COINTEGRATION (TIME SERIES)
1.OLS and cointegration
Professor Noman Arshed commented about OLS and cointegration as such >> If all
variables are I(0) no cointegration tests are required and OLS can be used.
2.Engle and Granger two step co-integration
Sayed Hossain commented about two steps spurious regression as such.
3. Cointegration relationship in VAR model
Professor Olasehinde Timilehin commented> If the number of variables are N,
cointegrating relationship would be maximum N-1. For example, if number of variables
in the VAR model are 5, cointegrating relationship can be maximum 4, can not be 5.
4. How to extract Johansen long run equation from VECM model?
Professor Sayed Hossain commented as such>
5. What does it mean number of cointegration And how we identified it?
DaifAllah Al-Houd has posted the following figure
Tahir Abubakar commented> Just put 1 in the no. of cointegrating equation box for
Sayed Hossain commented> I guess you have got 3 cointegration equation as per
Johansen test and that is why you have put 3 there. But I would prefer to put 1 there
instead of 3 so that I can extract Johansen long run equation.
Moulana Naykrasyvishyy Cholovik commented> This means ‘’number of cointegrating
equations’’ or simply, numbers of long-run equations. This option allows you to
choose how many co-integrating equations you wish to estimate at the same time
within 1 window. If you have 4 variables, for example, then you can choose maximum
3 cointegrating equations and you’ll get 3 different long-run equations for 3 different
dependent variables among your 4 variables. also you’ll get 3 different ECT. hope this
6. How to run Engle-Granger cointegration test using EVIEWS?
Seye Olasehinde-Williams commented>
1) select the variables and open as group.
2) under view select cointegration test
3) select single equation cointegration test
4) under test method choose Engle - Granger
Seye Olasehinde-Williams commented> Pls note that there are also Engle- Granger
based panel cointegration tests
Seye Olasehinde-Williams commente> Also note that Engle- Granger is not most
suitable when you have more than 2 variables in the model.
7. Interpret Johansen cointegration test
Ali Mohamed posted the Johansen image below.
Sayed Hossain commnted> As per Rank Test, "At Most 1" cointegration we fail to
reject null meaning we accept it. It means there is 1 cointegration relatonship. So
variables PRICE, FDI, INC and POP move together in the long run, So we can run
VECM model. But as per Max Eigen test, we faile to reject "none" meaning that there
is no cointegrating relationship so variables PRICE, FDI, INC and POP do not move
together in the long run. So we are unable to apply VECM model here but unrestricted
VAR model can be applied.
Now which one we should accept? Indeed, Trace and Max Eigen, both statistics are
equally efficient. So you can take decision on the basis of either one. But best would
be if both tests provide the same result.
8. Can you extract Johansen long run equation from the VECM result below?
Sayed Hossainposted the figure below:
Younes Azzouz commented> GDP=0.156 CONS - 0.544 EX + 47.79. Notify that EX
isn't significant. In this situation the equation is normalized to GDP which has the
coefficient 1. To make an equation with GDP as dependent variable, all other
variables go to the other side. Automatically, signs get inversed.
Saeed Aas Khan Meo GDP=47+0.15(COSN)-.54(EX)
Ghumro Niaz Hussain Just reverse the signs of normalized coefficients ur LR
Cointegration is GDP=47.78+0.15.69(COSN)-.54*(EX)
Sayed Hossain commented>I have just reversed the sign. So it comes up: GDP=0.
156CONS - 0.544 EX + 47.79. The bracket is t statistics. Here only EX isn't significant
while CONS is significant.
9. Types of cointegration tests
Seed Aas Khan Meo posted the below:
Meo School of Research
Hossain Academy Note