Johansen cointegration

[darryl]

Hey all, I am doing a research on speculative activity on commodity prices using Johansen's cointegration and vecm. However, I have a couple of queries to clarify.

1) I want to find the effect of macro, market level forces and speculation (in futures market) on spot crude prices and crude futures prices. Is it okay to perform Johansen cointegration on all these variables together, or do I have to do a separate one for spot and then future prices?

2) Similarly, I use several measures of speculation: net-positions of traders, total volume, open interest. Can I perform Johansen cointegration on them together, or do I have to perform it separately?

3) The data frequency is monthly (I transformed some annual data to monthly, but most others are monthly data). There are significant seasonal component in some of the series. Do I have to seasonally adjust those series before performing the cointegration?

Thanks in advance!

[Dim]

Hi,

You did a big mistake in your 3):

You CAN NOT transform annualy data into monthly data... Doing so, you speculate the values of your series. It's absolutely not permitted to do it if your not sure of the volatility of your series. For example, it is possible to transform annually dat into monthly data with population values. Because you know that - normally - the population grows at a constant rate. But some values like the GDP do not follow the same properties!

In conclusion: BE careful!

D.

[anca.dinut]

Hello,

I've got a question regarding the cointegration test: I've got an econometric model based on 6 variables: 5 of them are I(1) and the sixth vairable is I(2). in this case, can I still run the cointegration test (since one condition is that all the variables have the same integration order)?
if yes, how do I do this?
if not, how cand I continue if I can't find a stationary linear combination?

Thank you in advance!

[donihue]

You can run the test, but only after transforming the I(2) variable into an I(1) variable by first-differencing it
Regards
Donihue

[arkoudex]

if the series are all I(2). we can run the test? thanks

[nayana3012]

hi all!! i am tring to perform my cointegration test on my data but it continues givin me the error message ''near singular matrix''!! because of that i am not being able to proceed!!cn somebody please help me on this matter?? also, when giving the 6 options to choose from, i let it automatically choose option 3 since i dnt know what to choose really and i dnt know what to choose under the critical values box!!!can xomebody please help me??? :(

[darryl]

Yes you can perform cointegration test if all your variables are I(2). So long they are all of the same order.

[semsemkoko]

Hi every one. I would like to run cointegration test using johansen. my model consist of 6 series, 4 of them are stationary at 1st difference, and 2 of them are stationary in levels. So, can I use cointegration test in that case for all variables as they are not inegrated in the same order? What I should do in that case? also, what I should do after cointegration test if so as I don't have a good experience in that. please advise me as soon as possible. Thanks

[Farid]

you cannot use cointegration due to differences in the order. Just use normal regression with differenced data. or ARDL

[esha khanna]

Can any one please guide me..i am using panel data.all my variables are I(0) ..can i use johansen cointegration in panel data framework ..and secondly please suggest if any other panel cointegraion technique can be used..please help
thankyou so much

[startz]

If your variables are I(0) then they are not integrated. If they are not integrated they cannot be cointegrated.

[esha khanna]

thankyou so much startz