Lagged dependant variable (LDV) and AR

[Martin]

As we all know estimating an equation with LDV and AR using OLS will cause biased and inconsistent estimates of the parameters and inconsistent estimates of the error term and standard errors.

I was train to use a modified Cochrane-Orcutt procedure to estimate an equation with LDV and AR to obtain unbiased and consistent estimates. (Ramanathan 2002)

From EVIEWS6 user guide II it mentions that EVIEWS estimates AR model using nonlinear regression techniques.

My question is, if I have LDV and AR and select LS method in EVIEWS will it produce unbiased and consistent estimates? Assuming no other estimation problems.

[EViews Glenn]

Nonlinear estimation is valid using lagged dependent variables and ARs. Fair's (1984) book Specification, Estimation, and Analysis of Macroeconometric Models offers the best description that I have seen of why this is so...

The textbook methods that you have seen are, to a large extent, a relic of the sequential development of the literature, and the past lack of software for nonlinear estimation.