Beruflich Dokumente
Kultur Dokumente
Gloria Gonzlez-Rivera
University of California, Riverside
and
Jess Gonzalo U. Carlos III de Madrid
You are free to use and modify these slides for educational purposes, but please if you improve
this material send us your new version.
Spurious Regression
y t y t 1 u t ; u t iid (0, 2u )
Set-up:
x t x t 1 v t ; vt iid (0, 2v )
Regress
; E(u t , v s ) 0 t, s
E(u t u t k ) E( v t v t k ) 0 k
y t x t t
p
0
p
2
R 0
t t distribution
some distribution
2
R some distribution
p
DW 0
T 1 / 2 t some distribution
t t - distributi on , where (long - run varian ce of )1/2
Maybe the same thing can be done to solve the SPR problem with
I(1) variables.
*
Pt St Pt
An index of the price level
in the USA
In logs :
A weaker version
of the PPP:
$ per
*
pt st pt
p t s t p*t z t
If the three variables are I(1) and zt is I(0) then the PPP theory is
implying cointegrating between pt, st and p*t .
Yt (1 )
i E t ( y t i ) c
i 0
Stock Prices
dividends
Consumption
labor income
If yt has a unit root and the PVM holds then Yt and yt will be
cointegrated (see Campbell and Shiller (1987)
Z t Yt y t is I(0)
What is an ATTRACTOR?
Consider the price (over time) of a commodity that is traded in
two different locations i and j.
.
pit p jt
p jt
3
.
.
4
1 : ( pi1 , p j1 )
5
.
2 : ( pi 2 , p j 2 )
.
2
1
t : ( pit , p jt )
45
pit
Definition of Cointegration
y t x t z t ; y t , x t I(1)
Perform an ADF test on the residuals:
t z
t 1
z
t i error
i z
i 1
Ho : 0
This means that the residuals have a unit root and therefore yt and xt are not cointegrated.
If the residuals are I(0) then yt and xt are cointegrated
Result 2.
If X t , Yt are cointegrated, then exists an ECM representation.
Cointegration is a necessary condition for ECM and viceversa
(Granger Representation Theorem).
Yt
Zt 0
Yt AX t
Xt
Example 1.
N 3, h 2
Yt Wt ut
X t Wt vt Wt I (1) ut , vt , st I (0)
Z t Wt st
1 commonstochastic trend Wt
2 cointegrating vectors: (1 1 0)' (0 1 1)'
Example 2.
N 3, h 1
Yt Wt ut
X t Wt Rt vt Wt , Rt I (1) ut , vt , st I (0)
Z t Rt st
2 commonstochastic trends Wt , Rt
N 2, h 1
Yt AWt ut
Wt I (1) ut , vt , st I (0)
X t Wt vt
1 commonstochastic trend Wt
Example 4.
Framework
Consider a VAR(p)
[1 2 .... p ]
where i [ i 1 i 2 .... p ] i 1,...p 1
Yt 1Yt 1 2 Yt 2 .... p 1Yt p 1 Yt 1 t
Z
2
t 1
H1 : h n
2(1 0 )
H 0 : h
H : h 1 maximum eigenvalue test
1