Articles and Analysis


NY-23 Watch - Monday Morning

Topics: Bill Owens , Dede Scozzafava , Doug Hoffman , NY-23 , PPP , Siena Institute

We got two new pieces of polling news last night from New York's 23rd District. The first is the one and only survey from Public Policy Polling (PPP), a firm that does automated surveys for Democratic candidates but also conducts and releases surveys in high profile races like NY-23 as a marketing tool. PPP's poll, also the first conducted since Republican nominee Dede Scozzafava withdrew on Saturday morning, forecasts a different outcome than previous surveys, including the Siena Research Institute poll conducted last week: They show Conservative candidate Doug Hoffman with a huge lead over Democrat Bill Owens -- 17 points (51% to 34%) on a three-way matchup that includes Scozzafava, whose name will remain on the ballot), 16 points on a question that only asked about Hoffman and Owens (54% to 38%).

The second piece of news was a release from the Siena Research Institute announcing that they will release another new poll this morning a little after 9:30 a.m. Eastern time.

The new PPP survey suggests a significant acceleration of the trend evident in other polling including Siena's -- a collapse of Scozzafava's support while Hoffman's vote soars and Owens creeps up slowly. Strictly speaking, Scozzavfava's withdrawal and subsequent endorsement of Owens over the weekend render previous horse race results virtually useless as predictors of the outcome. We are probably best advised to throw out the previous polls (and the trend estimate based on them in the chart above) and simply examine the two post-withdrawal poll we'll have available later this morning.

If the Siena results are consistent with PPP, this discussion will be mostly academic. The Siena and PPP surveys use very methodologies, and given Charles Franklin's analysis of the last Siena survey on Saturday, I would not be surprised to see their update produce a closer result than PPP. I will update this post accordingly when we have those results.

UPDATE: Surprise, surprise, the Siena results show Hoffman narrowly ahead of Owens, but by a smaller, five-point margin (41% to 36%) than PPP, with Scozzafava getting only 6% of the vote and 18% undecided. With 606 interviews, that margin is not quite statistically significant given the usual 95% confidence interval.

Given that Siena uses a classic random digit dial (RDD) sample and live interviewers, while PPP uses a voter list sample and an automated recorded-voice method, some are going to want to ignore the PPP results and focus on the large number of undecided voters (18%) in the Siena survey. Even Siena pollster Steven Greenberg is arguing that Hoffman might not have the advantage going into tomorrow's election since "most voters are not political junkies like I am and didn't know" as of yesterday, that Scozzafava had endorsed Owens.

Apologies to my Democratic friends for the pessimism, but I don't see it. First of all, even if we focus only on the Siena survey, the crosstabs offer little hope of a decisive rebound among those undecided as of last night. Scozzafava's rating among the undecided is 28% favorable, 22% unfavorable, while half (50%) have no opinion. In case it's not obvious: If a voter doesn't like Scozzafava by now, there's not much chance her endorsement of Owens will mean much to them.

Keep in mind that before asking who they would support, the Siena question informed every voter that while Scozzafava's name would remain on the ballot she has "suspended her campaign" and "released those individuals supporting her campaign to transfer their support as they see fit." As such, nearly two-thirds (65%) of those still planning to vote for her say they are "absolutely certain" about their choice with "no chance I will change my mind."

Among the undecided, the ratings of Owens and Hoffman are similar, although Hoffman's negatives are slightly higher (by a margin that is far from statistically significant):

  • Owens - 26% favorable, 20% unfavorable, 54% don't know
  • Hoffman - 24% favorable, 25 unfavorable, 51% don't know

The overriding message from the big "don't know" numbers among the undecided is that most are not likely to vote. None of these numbers suggests a late, decisive break toward Owens.

There's also the matter of the trend. Scozzafava's withdrawal accelerated the trend to Hoffman's that was already quite strong. On the Siena surveys, for example, Hoffman's vote has increased from 16% to 23% to 35% to 41% on four surveys conducted since October 1. Count me as skeptical that the six-point bump in Hoffman's support seen in last night's poll fully captured the benefit to Hoffman from Scozzafava's departure.

And then, finally, we come back to the PPP poll and its bigger Hoffman margin. Rather than go on at length (again) about the differences between random digit dial (RDD) samples and voter lists and about the trade-offs between live interviewers and an automated method, let's consider it this way: Both are blunt instruments for sampling adults, selecting "likely voters" and measuring their preferences. Neither can be considered a gold-standard, a true random sample that perfectly covers, models or represents those who will vote tomorrow.

However, my experience conducting surveys for political campaigns, especially in Congressional districts in non-presidential year races, taught me the value of the vote history available on registered voter lists. More often than not, surveys I helped conduct based on such lists came closer representing the true likely electorate than media RDD samples which, like the Siena survey, disclose little to nothing about their likely voter screen or demographic composition.

Add to that the potential advantages of a self-administered automated survey in getting voters to provide more honest answers about whether they plan to vote and who they plan to vote for, and I find it difficult to ignore the PPP results. Hoffman looks like he's headed to a comfortable victory.




Good post. I don't disagree...this "feels" like it's going to Hoffmann.

One aspect of the cross tabs of both PPP and Siena that has not received much attention is the breakdown by gender...strange given that so many candidates in so many states have won recently with the (disproportionate) votes of women.

If I read the crosstabs correctly, if Owens is to have a chance, it would come through greater turnout of women.

In the latest Sienna (3 way) poll, Hoffmann leads Owen +11 among men but it is a TIE among women. Remaining Dede voters = 8% women and 4% male. 20% of women undecided versus 14% male.

So there is room for Owens to improve based on the votes of women.

The picture in PPP (two way) is less auspicious for Owens, but the pattern is the same: +27 for Hoffmann among men, but only +6 among women. Slightly more women undecided (10 versus 7 male undecided).

Bottom line: if it's close, it will be on the basis of the female vote...

I don't know if there is a source that shows the '08 or '06 turnout by gender, but it would be interesting to know.



"Strictly speaking, Scozzavfava's withdrawal and subsequent endorsement of Owens over the weekend render previous horse race results virtually useless as predictors of the outcome. We are probably best advised to throw out the previous polls (and the trend estimate based on them in the chart above) and simply examine the two post-withdrawal poll we'll have available later this morning."

I can't agree with that. In the first place we should take into account any mail in voting (absentee/military/early as appropriate for the area) which will have occurred before Dede's last minute drop out. Additionally we have to consider all those who are not even aware that Dede has dropped out. Striking as it is to us political junkies the majority just don't pay attention to politics. Since Dede will still be on the ballot we should assume some amount of her supporters will vote for her accidentally. Whether the post drop out polls capture this will depend in part on the wording (do they just give the three options or do they give two, do they mention the drop out).

For those reasons i do not think it is correct to simply ignore the previous data. Here's what I do recommend:

If we look at the pollster chart I think the best analysis to date is to use the whole data set minus the two tiny partisan surveys from Neighborhood and Basswood. Why those two? They used tiny samples (half or less of most of the surveys in the set), they came from partisan sources, and they hugely underestimated Owens support which plays a big role in the apparent gap between hoffman and owen if we use all data sets.

I didn't throw out the Kos data set because even though it is partisan it uses a decent sample size (suggesting they may not have done multiple polls and released the one they liked) and it also matches up well with the Sienna poll from close to the same time period.

If we do that we find Hoffman and Owens are precisely tied (35.6 to 35.8). That leaves the race a complete toss up.

Now is this wishful thinking? Maybe but I think its the best data set we have to date. It includes all the data sets that seem reasonable, does not ignore the historic data which informs us about early voting trends, and includes the reliable data from the post-Dede period (small though that window is). My prediction- Hoffman wins narrowly (by 3% or less).

That said I wouldn't bet on it given how chaotic special elections tend to be. :)


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.