Dear Statalists,
I have 3 questions on the use of VIF in a model with dummies or fixed effects.
The model is: y_it=x_ct+e_it+fe_t+fe_c+fe_industry.
The observations are uniquely defined by each firm i and year t. y_it is the profitability of overseas subsidiary i in year t at host country c from parent country A (across the sample, A is the same country). x_ct is a host country-year characteristic. Because x and y do not have the same dimension, the standard errors are clustered by host country c. I did not add unit fixed effects fe_i because my x's dimension is only country-year. I am worried that adding fe_i will absorb too much variation in the data. Please let me know if there is anything wrong with this specification.

I have two questions on the use of VIF in such kinds of 'fixed effect' models.
1. I notice in the forum and elsewhere that VIF should not be applied to the fixed-effect model, but have not found any formal reference. Do you agree with this view? If yes, do you have any references supporting this view?
2. VIF is very large after my regression as shown below. If I do not need to worry about it, do you have any references supporting this view?

Thanks!
Kailin

Code:
* Normally I use this commend: reghdfe y x `controls' ,a(year  countryid USSIC_code) vce(cluster countryid)
* But to calculate VIF, I have to use reg.     
reg  y x  `controls' i.year  i.countryid i.USSIC_code, cluster(countryid)
estat vif   
* The result is as follows, and control_ct means host country-year control variables.

    Variable |       VIF       1/VIF  
-------------+----------------------
     x_ct    |    162.24    0.006164
control1_it |      2.13    0.468649
control2_ct |     24.97    0.040050
control3_ct |  18592.15    0.000054
control4_ct |      2.73    0.366932
control5_ct |      6.22    0.160692
control6_ct |     94.34    0.010600
control7_ct |    265.49    0.003767

……（and a long list of VIF for fixed effects）

-------------+----------------------
    Mean VIF |    189.83

*Here is the correlation matrix of the variables calculated by pwcorr command. I think most are fine. I may be worried about the 0.6554 a little.

         |  x              con1      con2        con3    con4        con5      con6     con7
--------------------------------------------------------------------------------------------
x       |   1.0000 
con1 |   0.2398   1.0000 
con2 |   0.1508   0.1130   1.0000 
con3 |  -0.3978  -0.2978  -0.3933   1.0000 
con4 |   0.1723   0.2018   0.0507  -0.2647   1.0000 
con5 |  -0.3750   0.0711   0.1803   0.3199  -0.1466   1.0000 
con6 |  -0.4208  -0.3016  -0.5159   0.6554  -0.3842   0.1288   1.0000 
con7 |   0.4328   0.0842   0.1317  -0.3610   0.4450  -0.3889  -0.5276 1.0000