st: Re: number of lags

[Kit Baum <baum@bc.edu>]

The logic here is the same as underlying Arellano-Bond. If the errors  
are AR(1), then they are correlated with y_t-1, and if the model is  
correct, they are correlated with x_t-1, so that the minimum lag is  
2. Calculating a robust (or cluster-robust) covariance matrix does  
not change how the coefficients are calculated: if the instruments  
are not orthogonal to the error, the IV estimates will not be  
consistent.  But if 2-period lags are appropriate, so are 3-period  
lags, 4-period lags, etc. so there is no reason why the equation  
should be exactly identified: you can use several lags of the x  
variables.


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


On May 28, 2007, at 5:39 AM, alessia matano wrote:


> Dear Kit,
>
> I am estimating an IV using lagged values of my independent variables
> as instruments. if I know that the erros are AR(1), the minimum lag I
> have to start is 2, or if I put robust cluster(id) is can use also
> lag(1)??
> The equation is exactly identified, so I can not test the validity of
> instruments
>
> Thank you
>
> alessia
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/