Kao's panel cointegration tests

For requesting general information about EViews, sharing your own tips and tricks, and information on EViews training or guides.

Moderators: EViews Gareth, EViews Moderator

alessio74
Posts: 2
Joined: Mon Aug 10, 2009 11:28 pm

Kao's panel cointegration tests

Postby alessio74 » Mon Apr 11, 2011 10:12 am

Hi everybody,

is there anyone willing to explain me the eviews-7 output for the Kao's panel cointegration tests?

I know from the literature that there are four statistics df-rho, df-t, df-star-rho and df-star-t.

Eviews reports two lines (DF and DF-star) and two columns (rho and t statistics).

How does the latter suits with the former?

Thanks.

Alessio

EViews Glenn
EViews Developer
Posts: 2671
Joined: Wed Oct 15, 2008 9:17 am

Re: Kao's panel cointegration tests

Postby EViews Glenn » Mon Apr 11, 2011 10:47 am

We had hoped that it would be clear from the output, but perhaps not...

First line:
df/rho is df-rho
df/t-Statistic is df-t

Second line:
df*/rho is df-star-rho
df*/t-statistic is df-star-t

ccc_researcher
Posts: 3
Joined: Wed Jul 25, 2012 7:17 am

Re: Kao's panel cointegration tests

Postby ccc_researcher » Wed Jul 25, 2012 7:29 am

Hi everyone,

I am trying to run Kao and Pedroni's cointegration test for the following model in Eviews7:

LMOB LFIX LPOP LRDPG COMP TIME


Can someone tell me how many observations (from the time dimension) is necessary to conduct the tests? My panel dataset is on 14 cross-sections for the period 1996-2010 (15 years) and I keep getting the message that there are insufficient observations.

EViews Glenn
EViews Developer
Posts: 2671
Joined: Wed Oct 15, 2008 9:17 am

Re: Kao's panel cointegration tests

Postby EViews Glenn » Thu Jul 26, 2012 11:31 am

How many are required to run the test and how many you should have are two different things.

I just ran the Pedroni and Kao tests using the default settings on data the size of your workfile with no problem. Note that if there are any missings in the data, the number of available observations will drop.

Having said that, that is an awfully short panel to be used for cointegration testing, especially for the nonparametric estimation that is at the heart of the Pedroni tests.

ccc_researcher
Posts: 3
Joined: Wed Jul 25, 2012 7:17 am

Re: Kao's panel cointegration tests

Postby ccc_researcher » Thu Jul 26, 2012 12:02 pm

Thanks for the prompt response. I am really not sure why I am getting that message. I do not have any missing data (using a balanced panel). I know that the time period is short but I am dealing with both I(0) and I(1) variables. Before using applying a DOLS model to the data, I would have liked to test for cointegration. I am however getting a result when I drop the variables COMP (a dummy) and TIME and even so Pedroni indicates no cointegration and Kao indicates that there is.

ccc_researcher
Posts: 3
Joined: Wed Jul 25, 2012 7:17 am

Re: Kao's panel cointegration tests

Postby ccc_researcher » Thu Jul 26, 2012 12:17 pm

Ok. I believe it is the dummy variable then that may be giving me problems.

tgmoe
Posts: 3
Joined: Wed Dec 07, 2011 4:06 pm

Re: Kao's panel cointegration tests

Postby tgmoe » Wed Oct 15, 2014 4:48 pm

Is the insufficient number of observations statement in regards to the length of time series or the number of cross-sections? I am receiving the statement with the Kao tests but not with the Pedroni tests.


Return to “General Information and Tips and Tricks”

Who is online

Users browsing this forum: No registered users and 15 guests