Two-Stage Least Squares (2SLS)

[hussein bataineh]

My model has two endogenous variables and each has its own set of instrumental variables.


this is my model

ivregress 2sls EQUITY FSIZE OP_CF SD_OCF TAX BM LEV MA DPP RPP PSIZE DR Gov_score Sustain_Perf Sust_Commit i.id i.year (CSO= Instrumental) (CSOXSusCommittee = Instrumental Instrumentaltwo) , first robust

the first endogenous variable is CSO and the name of its instrumental variable is Instrumental
the second endogenous variable is CSOXSusCommittee and it has two instrumental variables Instrumental Instrumentaltwo

when i run the the code it gave me

invalid syntax
syntax is "(all instrumented variables = instrument variables)"
r(198);

Could you please advice me what to do ???

[Andrew Musau]

The error message tells you that the syntax needs to be:

ivregress 2sls ... (CSO CSOXSusCommittee= Instrumental Instrumental Instrumentaltwo) ...

[hussein bataineh]

Thanks Andrew Musau for help . Please the the problems i have two endogenous variables and each has its own set of instrumental variables . for CSO we have just one instrumental variable (instrumental) while for the second endogenous variable CSOXSusCommittee we have two instrumental variables . (Instrumental Instrumentaltwo )


if we run the code as you suggest the first endogenous variable will have two instrumental variable it should have one only


could you please help suggest how we can dealt with that?

Number of obs = 4,188
F(328, 3853) = 9598.92
Prob > F = 0.0000
R-squared = 0.6209
Adj R-squared = 0.5880
Root MSE = 0.2917


Robust
CSO Coefficient std. err. t P>t [95% conf. interval]

FSIZE .0106726 .020101 0.53 0.595 -.028737 .0500823
OP_CF -.2259394 .161462 -1.40 0.162 -.5424984 .0906197
SD_OCF 1.428631 .3206922 4.45 0.000 .7998882 2.057373
TAX .0239689 .0205146 1.17 0.243 -.0162516 .0641893
BM -.0453508 .0325633 -1.39 0.164 -.1091938 .0184921
LEV .0760251 .0775216 0.98 0.327 -.0759622 .2280124
MA -.0186127 .0387862 -0.48 0.631 -.0946562 .0574308
DPP -.0784279 .0581541 -1.35 0.178 -.1924436 .0355878
RPP -.1095729 .0822426 -1.33 0.183 -.2708161 .0516703
PSIZE .0537745 .0180921 2.97 0.003 .0183034 .0892456
DR -.0115439 .0167256 -0.69 0.490 -.0443359 .0212481
Gov_score .0003776 .0003411 1.11 0.268 -.0002912 .0010464
Sustain_Perf .001183 .0005215 2.27 0.023 .0001605 .0022055
Sust_Commit -.0512214 .0214162 -2.39 0.017 -.0932097 -.0092332
Instrumental .0022697 .0039373 0.58 0.564 -.0054498 .0099891
Instrumental 0 (omitted)
Instrumentaltwo .0122756 .0033237 3.69 0.000 .0057593 .0187919
_cons -.3918555 .2162017 -1.81 0.070 -.8157362 .0320252

[Andrew Musau]

There is no harm in including additional instruments. If the hypothesis is that the extra variables are irrelevant and do not predict your instrumented variable, then their coefficients will be zero. Consequently, only the coefficient of your "relevant" variable will have relevance in the first-stage regression.