Sujan, first take the log of all the variables and then check the stationarity of the variables. The variables integrated of order 2 are very rare and the variables you are saying to be I(2) can simply give you different result when log is applied.

Thanks for your suggestion. However, can you please see the attached file. and suggest me whether I go for Johansen or Engle-Gragner Cointegration test? @Nabya Tehreem

You can easily decide for the proper approach after applying unit root test on log of variables. If the order of integration is same, use johansen because your variables are greater than two. Engle Granger can only be used when their are two variables under study. If however, the order of integration is different for these variables, go for ARDL approach.

Thanks for your suggestion. Hope this will be also in future.

ARDL doesn't actually solve the problem of non-cointegration between the variables. There are higher order cointegration methods (which are more complicated). If all log variables are I(1), try Engle-Granger 3 step method - 2 step one has issues. E-G is still applicable if you are expecting one cointegrating vector among the (more than 2) variables. However, if you expect more than one (since there are 3 variables), then Johanssen is preferred. Having said that, Johanssen is not without its own problems either. Maddala and Kim's book on unit roots, cointegration and structural change is an excellent read. Good luck.

Cointegration is data testing that finds if there's a relationship between two or more time-related series.

Choose the one that suit you.

• The Engle Granger test is a test for cointegration. It constructs residuals (errors) based on the static regression.
• Johansen's test is a way to determine if three or more time series are cointegrated. it assesses the validity of a cointegrating relationship, using a maximum likelihood estimates (MLE) approach