How to interpret year dummies?

[Drishta Tengur]

Hi I am having some difficulties to interpret my results. Upon using the xtreg command, there is a dot for my f statistic. Please find my commands and let me know how to improve on my commands and any tests that I need to undertake.

xtset country year, yearly
panel variable: country (unbalanced)
time variable: year, 1995 to 2016
delta: 1 year

xtreg lnPATR1 l.lnPATR1 preelection lninflation lngdp lnurbanisation left right lndependenc
> y lndependencyold lndependencyyoung lnpopbelow14 lnpop1564 eurozone eu lngovtexp lnwage cor
> porate i.year, fe vce(robust)
note: lnpop1564 omitted because of collinearity

Fixed-effects (within) regression Number of obs = 284
Group variable: country Number of groups = 19

R-sq: within = 0.6416 Obs per group: min = 11
between = 0.9697 avg = 14.9
overall = 0.9248 max = 18

F(19,18) = .
corr(u_i, Xb) = 0.7596 Prob > F = .

(Std. Err. adjusted for 19 clusters in country)
-----------------------------------------------------------------------------------
| Robust
lnPATR1 | Coef. Std. Err. t P>|t| [95% Conf. Interval]
------------------+----------------------------------------------------------------
lnPATR1 |
L1. | .6587392 .0961799 6.85 0.000 .4566728 .8608056
|
preelection | -.0310178 .0136497 -2.27 0.036 -.0596948 -.0023408
lninflation | -.0325141 .0145187 -2.24 0.038 -.0630168 -.0020115
lngdp | -.0011804 .0055523 -0.21 0.834 -.0128454 .0104845
lnurbanisation | .1640505 .3152123 0.52 0.609 -.498186 .8262871
left | .03482 .0129105 2.70 0.015 .0076961 .0619439
right | .0190707 .0164828 1.16 0.262 -.0155583 .0536997
lndependency | 1.461228 1.916417 0.76 0.456 -2.565014 5.487471
lndependencyold | -.2222128 .3620672 -0.61 0.547 -.9828877 .5384621
lndependencyyoung | -3.404956 5.231889 -0.65 0.523 -14.39675 7.586835
lnpopbelow14 | 3.037522 5.050372 0.60 0.555 -7.572916 13.64796
lnpop1564 | 0 (omitted)
eurozone | .006002 .0239933 0.25 0.805 -.044406 .05641
eu | .0082717 .0235983 0.35 0.730 -.0413066 .0578499
lngovtexp | .046253 .1302887 0.36 0.727 -.2274733 .3199793
lnwage | -.0294403 .0482326 -0.61 0.549 -.1307733 .0718926
corporate | .05361 .033382 1.61 0.126 -.016523 .123743
|
year |
1998 | -.0244217 .0118786 -2.06 0.055 -.0493776 .0005343
1999 | .0011749 .0581403 0.02 0.984 -.1209735 .1233232
2000 | .0529274 .028335 1.87 0.078 -.0066022 .112457
2001 | .0031761 .0343156 0.09 0.927 -.0689182 .0752704
2002 | .0114473 .0298427 0.38 0.706 -.0512499 .0741445
2003 | .0013123 .0330321 0.04 0.969 -.0680857 .0707102
2004 | .0127499 .0395587 0.32 0.751 -.0703598 .0958596
2005 | .0081906 .0438625 0.19 0.854 -.083961 .1003423
2006 | .00978 .0471501 0.21 0.838 -.0892786 .1088386
2007 | .0228113 .0544768 0.42 0.680 -.0916402 .1372628
2008 | .0266165 .0580768 0.46 0.652 -.0953982 .1486313
2009 | -.013349 .0501775 -0.27 0.793 -.1187679 .09207
2010 | .0118372 .0588119 0.20 0.843 -.111722 .1353965
2011 | .0383978 .0643255 0.60 0.558 -.0967451 .1735407
2012 | .0174724 .0642887 0.27 0.789 -.1175932 .1525379
2013 | .0159212 .0665736 0.24 0.814 -.1239447 .1557872
2014 | -.0155703 .0623691 -0.25 0.806 -.1466028 .1154623
2015 | -.0237408 .0621758 -0.38 0.707 -.1543672 .1068857
|
_cons | -2.450586 3.901355 -0.63 0.538 -10.64703 5.745856
------------------+----------------------------------------------------------------
sigma_u | .08805032
sigma_e | .06522337
rho | .64569778 (fraction of variance due to u_i)
-----------------------------------------------------------------------------------



estat vce

Covariance matrix of coefficients of xtreg model

| L.
e(V) | lnPATR1 preelect~n lninflat~n lngdp lnurbani~n left
-------------+------------------------------------------------------------------------
L.lnPATR1 | .00925056
preelection | .00028303 .00018632
lninflation | .0008081 .00011593 .00021079
lngdp | .00009681 .00001412 .00004119 .00003083
lnurbanisa~n | .00544424 .00057364 .00128829 .00048724 .09935881
left | -.00082195 -.00006705 -.00008154 1.674e-06 -.00017171 .00016668
right | -.00092253 .00002546 9.942e-06 -.00001268 -.00164531 .00011422
lndependency | -.12673987 -.00498992 -.01130479 -.00331713 -.18911086 .0096124
lndependen~d | .00629079 .0013319 .0008094 -.00018274 .02413327 -.00064149
lndependen~g | .36958225 .01026206 .03027411 .01090291 .52694909 -.02668208
lnpopbelow14 | -.35793228 -.00962656 -.02937112 -.01088457 -.51575811 .02564723
o.lnpop1564 | 0 0 0 0 0 0
eurozone | .00194195 .00006412 .00021985 .00005502 .00171765 -.00017737
eu | -.00026631 -.00009371 -.00006516 -8.555e-06 -7.611e-06 .00013208
lngovtexp | .00081164 .00039566 .00061069 .00046558 .00913111 .00023142
lnwage | .00112654 .00028134 .00030564 .00015762 .00693591 -.00021183
corporate | -.00187491 -.00003567 -.00008875 -.00006321 -.0034997 .00010444
1998.year | .00017029 .00002397 .00009021 .00001898 .00112998 4.286e-07
1999.year | -.00427037 .0000159 -.00051914 -.00016272 -.00231565 .00032436
2000.year | -.0021096 -.00010572 -.00020808 -.00004742 -.00124726 .00023713
2001.year | -.00301369 -.00012852 -.00033477 -.00006053 -.00197125 .00025008
2002.year | -.00257488 -.00009529 -.00024134 -4.975e-06 -.00239876 .00024409
2003.year | -.0026942 -.00005391 -.00021533 -1.088e-06 -.00371932 .00032619
2004.year | -.00310612 -.00021894 -.00038798 -.00009621 -.00611263 .000328
2005.year | -.00334278 -.00025162 -.00040397 -.0000881 -.0080449 .00037705
2006.year | -.00329911 -.00028424 -.00042775 -.00013026 -.00706709 .0003862
2007.year | -.00365143 -.00032634 -.00050478 -.00017564 -.0098193 .00032952
2008.year | -.00336512 -.00034143 -.00045006 -.00013613 -.01273264 .00036455
2009.year | -.00336869 -.00036446 -.00053906 -.00012856 -.00915056 .00035083
2010.year | -.00353769 -.00044863 -.00052143 -.00013878 -.01204912 .00036307
2011.year | -.00409737 -.00044896 -.00059323 -.00018369 -.01173881 .00040223
2012.year | -.00436973 -.00034975 -.00050264 -.00014313 -.01279571 .00040469
2013.year | -.00329465 -.00039733 -.00032916 -.00011663 -.01359037 .00035859
2014.year | -.00340667 -.00033588 -.0003221 -.00012621 -.01278485 .00032806
2015.year | -.00346523 -.00023712 -.00020789 -.0000886 -.01100739 .00034053
_cons | .24042661 .00340934 .01530802 .00437671 -.07450756 -.01918052

| o.
e(V) | right lndepend~y lndepend~d lndepend~g lnpopbe~14 lnpop1564
-------------+------------------------------------------------------------------------
right | .00027168
lndependency | .0190229 3.6726546
lndependen~d | -.00153643 -.43527597 .13109263
lndependen~g | -.05481853 -9.8022833 .86738836 27.372663
lnpopbelow14 | .05242654 9.3423777 -.75431928 -26.381454 25.50626
o.lnpop1564 | 0 0 0 0 0 0
eurozone | -.00018273 -.02653169 .0016206 .07600528 -.07335694 0
eu | .00006746 -.00314496 .00312021 -.00004137 .00274918 0
lngovtexp | .00017341 .03255203 -.01356637 -.04986786 .04232329 0
lnwage | -.00026844 -.0161724 -.00290996 .0616608 -.06322536 0
corporate | .00032547 .0258133 -.000593 -.08137367 .07847253 0
1998.year | 6.625e-06 -.00554862 .00072924 .01294138 -.01232616 0
1999.year | .00039832 .05744091 .00079809 -.1771371 .17364561 0
2000.year | .00027131 .03371015 -.00357474 -.08952714 .08388609 0
2001.year | .00023273 .03937663 -.00226769 -.11471963 .11112928 0
2002.year | .00021023 .03558728 -.00314215 -.09878332 .09475978 0
2003.year | .00031544 .03421776 -.00091632 -.10413128 .10223435 0
2004.year | .00036729 .05100375 -.00341599 -.14738216 .14287464 0
2005.year | .00041286 .0559748 -.00350668 -.16361532 .15977221 0
2006.year | .00041554 .0503494 -.00220892 -.15227804 .1492299 0
2007.year | .00047384 .06958922 -.00531748 -.20101148 .19501956 0
2008.year | .00050492 .05968951 -.0047148 -.17426754 .16962338 0
2009.year | .00030224 .05908468 -.00448722 -.16920372 .16487998 0
2010.year | .00043027 .06842434 -.00699632 -.19208587 .18589402 0
2011.year | .00050491 .07424428 -.00679954 -.21237518 .20588224 0
2012.year | .00060864 .08277336 -.00835041 -.23351337 .22511036 0
2013.year | .00066324 .07041324 -.00912454 -.19350762 .184697 0
2014.year | .00062384 .07192719 -.00784606 -.20425439 .19689895 0
2015.year | .00074574 .08557085 -.01123501 -.23441057 .22360102 0
_cons | -.03173693 -6.6714525 .55972536 18.557187 -17.88497 0

| 1998.
e(V) | eurozone eu lngovtexp lnwage corporate year
-------------+------------------------------------------------------------------------
eurozone | .00057568
eu | .00004653 .00055688
lngovtexp | .00067822 .00035896 .01697513
lnwage | .00017452 -.00068882 .00327013 .00232638
corporate | -.00035992 -.0000474 -.00240466 -.00077452 .00111436
1998.year | .00005179 -.00002086 -.0001181 .00006535 .00002244 .0001411
1999.year | -.00113603 .00017369 -.000907 -.00073247 .00055901 -.00021877
2000.year | -.00047706 .00002055 -.00116903 -.00055669 .00067607 -.00003049
2001.year | -.0007258 .00001911 -.00071185 -.00043527 .00056138 -.0000513
2002.year | -.00056185 -.00005576 -.00015887 -.00020171 .00049255 -.00002826
2003.year | -.00054522 .00015073 .00030326 -.00048472 .00050459 -.00001363
2004.year | -.00068528 .00019415 -.00136938 -.00121938 .00084991 -.00008418
2005.year | -.00074108 .00027963 -.0009605 -.00129359 .00083568 -.00008247
2006.year | -.00079262 .00024363 -.00212753 -.00156007 .00097276 -.00003984
2007.year | -.00084945 .00019422 -.00276488 -.00185107 .00118438 -.00009994
2008.year | -.00073572 .00028651 -.0027333 -.0020785 .00126549 -.00005356
2009.year | -.000776 .00017662 -.00227035 -.00155291 .00092437 -.00011989
2010.year | -.00079626 .00015898 -.00278236 -.00189244 .00121162 -.00010167
2011.year | -.00099715 .00021246 -.00303374 -.00207766 .00128878 -.00009506
2012.year | -.00099075 .00003362 -.00270197 -.00182517 .00146703 -.00006661
2013.year | -.00065076 .00016536 -.00321618 -.00217174 .0016338 .000019
2014.year | -.00072546 .00011009 -.00260129 -.00192683 .00146622 .00002364
2015.year | -.00069797 .00005973 -.00066988 -.00150023 .00133122 .00003816
_cons | .04473751 -.00336196 -.15832917 -.00027982 -.02850309 .00703866

| 1999. 2000. 2001. 2002. 2003. 2004.
e(V) | year year year year year year
-------------+------------------------------------------------------------------------
1999.year | .0033803
2000.year | .0008921 .00080287
2001.year | .00161868 .00069366 .00117756
2002.year | .0010768 .00065027 .00090673 .00089059
2003.year | .0011651 .00059596 .00089169 .00087088 .00109112
2004.year | .00149768 .00088218 .00115219 .00095513 .00107325 .00156489
2005.year | .00150288 .00085279 .00120186 .00103365 .00124138 .00168476
2006.year | .00163742 .00093977 .00130149 .00101809 .00121084 .0018037
2007.year | .0018201 .0010762 .00143466 .0010915 .00120152 .00206522
2008.year | .00136667 .0010323 .00130484 .00110518 .0013372 .00212624
2009.year | .00149696 .00096585 .00131072 .00113119 .0012251 .0018884
2010.year | .00131266 .00109654 .00136608 .00119884 .00129481 .00214175
2011.year | .00190977 .0011831 .0016962 .00128154 .00141249 .00239406
2012.year | .00176553 .00129347 .00165108 .00143863 .00156288 .00239358
2013.year | .00074192 .00124516 .00114168 .0010896 .00123216 .00214569
2014.year | .00108558 .00110935 .0012805 .00112113 .00128368 .0021272
2015.year | .00113775 .00108892 .0012419 .00110038 .00126957 .00199673
_cons | -.11771106 -.05198049 -.06984541 -.05997404 -.06061522 -.07339917

| 2005. 2006. 2007. 2008. 2009. 2010.
e(V) | year year year year year year
-------------+------------------------------------------------------------------------
2005.year | .00192392
2006.year | .0019652 .00222313
2007.year | .00221861 .00245169 .00296772
2008.year | .00237529 .0025574 .00301453 .00337291
2009.year | .00208608 .0022225 .00259204 .0027203 .00251778
2010.year | .00238339 .00255355 .00305914 .00330564 .00285951 .00345884
2011.year | .00261465 .00290477 .00344409 .00358094 .00306081 .00365636
2012.year | .00262935 .00282292 .00337647 .00359434 .00303504 .00367885
2013.year | .00237882 .00256337 .00316389 .00360526 .00274777 .00362079
2014.year | .00237575 .00257321 .00313211 .00346323 .00269938 .0034745
2015.year | .00223481 .0023368 .00289054 .00313222 .00235973 .00311464
_cons | -.08029724 -.06828154 -.08968291 -.05575483 -.0746877 -.07357705

| 2011. 2012. 2013. 2014. 2015.
e(V) | year year year year year _cons
-------------+------------------------------------------------------------------------
2011.year | .00413777
2012.year | .00398939 .00413304
2013.year | .00375114 .00396366 .00443204
2014.year | .00372497 .00386357 .0040447 .0038899
2015.year | .00340625 .00362573 .0037826 .00369257 .00386582
_cons | -.08605384 -.09818809 -.05982791 -.0754702 -.11264465 15.220568

[Clyde Schechter]

This happens because you have 19 predictor variables in your model and you only have 19 countries in the data, so there are not enough degrees of freedom left to calculate the F-statistic when using the cluster robust VCE. The individual variable tests, however, are perfectly OK and you can interpret them just as you would regardless of the overall model F statistic.

However, the cluster robust variance estimator has nice properties with large numbers of clusters but performs poorly with small numbers of clusters. While there is no universally accepted threshold of how many clusters are needed, many would say that you shouldn't use vce(cluster) with only 19 clusters in any case.

[Drishta Tengur]

Clyde, thank you for replying.
My dataset actually consists of 27 countries over the period 1995-2016. However, I guess because of the country and year fixed effects, the sample size is reduced. Also, even if I use fe robust(cluster), the f statistic does not show up. I use the command robust as the Wooldridge test shows serial correlation among my variables. Can you please advise me how I should proceed in this case? Or should I ignore the f statistic and interpret my results as usual?

[Clyde Schechter]

No, you misunderstood me. I said not to use robust cluster. [Your command says vce(robust), but if you read the manual, you will see that for -xtreg, fe- (and not for other commands), -vce(robust)- is reinterpreted by Stata to mean -vce(cluster panelvar)-. So by specifying -vce(robust)- you are, whether you intend it or not, specifying the cluster robust standard error.]

What I'm telling you is that if you want that overall model F statistic, you have to just use the ordinary VCE. (That is, just don't specify any -vce()- option and go with the default.) Now, even if you don't really care about the overall model F statistic, I'm saying you probably should not use vce(cluster panelvar) in any case because your estimation sample only has 19 panels and that is probably too small to get valid results with. So I'm saying just remove the -vce()- option from your command and go with that. Alternatively, if serial correlation is a real concern, as you suggest it is, then consider using -vce(bootstrap)-. If you go that route, be sure to thoroughly read the manual chapter on -bootstrap- so that you understand what bootstrap options you need to specify in order for the bootstrap to properly respect the clustering of your observations within countries.

[Drishta Tengur]

Okay thank you for the clarification. I shall read the manual on -bootstrap-

[Drishta Tengur]

I get an error when I try -vce(bootstrap)-


xtreg lnPATR1 l.lnPATR1 preelection lninflation lngdp lnurbanisation left righ
> t lndependency lndependencyold lndependencyyoung lnpopbelow14 lnpop1564 eurozo
> ne eu lngovtexp lnwage corporate i.year, fe vce(bootstrap, nodots reps(200))
note: lnpop1564 omitted because of collinearity
insufficient observations to compute bootstrap standard errors
no results will be saved
r(2000);


Can you please tell me what is the problem or if the command is wrong?

[Clyde Schechter]

Your -vce(bootstrap)- option requires specification of the -cluster- and -idcluster- suboptions in order for it to work properly with panel data.

See the example on p. 230ff of the [R] manual. There they are llustrating it with the -bootstrap- command, but the -bootstrap- options are also used as suboptions in -vce(bootstrap)-.

[Drishta Tengur]

This is the results that I get:

xtreg lnPATR1 l.lnPATR1 preelection lninflation lngdp lnurbanisation left right lndependency lnd
> ependencyold lndependencyyoung lnpopbelow14 lnpop1564 lnpop65andabove eurozone eu lngovtexp lnwa
> ge plurality proportional lnpit lncorporate i.year, fe vce(bootstrap) cluster(country)
note: lnpop1564 omitted because of collinearity
note: lnpop65andabove omitted because of collinearity
(running xtreg on estimation sample)

Bootstrap replications (50)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx    50
insufficient observations to compute bootstrap standard errors
no results will be saved
r(2000);

Can you please tell me what is wrong? Is there something wrong with my data or I am using the wrong command?

[Clyde Schechter]

For starters, as I have said in earlier posts, you are not specifying -vce(bootstrap)- correctly. You need to specify suboptions to tell Stata how to handle the clustering in the data. So it should be something like -vce(bootstrap, cluster(country) idcluster(bs_country))-. Also, don't specify cluster(country) as a separate option in -xtreg-; it should only be inside the -vce(bootstrap,...)- option.

Try that. If that doesn't work, post some example data.

[Drishta Tengur]

Following what you suggested, these are the results:

. xtreg lnPATR1 l.lnPATR1 preelection lninflation lnunemployment lngdp lnurbanis
> ation left right lndependency lndependencyold lndependencyyoung lnpopbelow14 l
> npop1564 lnpop65andabove eurozone eu lngovtexp lnwage plurality proportional l
> npit lncorporate i.year, fe vce(bootstrap, cluster(country) idcluster(idcountr
> y))
note: lnpop1564 omitted because of collinearity
note: lnpop65andabove omitted because of collinearity
option cluster() not allowed
r(198);

[Clyde Schechter]

OK, then try this:

bootstrap, group(country) cluster(country) idcluster(newcountry: xtreg lnPATR1 l.lnPATR1...., fe

I'm sorry, I was under the impression that the cluster() etc. bootstrap options could be specified as suboptions in -vce(bootstrap)-, but that is apparently not the case.

[Drishta Tengur]

Do you think that there is something wrong with my data? Or I am not understanding how to use the code. The results are as follows:

generate newcountry = country

. xtset newcountry year, yearly
       panel variable:  newcountry (strongly balanced)
        time variable:  year, 1995 to 2016
                delta:  1 year

. bootstrap, group(country) cluster(country) idcluster(newcountry): xtreg lnPATR1 l.lnPATR1 preele
> ction lninflation lnunemployment lngdp lnurbanisation left right lndependency lndependencyold ln
> dependencyyoung lnpopbelow14 lnpop1564 lnpop65andabove eurozone eu lngovtexp lnwage plurality pr
> oportional lnpit lncorporate i.year, fe
(running xtreg on estimation sample)

Bootstrap replications (50)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx    50
insufficient observations to compute bootstrap standard errors
no results will be saved
r(2000);

end of do-file

r(2000);

[Clyde Schechter]

I think the code looks OK. If somebody else sees a problem with it, I hope they will chime in. There may be some problem with the data set, but not something that one could easily troubleshoot from afar.

Try adding the -noisily- option to -bootstrap-. That will give you separate output for each of the 50 attempted replications. That output may make it clear what's going wrong.

[Drishta Tengur]

Using -noisily-:

xtreg lnPATR1 l.lnPATR1 preelection lninflation lnunemployment lngdp lnurbanisation left right l
> ndependency lndependencyold lndependencyyoung lnpopbelow14 lnpop1564 lnpop65andabove eurozone eu
>  lngovtexp lnwage plurality proportional lnpit lncorporate i.year, fe 
note: lnpopbelow14 omitted because of collinearity
note: lnpop1564 omitted because of collinearity
note: plurality omitted because of collinearity
note: proportional omitted because of collinearity

Fixed-effects (within) regression               Number of obs      =       223
Group variable: newcountry                      Number of groups   =        19

R-sq:  within  = 0.7378                         Obs per group: min =        10
       between = 0.6651                                        avg =      11.7
       overall = 0.6379                                        max =        14

                                                F(32,172)          =     15.12
corr(u_i, Xb)  = 0.5122                         Prob > F           =    0.0000

-----------------------------------------------------------------------------------
          lnPATR1 |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
------------------+----------------------------------------------------------------
          lnPATR1 |
              L1. |   .5298407   .0654442     8.10   0.000     .4006636    .6590178
                  |
      preelection |   -.044028   .0133252    -3.30   0.001      -.07033    -.017726
      lninflation |  -.0218778    .016651    -1.31   0.191    -.0547444    .0109888
   lnunemployment |   .0517565     .02883     1.80   0.074    -.0051496    .1086626
            lngdp |  -.0042592   .0073415    -0.58   0.563    -.0187503    .0102319
   lnurbanisation |  -.4237397   .2974708    -1.42   0.156    -1.010903    .1634236
             left |   .0455591   .0283701     1.61   0.110    -.0104393    .1015575
            right |    .034423     .03043     1.13   0.260    -.0256414    .0944873
     lndependency |   5.379288   3.337524     1.61   0.109    -1.208491    11.96707
  lndependencyold |  -8.415692   10.13055    -0.83   0.407     -28.4119    11.58052
lndependencyyoung |  -1.880651   .6507028    -2.89   0.004    -3.165042     -.59626
     lnpopbelow14 |          0  (omitted)
        lnpop1564 |          0  (omitted)
  lnpop65andabove |   6.604941   10.21041     0.65   0.519    -13.54889    26.75877
         eurozone |   .0430169   .0278201     1.55   0.124    -.0118958    .0979297
               eu |   .0026391   .0312175     0.08   0.933    -.0589795    .0642578
        lngovtexp |   -.277875   .1516345    -1.83   0.069    -.5771791    .0214292
           lnwage |   -.070696   .0815346    -0.87   0.387    -.2316333    .0902413
        plurality |          0  (omitted)
     proportional |          0  (omitted)
            lnpit |   .0358701    .065669     0.55   0.586    -.0937508    .1654911
      lncorporate |   .1508051   .0392377     3.84   0.000     .0733556    .2282546
                  |
             year |
            1998  |          0  (empty)
            1999  |  -.0445785   .0270744    -1.65   0.101    -.0980194    .0088624
            2000  |   .1014618   .0274082     3.70   0.000      .047362    .1555616
            2001  |    .029823   .0297075     1.00   0.317    -.0288152    .0884613
            2002  |   .0506508   .0299201     1.69   0.092    -.0084072    .1097087
            2003  |   .0385666    .036197     1.07   0.288     -.032881    .1100142
            2004  |   .0765644   .0375634     2.04   0.043     .0024198    .1507091
            2005  |   .0841406   .0404896     2.08   0.039     .0042201     .164061
            2006  |   .0753827   .0435765     1.73   0.085    -.0106309    .1613963
            2007  |   .1039117   .0532412     1.95   0.053    -.0011785     .209002
            2008  |   .1354153   .0678329     2.00   0.047     .0015232    .2693074
            2011  |   .1431194   .0664564     2.15   0.033     .0119443    .2742944
            2012  |   .1091052   .0721911     1.51   0.133    -.0333893    .2515998
            2014  |   .1663344   .0744817     2.23   0.027     .0193185    .3133504
            2015  |   .1125414   .0641588     1.75   0.081    -.0140986    .2391815
                  |
            _cons |  -2.625406   8.348671    -0.31   0.754    -19.10445    13.85364
------------------+----------------------------------------------------------------
          sigma_u |  .24466719
          sigma_e |  .05040427
              rho |  .95928713   (fraction of variance due to u_i)
-----------------------------------------------------------------------------------
F test that all u_i=0:     F(18, 172) =     2.52             Prob > F = 0.0011
collinearity in replicate sample is not the same as the full sample, posting missing values

"collinearity in replicate sample is not the same as the full sample, posting missing values" repeats in all the replications.

[Clyde Schechter]

Aha! So, what's happening is that in each bootstrap replication, you are getting different years included, and that changes the colinearity relationships among the year indicator variables. I don't think there is any way to fix that problem, so the -bootstrap- approach is simply not going to work, I'm afraid. I'm sorry for leading you down this path and wasting your time trying to get it to work.

Putting that in the context of your original question. It seems that you would like to have all of the following at once:

1. Standard errors that are robust to serial autocorrelation.

2. Inclusion of year indicators in the model.

3. An F-test of the model as a whole.

A data set with more countries would make that possible Your data set, with only 18, means that you cannot have all three of these things. I think you will have to choose two out of the three. My own sense is that since the cluster robust standard errors are of dubious validity with only 18 countries in the first place, I would give up on that part, just use the ordinary -xtreg, fe-, and then you should be OK. If you are unwilling to give up errors robust to autocorrelation, then something else would have to go. I guess, if I were in your shoes, I'd give up the F-test of the model as a whole because I don't think it's a very useful statistic anyway. But that's your decision to make.