Interpretation of panel unit root test using -multipurt- command

[pankaj swain]

I am trying to find the presence of unit root or otherwise in my variables using the -multipurt- command, especially the CIPS test suggested by Pesaran (2007). However, I find that at lag 0, my variable is I(0), whereas it is I(1) at deeper lags. In this case, should I treat the variable as I(0) or I(1)?

This is going to be useful in deciding whether I proceed with a Johansen cointegration test (which requires all the series to be I(1) or a bounds-based Cointegration test as suggested by Pesaran, Sin, and Smith (2001), which is suitable in case we have a mix of I(0) and I(1) variables. The code and the results are as follows.

multipurt Stab_Low_T, lags(2)

First and Second Generation Panel Unit Root Tests

 Variables tested:           Stab_Low_T
 Group variable:             ID
 Number of groups:           39
 Total # of observations:    526+
 Average # of observations:  13.56+
 Panel is unbalanced and has gaps
 + Full sample statistics prior to testing.

----------------------------------------------------
 (A) Maddala and Wu (1999) Panel Unit Root test (MW)
----------------------------------------------------

----------------------------------------------------
             |     Specification without trend
-------------+--------------------------------------
    Variable |  lags   chi_sq   p-value
-------------+--------------------------------------
  Stab_Low_T |    0   634.984     0.000
  Stab_Low_T |    1   623.691     0.000
  Stab_Low_T |    2   190.263     0.000
----------------------------------------------------
----------------------------------------------------
             |     Specification with trend
-------------+--------------------------------------
    Variable |  lags   chi_sq   p-value
-------------+--------------------------------------
  Stab_Low_T |    0   318.359     0.000
  Stab_Low_T |    1   358.784     0.000
  Stab_Low_T |    2   211.316     0.000
----------------------------------------------------


----------------------------------------------------
 (B) Pesaran (2007) Panel Unit Root test (CIPS)
----------------------------------------------------

----------------------------------------------------
             |     Specification without trend
-------------+--------------------------------------
    Variable |  lags   Zt-bar   p-value     t-bar
-------------+--------------------------------------
  Stab_Low_T |    0    -6.126     0.000        .
  Stab_Low_T |    1    -1.321     0.093        .
  Stab_Low_T |    2     4.843     1.000        .
----------------------------------------------------
----------------------------------------------------
             |     Specification with trend
-------------+--------------------------------------
    Variable |  lags   Zt-bar   p-value     t-bar
-------------+--------------------------------------
  Stab_Low_T |    0    -7.658     0.000        .
  Stab_Low_T |    1    -0.664     0.253        .
  Stab_Low_T |    2     7.498     1.000        .
----------------------------------------------------
 Null for MW and CIPS tests: series is I(1).
 MW test assumes cross-section independence.
 CIPS test assumes cross-section dependence is in 
      form of a single unobserved common factor.

 -multipurt- uses Scott Merryman's -xtfisher- and 
      Piotr Lewandowski's -pescadf-.

Thanks in anticipation of an useful suggestion.

Regards
pankaj

[pankaj swain]

Any replies or suggestions, please.

Thanks and regards
pankaj

[Nick Cox]

Cross-posted at https://stats.stackexchange.com/ques...-test-in-stata

Please note our request to tell us about cross-posting.

[pankaj swain]

Dear Prof. Cox,

I do take note of your remark and will be open about cross-posting in the future. Further, in case my question gets answered on Stack Exchange, I would post a cross-reference to the same on Statalist.

Thanks
pankaj

[Nick Cox]

I don't work in this territory, as you may know. My guess is that a reply here is more likely.

[pankaj swain]

The link to my cross-posting of the issue to Stack Exchange is here.
stats.stackexchange.com/questions/663322/interpretation-of-panel-unit-root-test-in-stata?noredirect=1#comment1250142_663322

However, further insights and suggestions on the question would certainly be of immense help.

Regards