Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: How to run 2SLS twice

From   Christopher Baum <>
To   "" <>
Subject   Re: st: How to run 2SLS twice
Date   Tue, 15 Feb 2011 07:28:54 -0500

On Feb 15, 2011, at 2:33 AM, Mauro wrote:

> I am estimating a reduced-form model in which I need to perform 2SLS twice.
> In particular, this is the empirical model:
> 1) y = a + b1 * x1 + b2 * z1 + e
> 2) z1 = c + b3 * x2 + b4 * z2 + u
> 3) z2 = d + b5 * x3 + v
> e, u and v are classical error terms; x1, x2 and x3 are exogenous vectors.
> Provided that x3 is a valid and informative instrument, what is the
> best way to estimate this in Stata? I can think of different ways:
> 1. Estimate equation 3). Then use the fitted values to estimate 2)
> (correcting for the standard errors). And then again use the fitted
> values to estimate 1) (again correcting for the standard errors).
> 2. Instead of running 3 regressions, I could run ivreg2 or ivregress
> 2sls for the first two regressions or the last two and in either case
> use the fitted values from the remaining equation (once again
> correcting for the standard errors). But in this case, which
> regressions is it better to estimate with ivreg2 or ivregress 2sls?
> 3. Last, use reg3 with the iv option.

How to run 2SLS twice? Don't.

Just estimate each equation with ivregress or ivreg2, using as excluded instruments all exogenous variables in the system. In the case of the third equation, you can estimate it with OLS or with "HOLS" (see the ivreg2 help file). No reason to do 2SLS 'by hand' here. 

Forget reg3 with iv option; it is just an inferior subset of what you could do with ivregress or ivreg2 (i.e. it only handles i.i.d. errors).


Kit Baum   |   Boston College Economics & DIW Berlin   |
                              An Introduction to Stata Programming  |
   An Introduction to Modern Econometrics Using Stata  |

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index