财经节析 发表于 2018-5-19 16:46
首先,我说的就是在个体固定效应变截距面板数据模型下的结论(即我说的失效,是指reghdfe不能控制那些像行 ...
您好,老师。我分别做了三个回归xi:reg lnpatent c.index##i.hightech lnemployment assdebt curasslia fixass gdpg m2 i.industrycode i.citycode, r和xtreg lnpatent c.index##i.hightech lnemployment assdebt curasslia fixass gdpg m2 i.industrycode i.citycode, r以及reghdfe lnpatent c.index##i.hightech lnemployment assdebt curasslia fixass gdpg m2, a(citycode industrycode) vce(cl new_id)。
控制了行业固定效应industrycode,加入了高技术行业的交互项hightech,正常来说industrycode与hightech存在多重共线性,根据这个帖子hightech的系数应该omitted,但是回归结果只有reghdfe的hightech系数omitted,其余两个回归没有,并且回归系数也不相同,这是什么原因?非常感谢!
回归结果如下:
xi:reg lnpatent c.index##i.hightech lnemployment assdebt curasslia fixass gdpg m2 i.industrycode i.citycode, r
---------------------------------------------------------------------------------
| Robust
lnpatent | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
index | .0021595 .0001183 18.26 0.000 .0019277 .0023914
1.hightech | 1.501399 .1638783 9.16 0.000 1.180189 1.822608
hightech#c.index| -.0001405 .0001769 -0.79 0.427 -.0004872 .0002062
lnemployment | .4588536 .0075574 60.72 0.000 .4440407 .4736665
assdebt | -.0005759 .0008115 -0.71 0.478 -.0021665 .0010147
curasslia | .0021092 .0019901 1.06 0.289 -.0017915 .0060099
fixass | -.8463446 .0542857 -15.59 0.000 -.9527473 -.7399419
gdpg | -.096158 .0049903 -19.27 0.000 -.1059392 -.0863768
m2 | -.0135675 .002159 -6.28 0.000 -.0177993 -.0093357
_Iindustryc_2 | .3632323 .1788014 2.03 0.042 .0127727 .7136919
_Iindustryc_3 | .0604105 .1394887 0.43 0.665 -.2129943 .3338152
_Iindustryc_4 | .713872 .1598015 4.47 0.000 .4006531 1.027091
_Iindustryc_5 | .8429119 .2246513 3.75 0.000 .4025843 1.28324
xtreg lnpatent c.index##i.hightech lnemployment assdebt curasslia fixass gdpg m2 i.industrycode i.citycode, r
----------------------------------------------------------------------------------
| Robust
lnpatent | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
index | .0021649 .0001114 19.44 0.000 .0019466 .0023832
1.hightech | 1.368523 .3511785 3.90 0.000 .6802258 2.05682
hightech#c.index | .0002928 .000214 1.37 0.171 -.0001267 .0007123
lnemployment | .4364049 .0193669 22.53 0.000 .3984465 .4743633
assdebt | -.000465 .0012858 -0.36 0.718 -.002985 .0020551
curasslia | .0005733 .0026383 0.22 0.828 -.0045975 .0057442
fixass | -.4755166 .0941779 -5.05 0.000 -.660102 -.2909312
gdpg | -.0941485 .0050511 -18.64 0.000 -.1040485 -.0842485
m2 | -.0150203 .0016675 -9.01 0.000 -.0182885 -.011752
|
industrycode |
2 | .6245305 .2605092 2.40 0.017 .1139419 1.135119
3 | .1594951 .2964314 0.54 0.591 -.4214998 .7404901
4 | .8046571 .3948312 2.04 0.042 .0308022 1.578512
5 | .700001 .2642034 2.65 0.008 .1821718 1.21783
6 | .0890626 .3325467 0.27 0.789 -.5627169 .7408421
reghdfe lnpatent c.index##i.hightech lnemployment assdebt curasslia fixass gdpg m2, a(citycode industrycode) vce(cl new_id)
(dropped 9 singleton observations)
note: 1bn.hightech is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09)
(MWFE estimator converged in 11 iterations)
note: 1.hightech omitted because of collinearity
----------------------------------------------------------------------------------
| Robust
lnpatent | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------------+----------------------------------------------------------------
index | .0021595 .0001169 18.48 0.000 .0019304 .0023886
1.hightech | 0 (omitted)
hightech#c.index | -.0001405 .000233 -0.60 0.547 -.0005973 .0003163
lnemployment | .4588536 .0183455 25.01 0.000 .4228823 .4948249
assdebt | -.0005759 .0012194 -0.47 0.637 -.0029669 .0018151
curasslia | .0021092 .0028603 0.74 0.461 -.0034993 .0077177
fixass | -.8463446 .1109573 -7.63 0.000 -1.063906 -.6287827
gdpg | -.096158 .0054216 -17.74 0.000 -.1067885 -.0855275
m2 | -.0135675 .0017321 -7.83 0.000 -.0169637 -.0101713
_cons | -.8228936 .1505112 -5.47 0.000 -1.118012 -.5277755
----------------------------------------------------------------------------------
------------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
--------------+---------------------------------------|
citycode | 273 0 273 |
industrycode | 74 1 73 |
------------------------------------------------------+