How to export results from xtivreg using esttab

[Martina Barbaglia]

Hi everyone,
I'm running the following instrumental variable regression:

xtivreg sust (cash = mean_cash) cap i.year, fe first

Now, I'd like to export the results of both the first and second-stage regression (with F-stat from the first-stage regression) using esttab. Do you have any suggestions on how to do it?

Thank you so much for the help!

[Andrew Musau]

estout is from SSC (FAQ Advice #12).

webuse nlswork, clear
eststo second: xtivreg ln_w age c.age#c.age not_smsa (tenure = union south), fe first
eststo first: qui xtreg `e(instd)' `e(insts)' if e(sample), fe
esttab first second, drop(_cons) scalars(N F) label

Res.:

. eststo second: xtivreg ln_w age c.age#c.age not_smsa (tenure = union south), fe first

First-stage within regression

Fixed-effects (within) regression               Number of obs     =     19,007
Group variable: idcode                          Number of groups  =      4,134

R-squared:                                      Obs per group:
     Within  = 0.3019                                         min =          1
     Between = 0.0578                                         avg =        4.6
     Overall = 0.1289                                         max =         12

                                                F(5, 14868)       =    1285.83
corr(u_i, Xb) = -0.1871                         Prob > F          =     0.0000

------------------------------------------------------------------------------
      tenure | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
         age |   .0863031   .0343238     2.51   0.012     .0190241    .1535821
             |
 c.age#c.age |   .0039115    .000549     7.12   0.000     .0028353    .0049876
             |
    not_smsa |  -.3552488   .1269318    -2.80   0.005    -.6040507   -.1064468
       union |   .3896861   .0706568     5.52   0.000     .2511901    .5281821
       south |  -.4296172   .1349122    -3.18   0.001    -.6940618   -.1651726
       _cons |  -2.554764   .5293374    -4.83   0.000     -3.59233   -1.517197
-------------+----------------------------------------------------------------
     sigma_u |  3.1334845
     sigma_e |  2.5869529
         rho |  .59467598   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(4133, 14868) = 6.40                 Prob > F = 0.0000

Fixed-effects (within) IV regression            Number of obs     =     19,007
Group variable: idcode                          Number of groups  =      4,134

R-squared:                                      Obs per group:
     Within  =      .                                         min =          1
     Between = 0.1304                                         avg =        4.6
     Overall = 0.0897                                         max =         12

                                                Wald chi2(4)      =  147926.58
corr(u_i, Xb) = -0.6843                         Prob > chi2       =     0.0000

------------------------------------------------------------------------------
     ln_wage | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
      tenure |   .2403531   .0373419     6.44   0.000     .1671643    .3135419
         age |   .0118437   .0090032     1.32   0.188    -.0058023    .0294897
             |
 c.age#c.age |  -.0012145   .0001968    -6.17   0.000    -.0016003   -.0008286
             |
    not_smsa |  -.0167178   .0339236    -0.49   0.622    -.0832069    .0497713
       _cons |   1.678287   .1626657    10.32   0.000     1.359468    1.997106
-------------+----------------------------------------------------------------
     sigma_u |  .70661941
     sigma_e |  .63029359
         rho |  .55690561   (fraction of variance due to u_i)
------------------------------------------------------------------------------
 F test that all u_i=0: F(4133,14869) =     1.44          Prob > F    = 0.0000
------------------------------------------------------------------------------
Endogenous: tenure
Exogenous:  age c.age#c.age not_smsa union south

. 
. eststo first: qui xtreg `e(instd)' `e(insts)' if e(sample), fe

. 
. esttab first second, drop(_cons) scalars(N F) label

----------------------------------------------------
                              (1)             (2)   
                     Job tenure~s    ln(wage/GN~)   
----------------------------------------------------
Age in current year        0.0863*         0.0118   
                           (2.51)          (1.32)   

Age in current yea~r      0.00391***     -0.00121***
                           (7.12)         (-6.17)   

1 if not SMSA              -0.355**       -0.0167   
                          (-2.80)         (-0.49)   

1 if union                  0.390***                
                           (5.52)                   

1 if south                 -0.430**                 
                          (-3.18)                   

Job tenure, in years                        0.240***
                                           (6.44)   
----------------------------------------------------
Observations                19007           19007   
F                          1285.8                   
----------------------------------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

[Martina Barbaglia]

Thank you Andrew for your reply! I tried to run the same commands using my own regression:

est clear
eststo second: xtivreg sust (cash = mean_cash) cap i.year, fe first
eststo first: qui xtreg `e(instd)' `e(insts)' if e(sample), fe
esttab first second, drop(_cons) scalars(N F) label

The only problem is that the coefficients and the F-statistic from the first-stage regression printed in the table are different from the ones reported in the Stata output above. There is a mismatch between what I can see from the first-stage regression using xtivreg and what I find in the table from esttab. How is it possible?

Thank you!

[Andrew Musau]

It does not happen with my example in #2. Can you provide a reproducible example that illustrates this? If you can share your data, copy and paste the result of the following:

qui xtset
dataex `r(panelvar)' sust cash  mean_cash cap year

[Martina Barbaglia]

Dear Andrew,
I think the problem is related to the usage of time operators. Indeed, the regression I have to run is the following:

. est clear

. eststo second: xtivreg sust (l.cash = l.mean_cash) l.cap i.year, fe first

First-stage within regression

Fixed-effects (within) regression               Number of obs     =     18,574
Group variable: company_id                      Number of groups  =      3,903

R-squared:                                      Obs per group:
     Within  = 0.0315                                         min =          1
     Between = 0.2124                                         avg =        4.8
     Overall = 0.1553                                         max =         11

                                                F(12,14659)       =      39.74
corr(u_i, Xb) = 0.2820                          Prob > F          =     0.0000

------------------------------------------------------------------------------
    __000004 | Coefficient  Std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
         cap |
         L1. |  -.0407847   .0047118    -8.66   0.000    -.0500205    -.031549
             |
        year |
       2011  |          0  (empty)
       2012  |  -.0138531   .0023578    -5.88   0.000    -.0184747   -.0092315
       2013  |  -.0198712   .0024981    -7.95   0.000    -.0247677   -.0149747
       2014  |  -.0140905   .0024294    -5.80   0.000    -.0188523   -.0093286
       2015  |  -.0121639   .0023108    -5.26   0.000    -.0166933   -.0076345
       2016  |  -.0111176   .0022791    -4.88   0.000    -.0155849   -.0066502
       2017  |  -.0102835   .0021377    -4.81   0.000    -.0144736   -.0060933
       2018  |   -.007541   .0020314    -3.71   0.000    -.0115227   -.0035593
       2019  |  -.0155915   .0020651    -7.55   0.000    -.0196394   -.0115436
       2020  |   -.020013   .0020436    -9.79   0.000    -.0240187   -.0160072
       2021  |   .0005053   .0016177     0.31   0.755    -.0026657    .0036762
       2022  |          0  (omitted)
             |
   mean_cash |
         L1. |   .1416309   .0406277     3.49   0.000     .0619955    .2212662
             |
       _cons |   .1288699   .0077376    16.65   0.000     .1137031    .1440366
-------------+----------------------------------------------------------------
     sigma_u |  .12023968
     sigma_e |  .05146785
         rho |  .84515028   (fraction of variance due to u_i)
------------------------------------------------------------------------------
F test that all u_i=0: F(3902, 14659) = 16.73                Prob > F = 0.0000

Fixed-effects (within) IV regression            Number of obs     =     23,688
Group variable: company_id                      Number of groups  =      4,592

R-squared:                                      Obs per group:
     Within  =      .                                         min =          1
     Between = 0.0002                                         avg =        5.2
     Overall = 0.0001                                         max =         12

                                                Wald chi2(13)     =     431.32
corr(u_i, Xb) = -0.0091                         Prob > chi2       =     0.0000

------------------------------------------------------------------------------
        sust | Coefficient  Std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
        cash |
         L1. |  -155.8727   305.0177    -0.51   0.609    -753.6965    441.9511
             |
         cap |
         L1. |  -8.822369   16.82683    -0.52   0.600    -41.80235    24.15761
             |
        year |
       2012  |  -.7883762   3.491579    -0.23   0.821    -7.631744    6.054992
       2013  |  -2.038703   5.212085    -0.39   0.696     -12.2542    8.176796
       2014  |   -1.01614   3.811966    -0.27   0.790    -8.487455    6.455176
       2015  |  -.5365042   3.295733    -0.16   0.871    -6.996023    5.923015
       2016  |  -.4987909   3.206532    -0.16   0.876    -6.783478    5.785896
       2017  |  -.7228866   3.012198    -0.24   0.810    -6.626686    5.180913
       2018  |  -.3154553   2.836694    -0.11   0.911    -5.875274    5.244363
       2019  |  -1.508924   3.879447    -0.39   0.697    -9.112501    6.094653
       2020  |  -1.447044   5.009101    -0.29   0.773     -11.2647    8.370613
       2021  |   3.968255   4.071182     0.97   0.330    -4.011115    11.94762
       2022  |  -2.199229   4.163453    -0.53   0.597    -10.35945    5.960989
             |
       _cons |   33.19626   46.27593     0.72   0.473    -57.50289    123.8954
-------------+----------------------------------------------------------------
     sigma_u |   1462.871
     sigma_e |  71.025753
         rho |  .99764822   (fraction of variance due to u_i)
------------------------------------------------------------------------------
 F test that all u_i=0: F(4591,19083) =   846.49          Prob > F    = 0.0000
------------------------------------------------------------------------------
Instrumented: L.cash
 Instruments: L.cap 2012.year 2013.year 2014.year 2015.year 2016.year
              2017.year 2018.year 2019.year 2020.year 2021.year 2022.year
              L.mean_cash

. do "C:\Users\marti\AppData\Local\Temp\STD32e8_000000.tmp"

. eststo first: qui xtreg `e(instd)' `e(insts)' if e(sample), fe

. 
end of do-file

. esttab first second, drop(_cons) scalars(N F) label

----------------------------------------------------
                              (1)             (2)   
                           L.cash            sust   
----------------------------------------------------
L.cap                     -0.0518***       -8.822   
                         (-12.67)         (-0.52)   

year=2011                       0               0   
                              (.)             (.)   

year=2012                -0.00512*         -0.788   
                          (-2.12)         (-0.23)   

year=2013                 -0.0117***       -2.039   
                          (-4.73)         (-0.39)   

year=2014                -0.00547*         -1.016   
                          (-2.23)         (-0.27)   

year=2015                -0.00310          -0.537   
                          (-1.31)         (-0.16)   

year=2016                -0.00222          -0.499   
                          (-0.95)         (-0.16)   

year=2017                -0.00167          -0.723   
                          (-0.74)         (-0.24)   

year=2018                0.000119          -0.315   
                           (0.05)         (-0.11)   

year=2019                -0.00707**        -1.509   
                          (-3.25)         (-0.39)   

year=2020                 -0.0114***       -1.447   
                          (-5.21)         (-0.29)   

year=2021                 0.00863***        3.968   
                           (4.03)          (0.97)   

year=2022                 0.00760**        -2.199   
                           (3.25)         (-0.53)   

L.mean_cash                 0.152***                
                           (4.25)                   

L.cash                                     -155.9   
                                          (-0.51)   
----------------------------------------------------
Observations                23688           23688   
F                           46.64                   
----------------------------------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

As you can see, results are not consistent. However, if I eliminate lagged operators, the problem is solved. Is there a way to use your code when there are lagged variables in the regression?
Thank you!

[Andrew Musau]

See #7: https://www.statalist.org/forums/for...he-first-stage

[Martina Barbaglia]

Dear Andrew,
now it works, thank you so much! I have just one additional question: is it ok if the number of observations is different in the first and second-stage regression (17958 vs. 22993)?

Thank you!

[Andrew Musau]

Yes, that's fine. You may add an explanation that the difference is caused by lagging variables if you report both. I would just report the second stage \(N\).

[Martina Barbaglia]

Thank you Andrew for your help!