FMOLS EVIEWS 8

[lnp3]

dear forum,
I am doing a panel co-integration study in Eviews 8 using FMOLS. Data is imported and defined as "panel" (from excel). Eviews 8 says that "First, you should choose between the pooled, weighted, and group mean (averaged) FMOLS estimators:• Pooled estimation performs standard FMOLS on the pooled sample after removing the deterministic components from both the dependent variable and the regressors" .

I don;t understand this statement. In order to use FMOLS, should I then define data as "pooled"? or does FMOLS automatically change panel data into pooled data? What is pooled data?

i really need your help on this issue. I always thought that panel data is a special form of pooled data.

If I import data as panel, can I use FMOLS?

Thanks a lot as always.

lnp3,
Mine D.

[EViews Glenn]

Yes. You don't have to change anything in your data. The use of "pooled" in this context refers to the standard econometrics term for combining the data from all of the cross-sections to construct an estimator. You can estimate your FMOLS using the pooled data or using the between estimator which averages across the individual cross-section estimates. EViews will do this with your panel structured workfile. The use of pooled here is unrelated to the concept of a pool object in EViews.

[lnp3]

thanks a lot! I am relieved to hear.

[lnp3]

dear list,

For the final estimation in panel co-integration analysis in Eviews 8 (fully modified FMOLS, Dynamic Least Squares DOLS), should I enter the variables in their first difference forms or level forms in the estimation box where it tells me to list my variables---dependent variable followed by a list of regressor?

Given that my variables must be "non-stationary at level but rendered stationary at first difference" for panel co-integration analysis, is this right the estimation command?

D(X) C d(X1) D(x2)

OR

After, we prove integration at Levels and at 1st difference, all the rest analysis is made on raw data, ie at LEVEL DATA

X C X1 X2

I have seen some papers where they used Dynamic Least Squares (DOLS)-- a form of panel co-integration regression:
1) where variables enter the regression box in their level forms (they say results of "long run elasticities")
2) where variables enter the regression box in their first level forms (they say results of "short-run elasticities")

So this means that, only for "short-run" relationships in panel-cointegrating estimation, the analysis is made on first level data?

hope somebody can clear my confusion.. am I correct?

I really appreciate your help.

[EViews Glenn]

Levels.

[lnp3]

thanks

[lnp3]

in other words, my second regression command was correct.

After, we prove integration at Levels and at 1st difference, all the rest analysis is made on raw data, ie at LEVEL DATA.

Hope I cleared my confusion..:-)

best regards,
M.D.

[economist.pk]

Hi! Moderators and Other Forum Members.

I want to add common time dummies in FMOLS equation to account for cross section dependence. How I can add these dummies in model using Eviews 8? Please guide me

[EViews Glenn]

Time dummies are not built into EViews 8.

If you have a balanced panel then I believe that you can remove the common time effects yourselves prior to performing the testing. I believe that both Perron and Mark and Sul discuss this issue. I do vaguely recall, however, that these results implicitly assume a balanced panel.

[economist.pk]

Time dummies are not built into EViews 8.

II believe that both Perron and Mark and Sul discuss this issue.

Full reference please??? you mean Nelson C. Mark & Donggyu Sul, 2003. "Cointegration Vector Estimation by Panel DOLS and Long-run Money Demand," Oxford Bulletin of Economics and Statistics, Department of Economics, University of Oxford, vol. 65(5), pages 655-680, December. And which is other one?

[EViews Glenn]

Sorry, I was doing this from memory while on my cellphone. And obviously my memory isn't very good. :)

The Mark and Sul reference is correct, p. 667 discusses common time effects where they pre-transform the data to eliminate individual and common variation then apply the period fixed effects.

As for the other reference, I meant to point you to the Pedroni (I'm working on unit root stuff right now so I have "Perron" on the brain, sorry). Pedroni (2000) "Fully Modified OLS for Heterogeneous Cointegrated Panels", in Nonstationary Panels, Panel Cointegration, and Dynamic Panels, Volume 15, 93-130. (p. 114 has the *brief* discussion).

I haven't worked through all of the algebra on these two methods, but my instinct is that they both require balanced panels. In the typical fixed effects case, unbalanced panels implies that the two dimensional spectral decomposition doesn't have a simple form where you can do the cross and period effects sequentially. So my intuition is that these approaches will only work in fully balanced cases. But again, I haven't worked out the proofs for these methods in either the specific or the general cases which is the primary reason that we haven't added them to EViews.