I am running a panel data analysis of geographic diversification (quadratic) on company performance (winsorized ROA.) I noticed that the the effect of the quadratic geographic diversification term (cL.GSD#cL.GSD ) is significant (P>|t| of 0.014) if I run it without clustering SEs. However, if I cluster the SEs the effect of cL.GSD#cL.GSD is not significant (P>|t| of 0.063) I presume this is happening on account of heteroskedasticity and autocorrelation. My questions are as follows:
1. For checking for heteroskedasticity, I presume I run xttest3 after running the non-clustered xtreg fe and if Prob>chi2 is < 0.05, I should conclude that the data is heteroskedastic.
2. For autocorrelation, what would be the most appropriate test to use?
(I am neither a statistician nor an expert on STATA, so please pardon me if my question is too basic.)
For your reference, my output is pasted below:
1. For checking for heteroskedasticity, I presume I run xttest3 after running the non-clustered xtreg fe and if Prob>chi2 is < 0.05, I should conclude that the data is heteroskedastic.
2. For autocorrelation, what would be the most appropriate test to use?
(I am neither a statistician nor an expert on STATA, so please pardon me if my question is too basic.)
For your reference, my output is pasted below:
Code:
. xtreg ROA_win05 Ln_Revenue Ln_LTDTA CoAge TPSD Ind_GSD wGDPpc wCPI wDCF wExpgr wGDPgr wCon
> s c.l1.GSD##c.l1.GSD if Mfg_Ratio>80 &FOREIGNSALESTOTALSALES >10 & Year_ < YearInactive, f
> e
Fixed-effects (within) regression Number of obs = 1,099
Group variable: n_CUSIP Number of groups = 122
R-sq: Obs per group:
within = 0.0303 min = 1
between = 0.2820 avg = 9.0
overall = 0.1260 max = 18
F(13,964) = 2.31
corr(u_i, Xb) = 0.0634 Prob > F = 0.0050
---------------------------------------------------------------------------------
ROA_win05 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
Ln_Revenue | 2.422417 .7250379 3.34 0.001 .9995821 3.845251
Ln_LTDTA | -.5803542 .207356 -2.80 0.005 -.9872754 -.173433
CoAge | -.0860849 .1004581 -0.86 0.392 -.2832267 .111057
TPSD | -.4280329 1.105539 -0.39 0.699 -2.597573 1.741507
Ind_GSD_by_Year | 2.389643 2.10156 1.14 0.256 -1.734516 6.513803
wGDPpc | -.0000434 .0001677 -0.26 0.796 -.0003726 .0002857
wCPI | -.3764426 .2832035 -1.33 0.184 -.932209 .1793239
wDCF | 7.03e-13 1.28e-12 0.55 0.585 -1.82e-12 3.22e-12
wExpgr | .1760092 .1066144 1.65 0.099 -.0332138 .3852322
wGDPgr | -.4340272 .2912374 -1.49 0.136 -1.00556 .1375052
wCons | -3.26e-13 4.59e-13 -0.71 0.478 -1.23e-12 5.75e-13
|
GSD |
L1. | 8.412533 3.788138 2.22 0.027 .9785863 15.84648
|
cL.GSD#cL.GSD | -7.214446 2.919635 -2.47 0.014 -12.94402 -1.484873
|
_cons | -41.19891 14.33343 -2.87 0.004 -69.32723 -13.0706
----------------+----------------------------------------------------------------
sigma_u | 17.97445
sigma_e | 7.1515211
rho | .86333291 (fraction of variance due to u_i)
---------------------------------------------------------------------------------
F test that all u_i=0: F(121, 964) = 18.73 Prob > F = 0.0000
Code:
. xtreg ROA_win05 Ln_Revenue Ln_LTDTA CoAge TPSD Ind_GSD wGDPpc wCPI wDCF wExpgr wGDPgr wCon
> s c.l1.GSD##c.l1.GSD if Mfg_Ratio>80 &FOREIGNSALESTOTALSALES >10 & Year_ < YearInactive, f
> e cluster(n_CUSIP)
Fixed-effects (within) regression Number of obs = 1,099
Group variable: n_CUSIP Number of groups = 122
R-sq: Obs per group:
within = 0.0303 min = 1
between = 0.2820 avg = 9.0
overall = 0.1260 max = 18
F(11,121) = .
corr(u_i, Xb) = 0.0634 Prob > F = .
(Std. Err. adjusted for 122 clusters in n_CUSIP)
---------------------------------------------------------------------------------
| Robust
ROA_win05 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
Ln_Revenue | 2.422417 1.262198 1.92 0.057 -.0764376 4.921271
Ln_LTDTA | -.5803542 .2804897 -2.07 0.041 -1.135658 -.0250509
CoAge | -.0860849 .1138875 -0.76 0.451 -.3115552 .1393855
TPSD | -.4280329 1.173237 -0.36 0.716 -2.750766 1.8947
Ind_GSD_by_Year | 2.389643 2.990261 0.80 0.426 -3.530367 8.309653
wGDPpc | -.0000434 .0002331 -0.19 0.852 -.0005049 .0004181
wCPI | -.3764426 .3020473 -1.25 0.215 -.9744249 .2215398
wDCF | 7.03e-13 1.29e-12 0.55 0.586 -1.84e-12 3.25e-12
wExpgr | .1760092 .1363525 1.29 0.199 -.0939365 .4459549
wGDPgr | -.4340272 .3859219 -1.12 0.263 -1.198061 .330007
wCons | -3.26e-13 4.73e-13 -0.69 0.492 -1.26e-12 6.10e-13
|
GSD |
L1. | 8.412533 6.196508 1.36 0.177 -3.855088 20.68015
|
cL.GSD#cL.GSD | -7.214446 3.837988 -1.88 0.063 -14.81275 .3838632
|
_cons | -41.19891 23.68022 -1.74 0.084 -88.08015 5.682326
----------------+----------------------------------------------------------------
sigma_u | 17.97445
sigma_e | 7.1515211
rho | .86333291 (fraction of variance due to u_i)
---------------------------------------------------------------------------------
Comment