The pollsters continue to have rather different impressions of NY-23, where Siena shows the race Hoffman 41, Owens 36, with a huge number of undecideds, whereas PPP shows Hoffman with a much larger 54-38 lead given a two-way vote choice.
Mark Blumenthal has a very good, and very thorough, overview of the polls, and basically argues that if people want to concentrate on the Siena result while ignoring PPP, they're probably engaged in wishful thinking (provided they're rooting for Owens, anyway).
I'm not going to try to further dissect the particular polls in question.** The main thing, though, is that the true margins of error are much different in different sorts of contests. Presidential primary polls, for instance, have missed by an average of about 7 points in recent cycles, which equals a standard error of about 8.5 points. This race is more analogous to a primary than a general election in most ways, with multiple candidates, some of them relatively unknown to the voters, and what will probably be a low-ish turnout (as is the case in almost all special elections). And I'm not sure that either poll will fully capture the impact of Scozzafava's endorsement of Owens -- most of PPP's interviews were conducted before the endorsement took place (although they showed no real difference once they began informing voters of the endorsement), while Siena noted that she had dropped out, but not that she had endorsed her former rival. Plus, the polling was conducted over a holiday weekend.
The way I'm looking at this race is basically thusly. Take the polling average -- Hoffman +10.5 -- and give Owens a net of two points from the Scozzafava endorsement, which is probably not fully captured by the polling. That would make it Hoffman +8.5. But assume a very high standard error -- perhaps something like 10 points -- given the disagreement between the pollsters and the dynamic nature of the race.
Even with those assumptions, however, Owens would be about a 4:1 underdog. So I suppose I'm getting off the fence here and declaring Hoffman the favorite, although I wouldn't attach any precise probability estimate to it. On the basis of the polling evidence, indeed, 4:1 is fairly generous to Owens. But the polling evidence isn't everything, and on the basis of the fundamentals of the district (which is Republican but not so conservative) it is generous to Hoffman.
** I will say that the higher number of undecideds that Siena identifies -- 18 percent versus just 3 percent for PPP -- feels more "right" to me, since a large number of Scozzafava voters were quite literally released back into the electorate overnight.