Forum: help

RE: Instrument variable for Simultaneous Equations [ Reply ]
By: Arne Henningsen on 2013-07-18 05:52

Dear Rathish

The instrumental variables also include the exogenous explanatory variables. The order of the instrumental variables is arbitrary. Hence, you can use:

inst1 <- ~ x1 + x2 + x3 + x5
inst2 <- ~ x1 + x3 + x4 + x5
inst3 <- ~ x1 + x3 + x5 + x6

In the 3SLS estimation, all instrumental variable must be uncorrelated with all error terms. Therefore, there are hardly good reasons for specifying different instruments for the different equations. Hence, you could use:

inst <- ~ x1 + x2 + x3 + x4 + x5 + x6

For details, see:
http://www.jstatsoft.org/v23/i04/paper

... and please do not forget to cite the systemfit package in your publications. Thanks!

Best regards,
Arne

Instrument variable for Simultaneous Equations [ Reply ]
By: Rathish Bhatt on 2013-07-17 11:10

[forum:39726]

Dear Sir,

For my panel data, i have the following system of 3 simultaneous equation:
y1=y2+y3+x1+x2
y2=y1+y3+x3+x4
y3=y1+y2+x5+x6

the instrument variables for y1,y2 and y3 are x1,x3 and x5 respectively.

so when we are giving the input for 'inst' (while estimating the model using 2sls and 3sls), do we mention only the IV used in each equation as given below?:
inst1<- ~ x3+x5 #x3, x5 will replace the endogenous var y2 and y3 in eq1
inst2<- ~ x1+x5
inst3<- ~ x1+x3
If this is the right method, then how will the systemfit package know that y2 and y3 are to be replaced by IV x3 and x5 while estimation?

or do we need to give the full set of IV and exogenous variable present in each equation as below?..
inst1<- ~ x3+x5+x1+x2 #x3,x5 are IV and x1,x2 are exogenous var for eq1
inst2<- ~ x1+x5+x3+x4
inst3<- ~ x1+x3+x5+x6

i would estimate this system of simultaneous equation using 3sls and 2sls. i am not very clear as to which of the two ways of entry of instrument variable (mentioned above) is the right way.

Thanks and Regards,

Rathish Bhatt