serial correlation in panel data model

[treefarm]

Hello,

I am estimating a panel (balanced) least squares model with cross section fixed effects in EViews 7. I have 44 cross sections and 13 time periods. Serial correlation is indicated by running the following estimation on the residuals: residual = c + residual(t-1) + residual(t-2). The coefficient on the residual lagged by one time period is significant at the 1% level. I reran my original specification with the White period correction (d.f. corrected), but I received a warning from EViews that the estimated coefficient covariance matrix is of reduced rank. Assuming the White period correction cannot be used to address serial correlation in my case, what is the next solution I should try in EViews?

While investigating the serial correlation problem, I tried adding the dependent variable lagged by one time period as an explanatory variable in my equation. When I did this, the estimated coefficient (0.30) was highly significant (p value 0.0000), so I assume this means it should stay in the equation, and that I am now working with a dynamic model.

Thanks for any help.