Vector Autoregressions Model
a vector
Autoregressions:
could the model economic mariables such
· we use MR
forecast key
to as:
II
3
inflation Rate
Rate
ofunemployment
21
(3)
growth
Rate
ofGDP
(4) interest
R ates
model
lagged
set which the
vector
Autoregression VR time values
A a a a
regressions;
series
of
· series are
regressors of
is in
model extends the model a
A
autoregression rector time-seriesvariables.
w i re univariate to matrix
of
·
the and called
oflags
When the in each is the to this
number
equal MR(P)
same
equations system ofequations
or
·
of pc is a
consider the variables;4* the models
·
for simplicity; we case a-time
of
series
*s: mrcps consists 2
of equations:
a YPot Piet... pip 4,Xe+... Up*z-p kit
+ + + +
(2) XtPot
PE-st... Rpt-p
+ +
k, xt- s+...+Yp*t-p+Yat
the confficients ofaw a r
are
estimated by estimating each
equation by ous
words
I note:n o model
only works
Ix+34
are
endogenous C
cory , t 0,
=
in other
they are
determined win the model
I for ex in micro:
Ifwe
set
mis-mas we can obtain Q* c
endogenous
in looka tthe phillips curve;then Asy4=> unempv
& be 4 inflation"
ifhe
c macro -
=
model
· So A
R of 2 time-series mariables; and XI;consists ofequations.
. in
. all dependentmirble t
is
· in (2) the dependantmariable x=
is
The
·
regressors
in both
equations are
lagged values
of
both mariables.
general;where there atime
· in are seriesmriables
The model for
. . new will consist
ofa equations; one each
ofthe marbles
· And the
regressors
in All
equations Are
lagged values
of
All the mariables.
Determining Lag Lengths in vars:
intercept
M
have model model
c
if we a
wa r w scariables 2 Y
lags ( in one we estimate 3x4+1 al
conficients
to estimate
I since we have suriables, we have sequitions.
to estimate total.
I
so 21x5 105 =
coefficients in
c
estimating All these
coefficients vbf (number
of
obs.) I estimation error
ofaforecast
do determine which include the model have to thata ll the variables
a
practically mariables to in WAR we make some
Are related.
C.
for ex: economic
theory suggests thatt he
inflation rate, unemploymentrate, a short-term
related
interest
rate are
together
that these could help
suggesting variables
C..
forecastone Another in A
WA r.
include undated mar;introduces predictive
uriable
adding content;
if
o we an in A An estimation error wont
thereby u
forecastaccuracy.
model the Rates and
C A
R
o ofInflation unemployment:
and the
a
consider A 2 mriable
for inflation
VR rate,
Futy, unemploymentRate, Unempt
but has stochastic trend
a since
info non-stationary
is (it a
3. So iti s
Appropriate transform by computing
it its
firstdifference
to
into
the and
3 so hr
for it, unemp consists
of2 equations:
So one
for in.As the
dependentvariable
· And the other
for uneras the dependentvariable.
