A new poll from Public Opinion Strategies and GQR for NPR, which shows Democrats performing badly in vulnerable House seats, is making a lot of waves this morning. It certainly contains bad news for Democrats -- however -- it is the same bad news that was already implied by generic ballot polling.
The poll shows Democrats losing by an average of eight points -- 41 to 49 -- on average between the 70 districts that POS/GQR surveyed. However, ten of these districts are Republican-held seats. In the 60 House seats deemed to be most vulnerable by POS/GQR, the Democrats deficit is slightly milder at 42-47.
I'm not exactly sure which seats were contained among the 60. But I assume the list overlaps very heavily with the 64 Democrat-held seats that are rated by Cook Political as "lean Democrat" or worse.
Most of the House battlegrounds this year are on Republican turf, simply because that was the territory that Democrats "conquered" in 2006 and 2008. Specifically, these seats have, on average, a PVI of R+4. That means that Republican candidates for President there in 2004 and 2008 received about 4 more points than they did nationwide, and the Democratic candidate, 4 fewer points.
So, to "translate" the NPR poll to a nationwide, generic ballot poll, you'd probably want to add 4 points to the Democrats' total, and subtract 4 points from the Republican poll. This would imply that the POS/GQR poll is consistent with a generic ballot result of about Democrat 46, Republican 43. That's within the margin of error of an average of generic ballot polls.
Broadly speaking, this poll is consistent with the impression I have had of the House picture for almost a year now, which is that the over/under on the number of net Democratic losses is about 40 seats (i.e. they have about even odds of losing the House), with a 90 percent confidence interval of about +/- 20 seats. I also suspect that the race characterizations issued by Cook, etc., are about half a grade too optimistic for Democrats on average, e.g., a "toss-up" seat should be thought of as somewhere between toss-up and lean Republican, and a lean Democrat seat should be thought of as somewhere between lean Democrat and toss-up. But I'm looking forward to completing work on our House forecasting model so that we can be more exacting about this.