I am looking to analyse the impact (causal effect) of a conflict on firm valuation (using daily stock prices) for a sample of around 800 firms (I currently have this data over a sample period of around 365 days) so its panel data.
I want to analyse the overall impact but also identify the difference in impact between the industries.
However, i cannot decide on the most appropriate methodology to use. I have tried to combine both the event study methodology and the fixed effects one by generating the abnormal returns to be used as my dependent variable. The following would be my regression equation:
Yit = α + β1WarDummyit + ∑βk Controlk it + γj IndustryDummyjit+δj (WarDummyit×IndustryDummyjit) + ϵit
- Yit: dependent variable; abnormal returns for firm i at time t.
- IndustryDummyjit: dummies representing the 20 industries (manufacturing, IT, services etc)
- WarDummyit: 1 for the event window surrounding the start of the war and 0 otherwise.
- WarDummy×IndustryDummy: Interaction terms between the war dummy and each of the industry dummies; the idea would be to measure the differential impact of the war on firms in industry j, controlling for other factors.
Can this approach be considered valid? I have tried to run it on multiple occasions but i keep getting this error message: omitted because of collinearity.
Here's the output:
. xtreg AR event_window i.ggroup event_window#ggroup, fe vce(cluster gvkey)
note: event_window omitted because of collinearity.
note: 1510.ggroup omitted because of collinearity.
note: 2010.ggroup omitted because of collinearity.
note: 2020.ggroup omitted because of collinearity.
note: 2030.ggroup omitted because of collinearity.
note: 2510.ggroup omitted because of collinearity.
note: 2520.ggroup omitted because of collinearity.
note: 2530.ggroup omitted because of collinearity.
note: 2550.ggroup omitted because of collinearity.
note: 3010.ggroup omitted because of collinearity.
note: 3020.ggroup omitted because of collinearity.
note: 3030.ggroup omitted because of collinearity.
note: 3510.ggroup omitted because of collinearity.
note: 3520.ggroup omitted because of collinearity.
note: 4010.ggroup omitted because of collinearity.
note: 4020.ggroup omitted because of collinearity.
note: 4030.ggroup omitted because of collinearity.
note: 4510.ggroup omitted because of collinearity.
note: 4520.ggroup omitted because of collinearity.
note: 4530.ggroup omitted because of collinearity.
note: 5010.ggroup omitted because of collinearity.
note: 5020.ggroup omitted because of collinearity.
note: 5510.ggroup omitted because of collinearity.
note: 6010.ggroup omitted because of collinearity.
note: 6020.ggroup omitted because of collinearity.
note: 1.event_window#1010b.ggroup omitted because of collinearity.
note: 1.event_window#1510.ggroup omitted because of collinearity.
note: 1.event_window#2010.ggroup omitted because of collinearity.
note: 1.event_window#2020.ggroup omitted because of collinearity.
note: 1.event_window#2030.ggroup omitted because of collinearity.
note: 1.event_window#2510.ggroup omitted because of collinearity.
note: 1.event_window#2520.ggroup omitted because of collinearity.
note: 1.event_window#2530.ggroup omitted because of collinearity.
note: 1.event_window#2550.ggroup omitted because of collinearity.
note: 1.event_window#3010.ggroup omitted because of collinearity.
note: 1.event_window#3020.ggroup omitted because of collinearity.
note: 1.event_window#3030.ggroup omitted because of collinearity.
note: 1.event_window#3510.ggroup omitted because of collinearity.
note: 1.event_window#3520.ggroup omitted because of collinearity.
note: 1.event_window#4010.ggroup omitted because of collinearity.
note: 1.event_window#4020.ggroup omitted because of collinearity.
note: 1.event_window#4030.ggroup omitted because of collinearity.
note: 1.event_window#4510.ggroup omitted because of collinearity.
note: 1.event_window#4520.ggroup omitted because of collinearity.
note: 1.event_window#4530.ggroup omitted because of collinearity.
note: 1.event_window#5010.ggroup omitted because of collinearity.
note: 1.event_window#5020.ggroup omitted because of collinearity.
note: 1.event_window#5510.ggroup omitted because of collinearity.
note: 1.event_window#6010.ggroup omitted because of collinearity.
note: 1.event_window#6020.ggroup omitted because of collinearity.
Fixed-effects (within) regression Number of obs = 106,169
Group variable: gvkey Number of groups = 979
R-squared: Obs per group:
Within = 0.0000 min = 9
Between = 0.0044 avg = 108.4
Overall = . max = 110
F(0, 978) = .
corr(u_i, Xb) = . Prob > F = .
(Std. err. adjusted for 979 clusters in gvkey)
Robust
AR Coefficient std. err. t P>t [95% conf. interval]
event_window 0 (omitted)
ggroup
1510 0 (omitted)
2010 0 (omitted)
2020 0 (omitted)
2030 0 (omitted)
2510 0 (omitted)
2520 0 (omitted)
2530 0 (omitted)
2550 0 (omitted)
3010 0 (omitted)
3020 0 (omitted)
3030 0 (omitted)
3510 0 (omitted)
3520 0 (omitted)
4010 0 (omitted)
4020 0 (omitted)
4030 0 (omitted)
4510 0 (omitted)
4520 0 (omitted)
4530 0 (omitted)
5010 0 (omitted)
5020 0 (omitted)
5510 0 (omitted)
6010 0 (omitted)
6020 0 (omitted)
event_window#ggroup
1 1010 0 (omitted)
1 1510 0 (omitted)
1 2010 0 (omitted)
1 2020 0 (omitted)
1 2030 0 (omitted)
1 2510 0 (omitted)
1 2520 0 (omitted)
1 2530 0 (omitted)
1 2550 0 (omitted)
1 3010 0 (omitted)
1 3020 0 (omitted)
1 3030 0 (omitted)
1 3510 0 (omitted)
1 3520 0 (omitted)
1 4010 0 (omitted)
1 4020 0 (omitted)
1 4030 0 (omitted)
1 4510 0 (omitted)
1 4520 0 (omitted)
1 4530 0 (omitted)
1 5010 0 (omitted)
1 5020 0 (omitted)
1 5510 0 (omitted)
1 6010 0 (omitted)
1 6020 0 (omitted)
_cons -.0006182 9.87e-14 -6.3e+09 0.000 -.0006182 -.0006182
sigma_u .00848138
sigma_e .06421703
rho .0171444 (fraction of variance due to u_i)
Any help is appreciated!
I want to analyse the overall impact but also identify the difference in impact between the industries.
However, i cannot decide on the most appropriate methodology to use. I have tried to combine both the event study methodology and the fixed effects one by generating the abnormal returns to be used as my dependent variable. The following would be my regression equation:
Yit = α + β1WarDummyit + ∑βk Controlk it + γj IndustryDummyjit+δj (WarDummyit×IndustryDummyjit) + ϵit
- Yit: dependent variable; abnormal returns for firm i at time t.
- IndustryDummyjit: dummies representing the 20 industries (manufacturing, IT, services etc)
- WarDummyit: 1 for the event window surrounding the start of the war and 0 otherwise.
- WarDummy×IndustryDummy: Interaction terms between the war dummy and each of the industry dummies; the idea would be to measure the differential impact of the war on firms in industry j, controlling for other factors.
Can this approach be considered valid? I have tried to run it on multiple occasions but i keep getting this error message: omitted because of collinearity.
Here's the output:
. xtreg AR event_window i.ggroup event_window#ggroup, fe vce(cluster gvkey)
note: event_window omitted because of collinearity.
note: 1510.ggroup omitted because of collinearity.
note: 2010.ggroup omitted because of collinearity.
note: 2020.ggroup omitted because of collinearity.
note: 2030.ggroup omitted because of collinearity.
note: 2510.ggroup omitted because of collinearity.
note: 2520.ggroup omitted because of collinearity.
note: 2530.ggroup omitted because of collinearity.
note: 2550.ggroup omitted because of collinearity.
note: 3010.ggroup omitted because of collinearity.
note: 3020.ggroup omitted because of collinearity.
note: 3030.ggroup omitted because of collinearity.
note: 3510.ggroup omitted because of collinearity.
note: 3520.ggroup omitted because of collinearity.
note: 4010.ggroup omitted because of collinearity.
note: 4020.ggroup omitted because of collinearity.
note: 4030.ggroup omitted because of collinearity.
note: 4510.ggroup omitted because of collinearity.
note: 4520.ggroup omitted because of collinearity.
note: 4530.ggroup omitted because of collinearity.
note: 5010.ggroup omitted because of collinearity.
note: 5020.ggroup omitted because of collinearity.
note: 5510.ggroup omitted because of collinearity.
note: 6010.ggroup omitted because of collinearity.
note: 6020.ggroup omitted because of collinearity.
note: 1.event_window#1010b.ggroup omitted because of collinearity.
note: 1.event_window#1510.ggroup omitted because of collinearity.
note: 1.event_window#2010.ggroup omitted because of collinearity.
note: 1.event_window#2020.ggroup omitted because of collinearity.
note: 1.event_window#2030.ggroup omitted because of collinearity.
note: 1.event_window#2510.ggroup omitted because of collinearity.
note: 1.event_window#2520.ggroup omitted because of collinearity.
note: 1.event_window#2530.ggroup omitted because of collinearity.
note: 1.event_window#2550.ggroup omitted because of collinearity.
note: 1.event_window#3010.ggroup omitted because of collinearity.
note: 1.event_window#3020.ggroup omitted because of collinearity.
note: 1.event_window#3030.ggroup omitted because of collinearity.
note: 1.event_window#3510.ggroup omitted because of collinearity.
note: 1.event_window#3520.ggroup omitted because of collinearity.
note: 1.event_window#4010.ggroup omitted because of collinearity.
note: 1.event_window#4020.ggroup omitted because of collinearity.
note: 1.event_window#4030.ggroup omitted because of collinearity.
note: 1.event_window#4510.ggroup omitted because of collinearity.
note: 1.event_window#4520.ggroup omitted because of collinearity.
note: 1.event_window#4530.ggroup omitted because of collinearity.
note: 1.event_window#5010.ggroup omitted because of collinearity.
note: 1.event_window#5020.ggroup omitted because of collinearity.
note: 1.event_window#5510.ggroup omitted because of collinearity.
note: 1.event_window#6010.ggroup omitted because of collinearity.
note: 1.event_window#6020.ggroup omitted because of collinearity.
Fixed-effects (within) regression Number of obs = 106,169
Group variable: gvkey Number of groups = 979
R-squared: Obs per group:
Within = 0.0000 min = 9
Between = 0.0044 avg = 108.4
Overall = . max = 110
F(0, 978) = .
corr(u_i, Xb) = . Prob > F = .
(Std. err. adjusted for 979 clusters in gvkey)
Robust
AR Coefficient std. err. t P>t [95% conf. interval]
event_window 0 (omitted)
ggroup
1510 0 (omitted)
2010 0 (omitted)
2020 0 (omitted)
2030 0 (omitted)
2510 0 (omitted)
2520 0 (omitted)
2530 0 (omitted)
2550 0 (omitted)
3010 0 (omitted)
3020 0 (omitted)
3030 0 (omitted)
3510 0 (omitted)
3520 0 (omitted)
4010 0 (omitted)
4020 0 (omitted)
4030 0 (omitted)
4510 0 (omitted)
4520 0 (omitted)
4530 0 (omitted)
5010 0 (omitted)
5020 0 (omitted)
5510 0 (omitted)
6010 0 (omitted)
6020 0 (omitted)
event_window#ggroup
1 1010 0 (omitted)
1 1510 0 (omitted)
1 2010 0 (omitted)
1 2020 0 (omitted)
1 2030 0 (omitted)
1 2510 0 (omitted)
1 2520 0 (omitted)
1 2530 0 (omitted)
1 2550 0 (omitted)
1 3010 0 (omitted)
1 3020 0 (omitted)
1 3030 0 (omitted)
1 3510 0 (omitted)
1 3520 0 (omitted)
1 4010 0 (omitted)
1 4020 0 (omitted)
1 4030 0 (omitted)
1 4510 0 (omitted)
1 4520 0 (omitted)
1 4530 0 (omitted)
1 5010 0 (omitted)
1 5020 0 (omitted)
1 5510 0 (omitted)
1 6010 0 (omitted)
1 6020 0 (omitted)
_cons -.0006182 9.87e-14 -6.3e+09 0.000 -.0006182 -.0006182
sigma_u .00848138
sigma_e .06421703
rho .0171444 (fraction of variance due to u_i)
Any help is appreciated!
