Refining my model and Serial Correlation

[axiom]

I'm modeling national GDP, and I have maybe 2 dozen different X variables.

As I refine the model (add appropriate lags to variables, etc...) my R-squared, regression errors, and Akaike and Schwarz all get better.... but my serial correlation seems to always get worse, according to my DW. My DW hangs around 3, despite the fact that all my other indicators suggest that the equation is doing well.

When I do an ex-post forecast, it too seems good and remains stable.

I'll add an AR1 AR2 and AR3, but that doesn't seem to help either, as the DW remains at around 3.

So what is one to do? Is it because of the nature of what I am attempting to forecast (GDP) that I will always be plagued by this problem, or is there something that can be done about it?