Panel Co-integration tests

[lnp3]

dear forum,

i need your help. I am doing a panel co-integration study of some firms using Eviews 8 (latest). Before final estimation (Dynamic OLS, and Fully-Modified OLS), I first do panel unit root tests to verify stationary. if some of the variables are not stationary (contain unit root), I render them stationary by taking first difference operator (or second difference, if necessary).

Second, on the same variables, I do panel co-integration tests (Pedroni) in order select the ones with co-integrating (long run) relationships.

However, how do variables enter the combo box in Pedroni co-integration test (or any co-integration test)? in their level forms or first difference (if unit root exists) ? PLEASE HELP IF YOU CAN.

best,
lnp3

[wilshire]

I have the same problem.

How can I reproduce the same results in the paper Pedroni(2001)? I am attaching the data file.

[EViews Glenn]

Levels.

[wilshire]

Levels? Could you just tell me the commands to generate the same results? Thanks.

[EViews Glenn]

(Levels was the answer to the earlier question).

Estimate the panel cointegration model using the between estimators for either FMOLS and DOLS (using the COINTREG estimation method). The t-statistics are presented for you as part of the output. Note however that the standard errors are calculated slightly differently from those described in the paper. We describe this difference in the EViews manual.

[lnp3]

thanks Eviews Glenn.