EVI EWS tutorial:

Co integration and Error Correction Modal

EVI EWS
Start/ Programs/E views 5.1
Analyzing stationary in a single variable.
Analyzing co integration among a group of variables
Estimating an ECM model

Unit Root Test / Stationarity Test

To begin, double click on the series name to open the series window
Make a graph to check whether the series has trend or not
- Using view/Graph/Line

Graph (M)





0
40
80
120
160
200
1970 1975 1980 1985 1990 1995 2000 2005
M
Test for stationarity
- using View /Unit Root Test..
- First, use the topmost combo box to select the type of unit root test that you
wish to perform. You may choose one of six tests: ADF, DFGLS, PP, KPSS,
ERS, and NP.
- Next, specify whether you wish to test for a unit root in the level, first
difference, or second difference of the series.
- Lastly, choose exogenous regressors .You can choose to include a constant,
a constant and linear trend, or neither
-
Augmented Dickey-Fuller (ADF) Test





ADF Test Results: Level (I ntercept and Trend)

The first part of the unit root output provides information about the form of the test
(the type of test, the exogenous variables, and lag length used), and contains the
test output, associated critical values, and in this case, the p-value.

Null Hypothesis: M has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 1 (Fixed)


t-Statistic Prob.*


Augmented Dickey-Fuller test statistic -1.286370 0.8749
Test critical values: 1% level -4.243644
5% level -3.544284
10% level -3.204699


*MacKinnon (1996) one-sided p-values.


- The ADF statistic value is -1.286370 and the associated one-sided p-value is
0.8740which mean the statistical value is greater than the critical values so
that we do not reject the null at conventional test sizes.
- The second part of the output shows the intermediate test equation that
EViews used to calculate the ADF statistic,

ADF test results: first difference (I ntercept and Trend)


Null Hypothesis: D(M) has a unit root
Exogenous: Constant, Linear Trend
Lag Length: 1 (Fixed)


t-Statistic Prob.*


Augmented Dickey-Fuller test statistic -4.889876 0.0020
Test critical values: 1% level -4.252879
5% level -3.548490
10% level -3.207094


*MacKinnon (1996) one-sided p-values.








- The hypothesis that the first difference of the series M has a unit root can be
rejected because statistic value is less than the critical value. So series M is
I(1).

If you want to perform any of the other unit root tests (PP, KPSS, ERS, NP),
the right side of the dialog will show the different options associated with
the specified test.

Philips- Perron (PP)Test



Choosing the appropriate settings for your test, click on the OK button.





PP test results: Level (I ntercept and Trend)

Running a PP test using the M series yields:

Null Hypothesis: M has a unit root
Exogenous: Constant, Linear Trend
Bandwidth: 1 (Fixed using Bartlett kernel)


Adj. t-Stat Prob.*


Phillips-Perron test statistic -1.661058 0.7477
Test critical values: 1% level -4.234972
5% level -3.540328
10% level -3.202445


*MacKinnon (1996) one-sided p-values.



Residual variance (no correction) 40.48252
HAC corrected variance (Bartlett kernel) 41.24279
----------------------------------------------------------------------------------


- As with the ADF test, we fail to reject the null hypothesis of a unit root in
the M series at conventional significance levels.
- The second part of the output shows the intermediate test equation that
EViews used to calculate the PP statistic.

After performing different tests, the next step is how to report the results of these
tests.




How to report the estimated Results

The results of ADF test using series M, PM, PD and GDP are given in table 1
below:

Table 1: Results of the Unit Root Test

Series Level Difference Decision

-1.28637 (1) -4.88987 (1)** I(1)

-0.51901 (1) -3.80045 (1)** I(1)

-0.11595 (1) -3.63854 (4)** I(1)

2.09726 (1) -3.75127 (1)** I(1)

Note: The regressions include a constant and trend. The numbers in parentheses exhibits the
augmentation lags whereas **show significance at 5% and 1% level of significance respectively.

(Data relating to these series are given at the end of this titurial)
Other tests can also be reported in the similar way.















How to Perform a Cointegration Test
To carry out the Johansen cointegration test,
- Select View/Cointegration Test... from the group or VAR window toolbar.




Series may have nonzero means and deterministic trends as well as
stochastic trends.Similarly, the cointegrating equations may have intercepts and
deterministic trends. The asymptotic distribution of the LR test statistic for
cointegration does not have the usual distribution and depends on the assumptions
made with respect to deterministic trends. Therefore, in order to carry out the test,
must need to make an assumption regarding the trend underlying your data.




Results of a Cointegration Test:

The first part of the cointegration test output for the four-variable system is shown
below:


Date: 01/12/10 Time: 16:09
Sample (adjusted): 1972 2006
Included observations: 35 after adjustments
Trend assumption: Linear deterministic trend
Series: M GDP PD PM
Lags interval (in first differences): 1 to 1

Unrestricted Cointegration Rank Test (Trace)


Hypothesized Trace 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**


None * 0.617211 57.72883 47.85613 0.0045
At most 1 0.373971 24.11935 29.79707 0.1954
At most 2 0.135384 7.726795 15.49471 0.4951
At most 3 0.072531 2.635365 3.841466 0.1045


Trace test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level
**MacKinnon-Haug-Michelis (1999) p-values

Unrestricted Cointegration Rank Test (Maximum Eigenvalue)


Hypothesized Max-Eigen 0.05
No. of CE(s) Eigenvalue Statistic Critical Value Prob.**


None * 0.617211 33.60948 27.58434 0.0074
At most 1 0.373971 16.39256 21.13162 0.2028
At most 2 0.135384 5.091429 14.26460 0.7302
At most 3 0.072531 2.635365 3.841466 0.1045


Max-eigenvalue test indicates 1 cointegrating eqn(s) at the 0.05 level
* denotes rejection of the hypothesis at the 0.05 level.









Number of Cointegrating Relations
The first part of the table reports results for testing the number of
cointegrating relations.Two types of test statistics are reported i.e trace
statistics and the maximum eigenvalue statistics.
The (nonstandard) critical values are taken from MacKinnon-Haug-Michelis
(1999), and differ slightly from those reported in Johansen and Juselius
(1990)
To determine the number of cointegrating relations proceed sequentially
from r =0 to r = k-1 until fail to reject the null hypothesis.

Cointegrating vector (First stage regression)

Normalized cointegrating coefficients (standard error in parentheses)

M GDP PD PM C
1.000000 0.576596 -1.856109 0.273574 -30.77634
(0.40692) (0.62362) (0.44873)


How to report the estimated Results

Table 2: Johansen-Juselius Cointeration Test
Variables Hypothesized
No. of CE(s)
Eigenvalues -max 5%CV -trace 5%CV
M None 0.617211 33.60948* 27.58434 57.72883* 47.85613
GDP At most 1 0.373971 16.39256 21.13162 24.11935 29.79707
PM At most 2 0.135384 5.091429 14.26460 7.726795 15.49471
PD At most 2 0.072531 2.635365 3.841466 2.635365 3.841466

Note: The trace statistic and Max-Eigen statistics shows a unique long run relationship at 5%
level of significance.


The estimated cointegrating relationship and standard errors are given in equation:


M= 30.77634 0.576596 GDP + 1.856109PD 0.273574PM
S.E = (0.40692) (0.62362) (0.44873)

Estimation of Error correction Model
Select all variables.
Estimate OLS equation using Quick/Estimate equation.



Results of estimated equation

Dependent Variable: M
Method: Least Squares
Date: 01/12/10 Time: 17:10
Sample: 1970 2006
Included observations: 37


Variable Coefficient Std. Error t-Statistic Prob.


GDP 1.366056 0.235618 5.797759 0.0000
PD -0.768996 0.308138 -2.495623 0.0177
PM 0.475334 0.173069 2.746499 0.0097
C 2.646157 5.809106 0.455519 0.6517


R-squared 0.952927 Mean dependent var 74.78995
Adjusted R-squared 0.948648 S.D. dependent var 37.09991
S.E. of regression 8.407214 Akaike info criterion 7.197863
Sum squared resid 2332.481 Schwarz criterion 7.372017
Log likelihood -129.1605 F-statistic 222.6806
Durbin-Watson stat 0.555655 Prob(F-statistic) 0.000000

Save first-stage residuals (ut =RES)

Using proc/Make residual series.






Take first difference of all the variables.
Take lag 1 of residual term.
Again estimate the equation just as describe above.

Results of ECM

Dependent Variable: D(M)
Method: Least Squares
Date: 01/12/10 Time: 17:48
Sample (adjusted): 1971 2006
Included observations: 36 after adjustments


Variable Coefficient Std. Error t-Statistic Prob.


D(GDP) 1.954820 0.490876 3.982312 0.0004
D(PD) -1.212164 0.545322 -2.222840 0.0337
D(PM) 0.269950 0.234575 1.150803 0.2586
ECM(-1) -0.353499 0.119889 -2.948555 0.0060
C 1.048710 1.613422 0.649991 0.5205


R-squared 0.418107 Mean dependent var 3.900158
Adjusted R-squared 0.343024 S.D. dependent var 6.761994
S.E. of regression 5.480872 Akaike info criterion 6.368652
Sum squared resid 931.2387 Schwarz criterion 6.588585
Log likelihood -109.6357 F-statistic 5.568592
Durbin-Watson stat 1.944575 Prob(F-statistic) 0.001694


Error term must be negative and significant.

How to report the estimated Results

Table 3:Dependent variable (M)





D


Variables Coefficients Standard Error T-value
GDP
1.954820 0.490876 3.982312
PD
-1.212164 0.545322 -2.222840
PM
0.269950 0.234575 1.150803
EC(-1)
-0.353499 0.119889 -2.948555
Constant
1.048710 1.613422 0.649991