One downside for this fall's election is that respondents indicate they will vote for Democrats in their district this fall (44% for Democrats, 40% for Republicans). Last January it was 39%D, 41%R. See chart question Q16 for this result. No question I liked the January numbers better, but do you remember what they were before the '94 elections? Don't feel too bad. The generic numbers are next to irrelevant since there are only a hadfull of districts that are even marginally competitive. I've even heard that McAuwful has given up on re-taking the House this cycle (and 2004 won't be much prettier with Bush on the ticket and Texas finally redrawing it's lines the Right way).
Actually, the only accurate predictor of midterm success is presidential approval ratings. Clinton's were bad (equal parts approval and disapproval) in 1994 and his party was decimated. His numbers were great (66% job approval) in 1998 and he picked up seats (putting him in the league with FDR as the only president to pick-up midterm seats). A Bush with approval numbers in the high 60's or low 70's this fall will be very bad news for the Democrats.
2004 will see the pick-up of Ralph Hall's (who's basicly promised to retire) and Charlie Stenholm's (who's likely to retire) seats for starters.