cross sectional regression

[MARIA CHRISTOFIDOU]

cross sectional regression across countries: My sample consists of 150 bank observations for 2023, across 30 countries.
My question is which one of the below is best to use to account for across country variations?
1) country fixed effects by using i.country command in stata
2) use country control variables such as GDP, Rule of Law, control of corruption, Enviromental Performance inded
3) Should I do both 1) and 2) above - if not which one is better?
4) if I do 1) or 2) do I need to cluster standard errors across country?

I have tried to do 1) and 4) and my F statitstic disappears and suddenly none of my independent variables are significant, so I suppose this is inappropriate.
thank you so much

[Carlo Lazzaro]

Maria:
1) correct;
2) correct;
3) 1)+2):
4) yes.

As far as your last point is concerned, please share what you typed and what Stata have you back (as per FAQ). Thanks.

[MARIA CHRISTOFIDOU]

Thank you so much for the prompt response. I am copying below my output without clustering standard errors. The Homoskedasticity tests indicate no heterosckedastisticity, Does this mean that I do not need to do robust or cluster standard errors:

. . regress GERQS ESGCom PERF REM WOMB WOME PCAF NZBA SIZE LEV ROAA CGF i.CCODE
note: 30.CCODE omitted because of collinearity.

Source | SS df MS Number of obs = 176
-------------+---------------------------------- F(39, 136) = 7.41
Model | 9562.19251 39 245.184423 Prob > F = 0.0000
Residual | 4501.65088 136 33.1003741 R-squared = 0.6799
-------------+---------------------------------- Adj R-squared = 0.5881
Total | 14063.8434 175 80.3648194 Root MSE = 5.7533

------------------------------------------------------------------------------
GERQS | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
ESGCom | .5936342 1.535091 0.39 0.700 -2.442101 3.629369
PERF | 1.664053 2.204215 0.75 0.452 -2.694915 6.023022
REM | 8.458434 2.397928 3.53 0.001 3.716386 13.20048
WOMB | .6919396 2.801889 0.25 0.805 -4.848966 6.232845
WOME | -3.238755 3.221254 -1.01 0.316 -9.608981 3.131472
PCAF | 3.509845 1.505327 2.33 0.021 .5329691 6.486721
NZBA | 3.76366 1.991068 1.89 0.061 -.1737974 7.701118
SIZE | .1634796 .5767978 0.28 0.777 -.977173 1.304132
LEV | -.0783147 .1685003 -0.46 0.643 -.4115343 .2549049
ROAA | .6812615 .4526264 1.51 0.135 -.2138347 1.576358
CGF | 1.574039 1.268099 1.24 0.217 -.9337043 4.081783
|
CCODE |
2 | 10.24028 8.582243 1.19 0.235 -6.731632 27.21218
3 | 23.98475 25.27272 0.95 0.344 -25.99359 73.96309
4 | 15.90648 18.83823 0.84 0.400 -21.34726 53.16021
5 | 17.87123 17.72687 1.01 0.315 -17.18472 52.92719
6 | 4.260802 6.810191 0.63 0.533 -9.206764 17.72837
7 | -3.14963 7.383645 -0.43 0.670 -17.75124 11.45198
8 | 2.209857 4.424958 0.50 0.618 -6.540767 10.96048
9 | -2.788007 6.409615 -0.43 0.664 -15.46341 9.887396
10 | 8.412635 9.936375 0.85 0.399 -11.23715 28.06242
11 | 2.50148 3.070251 0.81 0.417 -3.570128 8.573087
12 | 27.28537 20.04259 1.36 0.176 -12.35006 66.92081
13 | 23.52056 21.06102 1.12 0.266 -18.12889 65.17001
14 | 6.426802 4.208285 1.53 0.129 -1.895338 14.74894
15 | -.6573553 3.456969 -0.19 0.849 -7.493721 6.17901
16 | 23.47177 18.11815 1.30 0.197 -12.35798 59.30151
17 | 10.90944 14.07277 0.78 0.440 -16.92032 38.73919
18 | 1.865625 4.528703 0.41 0.681 -7.090161 10.82141
19 | 5.031645 8.53808 0.59 0.557 -11.85293 21.91622
20 | -12.23272 8.340597 -1.47 0.145 -28.72676 4.261314
21 | 8.568783 13.94674 0.61 0.540 -19.01175 36.14932
22 | 2.883075 3.375749 0.85 0.395 -3.792674 9.558824
23 | -8.03483 6.910293 -1.16 0.247 -21.70035 5.630693
24 | 17.02607 16.59738 1.03 0.307 -15.79625 49.8484
25 | 16.38382 11.89338 1.38 0.171 -7.136062 39.9037
26 | 25.82149 23.8293 1.08 0.280 -21.3024 72.94538
27 | 15.0263 18.3246 0.82 0.414 -21.21171 51.26431
28 | 9.560482 10.94494 0.87 0.384 -12.08381 31.20477
29 | 21.73535 15.7669 1.38 0.170 -9.444647 52.91534
30 | 0 (omitted)
|
_cons | -28.09633 27.20939 -1.03 0.304 -81.90456 25.7119
------------------------------------------------------------------------------

. estat hettest

Breusch–Pagan/Cook–Weisberg test for heteroskedasticity
Assumption: Normal error terms
Variable: Fitted values of GERQS

H0: Constant variance

chi2(1) = 0.01
Prob > chi2 = 0.9178

. estat imtest, white

White's test
H0: Homoskedasticity
Ha: Unrestricted heteroskedasticity

chi2(176) = 176.00
Prob > chi2 = 0.4858

Cameron & Trivedi's decomposition of IM-test

--------------------------------------------------
Source | chi2 df p
---------------------+----------------------------
Heteroskedasticity | 176.00 176 0.4858
Skewness | 38.25 39 0.5038
Kurtosis | 0.58 1 0.4475
---------------------+----------------------------
Total | 214.83 216 0.5097
--------------------------------------------------

[MARIA CHRISTOFIDOU]

sorry the copy did not work very well. . I tried again and it is better now.

egress GERQS ESGCom PERF REM WOMB WOME PCAF NZBA SIZE LEV ROAA CGF, robust

Linear regression Number of obs = 176
F(11, 164) = 35.77
Prob > F = 0.0000
R-squared = 0.5950
Root MSE = 5.8933

------------------------------------------------------------------------------
| Robust
GERQS | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
ESGCom | 1.618787 1.310701 1.24 0.219 -.969237 4.206811
PERF | 1.27092 1.93083 0.66 0.511 -2.541571 5.083411
REM | 8.794508 1.915623 4.59 0.000 5.012044 12.57697
WOMB | .4619205 2.490686 0.19 0.853 -4.456025 5.379866
WOME | -4.275162 2.614473 -1.64 0.104 -9.437529 .8872048
PCAF | 3.771934 1.237289 3.05 0.003 1.328864 6.215004
NZBA | 5.929158 1.215245 4.88 0.000 3.529613 8.328702
SIZE | .4336076 .3779865 1.15 0.253 -.3127398 1.179955
LEV | .0564408 .1771497 0.32 0.750 -.2933474 .4062291
ROAA | .7406547 .2825804 2.62 0.010 .1826898 1.29862
CGF | .2361721 .07721 3.06 0.003 .0837183 .3886258
_cons | -5.265707 5.902145 -0.89 0.374 -16.9197 6.388282
------------------------------------------------------------------------------

.
. . regress GERQS ESGCom PERF REM WOMB WOME PCAF NZBA SIZE LEV ROAA CGF i.CCODE
note: 30.CCODE omitted because of collinearity.

Source | SS df MS Number of obs = 176
-------------+---------------------------------- F(39, 136) = 7.41
Model | 9562.19251 39 245.184423 Prob > F = 0.0000
Residual | 4501.65088 136 33.1003741 R-squared = 0.6799
-------------+---------------------------------- Adj R-squared = 0.5881
Total | 14063.8434 175 80.3648194 Root MSE = 5.7533

------------------------------------------------------------------------------
GERQS | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
ESGCom | .5936342 1.535091 0.39 0.700 -2.442101 3.629369
PERF | 1.664053 2.204215 0.75 0.452 -2.694915 6.023022
REM | 8.458434 2.397928 3.53 0.001 3.716386 13.20048
WOMB | .6919396 2.801889 0.25 0.805 -4.848966 6.232845
WOME | -3.238755 3.221254 -1.01 0.316 -9.608981 3.131472
PCAF | 3.509845 1.505327 2.33 0.021 .5329691 6.486721
NZBA | 3.76366 1.991068 1.89 0.061 -.1737974 7.701118
SIZE | .1634796 .5767978 0.28 0.777 -.977173 1.304132
LEV | -.0783147 .1685003 -0.46 0.643 -.4115343 .2549049
ROAA | .6812615 .4526264 1.51 0.135 -.2138347 1.576358
CGF | 1.574039 1.268099 1.24 0.217 -.9337043 4.081783
|
CCODE |
2 | 10.24028 8.582243 1.19 0.235 -6.731632 27.21218
3 | 23.98475 25.27272 0.95 0.344 -25.99359 73.96309
4 | 15.90648 18.83823 0.84 0.400 -21.34726 53.16021
5 | 17.87123 17.72687 1.01 0.315 -17.18472 52.92719
6 | 4.260802 6.810191 0.63 0.533 -9.206764 17.72837
7 | -3.14963 7.383645 -0.43 0.670 -17.75124 11.45198
8 | 2.209857 4.424958 0.50 0.618 -6.540767 10.96048
9 | -2.788007 6.409615 -0.43 0.664 -15.46341 9.887396
10 | 8.412635 9.936375 0.85 0.399 -11.23715 28.06242
11 | 2.50148 3.070251 0.81 0.417 -3.570128 8.573087
12 | 27.28537 20.04259 1.36 0.176 -12.35006 66.92081
13 | 23.52056 21.06102 1.12 0.266 -18.12889 65.17001
14 | 6.426802 4.208285 1.53 0.129 -1.895338 14.74894
15 | -.6573553 3.456969 -0.19 0.849 -7.493721 6.17901
16 | 23.47177 18.11815 1.30 0.197 -12.35798 59.30151
17 | 10.90944 14.07277 0.78 0.440 -16.92032 38.73919
18 | 1.865625 4.528703 0.41 0.681 -7.090161 10.82141
19 | 5.031645 8.53808 0.59 0.557 -11.85293 21.91622
20 | -12.23272 8.340597 -1.47 0.145 -28.72676 4.261314
21 | 8.568783 13.94674 0.61 0.540 -19.01175 36.14932
22 | 2.883075 3.375749 0.85 0.395 -3.792674 9.558824
23 | -8.03483 6.910293 -1.16 0.247 -21.70035 5.630693
24 | 17.02607 16.59738 1.03 0.307 -15.79625 49.8484
25 | 16.38382 11.89338 1.38 0.171 -7.136062 39.9037
26 | 25.82149 23.8293 1.08 0.280 -21.3024 72.94538
27 | 15.0263 18.3246 0.82 0.414 -21.21171 51.26431
28 | 9.560482 10.94494 0.87 0.384 -12.08381 31.20477
29 | 21.73535 15.7669 1.38 0.170 -9.444647 52.91534
30 | 0 (omitted)
|
_cons | -28.09633 27.20939 -1.03 0.304 -81.90456 25.7119
------------------------------------------------------------------------------

. estat hettest

Breusch–Pagan/Cook–Weisberg test for heteroskedasticity
Assumption: Normal error terms
Variable: Fitted values of GERQS

H0: Constant variance

chi2(1) = 0.01
Prob > chi2 = 0.9178

. estat imtest, white

White's test
H0: Homoskedasticity
Ha: Unrestricted heteroskedasticity

chi2(176) = 176.00
Prob > chi2 = 0.4858

Cameron & Trivedi's decomposition of IM-test

--------------------------------------------------
Source | chi2 df p
---------------------+----------------------------
Heteroskedasticity | 176.00 176 0.4858
Skewness | 38.25 39 0.5038
Kurtosis | 0.58 1 0.4475
---------------------+----------------------------
Total | 214.83 216 0.5097
--------------------------------------------------

.

[MARIA CHRISTOFIDOU]

and when I cluster standard errors by country I get

. regress GERQS ESGCom PERF REM WOMB WOME PCAF NZBA SIZE LEV ROAA CGF i.CCODE,	cluster(CCODE)
note: 30.CCODE omitted because of collinearity.
Linear regression Number of obs = 176
F(9, 29) = .
Prob > F = .
R-squared = 0.6799
Root MSE = 5.7533
(Std. err. adjusted for 30 clusters in CCODE)
|Robust
GERQS | Coefficient std. err. t P>|t| [95% conf. interval]
ESGCom | .5936342 1.158382 0.51 0.612 -1.775522 2.962791
PERF | 1.664053 2.23379 0.74 0.462 -2.90456 6.232667
REM | 8.458434 2.7045 3.13 0.004 2.92711 13.98976
WOMB | .6919396 3.036366 0.23 0.821 -5.518125 6.902005
WOME | -3.238755 2.663609 -1.22 0.234 -8.686448 2.208938
PCAF | 3.509845 1.864528 1.88 0.070 -.303544 7.323234
NZBA | 3.76366 2.689615 1.40 0.172 -1.73722 9.26454
SIZE | .1634796 .814418 0.20 0.842 -1.502192 1.829151
LEV | -.0783147 .3044736 -0.26 0.799 -.7010332 .5444038
ROAA | .6812615 .4375015 1.56 0.130 -.2135295 1.576052
CGF | 1.574039 .7503912 2.10 0.045 .0393172 3.108762
|
CCODE |
2 | 10.24028 4.014636 2.55 0.016 2.029425 18.45113
3 | 23.98475 13.70314 1.75 0.091 -4.041329 52.01082
4 | 15.90648 10.05387 1.58 0.124 -4.655998 36.46895
5 | 17.87123 9.155598 1.95 0.061 -.8540667 36.59653
6 | 4.260802 3.052275 1.40 0.173 -1.981802 10.50341
7 | -3.14963 3.933057 -0.80 0.430 -11.19363 4.8943

[George Ford]

ssc install reghdfe

reghdfe GERQS ESGCom PERF REM WOMB WOME PCAF NZBA SIZE LEV ROAA CGF , absorb(CCODE YEAR) cluster(CCODE)

[MARIA CHRISTOFIDOU]

Thank you George. Please note that all my observations are for 2023 as I am checking for early adoption of climate related disclosures.Thus YEAR is not relevant.
I have run the regression and I copy the output below.
1) One of my variables CGF which is key for country governance quality - it says below because it is probably collinear wiht fixed effects. Can I run a second regression wihtout country fixed effects and include this variable, as a robusness check?
2) Do you think the below results are good and robust?
3)Do I need to do any robustness checks?

reghdfe GERQS ESGCom PERF REM WOMB WOME PCAF NZBA SIZE LEV ROAA CGF , absorb(CCODE) cluster(CCODE)
(MWFE estimator converged in 1 iterations)
note: CGF is probably collinear with the fixed effects (all partialled-out values are close to zero;	tol	=	1.0e-09)
HDFE Linear regression Number of obs = 176
Absorbing 1 HDFE group F( 10, 29) = 10.75
Statistics robust to heteroskedasticity Prob > F = 0.0000
R-squared = 0.6799
Adj R-squared = 0.5881
Within R-sq. = 0.3529
Number of clusters (CCODE) = 30 Root MSE = 5.7533
(Std. err. adjusted for 30 clusters in CCODE)
|Robust
GERQS | Coefficient std. err. t P>|t| [95% conf. interval]
ESGCom | .5936342 1.051669 0.56 0.577 -1.557271 2.744539
PERF | 1.664053 2.028009 0.82 0.419 -2.483691 5.811798
REM | 8.458434 2.455356 3.44 0.002 3.436666 13.4802
WOMB | .6919396 2.75665 0.25 0.804 -4.946042 6.329921
WOME | -3.238755 2.418232 -1.34 0.191 -8.184595 1.707086
PCAF | 3.509845 1.692764 2.07 0.047 .0477531 6.971937
NZBA | 3.76366 2.441842 1.54 0.134 -1.230467 8.757788
SIZE | .1634796 .7393921 0.22 0.827 -1.348747 1.675706
LEV | -.0783147 .2764249 -0.28 0.779 -.6436671 .4870377
ROAA | .6812615 .397198 1.72 0.097 -.1310996 1.493623
CGF | 0 (omitted)
_cons | 3.536742 12.21612 0.29 0.774 -21.44802 28.5215
Absorbed degrees of freedom:
Absorbed FE | Categories - Redundant = Num. Coefs |
-|
CCODE | 30 30 0 *|
* = FE nested within cluster; treated as redundant for DoF computation

[Carlo Lazzaro]

Maria:
you can go -regress- with -vce(cluster CCODE)- standard errors.
Please share what you typed and what Stata gave you back using CODE delimiters (see the FAQ). Thanks.

[MARIA CHRISTOFIDOU]

Thank you Carlo. It is not clear to me what the command in stata should be, afer I write my regress command. This is my regress equation. Can you please indicate what I should change?

. regress GERQS ESGCom PERF REM NZBA PCAF WOMB WOME SIZE LEV ROAA i.CCODE, cluster(CCODE)

Linear regression Number of obs = 176
F(9, 29) = .
Prob > F = .
R-squared = 0.6799
Root MSE = 5.7533

(Std. err. adjusted for 30 clusters in CCODE)

[MARIA CHRISTOFIDOU]

1) Also I need some help on how to run endogeneity test. My dependent variable can in no way impact the independent variables.
2) what do you think about the suggestion to run reghdfe as suggested by Geoge Ford above. My results can be viewed above.

[Carlo Lazzaro]

Maria:
please read (and act on) Help - Statalist , point 12.3.
About the reason why your F test does not appear, see -help j_robustsingular-.
As far as the command -estat endogenous- is concerned, see ivregress postestimation##syntax_estat.
George's helpful hint is one of the possible approaches to deal with what you're planning to do.

[MARIA CHRISTOFIDOU]

.    regress    GERQS    ESGCom    PERF    REM    NZBA    PCAF    WOMB    WOME    SIZE    LEV    ROAA    i.CCODE

[MARIA CHRISTOFIDOU]

. regress GERQS ESGCom PERF REM NZBA PCAF WOMB WOME SIZE LEV ROAA i.CCODE

 regress GERQS ESGCom PERF REM NZBA PCAF WOMB WOME SIZE LEV ROAA i.CCODE

      Source |       SS           df       MS      Number of obs   =       176
-------------+----------------------------------   F(39, 136)      =      7.41
       Model |  9562.19251        39  245.184423   Prob > F        =    0.0000
    Residual |  4501.65088       136  33.1003741   R-squared       =    0.6799
-------------+----------------------------------   Adj R-squared   =    0.5881
       Total |  14063.8434       175  80.3648194   Root MSE        =    5.7533

------------------------------------------------------------------------------
       GERQS | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
      ESGCom |   .5936342   1.535091     0.39   0.700    -2.442101    3.629369
        PERF |   1.664053   2.204215     0.75   0.452    -2.694915    6.023022
         REM |   8.458434   2.397928     3.53   0.001     3.716386    13.20048
        NZBA |    3.76366   1.991068     1.89   0.061    -.1737974    7.701118
        PCAF |   3.509845   1.505327     2.33   0.021     .5329691    6.486721
        WOMB |   .6919396   2.801889     0.25   0.805    -4.848966    6.232845
        WOME |  -3.238755   3.221254    -1.01   0.316    -9.608981    3.131472
        SIZE |   .1634796   .5767978     0.28   0.777     -.977173    1.304132
         LEV |  -.0783147   .1685003    -0.46   0.643    -.4115343    .2549049
        ROAA |   .6812615   .4526264     1.51   0.135    -.2138347    1.576358
             |
       CCODE |
          2  |   1.693242   3.190253     0.53   0.596    -4.615678    8.002162
          3  |  -5.701635   3.306326    -1.72   0.087     -12.2401     .836825
          4  |  -5.689344    3.36573    -1.69   0.093    -12.34528    .9665906
          5  |  -2.370914     3.5371    -0.67   0.504    -9.365744    4.623916
          6  |   -1.81499   3.311099    -0.55   0.584    -8.362889    4.732909
          7  |   7.191809   3.189708     2.25   0.026     .8839665    13.49965
          8  |   1.296914   4.164011     0.31   0.756    -6.937671    9.531499
          9  |   5.522921    4.38155     1.26   0.210     -3.14186     14.1877
         10  |  -.8112354   3.809942    -0.21   0.832    -8.345628    6.723157
         11  |   2.265374   2.974989     0.76   0.448    -3.617848    8.148596
         12  |   4.571983   3.713791     1.23   0.220    -2.772264    11.91623
         13  |  -1.002972   3.257435    -0.31   0.759    -7.444747    5.438803
         14  |   7.087898   4.372768     1.62   0.107    -1.559517    15.73531
         15  |   2.254618   3.567351     0.63   0.528    -4.800036    9.309271
         16  |   3.670351   4.530704     0.81   0.419    -5.289391    12.63009
         17  |  -4.610592    3.57599    -1.29   0.199    -11.68233    2.461146
         18  |   2.306356   4.639536     0.50   0.620    -6.868608    11.48132
         19  |  -3.074658   3.815734    -0.81   0.422     -10.6205    4.471188
         20  |  -.9940828   3.549297    -0.28   0.780    -8.013033    6.024868
         21  |   -6.73088   3.947704    -1.71   0.090     -14.5377    1.075945
         22  |   5.795048   3.760546     1.54   0.126    -1.641661    13.23176
         23  |   .9529345   4.553552     0.21   0.835    -8.051991     9.95786
         24  |    -1.8624   2.991911    -0.62   0.535    -7.779085    4.054286
         25  |   3.791504   3.545848     1.07   0.287    -3.220626    10.80363
         26  |  -1.850121   3.267341    -0.57   0.572    -8.311486    4.611244
         27  |  -5.892686   3.449826    -1.71   0.090    -12.71493    .9295544
         28  |  -1.914265   3.565225    -0.54   0.592    -8.964715    5.136185
         29  |   4.861646    4.03239     1.21   0.230     -3.11265    12.83594
         30  |   5.162849   4.159366     1.24   0.217     -3.06255    13.38825
             |
       _cons |   3.384457   9.967332     0.34   0.735    -16.32655    23.09546
------------------------------------------------------------------------------

.

[MARIA CHRISTOFIDOU]

Struggled a bit but managed to properly copy my data. Thank you and apologies

[Carlo Lazzaro]

Maria:
try:

testparm i.CCODE

and see if unit fixed effect is worth considering for -reghdfe- (or -areg-).