Samuel,
You are describing a Hausman Taylor type model, with a time invariant explanatory variable. Traditional fixed effects, or first differencing, will eliminate time invariant variables.
There is an xthtaylor command that might work for your problem. Or you can code the generalized IV estimator described on p.327 of Woolridge's Econometric Analysis of Cross Section and Panel Data.
Carl Nelson
---- Original message ----
>Date: Sat, 10 Jan 2009 08:08:50 -0500
>From: "Samuel Finkelstein" <[email protected]>
>Subject: st: Appropriate Use of Two Way Fixed Effects
>To: [email protected]
>
>I am trying to figure out whether or not it is appropriate to use
>two-way fixed effects given the data I am currently using.
>
>My dataset consists of 150 firms with single yearly observations for
>each firm from 1990 to 2005. The primary independent variable of
>interest is a binary variable. Thus, for any given year, Firm A (or
>any other firm in the set) will only have a single observation. I was
>originally using a yearly fixed effect and clustering by firm.
>However, at a recent conference I was told that I may need to use
>two-way fixed effects (i.e. a year and firm fixed effect) instead of
>clustering.
>
>I have discussed this with a number of people and some have told me
>that this is appropriate, while others have told me that I cannot use
>two-way fixed effects with only a single observation for each firm in
>each year. The two different methods give me very different results.
>Can someone please tell me whether it is or is not appropriate to use
>two-way fixed effects given my data?
>
>Any response would be greatly appreciated.
>
>Thanks,
>
>Samuel
>*
>* 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/
*
* 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/
