Scribd
Upload
53 views17 pages

Stationary and Nonstationary Series: T y y E T y S S T y T y S T y T y T y

This document discusses concepts related to stationary and nonstationary time series. It defines a stationary time series as one where the mean and variance are constant over time. A nonstationary series is one where the mean and/or variance change over time. It provides examples of economic time series that tend to be nonstationary, like GDP. It also discusses the risk of spurious regressions if nonstationary variables are used in estimation. Various tests for stationarity are presented, including unit root tests like the Dickey-Fuller test and the concept of cointegration between nonstationary variables.

Uploaded by

Prof B Sudhakar Reddy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
53 views17 pages

Stationary and Nonstationary Series: T y y E T y S S T y T y S T y T y T y

This document discusses concepts related to stationary and nonstationary time series. It defines a stationary time series as one where the mean and variance are constant over time. A nonstationary series is one where the mean and/or variance change over time. It provides examples of economic time series that tend to be nonstationary, like GDP. It also discusses the risk of spurious regressions if nonstationary variables are used in estimation. Various tests for stationarity are presented, including unit root tests like the Dickey-Fuller test and the concept of cointegration between nonstationary variables.

Uploaded by

Prof B Sudhakar Reddy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 17

Stationary and Nonstationary Series

A given time series  yt  is stationary when mean and variance are


constant or independent of time.
E  yt    constant mean
 
var yt    2 constant variance
 
cov yt yt  s   cov yt yt  s    s time independent covariance
   
Time series yt is non-stationary if the mean and variance is not
constant or is changing over time.

Many economic variables such as GDP, GDP components,


inflation, exchange rates, labour force evolve over time. It is
important to check whether these series are stationary or non
stationary before any econometric estimation because estimation
using non-stationary variables may generate a spurious
relationship: reported to have relationship when there is no
relationship.
1
Spurious Regression

C o n s id e r a s to c h a s tic a u to re g r e s s iv e A R (1 ) s e rie s :
yt   y  vt
t1

A p p lic a tio n o f O L S in th is m a y g e n e ra te a s p u rio u s


re g re s s io n , w ith a h ig h R 2 a n d v e ry lo w D u rb in -
W a ts o n s ta tis tic s (R 2> d ). O L S g e n e ra te s s p u rio u s
re g r e s s io n if v a ria b le s in v o lv e d a r e n o n s ta tio n a r y .

2
Four ways of checking stationarity

 Partial autocorrelation function and Ljung and Box


statistics

 Unit Root Test: Dicky-Fuller Test

 Cointegration Test

 Test for Error Correction


3
Unit Root and Non-stationarity
L ik e a u to c o rre la te d e rro rs c o n s id e r a s to c h a s tic
a u to re g re s s iv e A R (1 ) s e rie s :

yt   y  vt (4 )
t1

H e re 1    1 . It is a u n it-ro o t p ro c e s s if   1 . T h e n
(4 ) b e c o m e s a ra n d o m w a lk y t  y v .
t1 t

A p p lic a tio n o f O L S in th is m a y g e n e ra te a
s p u rio u s re g re s s io n , w ith a h ig h R 2 a n d v e ry lo w
D u r b in -W a ts o n s ta tis tic s ( R 2> d ).
T o s e e h o w (4 ) is n o n s ta tio n a ry , le t u s a s s u m e
t h a t v t ~ N  0 ,  v2  .
 

4
DF TEST

• Consider the hypothesis


H 0 :   1  Yt ~ I 1
H1 :   1  Yt ~ I  0 

• The hypothesis is the reverse of KPSS test.

5
UNIT ROOTS IN AUTOREGRESSION
1. DICKEY-FULLER (DF) TEST: The simplest
approach to test for a unit root begins with
AR(1) model
Yt   0  Yt 1  at
where at ~ Normal _ WN  0, .
2
a

• DF test actually does not consider 0 in the


model, but actually model with 0 and not 0
gives different results.

6
DF TEST
• To simplify the computation, subtract Yt-1 from
both sides of the AR(1) model;
Yt  Yt 1   0  1   Yt 1  at
Yt   0  Yt 1  at

• If =0, system has a unit root.


H0 :   0
H1 :   0

7
DF TEST
• DF (1979)
Yt  Yt 1  at  Pure RW
Yt   0  Yt 1  at  RW with drift
Yt   0  1t  Yt 1  at  adds drift and a linear time trend

8
DF TEST
• With a constant term:
The test regression is
Yt   0  Yt 1  at
and includes a constant to capture the nonzero mean
under the alternative. The hypotheses to be tested

H 0 :   1, 0  0  Yt ~ I 1 without drift


This formulation is appropriate for non-trending
 financial
H 1 :and
economic I  0 with
1  Yt ~series mean rates,
like interest
exchange rates and spreads.

9
DF TEST
• With constant and trend term
The test regression is
Yt   0  t  Yt 1  at
and includes a constant and deterministic
time trend to capture the deterministic trend
under the alternative. The hypotheses to be
tested
H 0 :   1,   0  Yt ~ I 1 with drift
H 1 :   1  Yt ~ I  0 with deterministic time trend
10
DF TEST
• What do we conclude if H0 is not rejected? The
series contains a unit root, but is that it? No!
What if Yt∼I(2)? We would still not have rejected.
So we now need to test
H0: Yt∼I(2) vs. H1: Yt∼I(1)
• We would continue to test for a further unit root
until we rejected H0.
11
ADF TEST
• Hence, testing for a unit root is equivalent to
testing =1 in the following model
p 1
ADF test equation : Yt  Yt 1    j Yt  j   0  at
j 1
or p 1
Yt     1 Yt 1    j Yt  j   0  at
 j 1

p 1
ADF test equation : Yt  Yt 1    j Yt  j   0  at
j 1

12
ADF TEST
• Hypothesis
H0 :  1 H0 :   0
H1 :   1 H1 :   0

Reject H0 if t=1<CV Reject H0 if t=0<CV

• We can also use the following test statistics:


 n  p  ˆ  1  1
where  1  1  1  2   obtained from RSF of the model.
13
ADF TEST
• EXAMPLE: n=54
Examine the original model and the
differenced one to determine the order of AR
parameters. For this example, p=3.
Yt  0  Yt 1  1Yt 1  2Yt 2  at
Fit the model with t = 4, 5,…, 54.
OLS equation :
Yt  139  0.856 Yt 1  0.141 Yt 1  0.326 Yt  2
 87.97   0.08699   0.1353   0.1369 
14
ADF TEST
• EXAMPLE (contd.) Under H0,
1  B  Yt      B  at
 
where    0 / 1  1B  2 B 2 and   B  
1
 p 1  B 

1
1  1B  2 B 2
.

1 1
  1    0.843
1  0.141  0.326 1.185
 n  p  ˆ  1 1  6.19  13.3
n=50, CV=-13.3

• H0 cannot be rejected. There is a unit root.


The series should be differenced. 15
Co-integration
If two economic variables have long-runequilibrium
relationshiplinear combinationof these variables may be
stationary evenif the individual series maybe non
stationary. These two variables are saidtobe co-integrated
toeachother.
Suppose yt is consumptionand Xt is disposable income.

et Yt   Xt
1 2

Evenif yt and Xt are I(1) et is I(0).

et  e vt
0 t1
If  is zerothenseries et is stationary and yt and Xt are
I(1).
16
Error Correction Model


y
 Y


X v
t 1 2t 1 2t t





T
het
erm i
n t
hepa
re
nt
hes
is
ist
hee
rr
ort
erma
ndthe
c
oe
ff
ici
ent2
g
ov
er
nsth
esp
eedo
fad
ju
stme
ntt
owa
rds
l
ong
-
run
eq
ui
li
bri
um.

17

You might also like