Hi all,
I'm very confused by a set of results I have for a Fixed-Effects linear model. I am predicting the effect of job mobility on log wages. I control for a number of variables including age, industry, occupation, the size of the firm etc etc. For some reason the estimates for job change are always negative, suggesting a promotion or a quit brings a decline in wages. This doesn't really make sense, theory-wise. I've followed the -xtreg, fe- examples in the Stata syntax and I specify the models in exactly the same way. Am I doing something wrong?
I have estimated several models like the one below using objective measures (like satisfaction), these always predict positive effects. It's very confusing.
I'm very confused by a set of results I have for a Fixed-Effects linear model. I am predicting the effect of job mobility on log wages. I control for a number of variables including age, industry, occupation, the size of the firm etc etc. For some reason the estimates for job change are always negative, suggesting a promotion or a quit brings a decline in wages. This doesn't really make sense, theory-wise. I've followed the -xtreg, fe- examples in the Stata syntax and I specify the models in exactly the same way. Am I doing something wrong?
I have estimated several models like the one below using objective measures (like satisfaction), these always predict positive effects. It's very confusing.
Code:
. xtreg ln_ z_worksat i.M_ ib2.agecat ib1.perm i.qual ///
> i.child ib3.jbsize1 i.wave ///
> ib9.industry i.isco10 ///
> , fe
Fixed-effects (within) regression Number of obs = 32,974
Group variable: pid Number of groups = 6,963
R-sq: Obs per group:
within = 0.0695 min = 1
between = 0.1028 avg = 4.7
overall = 0.0865 max = 8
F(44,25967) = 44.08
corr(u_i, Xb) = 0.1246 Prob > F = 0.0000
----------------------------------------------------------------------------------------------------------
ln_income | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-----------------------------------------+----------------------------------------------------------------
z_worksat | -.0031597 .0028432 -1.11 0.266 -.0087324 .002413
|
M_event |
Changed employer- voluntary | -.1155245 .0086815 -13.31 0.000 -.1325407 -.0985084
Changed employer- involuntary | -.081976 .0170219 -4.82 0.000 -.1153399 -.0486121
Changed employer- other reason | -.0611439 .0122549 -4.99 0.000 -.0851642 -.0371236
Changed job, kept employer- voluntary | -.1012232 .0103597 -9.77 0.000 -.1215288 -.0809176
Changed job, kept employer- involuntary | -.0427563 .0369454 -1.16 0.247 -.1151713 .0296586
Changed job, kept employer- other | -.0246434 .0164745 -1.50 0.135 -.0569343 .0076476
|
agecat |
1 | -.0531777 .0124832 -4.26 0.000 -.0776454 -.0287099
3 | .0275001 .0109713 2.51 0.012 .0059957 .0490045
4 | -.6151886 .0444107 -13.85 0.000 -.7022361 -.5281412
|
permanent |
Temporary | -.0239266 .0164654 -1.45 0.146 -.0561997 .0083464
|
qualification |
2 | -.0360106 .0507203 -0.71 0.478 -.1354251 .0634039
3 | -.1045939 .0569402 -1.84 0.066 -.2161998 .007012
4 | -.0927125 .0590169 -1.57 0.116 -.208389 .022964
5 | -.13572 .0595599 -2.28 0.023 -.2524608 -.0189793
6 | -.0750151 .0613486 -1.22 0.221 -.1952618 .0452316
|
child |
1 | -.0591078 .0096059 -6.15 0.000 -.0779358 -.0402798
2 | -.1081578 .012247 -8.83 0.000 -.1321626 -.084153
3 | -.1609545 .0197324 -8.16 0.000 -.199631 -.122278
................................................omitted...........................................................................................
|
_cons | 10.46623 .0550627 190.08 0.000 10.3583 10.57416
-----------------------------------------+----------------------------------------------------------------
sigma_u | .6242047
sigma_e | .33327824
rho | .77816442 (fraction of variance due to u_i)
----------------------------------------------------------------------------------------------------------
F test that all u_i=0: F(6962, 25967) = 11.08 Prob > F = 0.0000
.