Dear all:
I would like to ask for help on the following.
>
> When one uses panel data and needs to resort to Fixed Effects
Estimation (FEE)
> the problem arises that some of the regressors might not vary in time
and,
> therefore, they are "wiped out" by the transformations involved
> in the FEE (as these consist of finding the "within group" variation
of the
> regressors).
A solution proposed by Hausman and Taylor (1981), HT, suggest
substituting in a OLS
model the problematic variables with instrumental variables that are
assumed not to
be correlated with the fixed effects. The HT estimator requires that the
number of
exogenous time-varying variables to be greater than the number of
variables X1,
that are both endogenous (and are therefore correlated with the fixed
effects,
which invalidates OLS analysis) and static (so they are swept away when
more
conventional methods are used to isolate them from the fixed effects).
This problem happens often when we analyse labour supply data and we
find that the characteristics
of people (race, schooling, gender) do not change for each individual,
but whose effects are
interesting to estimate.
Amemiya and MaCurdy (1986) (AM), and Breusch, Mizon and Schmidt (1989)
(BMS)
suggest the use of additional instruments that are valid under
increasingly
restrictive exogeneity conditions. These additional instruments add
explanatory
power if there is a variation in time of the correlation between the X1
= exogenous
and time-varying regressors and Z2 = static and endogenous regressors.
My understanding is that there would be no gain if the set Z2 is empty.
Is there?
Should I expect the estimates to be very similar but different... should
I expect
the t-ratios to decrease ...as I go from HT to AM and BMS? or should the
three
estimations yield exactly the same result if Z2 is empty so that it
cannot play any
role?
I would appreciate any comments on this.
Rober
The references mentioned are:
Hausman, J. and W.E. Taylor (1981) “Panel Data and Unobservable
Individual
Effects.” Econometrica. Vol. 49, No. 6, p. 1377-1398
Amemiya, T. and T.E. MaCurdy (1986) “Instrumental-Variable Estimation of
an
Error-Components Model”. Econometrica. Vol. 54, No 4, p. 869-880
Breusch, T.S., G.E. Mizon and P. Schmidt (1989) ”Efficient Estimation
Using Panel
Data.” Econometrica 57, No. 3, p. 695-700
--
======================================================================
Roberto Martinez-Espineira
Environment Department
University of York
Heslington
York YO1 5DD
Tel: +44 (0)1904 434067
E-mail: [log in to unmask]
Webpage: http://www-users.york.ac.uk/~rme103/webpage1_htm.html
======================================================================
Quote of the day:
"Let us so live that when we come to die even the undertaker
will be sorry"
--Mark Twain
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|