I'm not clear on precisely what he's doing wrong, but I think he might be using a static model where he should be using a dynamic one. When I do a quick and dirty run on what I think he did, I get autocorrelated residuals, which usually means that you've forgotten a lagged dependent variable in your specification.

I'm not totally clear in how that would give him the sort of results he's getting, but models with autocorrelated residuals take great delight in producing interesting and innovative forms of gibberish.

- Jake

Friends come and go. Enemies accumulate.