Re: st: Clustering of Standard Errors in a fixed effect model.

[Austin Nichols <austinnichols@gmail.com>]

natasha agarwal <agarwana2@googlemail.com> :
The number of clusters and how balanced they are determine the
tradeoff--see e.g. http://www.stata.com/meeting/13uk/nichols_crse.pdf
and refs therein, and for the follow-up see
http://www.stata.com/meeting/boston10/abstracts.html#baum

On Mon, Jun 21, 2010 at 7:22 AM, natasha agarwal
<agarwana2@googlemail.com> wrote:
> Dear Everyone,
>
> I am trying to estimate an augmented production function where am
> attempting to measure the effect of aggregate variable (FDI in a
> region) on micro units (productivity of domestic firms). I use an
> unbalanced panel data which spans for 5 years.
>
> According to Moultan (1986) paper titled "An illustration of a pitfall
> in estimating the effects of aggregate variables on micro units", it
> mentions that is is essential to cluster at the aggregate level when
> an aggregate regressor is specified in the right-hand side.
>
> However, Woolridge 2002 in his book Cross-Section and Panel Data
> mentions that clustering at the aggregate level when the number of
> groups is less than the group size might not be correct.
>
> I was wondering, if anyone could explain me what happens if one tries
> to cluster at the aggregate level in a fixed effect model when the
> group size (cluster size) is greater than number of groups (number of
> clusters)?
>
> Thanks
> Natasha
*
*   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/