Large-N and Large-T Properties of Panel Data Estimators and the Hausman Test



while the within estimator βw is computed based on N(T - 1) within-individual

variations. Accordingly, (βb - β) converges to a zero vector in probability much
slower than (β
w - β) does. Thus, we can conjecture that the convergence rate
of (β
bw - βbg) will depend on that of (βbb - β), not (βbw - β). Indeed, we below

justify this conjecture.


In this subsection, we only consider a simple model which has a single time-
varying regressor (x
it) and a single time-invariant regressor (zi). Accordingly,
all of the terms defined in (7)-(11) are scalars. We consider asymptotics under
the RE assumption (3). To save space, this section only considers the estimators
of β and the Hausman test. The asymptotic distributions of the estimators of
γ will be discussed in Section 4. Throughout the examples below, we assume
that the z
i are i.i.d. over different i with N (0, σ2z). In addition, we introduce
a notation e
it to denote a white noise component in the time-varying regressor
x
it. We assume that the eit are i.i.d. over different i and t with N(0, σe2), and
are uncorrelated with the z
i .

CASE 1: We here consider a case in which the time-varying regressor xit
contains a time trend of order m. Specifically, we assume:

xit = Θ1,itm + eit.                              (12)

We assume the parameters Θ1,i are fixed with finite IimN→∞ -N Pi Θ1,i =
IimN→∞Θι p1,1 and IimN→∞-N Pi Θ^ i p1,2. We can allow them to be
random without changing our results, but at the cost of analytical complexity.
6
We consider two possible cases: one in which the parameters Θ1,i are hetero-
geneous, and the other in which they are constant over different individuals.
Allowing the Θ
1,i to be different across different individuals, we allow the xit
be cross-sectionally heteroskedastic. In contrast, if the Θ1,i are constant over
different
i, the xit become cross-sectionally homogeneous. As we show below,
the convergence rates of the between estimator and Hausman test statistic are
different in the two cases. Furthermore, whether or not the model is estimated
with an overall intercept could matter for convergence rates.

To be more specific, consider the three terms BNT, CNT, and bNT defined
below (8). A straightforward algebra reveals that with
rt t/T ,

BNT = Pi1,i - Θ1)2 mTT Pt rm 2

+2 Pi1,i Θι) TmmT Pt rm (ei - e) + Pi(ei - e)2;

CNT = Pi1,i - Θ1) TmmT Pt rm) (zi - z)+Pi(ei - e)(zi - z);

6We can consider a more general case: for example, xit = aitm + Θi + Θt + bizi + eit .
However, the same asymptotic results apply to this general model. This is so because the
trend term (
tm) dominates asymptotics.



More intriguing information

1. Der Einfluß der Direktdemokratie auf die Sozialpolitik
2. Accurate and robust image superresolution by neural processing of local image representations
3. Industrial districts, innovation and I-district effect: territory or industrial specialization?
4. BARRIERS TO EFFICIENCY AND THE PRIVATIZATION OF TOWNSHIP-VILLAGE ENTERPRISES
5. THE ECONOMICS OF COMPETITION IN HEALTH INSURANCE- THE IRISH CASE STUDY.
6. The name is absent
7. Licensing Schemes in Endogenous Entry
8. Educational Inequalities Among School Leavers in Ireland 1979-1994
9. The name is absent
10. The name is absent