xtscc command for robust standard errors in unbalanced panels (autoregressive model)

[Colm Bairead]

Dear all,

I am currently trying to estimate a dynamic panel model using the xtscc command to provide robustness in terms of heteroscedasticity and correlation however every time I try to execute the command I get the following message:

xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta yr*
invalid syntax
r(197);

mllev is dependent variable. Does anyone know what the issue could be?? Has anyone had this experience? (The data is set as xtset).

Thanks in advance.

Colm.

[Joro Kolev]

If you want provide some minimal data subsample (see -dataex-) which generates the error, and we can toy around and see what goes wrong.

My first guess would be that some combination of your -xtset- ing the data, and your use of lag operator causes the problem.

[Colm Bairead]

Hi Joro, appreciate your response, I am new to Stata and especially the forum and hence not sure if I am in compliance. I used dataex to generate the below data as you requested however it does not allow me to use the dependent variable lag. An example of code is

xtreg mllev l.mllev rate ebitta size tangta infl medlev cashassets, fe. When I type in xtscc followed by these variables I get the error message i.e. invalid syntax. I have downloaded xtscc and is up to date. Bit of background is I am estimating the effects of credit ratings on leverage however rating is an endogenous variable and hence the lagged regressor. Would be grateful if you would have any recommendations. Colm.

dataex mllev rate cashassets ebitta rate

----------------------- copy starting from the next line -----------------------
[CODE]
* Example generated by -dataex-. To install: ssc install dataex
clear
input float mllev byte rate float(cashassets ebitta)
.473 15 .04 .081
.421 15 .025 .148
.535 15 .021 .11
.466 14 .059 .048
.268 14 .028 .064
.362 14 .082 .134
.413 14 .033 .031
.383 13 .016 .038
.321 13 .029 .116
. 17 .161 .
.877 17 .101 .115
.771 17 .137 .157
.856 17 .147 .166
.791 17 .162 .154
.762 17 .128 .153
.754 17 .12 .151
.797 17 .116 .144
. 7 .11 .029
. 7 .046 .013
.577 8 .053 .032
.6 9 .093 .071
.584 8 .083 .009
.574 11 .065 .036
.327 11 .229 .
--more--

[Joro Kolev]

I am also not a big specialist on this -dataex-, but apparently something did not work. Your data is truncated, does not import properly (when I copy and paste I get missing values for your dep var mllev) , and more importantly it does not include the two variables which you use to

xtset panelvar timevar

your data.

[Colm Bairead]

Joro,
I have the first 100 observations from 493 observations posted here, I can paste them all if needed. I used -xtset- code year. Note that the panels are highly unbalanced.
Appreciate your effort with this, Colm.
----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. To install: ssc install dataex
clear
input long code int year float mllev byte rate float(cashassets ebitta size tangta infl medlev)
 15322 2006 .473 15  .04  .081 16.998 .638 2.334 20.52
 15322 2007 .421 15 .025  .148 17.086 .638 2.321 21.95
 15322 2008 .535 15 .021   .11  17.45 .654 3.613 26.52
 15322 2009 .466 14 .059  .048 17.489  .62 2.166 25.28
 15322 2010 .268 14 .028  .064 17.611 .661 3.286 20.27
 26923 2003 .362 14 .082  .134 15.479 .748 1.363 25.14
 26923 2004 .413 14 .033  .031 15.379 .791 1.345  25.5
 26923 2009 .383 13 .016  .038 14.366 .656 2.166 26.19
 26923 2010 .321 13 .029  .116 14.361 .627 3.286 21.28
 50287 2008   .6  9 .093  .071 15.883 .394 3.613 31.49
 50287 2010 .574 11 .065  .036 15.871 .397 3.286 23.73
 74190 2010 .397 10  .19  .004 16.372 .627 3.286 32.08
 77128 2009 .464 13 .107  .043 17.529 .103 2.166 25.28
 77128 2010 .438 13 .095  .034 17.744 .152 3.286 20.27
135229 2002 .514 14 .058  .106 14.379 .454 1.256 34.74
135229 2003 .475 14 .047   .09 14.363 .449 1.363 33.51
135229 2004 .467 14 .069  .108 14.411  .44 1.345 30.35
135229 2005  .48 14  .08  .105 14.469 .437  2.05 34.64
135229 2006 .479 14  .09  .095 14.553 .443 2.334 29.18
135229 2007 .477 13 .165  .094 14.749 .417 2.321 28.75
135229 2008 .575 13 .067  .059 15.388 .399 3.613 34.25
135229 2009 .489 13 .056  .062 15.585 .409 2.166  30.4
135229 2010 .538 13 .066  .072 15.531 .411 3.286 31.22
135565 2010 .293 13 .243  .066 13.799 .423 3.286 31.22
137668 2006 .765 12 .019  .107 15.609 .889 2.334 43.88
255049 2001 .443 15 .082  .056 15.969 .242 1.236 40.33
255049 2002 .295 15 .045  .059 16.024  .26 1.256 36.03
255049 2003 .403 15 .036  .058 15.953 .204 1.363 36.04
255049 2004 .429 15 .032   .06 15.939 .216 1.345 36.75
255049 2005 .415 15 .038  .037 15.928 .215  2.05 39.82
255049 2006 .386 15 .122  .074 15.768 .107 2.334 43.88
255049 2007 .359 15 .136  .088 15.639 .093 2.321 44.21
255049 2008  .27 15 .087  .099 15.712 .069 3.613 44.85
255049 2009 .265 15 .084  .118 15.809 .072 2.166 44.22
255049 2010 .245 15 .082  .123  15.89 .073 3.286 40.27
258092 2005 .522 13 .136  .212 12.866 .299  2.05 24.03
258107 2009 .272 12 .066  .186 14.418 .108 2.166 41.65
258107 2010 .566 12  .06  .088 14.409 .109 3.286 32.51
265148 2001 .619 14 .025  .073 15.022 .881 1.236 21.88
265148 2002 .714 14 .026  .073 15.184 .889 1.256 23.36
265148 2003 .646 16 .055  .047 16.471 .715 1.363 15.72
265148 2004 .529 16 .024  .083 16.486 .753 1.345 41.45
265148 2005 .499 16 .042  .084 16.617  .75  2.05 42.88
265148 2006 .515 16 .038  .092 16.558 .768 2.334 46.12
265148 2007 .444 16 .028  .086 16.626 .779 2.321 31.66
265148 2008 .541 16 .019  .064 16.893 .792 3.613 31.73
265148 2009 .614 15 .015  .069 16.924 .811 2.166 30.09
265148 2010 .554 15 .011  .065 16.998 .826 3.286 26.83
319410 2000 .526 14 .242  .094  15.03 .335  .785  39.2
319410 2001 .549 14 .064 -.087 14.727 .465 1.236 36.59
319410 2002 .519 13 .061  .042 14.718  .45 1.256 34.74
319410 2004 .247 13 .308  .081 14.253 .399 1.345 30.35
319410 2005 .238 14 .119  .123 13.976 .545  2.05 34.64
319410 2006 .272 14 .158  .115 14.063 .554 2.334 29.18
319410 2008 .399 13 .215  .134 14.135 .474 3.613 34.25
319410 2009  .23 13 .186  .115 14.224 .522 2.166  30.4
319410 2010  .39 13 .248  .081 14.298 .491 3.286 31.22
319608 2000 .072 21 .244  .108 16.319 .272  .785 27.06
319608 2001 .078 21 .214  .204 16.321 .303 1.236 15.34
319608 2004 .093 21 .202  .194   16.4 .318 1.345 12.43
319608 2005 .109 21 .278  .265 16.441 .294  2.05 13.72
319608 2007 .338 19 .129  .195 16.979 .177 2.321 24.59
319608 2008 .366 19 .103  .163 17.256 .155 3.613 31.17
319608 2009 .277 19 .213   .22 17.318 .136 2.166 26.69
319608 2010 .289 19  .23  .217 17.383 .127 3.286 23.63
410124 2006 .516 15 .032  .073  15.78 .135 2.334 23.87
410124 2007 .366 15 .166    .1 14.891  .09 2.321 24.43
410124 2008 .512 15 .021  .091 15.137 .081 3.613 27.72
410124 2009 .461 15 .024  .102 15.276 .078 2.166 25.85
410124 2010 .434 15 .031  .087 15.271 .069 3.286 24.67
673176 2001 .472 15 .136  .087 16.193 .478 1.236 21.11
673176 2002 .518 15 .145  .095 16.172 .474 1.256 15.48
673176 2003  .49 15 .117  .085 16.219 .465 1.363 16.33
673176 2004 .503 15 .111  .092 16.289 .448 1.345 22.64
673176 2005 .481 15 .098  .086 16.562 .438  2.05 16.29
673176 2006 .492 15 .083  .092 16.698 .419 2.334 22.22
673176 2007 .515 15 .069  .105 16.783 .423 2.321 25.67
673176 2008 .494 15 .045  .104  16.85 .428 3.613    31
673176 2009 .479 15 .072  .074 16.809 .428 2.166  22.7
673176 2010 .437 15 .085  .048 16.864 .422 3.286 21.64
870181 2000 .512 18 .116  .194 15.985 .564  .785 38.82
870181 2001 .488 18 .027  .055 16.115 .563 1.236 40.28
870181 2002 .492 17 .012  .018  16.67 .525 1.256 44.91
870181 2003 .514 17 .024  .087 17.031 .676 1.363 49.22
870181 2004  .52 17 .026  .087 16.968 .714 1.345 53.18
870181 2005 .469 17 .027  .093 17.023 .718  2.05 51.67
870181 2006 .461 17 .069  .101 17.065 .735 2.334 49.76
870181 2007 .605 16  .14  .077 17.162 .666 2.321 52.76
870181 2008 .527 16 .072  .076 17.448 .643 3.613 52.64
870181 2009 .583 16  .08  .064 17.607 .666 2.166 52.36
870181 2010 .567 16 .054  .074 17.589 .708 3.286 51.25
871674 2009 .473 14 .079  .075 15.422  .11 2.166 25.85
871674 2010 .488 14 .083  .098 15.466 .111 3.286 24.67
888276 2001 .107 17 .074  .069 15.772 .291 1.236 10.15
888276 2002 .121 17  .04  .076 15.976 .319 1.256 14.61
888276 2003 .127 17 .119  .092 15.973 .315 1.363 12.75
888276 2004 .105 17  .14  .098 15.969 .329 1.345 10.72
888276 2005 .214 17 .265   .16  16.38 .282  2.05  8.17
888276 2006 .275 17 .134  .061   16.2 .339 2.334  9.45
888276 2007  .21 17 .177  .133 16.286 .331 2.321  9.86
end
label var code "Code"
label var year "Year"
label var mllev "MLLEV"
label var rate "Rate"
label var cashassets "Cash/Assets"
label var ebitta "EBIT/TA"
label var size "Size"
label var tangta "Tang/TA"
label var infl "Infl"
label var medlev "Medlev (%)"

[Joro Kolev]

I am not able to replicate the error. The data seems to have been imported nicely this time. But there is not error:

. xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta

Regression with Driscoll-Kraay standard errors   Number of obs     =        75
Method: Pooled OLS                               Number of groups  =        14
Group variable (i): code                         F(  8,     9)     =   5760.81
maximum lag: 2                                   Prob > F          =    0.0000
                                                 R-squared         =    0.7462
                                                 Root MSE          =    0.0736

------------------------------------------------------------------------------
             |             Drisc/Kraay
       mllev |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       mllev |
         L1. |   .6869463   .0921094     7.46   0.000     .4785802    .8953123
             |
        rate |  -.0039528   .0060886    -0.65   0.532    -.0177261    .0098206
  cashassets |  -.0394985   .1958415    -0.20   0.845    -.4825227    .4035258
      ebitta |  -.2986732   .1302197    -2.29   0.047    -.5932506   -.0040958
        size |   -.003162   .0126859    -0.25   0.809    -.0318596    .0255356
      medlev |   .0007729   .0007527     1.03   0.331    -.0009299    .0024758
        infl |   .0159977   .0100178     1.60   0.145    -.0066641    .0386596
      tangta |   .0615299   .0419264     1.47   0.176    -.0333143    .1563741
       _cons |   .1859049   .1362775     1.36   0.206    -.1223763    .4941861
------------------------------------------------------------------------------

[Joro Kolev]

And here it is with the year variable:

. xtset code year
       panel variable:  code (unbalanced)
        time variable:  year, 2000 to 2010, but with gaps
                delta:  1 unit

. xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta i.year

Regression with Driscoll-Kraay standard errors   Number of obs     =        75
Method: Pooled OLS                               Number of groups  =        14
Group variable (i): code                         F( 16,     9)     =    127.00
maximum lag: 2                                   Prob > F          =    0.0000
                                                 R-squared         =    0.7573
                                                 Root MSE          =    0.0768

------------------------------------------------------------------------------
             |             Drisc/Kraay
       mllev |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       mllev |
         L1. |   .6943744   .0852869     8.14   0.000      .501442    .8873068
             |
        rate |  -.0110723   .0080324    -1.38   0.201    -.0292429    .0070983
  cashassets |     .03596    .215608     0.17   0.871    -.4517793    .5236993
      ebitta |  -.1544263   .1582091    -0.98   0.355    -.5123201    .2034675
        size |   .0042529   .0162563     0.26   0.800    -.0325213    .0410272
      medlev |   .0009085   .0007565     1.20   0.260    -.0008029    .0026198
        infl |   .0640381   .0476532     1.34   0.212    -.0437609    .1718372
      tangta |   .0561207   .0448327     1.25   0.242    -.0452979    .1575393
             |
        year |
       2000  |          0  (empty)
       2001  |   .1167228   .0920622     1.27   0.237    -.0915364     .324982
       2002  |   .1080642   .1052883     1.03   0.332    -.1301144    .3462428
       2003  |   .1117307   .1017752     1.10   0.301    -.1185007    .3419621
       2004  |   .0933072    .102148     0.91   0.385    -.1377676    .3243821
       2005  |   .0524274   .0674885     0.78   0.457    -.1002421     .205097
       2006  |   .0415913   .0539446     0.77   0.460    -.0804399    .1636225
       2007  |   .0246245   .0624939     0.39   0.703    -.1167464    .1659955
       2008  |          0  (omitted)
       2009  |   .0211258   .0714686     0.30   0.774    -.1405473     .182799
       2010  |  -.0210113   .0196406    -1.07   0.313    -.0654415    .0234189
             |
       _cons |          0  (omitted)
------------------------------------------------------------------------------


. xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta yr*
variable yr* not found
r(111);

. tab year, gen(yr)

       Year |      Freq.     Percent        Cum.
------------+-----------------------------------
       2000 |          3        3.00        3.00
       2001 |          7        7.00       10.00
       2002 |          7        7.00       17.00
       2003 |          7        7.00       24.00
       2004 |          9        9.00       33.00
       2005 |          9        9.00       42.00
       2006 |         10       10.00       52.00
       2007 |          9        9.00       61.00
       2008 |         10       10.00       71.00
       2009 |         13       13.00       84.00
       2010 |         16       16.00      100.00
------------+-----------------------------------
      Total |        100      100.00

. xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta yr*

Regression with Driscoll-Kraay standard errors   Number of obs     =        75
Method: Pooled OLS                               Number of groups  =        14
Group variable (i): code                         F( 16,     9)     =    127.00
maximum lag: 2                                   Prob > F          =    0.0000
                                                 R-squared         =    0.7573
                                                 Root MSE          =    0.0768

------------------------------------------------------------------------------
             |             Drisc/Kraay
       mllev |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       mllev |
         L1. |   .6943744   .0852869     8.14   0.000      .501442    .8873068
             |
        rate |  -.0110723   .0080324    -1.38   0.201    -.0292429    .0070983
  cashassets |     .03596    .215608     0.17   0.871    -.4517793    .5236993
      ebitta |  -.1544263   .1582091    -0.98   0.355    -.5123201    .2034675
        size |   .0042529   .0162563     0.26   0.800    -.0325213    .0410272
      medlev |   .0009085   .0007565     1.20   0.260    -.0008029    .0026198
        infl |   .0640381   .0476532     1.34   0.212    -.0437609    .1718372
      tangta |   .0561207   .0448327     1.25   0.242    -.0452979    .1575393
         yr1 |          0  (omitted)
         yr2 |   .1167228   .0920622     1.27   0.237    -.0915364     .324982
         yr3 |   .1080642   .1052883     1.03   0.332    -.1301144    .3462428
         yr4 |   .1117307   .1017752     1.10   0.301    -.1185007    .3419621
         yr5 |   .0933072    .102148     0.91   0.385    -.1377676    .3243821
         yr6 |   .0524274   .0674885     0.78   0.457    -.1002421     .205097
         yr7 |   .0415913   .0539446     0.77   0.460    -.0804399    .1636225
         yr8 |   .0246245   .0624939     0.39   0.703    -.1167464    .1659955
         yr9 |          0  (omitted)
        yr10 |   .0211258   .0714686     0.30   0.774    -.1405473     .182799
        yr11 |  -.0210113   .0196406    -1.07   0.313    -.0654415    .0234189
       _cons |          0  (omitted)
------------------------------------------------------------------------------


.

[Colm Bairead]

Joro, thanks for your swift response! I just pasted your code and Im still getting the same error message!! Im using Stata10. Would that make a difference??

[Joro Kolev]

I do not have Stata 10, the oldest Stata that I have on this laptop is Stata 11.

And under Stata 11 my code works all fine. Then I tried version control under Stata 11.

version 10

Code still works without errors.

I tried version control under Stata 15,

version 10

code still works.


Can it be that your installation of -xtscc - is garbled up? Try uninstalling it, searching in windows and manually deleting the .ado file if it still exists, and then installing it from scratch.

. xtset code year
       panel variable:  code (unbalanced)
        time variable:  year, 2000 to 2010, but with gaps
                delta:  1 unit

. version 10

. xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta i.year

Regression with Driscoll-Kraay standard errors   Number of obs     =        75
Method: Pooled OLS                               Number of groups  =        14
Group variable (i): code                         F( 16,     9)     =    127.00
maximum lag: 2                                   Prob > F          =    0.0000
                                                 R-squared         =    0.7573
                                                 Root MSE          =    0.0768

------------------------------------------------------------------------------
             |             Drisc/Kraay
       mllev |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       mllev |
         L1. |   .6943744   .0852869     8.14   0.000      .501442    .8873068
             |
        rate |  -.0110723   .0080324    -1.38   0.201    -.0292429    .0070983
  cashassets |     .03596    .215608     0.17   0.871    -.4517793    .5236993
      ebitta |  -.1544263   .1582091    -0.98   0.355    -.5123201    .2034675
        size |   .0042529   .0162563     0.26   0.800    -.0325213    .0410272
      medlev |   .0009085   .0007565     1.20   0.260    -.0008029    .0026198
        infl |   .0640381   .0476532     1.34   0.212    -.0437609    .1718372
      tangta |   .0561207   .0448327     1.25   0.242    -.0452979    .1575393
             |
        year |
       2000  |          0  (empty)
       2001  |   .1167228   .0920622     1.27   0.237    -.0915364     .324982
       2002  |   .1080642   .1052883     1.03   0.332    -.1301144    .3462428
       2003  |   .1117307   .1017752     1.10   0.301    -.1185007    .3419621
       2004  |   .0933072    .102148     0.91   0.385    -.1377676    .3243821
       2005  |   .0524274   .0674885     0.78   0.457    -.1002421     .205097
       2006  |   .0415913   .0539446     0.77   0.460    -.0804399    .1636225
       2007  |   .0246245   .0624939     0.39   0.703    -.1167464    .1659955
       2008  |          0  (omitted)
       2009  |   .0211258   .0714686     0.30   0.774    -.1405473     .182799
       2010  |  -.0210113   .0196406    -1.07   0.313    -.0654415    .0234189
             |
       _cons |          0  (omitted)
------------------------------------------------------------------------------

[Colm Bairead]

Joro,
Some progress! I have uninstalled the package and installed via another provider and it works however does not allow to include the lagged dependent variable! Code and Error message below. Works fine without the lagged regressor. Any tips??

xtscc mllev l.mllev rate cashassets ebitta size medlev infl tangta yr*
time-series operators not allowed
r(101);

[Joro Kolev]

If this is the only problem, we can easily resolve it:

gen laggedmllev = l.mllev


xtscc mllev laggedmllev rate cashassets ebitta size medlev infl tangta yr*

[Colm Bairead]

Joro,
It works! Apologies for the messy presentation. Thank you very much for your help. Colm.

xtscc mllev laggedmllev rate cashassets ebitta size medlev infl tangta yr*

Regression with Driscoll-Kraay standard errors Number of obs = 423
Method: Pooled OLS Number of groups = 57
Group variable (i): code F( 19, 56) = 4189.72
maximum lag: 2 Prob > F = 0.0000
R-squared = 0.8222
Root MSE = 0.0747

------------------------------------------------------------------------------
| Drisc/Kraay
mllev | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
laggedmllev | .7421992 .0198499 37.39 0.000 .7024351 .7819633
rate | -.0078542 .002041 -3.85 0.000 -.0119429 -.0037656
cashassets | -.177286 .0265429 -6.68 0.000 -.2304578 -.1241142
ebitta | .1024929 .0401424 2.55 0.013 .0220781 .1829078
size | .0009425 .0042348 0.22 0.825 -.0075408 .0094259
medlev | .00148 .0003529 4.19 0.000 .000773 .002187
infl | .0425557 .0174883 2.43 0.018 .0075224 .077589
tangta | .0400027 .0163387 2.45 0.018 .0072723 .0727331
yr01 | (dropped)
yr02 | .1127904 .0418248 2.70 0.009 .0290052 .1965756
yr03 | .1084367 .0406054 2.67 0.010 .0270944 .189779
yr04 | .0945705 .0387906 2.44 0.018 .0168636 .1722775
yr05 | .098886 .0378158 2.61 0.011 .0231319 .17464
yr06 | .0559968 .0250691 2.23 0.030 .0057774 .1062163
yr07 | .0276938 .0201795 1.37 0.175 -.0127306 .0681182
yr08 | .0604141 .0206145 2.93 0.005 .0191184 .1017098
yr09 | (dropped)
yr10 | .0789085 .0257259 3.07 0.003 .0273732 .1304438
yr11 | .0078355 .0053333 1.47 0.147 -.0028485 .0185195
_cons | (dropped)
------------------------------------------------------------------------------

[Joro Kolev]

Well, very good for us then, Colm. We slaughtered one mysterious bug today. We can be proud of ourselves (or a bit ashamed if this was the last bug of its kind, and we just killed it )

If you do not want to spend your time on slaughtering more mysterious bugs in the future, update your Stata a bit.

I use old Stata sometimes if I feel that it does something computationally intensive faster than the newer versions...

But you re in the pre-historical age with this Stata 10. Last time I used Stata 10 was for a paper that got published in year 2011. Probably even then I was a bit out of date.

Stata 10 is way too old from today's vintage of updates, you are just calling for trouble when you re using that old version.

[Colm Bairead]

I am currently completing my masters dissertation and am using the university`s software packages, hence the reason for using Stata10! There were a few convenient commands I couldn't execute because of this. Also, I am just learning Stata just to enable me to finish the project however am getting very fond of it!! I definitely would like to stay abreast of the future developments in the software! Thanks once again, Colm.

[Nurfadhilah Ichdar]

i have to try this suggestion but in my stata but factor-variable and time-series operators not allowed.