help with Xtscc
Hi, i am currently completing regression analysis as part of my dissertation, looking at the effects of board diversity on company performance, using an unbalance panel with n=1894 and t=8 years. I initally used fixed effects for my model (after looking at the F-test and Hausman test), with clustered standard errors.
However my data appeared to have 2nd and 3rd order autocorrelation, and was strongly cross sectional dependent according to the pesaran test.
I therefore re-ran my model using xtscc with fixed effects. For most of my variables, this appears to have had the desired effect in terms of reducing SE's and adjusting the t-stats/p-values accordingly.
However, the coefficient for the variable firm age has changed drastically, along with its p values. I was just wondering if anyone knows why this may be the case. Firm age measure the number of years between t and the year the company was listed on the London Stock Exchange, so i was wondering if this may be because it varies by increases by 1 in line with t for each firm.
My regression ouput for each is below, many thanks for the help.
xtreg ROA BGD Lnfirmsize nodirectors FirmAge i.sector i.Year, fe cluster(id)
note: 2.sector omitted because of collinearity
note: 3.sector omitted because of collinearity
note: 4.sector omitted because of collinearity
note: 5.sector omitted because of collinearity
note: 6.sector omitted because of collinearity
note: 7.sector omitted because of collinearity
note: 8.sector omitted because of collinearity
note: 9.sector omitted because of collinearity
note: 10.sector omitted because of collinearity
note: 11.sector omitted because of collinearity
note: 12.sector omitted because of collinearity
note: 13.sector omitted because of collinearity
note: 14.sector omitted because of collinearity
note: 15.sector omitted because of collinearity
note: 16.sector omitted because of collinearity
note: 17.sector omitted because of collinearity
note: 18.sector omitted because of collinearity
note: 19.sector omitted because of collinearity
note: 2018.Year omitted because of collinearity
Fixed-effects (within) regression Number of obs = 1,894
Group variable: id Number of groups = 287
R-sq: Obs per group:
within = 0.0470 min = 1
between = 0.1558 avg = 6.6
overall = 0.1020 max = 8
F(10,286) = 4.50
corr(u_i, Xb) = -0.7558 Prob > F = 0.0000
(Std. Err. adjusted for 287 clusters in id)
------------------------------------------------------------------------------
| Robust
ROA | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
BGD | 4.110436 3.639013 1.13 0.260 -3.052209 11.27308
Lnfirmsize | -5.558086 1.218856 -4.56 0.000 -7.957152 -3.15902
nodirectors | .2232078 .1897501 1.18 0.240 -.150276 .5966915
FirmAge | -.0736924 .1804944 -0.41 0.683 -.4289584 .2815735
|
sector |
2 | 0 (omitted)
3 | 0 (omitted)
4 | 0 (omitted)
5 | 0 (omitted)
6 | 0 (omitted)
7 | 0 (omitted)
8 | 0 (omitted)
9 | 0 (omitted)
10 | 0 (omitted)
11 | 0 (omitted)
12 | 0 (omitted)
13 | 0 (omitted)
14 | 0 (omitted)
15 | 0 (omitted)
16 | 0 (omitted)
17 | 0 (omitted)
18 | 0 (omitted)
19 | 0 (omitted)
|
Year |
2012 | -.1487878 .563834 -0.26 0.792 -1.258578 .9610028
2013 | .2788925 .6452646 0.43 0.666 -.9911775 1.548963
2014 | .0247799 .6712566 0.04 0.971 -1.29645 1.34601
2015 | -.5558038 .5326529 -1.04 0.298 -1.604221 .4926134
2016 | .032129 .5087251 0.06 0.950 -.9691913 1.033449
2017 | .7315082 .4921294 1.49 0.138 -.2371467 1.700163
2018 | 0 (omitted)
|
_cons | 128.953 25.06172 5.15 0.000 79.62413 178.2818
-------------+----------------------------------------------------------------
sigma_u | 10.070825
sigma_e | 6.6679533
rho | .6952245 (fraction of variance due to u_i)
------------------------------------------------------------------------------
xtscc ROA BGD Lnfirmsize nodirectors FirmAge i.sector i.Year, fe lag (3)
Regression with Driscoll-Kraay standard errors Number of obs = 1894
Method: Fixed-effects regression Number of groups = 287
Group variable (i): id F( 10, 7) = 490.17
maximum lag: 3 Prob > F = 0.0000
within R-squared = 0.0470
------------------------------------------------------------------------------
| Drisc/Kraay
ROA | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
BGD | 4.110436 1.559067 2.64 0.034 .4238274 7.797044
Lnfirmsize | -5.558086 1.095728 -5.07 0.001 -8.149071 -2.967101
nodirectors | .2232078 .2173903 1.03 0.339 -.2908385 .7372541
FirmAge | 5.370975 1.052568 5.10 0.001 2.882048 7.859902
|
sector |
1 | 0 (empty)
2 | 0 (omitted)
3 | 0 (omitted)
4 | 0 (omitted)
5 | 0 (omitted)
6 | 0 (omitted)
7 | 0 (omitted)
8 | 0 (omitted)
9 | 0 (omitted)
10 | 0 (omitted)
11 | 0 (omitted)
12 | 0 (omitted)
13 | 0 (omitted)
14 | 0 (omitted)
15 | 0 (omitted)
16 | 0 (omitted)
17 | 0 (omitted)
18 | 0 (omitted)
19 | 0 (omitted)
|
Year |
2011 | 0 (empty)
2012 | -5.593455 .9827992 -5.69 0.001 -7.917406 -3.269504
2013 | -10.61044 2.001369 -5.30 0.001 -15.34293 -5.877956
2014 | -16.30922 2.954217 -5.52 0.001 -23.29483 -9.32361
2015 | -22.33447 4.051606 -5.51 0.001 -31.915 -12.75395
2016 | -27.19121 5.124306 -5.31 0.001 -39.30827 -15.07415
2017 | -31.9365 6.198879 -5.15 0.001 -46.59452 -17.27848
2018 | -38.11267 7.097793 -5.37 0.001 -54.89629 -21.32906
|
_cons | 0 (omitted)
------------------------------------------------------------------------------
However my data appeared to have 2nd and 3rd order autocorrelation, and was strongly cross sectional dependent according to the pesaran test.
I therefore re-ran my model using xtscc with fixed effects. For most of my variables, this appears to have had the desired effect in terms of reducing SE's and adjusting the t-stats/p-values accordingly.
However, the coefficient for the variable firm age has changed drastically, along with its p values. I was just wondering if anyone knows why this may be the case. Firm age measure the number of years between t and the year the company was listed on the London Stock Exchange, so i was wondering if this may be because it varies by increases by 1 in line with t for each firm.
My regression ouput for each is below, many thanks for the help.
xtreg ROA BGD Lnfirmsize nodirectors FirmAge i.sector i.Year, fe cluster(id)
note: 2.sector omitted because of collinearity
note: 3.sector omitted because of collinearity
note: 4.sector omitted because of collinearity
note: 5.sector omitted because of collinearity
note: 6.sector omitted because of collinearity
note: 7.sector omitted because of collinearity
note: 8.sector omitted because of collinearity
note: 9.sector omitted because of collinearity
note: 10.sector omitted because of collinearity
note: 11.sector omitted because of collinearity
note: 12.sector omitted because of collinearity
note: 13.sector omitted because of collinearity
note: 14.sector omitted because of collinearity
note: 15.sector omitted because of collinearity
note: 16.sector omitted because of collinearity
note: 17.sector omitted because of collinearity
note: 18.sector omitted because of collinearity
note: 19.sector omitted because of collinearity
note: 2018.Year omitted because of collinearity
Fixed-effects (within) regression Number of obs = 1,894
Group variable: id Number of groups = 287
R-sq: Obs per group:
within = 0.0470 min = 1
between = 0.1558 avg = 6.6
overall = 0.1020 max = 8
F(10,286) = 4.50
corr(u_i, Xb) = -0.7558 Prob > F = 0.0000
(Std. Err. adjusted for 287 clusters in id)
------------------------------------------------------------------------------
| Robust
ROA | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
BGD | 4.110436 3.639013 1.13 0.260 -3.052209 11.27308
Lnfirmsize | -5.558086 1.218856 -4.56 0.000 -7.957152 -3.15902
nodirectors | .2232078 .1897501 1.18 0.240 -.150276 .5966915
FirmAge | -.0736924 .1804944 -0.41 0.683 -.4289584 .2815735
|
sector |
2 | 0 (omitted)
3 | 0 (omitted)
4 | 0 (omitted)
5 | 0 (omitted)
6 | 0 (omitted)
7 | 0 (omitted)
8 | 0 (omitted)
9 | 0 (omitted)
10 | 0 (omitted)
11 | 0 (omitted)
12 | 0 (omitted)
13 | 0 (omitted)
14 | 0 (omitted)
15 | 0 (omitted)
16 | 0 (omitted)
17 | 0 (omitted)
18 | 0 (omitted)
19 | 0 (omitted)
|
Year |
2012 | -.1487878 .563834 -0.26 0.792 -1.258578 .9610028
2013 | .2788925 .6452646 0.43 0.666 -.9911775 1.548963
2014 | .0247799 .6712566 0.04 0.971 -1.29645 1.34601
2015 | -.5558038 .5326529 -1.04 0.298 -1.604221 .4926134
2016 | .032129 .5087251 0.06 0.950 -.9691913 1.033449
2017 | .7315082 .4921294 1.49 0.138 -.2371467 1.700163
2018 | 0 (omitted)
|
_cons | 128.953 25.06172 5.15 0.000 79.62413 178.2818
-------------+----------------------------------------------------------------
sigma_u | 10.070825
sigma_e | 6.6679533
rho | .6952245 (fraction of variance due to u_i)
------------------------------------------------------------------------------
xtscc ROA BGD Lnfirmsize nodirectors FirmAge i.sector i.Year, fe lag (3)
Regression with Driscoll-Kraay standard errors Number of obs = 1894
Method: Fixed-effects regression Number of groups = 287
Group variable (i): id F( 10, 7) = 490.17
maximum lag: 3 Prob > F = 0.0000
within R-squared = 0.0470
------------------------------------------------------------------------------
| Drisc/Kraay
ROA | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
BGD | 4.110436 1.559067 2.64 0.034 .4238274 7.797044
Lnfirmsize | -5.558086 1.095728 -5.07 0.001 -8.149071 -2.967101
nodirectors | .2232078 .2173903 1.03 0.339 -.2908385 .7372541
FirmAge | 5.370975 1.052568 5.10 0.001 2.882048 7.859902
|
sector |
1 | 0 (empty)
2 | 0 (omitted)
3 | 0 (omitted)
4 | 0 (omitted)
5 | 0 (omitted)
6 | 0 (omitted)
7 | 0 (omitted)
8 | 0 (omitted)
9 | 0 (omitted)
10 | 0 (omitted)
11 | 0 (omitted)
12 | 0 (omitted)
13 | 0 (omitted)
14 | 0 (omitted)
15 | 0 (omitted)
16 | 0 (omitted)
17 | 0 (omitted)
18 | 0 (omitted)
19 | 0 (omitted)
|
Year |
2011 | 0 (empty)
2012 | -5.593455 .9827992 -5.69 0.001 -7.917406 -3.269504
2013 | -10.61044 2.001369 -5.30 0.001 -15.34293 -5.877956
2014 | -16.30922 2.954217 -5.52 0.001 -23.29483 -9.32361
2015 | -22.33447 4.051606 -5.51 0.001 -31.915 -12.75395
2016 | -27.19121 5.124306 -5.31 0.001 -39.30827 -15.07415
2017 | -31.9365 6.198879 -5.15 0.001 -46.59452 -17.27848
2018 | -38.11267 7.097793 -5.37 0.001 -54.89629 -21.32906
|
_cons | 0 (omitted)
------------------------------------------------------------------------------
