DOLS estimation with and without lags and lead

[amkh8857]

Dear,

I am testing a model by Panel Co-integrated DOLS method. I tested the model by specifying the values in Nonstationary estimation settings window. I get two different results (everything including the coefficients and the R2-adjusted) for these settings differing only in specification of "lags & leads methods". What is the difference of setting lags and leads equal to zero, or setting them to None? The E-Views Manual on page 802 (See, Chapter 42. Panel Cointegration Estimation) states that: "If you select None, EViews will estimate static OLS.". Then is it really the case that there is a difference between a Static OLS (None) and a DOLS with Fixed leads and lags specification (lead=0, lag=0)?

See these settings:

Setting 1:
Method: Dynamic OLS (DOLS)
Panel method: Weighted estimation
Fixed leads and lags specification (lead=0, lag=0)
Long run Variance Option: [Weighting options\Lag Specification: None, Kernel Options\Kernel: Bohman, Bandwith method: Newey-West Fixed, Crossed the truncated to integer]
Adjusted R-squared 0.651241

Setting 2:
Method: Dynamic OLS (DOLS)
Panel method: Weighted estimation
Static OLS leads and lags specification
Long run Variance Option: [Weighting options\Lag Specification: None, Kernel Options\Kernel: Bohman, Bandwith method: Newey-West Fixed, Crossed the truncated to integer]
Adjusted R-squared 0.276911


I look forward to get an answer on this issue.

Many thanks.

Best regards,
AMKH

[EViews Glenn]

The former includes the difference at a lag of zero. The latter includes no differences.

[amkh8857]

Many thanks for the timely reply. I need some clearance. I appreciate if you consider it.

First, by the former (includes the difference at a lag of zero) you meant, DOLS with Fixed leads and lags specification (lead=0, lag=0), and by the latter includes no differences you meant the Static OLS leads and lags specification.

Secondly, what is the DOLS with lag and leads zero? Isn't it the same as with no lags and leads? If so, then what is the difference between the two case I mentioned in the earlier post. I am referring to the literature, and it is stated that SOLS is a special case of DOLS, then in what way? Consistently the output of SOLS in Eviews is still titled DOLS estimation.

Many thanks again.

Best regards,
AMKH

[amkh8857]

The former includes the difference at a lag of zero. The latter includes no differences.

What does the "difference at a lag zero" means? Is it possible to have a dynamic model with lag zero?!

[EViews Glenn]

DOLS adds differences of the cointegrating variables on the RHS at various lags and leads (-p1 to p2) in the long-run regression.

Thus it is possible to have no lags and leads in the regression (p1=0, p2=0), but to still include the differences at a lag of 0.

The static option allows you to remove even the differences at lag 0.

Since the variables are assumed to be cointegrated, the latter is superconsistent by the usual Phillips-Moon arguments.