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



is possible to show that

_1 S2
NSNT


σv2ι0E


11X1X
UγTγ


GxT (xit


- Xi) (Xit - xi)0 GxT


ιk


σv2ι0Ψxι > 0,

(80)


as (N, T →∞) . So, the asymptotic normality in (79) holds if

ι0Qi,τ
Snt


N(0,1),


(81)


as (N, T → ∞). Let Pi,Nτ = 1q,t. Then, E (Pi,Nτ) = 0 and Pi EPi2NT = 1.
Thus, by the central limit theorem of the double indexed process (e.g., see
Theorem 2 in Phillips and Moon, 1999), we can claim that (81) holds, if we can
show that

^X EPi,NT1 ©|Pi2NT I ε} 0 for all ε0,             (82)

i

as (N, T →∞).

Now, in view of (80) , condition (82) holds if

supEkι0Qi2Tk4 ≤ supE kQi2T k4 < ∞.                (83)

Note for some constant M1 that

sup E kQi2T k


iT


sup E


iT


sup E


iT


-1T X Gx,T (Xit

Tt


tr


M1 sup


iT


Xi) vit


/1 X μ Gχ,τ (

τ2Λ ⅛Gx,τ∙
t2s2p2q


Xit


(Xip


xi) (Xis


Xi) (xiq


Xi)0 Gχ,τ ʌ
- Xxi ) Gx2T


+M1 sup


iT


+M1 sup


iT


×E (vitvisvipviq)


f'2 X tr (t2Gx∙τ (Xit — xi) (x


Gx,T (Xit   Xi) (Xis


×E vi2tvi2s


is


Xi)0 Gχ,τ


Xi)0 Gχ,τ


(T2 Xtr (βC,.τ (x


Gx,T (Xit   Xi) (Xis


is


Xi) (Xit


Xi)0 Gχ,τ


Xi)0 GxT


×E vi2tvi2s


(T2 Xtr (βC,.τ (x


Gχ,τ (Xit - Xi) (Xit


is


Xi) (Xis


Xi)0 Gχ,τ


Xi)0 GxT


. (84)


52


×E vi2tvi2s




More intriguing information

1. The name is absent
2. Knowledge, Innovation and Agglomeration - regionalized multiple indicators and evidence from Brazil
3. Ex post analysis of the regional impacts of major infrastructure: the Channel Tunnel 10 years on.
4. The name is absent
5. The Clustering of Financial Services in London*
6. Public infrastructure capital, scale economies and returns to variety
7. Standards behaviours face to innovation of the entrepreneurships of Beira Interior
8. The name is absent
9. Can genetic algorithms explain experimental anomalies? An application to common property resources
10. TRADE NEGOTIATIONS AND THE FUTURE OF AMERICAN AGRICULTURE