st: Re: standard errors in ivreg2

[Christopher Baum <kit.baum@bc.edu>]

No, quite incorrect. ivreg2 works perfectly well with OLS regressions. It does, by default, report asymptotic standard errors (and z-statistics). If you use the -small- option, it will report results (and standard errors) identical with those of -regress-, and t-statistics. Without the correction for finite sample size, asy standard errors will be smaller, and test statistics larger. I presume the sample size in these regressions is not very large, as with N > 125 it doesn't make much difference.

Kit


On Jan 7, 2011, at 10:27 AM, Shanker Satyanath wrote:

> Dear Kit,
>  
> I hope you are well.  I had a quick question for you.  I am reviewing a paper in which the authors are using ivreg2 to run  OLS regressions. (I have encoutered this a couple of times now.)  The standard errors are substantially lower with ivreg2 than with the standard reg command.  In this case it makes an insignificant result significant.  I assume that ivreg2 should not be used for standard OLS regressions.  Could you give me an idea/intuition as to why ivreg2 may yield standard errors that are biased downward when used for OLS?
>  
> Thanks and sorry for the bother.
>  
> Best,
>  
> Shanker Satyanath (NYU)




Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/