The construction of panel unit root

[guo]

Dear Tom,

Always thank you so much for your kind reply.

After rearranging the data, right now I am doing panel unit root test in terms of Constant and Constant+Trend.

Code: Select all

OPEN DATA "D:\Electro New\panel rearragment 2012-7-8.xlsx"
CALENDAR(M) 2003:2
DATA(FORMAT=XLSX,ORG=COLUMNS,right=100) 2003:02 2012:04
pform x
# individual_1 to individual_96
calendar(m,panel=2012:3) 2003:2
all 96//2012:3
pform(repeat) p_new
# p
pform(repeat) n_new
# n
* panel unit root test

* Supplementary card:
*   If you list just one variable, it will do a raw panel unit root test
@pancoint X
# X
@pancoint P
# P
@pancoint  N
# N

Then I got the results:

Code: Select all

     RESULTS:
         ********************************************
             -raw panel unit root test results-

             Levin-Lin rho-stat   =     -371.05071
             Levin-Lin t-rho-stat =     -143.86232
             Levin-Lin ADF-stat   =      -72.30250

                 IPS ADF-stat   =      -97.65569
             (using large sample adjustment values)

        Nsecs = 96 , Tperiods = (unbalanced) , no. regressors = 0

            All reported values are distributed N(0,1)
            under null of unit root or no cointegration

            Panel stats are weighted by long run variances
                (see OBES reference for details)
         ********************************************

      Currently computing panel statistics. Please wait.

## SR10. Missing Values And/Or SMPL Options Leave No Usable Data Points
The Error Occurred At Location 2830, Line 236 of PANCOINT

      Currently computing panel statistics. Please wait.

## SR10. Missing Values And/Or SMPL Options Leave No Usable Data Points
The Error Occurred At Location 2830, Line 236 of PANCOINT

My questions are:
1. How to do panel unit root test in terms of Constant and Constant+Trend? Note:Options in @pancoit do not work here.
2. How to get the panel unit root for P and N also in terms of Constant and Constant+Trend?

Thank you very much for your precious time and attention.

Best Regards

guo

[moderator]

First of all, you don't want to apply a panel unit root test to P and N because they aren't panel data series---they are single time series and should be tested using standard unit root tests.

The other question though is whether there is any reason to test any of these for unit roots and why you would want to allow for a clearly non-existent trend.

[guo]

Dear Tom,

Always thank you very much for your precious time and attention.

I ran

Code: Select all

@pancoint X
# X

in panel analysis.

The result is:

Code: Select all

   RESULTS:
         ********************************************
             -raw panel unit root test results-

             Levin-Lin rho-stat   =      -29.90221
             Levin-Lin t-rho-stat =      -10.76833
             Levin-Lin ADF-stat   =       -2.50490

                 IPS ADF-stat   =      -15.28625
             (using large sample adjustment values)

        Nsecs = 96 , Tperiods = (unbalanced) , no. regressors = 0

            All reported values are distributed N(0,1)
            under null of unit root or no cointegration

            Panel stats are weighted by long run variances
                (see OBES reference for details)
         ********************************************

Could you please tell me how to explain the result? The result means the panel X is stationary or not? What does OBES mean?

I am very looking forward to your kind reply.

guo

[TomDoan]

"All reported values are distributed N(0,1) under null of unit root or no cointegration"

The test statistics are WAY out in the left tail. These are about as far from I(1) as a series can be.

OBES is Oxford Bulletin of Economics and Statistics.