Dear Stata Community
I am aware that this question has come up a lot in this forum, yet I am not sure how to handle this problem in my case.
In my master's thesis I am investigating whether import competition influences the prevalence of zombie firms. The regression model is as follows: Share Zombie Firms (by industry) = b*Import Penetration (by industry) + control variables (firm and industry level) +industry and year fixed effects. I do the analysis only for the manufacturing sector (two-digit sic code 20-38), so I have 19 different industries in total. Now the question of clustered standard errors arises. If I use clustered standard errors at the industry level, I get the following result:
If I, on the other hand, use clustered standard errors at the firm level, I get totally different results. As it can be seen t-values are much higher and I get more significant results. What would you recommend? Thank you very much for your help.
I am aware that this question has come up a lot in this forum, yet I am not sure how to handle this problem in my case.
In my master's thesis I am investigating whether import competition influences the prevalence of zombie firms. The regression model is as follows: Share Zombie Firms (by industry) = b*Import Penetration (by industry) + control variables (firm and industry level) +industry and year fixed effects. I do the analysis only for the manufacturing sector (two-digit sic code 20-38), so I have 19 different industries in total. Now the question of clustered standard errors arises. If I use clustered standard errors at the industry level, I get the following result:
Code:
areg share_zombiesBH2 L3.penetration L3.ln_at L3.age L3.F_E L3.tnic3hhi L3.dtfp4 i.year, absorb(sic) vce(cluster sic)
Linear regression, absorbing indicators Number of obs = 10,760
F( 18, 18) = .
Prob > F = .
R-squared = 0.5429
Adj R-squared = 0.5411
Root MSE = 0.0306
(Std. Err. adjusted for 19 clusters in sic)
------------------------------------------------------------------------------
| Robust
share_zomb~2 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
penetration |
L3. | .3198444 .1800509 1.78 0.093 -.0584284 .6981172
|
ln_at |
L3. | .0006393 .0004026 1.59 0.130 -.0002066 .0014852
|
age |
L3. | -.0002218 .0001504 -1.48 0.157 -.0005377 .0000941
|
F_E |
L3. | .0000837 .0000515 1.63 0.121 -.0000244 .0001918
|
tnic3hhi |
L3. | -.0023026 .0011706 -1.97 0.065 -.0047619 .0001567
|
dtfp4 |
L3. | -.0112122 .0162087 -0.69 0.498 -.0452654 .0228409
|
year |
1993 | -.0084138 .00608 -1.38 0.183 -.0211874 .0043598
1994 | -.0195333 .0046546 -4.20 0.001 -.0293122 -.0097545
1995 | -.0151116 .004606 -3.28 0.004 -.0247884 -.0054348
1996 | -.0346192 .0071837 -4.82 0.000 -.0497115 -.0195269
1997 | -.0340963 .009619 -3.54 0.002 -.054305 -.0138877
1998 | -.0305902 .0137055 -2.23 0.039 -.0593843 -.0017961
1999 | -.017059 .0133948 -1.27 0.219 -.0452005 .0110825
2000 | -.0232423 .0109319 -2.13 0.048 -.0462093 -.0002753
2001 | -.0088355 .0147003 -0.60 0.555 -.0397197 .0220486
2002 | .0187259 .0149259 1.25 0.226 -.0126323 .050084
2003 | .019577 .0178908 1.09 0.288 -.0180101 .0571641
2004 | .0085283 .0183033 0.47 0.647 -.0299255 .0469821
2005 | -.0092883 .0166647 -0.56 0.584 -.0442996 .025723
2006 | -.0155196 .020882 -0.74 0.467 -.059391 .0283518
2007 | -.0187548 .0259719 -0.72 0.479 -.0733198 .0358103
2008 | -.0157858 .023747 -0.66 0.515 -.0656763 .0341048
2009 | -.0136269 .0212042 -0.64 0.529 -.0581753 .0309214
2010 | -.0205308 .0182965 -1.12 0.277 -.0589703 .0179087
2011 | -.0204554 .0232057 -0.88 0.390 -.0692087 .0282978
|
_cons | .1313362 .0277438 4.73 0.000 .0730486 .1896237
-------------+----------------------------------------------------------------
sic | absorbed (19 categories)
Code:
areg share_zombiesBH2 L3.penetration L3.ln_at L3.age L3.F_E L3.tnic3hhi L3.dtfp4 i.year, absorb(sic) vce(cluster gvkey)
Linear regression, absorbing indicators Number of obs = 10,760
F( 25, 1445) = 198.89
Prob > F = 0.0000
R-squared = 0.5429
Adj R-squared = 0.5411
Root MSE = 0.0306
(Std. Err. adjusted for 1,446 clusters in gvkey)
------------------------------------------------------------------------------
| Robust
share_zomb~2 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
penetration |
L3. | .3198444 .0820098 3.90 0.000 .1589734 .4807154
|
ln_at |
L3. | .0006393 .0002082 3.07 0.002 .0002308 .0010478
|
age |
L3. | -.0002218 .0000835 -2.66 0.008 -.0003855 -.0000581
|
F_E |
L3. | .0000837 .0000641 1.31 0.192 -.0000421 .0002094
|
tnic3hhi |
L3. | -.0023026 .0013611 -1.69 0.091 -.0049725 .0003673
|
dtfp4 |
L3. | -.0112122 .0032899 -3.41 0.001 -.0176656 -.0047588
|
year |
1993 | -.0084138 .0010055 -8.37 0.000 -.0103862 -.0064414
1994 | -.0195333 .0010636 -18.37 0.000 -.0216197 -.017447
1995 | -.0151116 .0010935 -13.82 0.000 -.0172566 -.0129666
1996 | -.0346192 .0016817 -20.59 0.000 -.037918 -.0313204
1997 | -.0340963 .0023852 -14.30 0.000 -.0387751 -.0294176
1998 | -.0305902 .0030292 -10.10 0.000 -.0365323 -.024648
1999 | -.017059 .003114 -5.48 0.000 -.0231674 -.0109506
2000 | -.0232423 .0033798 -6.88 0.000 -.029872 -.0166125
2001 | -.0088355 .0038973 -2.27 0.024 -.0164805 -.0011906
2002 | .0187259 .0050066 3.74 0.000 .0089049 .0285469
2003 | .019577 .0060633 3.23 0.001 .0076831 .0314708
2004 | .0085283 .0059876 1.42 0.155 -.003217 .0202736
2005 | -.0092883 .0069266 -1.34 0.180 -.0228755 .0042989
2006 | -.0155196 .0076394 -2.03 0.042 -.0305051 -.0005342
2007 | -.0187548 .0091819 -2.04 0.041 -.036766 -.0007436
2008 | -.0157858 .0096391 -1.64 0.102 -.0346938 .0031223
2009 | -.0136269 .0101453 -1.34 0.179 -.0335281 .0062742
2010 | -.0205308 .008291 -2.48 0.013 -.0367946 -.004267
2011 | -.0204554 .0085807 -2.38 0.017 -.0372875 -.0036234
|
_cons | .1313362 .0116397 11.28 0.000 .1085038 .1541686
-------------+----------------------------------------------------------------
sic | absorbed (19 categories)
Comment