0 .    13.255058  13.63851
"nj" 703 1928 "c l howard"     0 .     13.295777 15.287073
"nj" 703 1929 "c l howard"     0 .    13.321268 15.258593
"nj" 703 1930 "c l howard"     0 .    13.332827 15.251134
"nj" 703 1931 "c l howard"     0 .    13.047334  15.16888
"nj" 703 1932 "c l howard"     0 .      12.56793 15.107882
"nj" 703 1933 "c l howard"     0 . 12.025054 15.103624
"nj" 703 1934 "c lhoward"      0 .  11.49074 15.207767
"al" 704 1928 "c e hulbert"    0 . 10.054706  12.52019
"al" 704 1929 "e j turnbull"   0 . 10.053415 12.560487
"al" 704 1930 "e j turnbull"   0 . 10.071372 12.590498
"al" 704 1931 "e j turnbull"   0 .  9.836439 12.479259
"al" 704 1932 "e j turnbull"   0 .  9.644393  12.41951
"al" 704 1933 "e j turnbull"   0 .  9.432684 12.495944
"al" 704 1934 "e j tumbull"    0 .  8.903272 12.518402
"ga" 705 1928 "j s phillips"   0 . 11.158662 12.925046
"ga" 705 1929 "j s phillips"   1 0  11.21182 12.909145
"ga" 705 1930 "j e cannon"     0 . 11.274262  12.85427
end

Thanks a lot!

[Carlo Lazzaro]

Tomas:
- please note that -string- variable are not suitable for -xtreg-; you have to remove inverted commas first (see -help string functions-);
- I assume that you have already ruled out -re- specification for your data;
- due to the -fe-machinery, all time-invariant predictors (such as -state- and -ruralyesno-) will be omitted from the regression;
- you may want to try interacting -year- with -post_death-:

i.year##i.post_death