Should the Johansen Cointegration test by conducted on variables at levels or in their transformed state?
Johansen test should be conducted with variables that are stationary. If not, the likelihood is that you will be dealing with explosive series that do not exhibit mean reversion. To your question, usually, most economic variables are stationary at first difference. But it could also be possible that you are dealing with a series that is stationary at level. Do your stationarity test first. Good luck.
Sorry, in the Johansen framework, I suggest you log transform your variables and use them at level. The transformation at difference will be done by Eviews. But be sure that the variables are stationary at first difference since some variables are stationary at 2nd difference or higher.
Yes!, the variables must be at their transformed state, this is because cointegrated variables are generally unstable in their levels. Therefore Johansen cointegration test cannot be applied to non-stationary variables, as this method requires all the variables to be I(1).
*
*
*
Goodluck
Johansen Cointegration technique shows the long run association between the variables. But if variables are stationary at level then it implies no long run association. Thus to estimate this test all variables must be integrated of same order i.e. both i (1) or both i (2). So first check stationary property of the variables using ADF or PP test and then estimate accordingly.
Johansen Co-integration test should be conducted on variables at levels.The reason is quite clear: two or more variables (that are non-stationary) are said to be co-integrated if they have a linear combination that is stationary, that is, if they grow up at roughly the rate. If they are stationary at levels, or if you conduct the test after differencing them, then you will never find co-integrating relationships.
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory