When that Newsweek poll came out last week showing Obama over McCain by a whopping 51%-36%, I don't think I was alone in assuming it was a kooky outlier. But here comes Bloomberg/L.A. Times today saying basically the same thing -- they have Obama up by 12%, and if you add wildcards Ralph Nader and Bob Barr, Obama leads by 15 -- 48% to 33%, the same margin as Newsweek.

Folks, there are bounces and then there are bounces. You rarely see a presidential race poll this lopsided -- sure, current performance is not indicative of future resuts (see: Dukakis '88), but even with maximum skepticism there's no way to read these results as anything but a big advantage for Obama.

Pollster.com's national averages are here. As of this writing, they stand at Obama 48%, McCain 42%, without the aforementioned Bloomberg number.

UPDATE (June 26): Some questions have been raised by McCain partisans over Bloomberg/LAT's party identification numbers, arguing that this poll underrepresents Republicans. This actually is a valid concern -- though Republican identification has been declining drastically lately, it's hard to believe it's down to 22% of the country. But this goes to one of the most controversial questions in polling: do you weight for party identification, or not? That is, do you standardize your results to a known level of partisanship (as you do with things like race, gender, income, etc) or do you take what you get as a reflection of the national mood? Evidence has not answered that question in the affirmative or the negative, and polls using either methodology have been variously accurate, so this issue isn't enough to discredit the Bloomberg/LAT poll. Mark Blumenthal has more on this.

I say the real culprit, in both the matchup results and the party ID, is probably that both Bloomberg and Newsweek polled a universe of registered voters (as vs. likely voters, which is what many other pollsters use). Polls of registered voters always favor Democrats disproportionately, because our supporters -- being predominantly young, lower-income, and minorities -- tend to vote less regularly. But I think an RV poll is actually a much better indicator for the 2008 race at this point, both because we're trying to measure the state of the race as a whole (not predict its outcome), and because Obama will probably fuck up everybody's turnout models (thanks to his strength among African-Americans and young people, plus his legendary field operation). So for those reasons I'd give the Bloomberg and Newsweek numbers more credence than the McCain people want us to, and less to polls that use a likely-voter screen.

(NB: One criticism of the RV/LV argument for this discrepancy points out that the Gallup Daily Tracking Poll, which uses an RV universe, shows a much closer race than anybody else. But that's a tracking poll, which is really apples and oranges; it's designed to measure day-by-day trend shifts, not the overall picture.)