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: Fixed effects - not quite invariant variables?

From   <>
To   <>
Subject   RE: st: Fixed effects - not quite invariant variables?
Date   Mon, 15 Oct 2012 11:33:56 +0000

Hello Guo,
Many thanks. I know it might not be consistent, but I wonder if there was a way of overcoming this problem. When I use Fixed Effects Vector Decomposition I have significant results, but I've been advised not to use such method as apparently it is not consistent. The coefficients under FEVD were in fact much larger.

-----Original Message-----
From: [] On Behalf Of Guo Xu
Sent: 15 October, 2012 11:35 AM
Subject: Re: st: Fixed effects - not quite invariant variables?


When regressors are not changing much (as in your case), FE is prone
to measurement errors - have a look at the discussion in these
applications, it may help:

Barro, Robert J. (2000) 'Inequality and growth in a panel of
countries.' Journal of Economic
Growth 5(1), 5-32

Banerjee and Duflo (2003) 'Inequality and Growth: What Can the Data Say?'


On 15 October 2012 09:17, fernando luiz mistura <> wrote:
> Hi,
> I have an unbalanced panel data with 180 observations. Data is for 56 countries for the years 1997, 2003, 2006, 2010. I want to see what is the effect of an Index variable of FDI restrictions that ranges between 0 and 1 in the level of countries' FDI. Under Pooled OLS and Random Effects, the index variable is significant as expected.
> My problem is with fixed effects. The index variable for some contries do not change, as countries have liberalised in a period earlier that the one covered. For other countries, there are either small or large changes. Hence, under Fixed Effects however, the Index variable is not significant and the coefficients smaller.
> Could someone please explain me if the FE would still be appropriate even with this type of semi-variant variable? And if this issue could be addressed with an Instrumentral variable approach (hausmann taylor) even if its not a endogenous variable?
> Many thanks
> Fernando
> *
> *   For searches and help try:
> *
> *
> *

*   For searches and help try:

*   For searches and help try:

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