Loop with 2SLS (Xtivreg2)

[Jala Youssef]

Hi everyone,

I have panel data with a list of 60 endogeneous variables that I would like to instrument with 60 corresponding instruments (I want to run 60 separate regressions). Please find below a simplified example of what I would like to do:
xtivreg2 depvar control1 control2 (endogeneous1= instrument_endogeneous1 variablex), fe endog( endogeneous1)

xtivreg2 depvar control1 control2 (endogeneous2= instrument_endogeneous2 variablex ), fe endog( endogeneous2)

xtivreg2 depvar control1 control2 (endogeneous3= instrument_endogeneous3 variablex ), fe endog( endogeneous3)

and so on for the remaining ones.

I tried creating loops for endogenous variables and instruments. Please find below a simplification:
local ivar1 endogeneous1
local ivar2 endogeneous2

local dvar1 instrument_endogeneous1
local dvar2 instrument_endogeneous2


forv i=1/60 {
forv d=1/60 {
xtivreg2 depvar control1 control2 (`ivar`i''=`dvar`d'' variablex), fe endog(`ivar`i'') first savefirst
}
}


The problem here is that it runs regressions for each endogenous variable instrumented once by each instrument in the list. This is not what I intend to do. I want to instrument each variable with its corresponding instrument (not the whole list of instruments).

I would be grateful if you could advise on this. I am sorry in advance if this has been asked before and I couldn't locate the answer.

Many thanks in advance.

Jala Youssef

[Andrew Musau]

The index is the same, so no need to define it separately for each set of locals. xtivreg2 is from SSC (FAQ Advice #12).

forv i=1/60 {
    xtivreg2 depvar control1 control2 (`ivar`i''=`dvar`i'' variablex), fe endog(`ivar`i'') first savefirst
}

[Jala Youssef]

Dear Andrew Musau

Many thanks for the prompt and very useful reply, much appreciated. The code you suggested works perfectly.

Jala

[Jala Youssef]

Dear Andrew Musau

Sorry for the too many questions. Relative to the same code discussed above, I would be grateful if you could advise if there is a way to report the F-statistic of the first-stage results. I tried using outreg2 but I am not sure if it would allow me to report it.

I read this relevant thread but it is more about the first stage results in general:
https://www.statalist.org/forums/for...eg2-by-outreg2

And this one raising the same question (though unanswered unfortunately):
https://www.statalist.org/forums/for...after-xtivreg2

Many thanks in advance for your help.

Jala

[Andrew Musau]

outreg2 is from SSC (FAQ Advice #12). As I show here if using estout from SSC, you need to estimate the first stage regression to retrieve the statistic. Here is how you would do it using outreg2.

use http://fmwww.bc.edu/ec-p/data/wooldridge/mroz.dta, clear
ivreg2 lwage exper expersq (educ=age kidslt6 kidsge6), first savefirst
qui reg `e(instd)' `e(insts)' if e(sample)
local F= e(F)
est restore _ivreg2_educ
outreg2 using "ivreg2-12.rtf", replace noobs
ivreg2 lwage exper expersq (educ=age kidslt6 kidsge6)
outreg2 using "ivreg2-12.rtf", append addstat(First stage F, `F')

Res.:

. ivregress 2sls lwage exper expersq (educ=age kidslt6 kidsge6), first

First-stage regressions
-----------------------

                                                Number of obs     =        428
                                                F(   5,    422)   =       3.04
                                                Prob > F          =     0.0105
                                                R-squared         =     0.0347
                                                Adj R-squared     =     0.0233
                                                Root MSE          =     2.2586

------------------------------------------------------------------------------
        educ |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       exper |   .0501414   .0452026     1.11   0.268    -.0387088    .1389916
     expersq |  -.0017657   .0013568    -1.30   0.194    -.0044326    .0009012
         age |   -.012015   .0178791    -0.67   0.502    -.0471582    .0231281
     kidslt6 |   .7254425   .2979333     2.43   0.015     .1398244    1.311061
     kidsge6 |  -.2219447    .093675    -2.37   0.018    -.4060725   -.0378169
       _cons |   13.12194   .8598407    15.26   0.000     11.43184    14.81204
------------------------------------------------------------------------------


Instrumental variables (2SLS) regression          Number of obs   =        428
                                                  Wald chi2(3)    =      22.69
                                                  Prob > chi2     =     0.0000
                                                  R-squared       =     0.1556
                                                  Root MSE        =     .66378

------------------------------------------------------------------------------
       lwage |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        educ |   .0964002   .0814278     1.18   0.236    -.0631952    .2559957
       exper |    .042193   .0138831     3.04   0.002     .0149827    .0694033
     expersq |  -.0008323   .0004204    -1.98   0.048    -.0016563   -8.33e-06
       _cons |  -.3848718   1.011551    -0.38   0.704    -2.367476    1.597732
------------------------------------------------------------------------------
Instrumented:  educ
Instruments:   exper expersq age kidslt6 kidsge6