Articles and Analysis


NJ-Gov: 42.0% to 42.0%. Really.

Topics: Chris Christie , Incumbent Rule , IVR Polls , Jon Corzine , New Jersey 2009

With all but perhaps or or two final polls logged, our trend estimate in New Jersey, as of this writing, stands at a 42.0% to 42.0% deadlock between Corzine and Christie, with Chris Daggett falling to 10.1%. That amazingly close result will likely change if we add another poll or two tomorrow, but a shift of a half point or so in either direction will have little meaning. The polling on this race is as close as it every gets, and as our standard trend line (below) shows, has been for the last few weeks.

The bottom line is that our final estimate is too close, in and of itself, to forecast winner. As I noted last Friday, our final estimates for the 2008 election included four states with final Obama-McCain margins that rounded to a percentage point or less. The nominal leader won in two of these states (North Carolina and North Dakota) but lost in two others (Missouri and Indiana).

But wait. Does our standard trend line ignore a last minute trend to Christie, analogous to the presumed movement to Hillary Clinton in the New Hampshire primary? If you use our chart's Smoothing Tool to change to the "more sensitive" setting (as illustrated below), you will see a hint of a trend toward Christie. The Republican challenger's support ticks up slightly (to 42.5% as of this writing), while Corzine's line moves slightly (to 41.4%).   

Intriguing as it seems, given the mix of different methodologies and field periods on our chart, we cannot be sure that the twitch in the more sensitive line represents a real change and not just random noise. The slight move to Christie is because three of the five surveys released today show nominal movement to Christie, while only one shows a nominal shift to Corzine and one shows no change in the margin. I'm not certain of the odds calculation on that outcome, but the probability that it occurred by chance alone is far more than the usual 5% we usually require to say it is statistically significant (using different calculations, Monmouth University pollster Patrick Murray reaches a similar conclusion).

Even if real, the 1.1 point margin favoring Christie on the more sensitive trend estimate is still not large enough to characterize the race as "leaning" Christie's way. So, for better or worse, if you are looking for a purely objective, empirical "call" of the New Jersey race, our trend estimates are not much help. The finals snapshot is just too close.

If, on the other hand, you are interested in some purely subjective speculation -- and that's what all of the various predictions of the outcome amount to -- read on.

Let's start with a review of the biggest New Jersey polling puzzle, the consistent difference in results between automated and live interviewer surveys, and then consider what it may imply about two things that polls are least able to measure consistently and accurately: Who will turn out to vote and what voters really mean when they say they are undecided.

The automated-vs-live-interviewer puzzle. As reviewed here on Friday, the three pollsters that use an automated, recorded voice methodology -- SurveyUSA, Rasmussen Reports and PPP -- have produced results consistently better for Christie and worse for Corzine than the other live-interviewer telephone surveys.


As the chart above shows, that difference persists through the final round of surveys released today and late last week. On their final polls PPP, SurveyUSA and Rasmussen have Christie leading by an average of 4 points (46% go 42%) while the three live-interviewer surveys released earlier today by Quinnipiac University, Monmouth University and Democracy Corps had Corzine leading by an average of one point (41% to 40%).

Modeling turnout: Probably the most important result in today's data comes from a single table buried in the cross-tabulations of the poll released by the Democratic-affilliated Democracy Corps. Their poll puts Corzine ahead by the largest margin 41% to 36%. The cross-tab shows that virtually all of Corzine's lead on the Democracy Corps poll comes from voters who did not cast a ballot in the 2005 election (and since Democracy Corps samples from a voter list, this classification is based on actual vote history, not a self-report):

2009-11-02_DemocracyCorpsVote History.png

Among those who voted in 2005 (84% of the Democracy Corps sample), Corzine leads by only a single percentage point (39% to 38%) in the Democracy Corps poll. Among those who have voted in other elections but not 2005 (and every respondent to this poll self-reported having voted in 2008), Corzine leads by more than two-to-one (54% to 26%). Thus the much discussed "all out push" by the Corzine campaign to win over Obama backers.

Corzine backers were cheered today by tabulations on two of the new polls showing the Governor with a significant lead among early voters. SurveyUSA's new poll found 14% had already voted and Corzine led among these voters by 12 points (50% to 38%). Monmouth found half as many early voters (6%) but an even wider Corzine lead among them (51% to 31%). Tantalizing as it is, we will not know until this time tomorrow whether the Corzine campaign is truly mobilizing the new Obama voters from 2008 or whether they are simply getting a lot of hard core Democrats that would have voted anyway to cast their ballots early. If Corzine wins, it will surely be because of this organizations advantage.

Last week, Nate Silver speculated better Corzine performance on automated polls might be due to an effectively "tighter screen" on those surveys:

An automated poll tends to be associated with lower response rates, since an automated script can't do as much a human to coax someone into an interview, and therefore sometimes tends to reach a more enthusiastic set of respondents (in effect, it may serve some of the same functions as a very tight likely voter screen).

Since Republicans tend to be more enthusiastic right now, that may be what's causing the automated polls to be more favorable to them. But since none of us yet know how the enthusiasm gap is going to play out in practice, it would be premature to come to any conclusion about whether the voter universe that Rasmussen and PPP are coming up with is "too tight" or "just right.

Perhaps, although as I noted on Friday, the automated-vs-live-inteviewer gap is significantly smaller in Virginia. On the final round of polls, the lead by Republican Bob McDonnell is only two points greater on the automated polls (56% to 41%) than the last five live interviewer polls.

Either way, my sense is that overall turnout will be driven less by the respective campaign field organizations than by the underlying enthusiasm gap driving voter decisions. It is one thing to help an already enthusiastic voter cast an early ballot. It is something else to convince a complacent voter to get excited about a candidate for whom they have mixed feelings.

Measuring those who are undecided. If you spend time with horse race polling numbers, it doesn't take long to discover that some of the biggest differences among pollsters involve the undecided percentage. One reason is that voters do not fall neatly into "decided" and "undecided" categories. Vote decisions fall along a continuum from completely committed to totally undecided, with most voters falling somewhere in between. The size of the undecided category on a poll may depend on the wording or structure of the question or how hard interviewers pushes for a decision. Automated surveys frequently obtain a smaller undecided percentage, and one reason may be that voters feel less comfortable revealing their "secret ballot" choice with a live interviewer.

Complicating this issue further is that saying "I'm undecided" on a survey may imply something other than total indecision. In a three-way race, it may mean that the voter has decided against voting for one candidate, but has not settled on which of the two alternatives deserves their choice (see some evidence of this sort of uncertainty in the focus group conducted by Monmouth University pollster Patrick Murray).

The theory behind the incumbent rule (that I spent a lot of time speculating about in 2004) is that the undecided category swells with voters that have decided against supporting the incumbent but are not yet ready to embrace the challenger.

Over the last eight to ten years, it has been hard to find much evidence of an automatic "break" of undecided voters toward challengers especially in highly competitive races, probably because incumbents have become so much more adept at turning the tables and "going negative" on their opponents.

However, if we take a closer look at the automated-vs-live-interviewer puzzle in New Jersey, we see a pattern that a few years ago I would have treated as clear evidence of the incumbent rule in action. All of the surveys, regardless of their methods, are yielding consistent results for Corzine -- most have him within a point of 42%. But the final automated surveys show a much smaller undecided vote and a consistently higher percentage for Christie (45-47%) than the live interviewer surveys (36-42%). (And NRO's Jim Gerhaghty notices that this pattern extends back to far more surveys than those in the table below).


So what does all this tell us about the too-close-to-call final estimate we are showing for New Jersey? This now nearly three-year-old comment from Republican pollster Neil Newhouse sums up my feeling:

[N]ewhouse noted the example of his client, incumbent Republican Jim Gerlach (Pennsylvania-6), who was in a 44% to 44% tie on their final internal poll conducted a week before the [2006] election. In the "old days," Newhouse said, we would have assumed an easy Murphy victory. However, Gerlach ultimately prevailed (51% to 49%) after a closing with a final television ad featuring a personal appeal by Gerlach that Newhouse credited for the victory. As for the incumbent rule, Newhouse said, "we are seeing a bit of a change, but not much consistency." While he still tends to give challengers the "benefit of the doubt" when incumbents are under 50%, Newhouse believes it is no longer "carte blanche automatic" that the undecided vote on the final poll will all go to the challenger.

All other things being equal -- and 42.0% to 42.0% is about as equal as they get -- I still tend to give a challenger like Christie the "benefit of the doubt" when up against an incumbent like Corzine even though recent examples of the "incumbent rule" are few and far between. That instinct is reinforced by the large number of voters that are either undecided or still leaning to independent Daggett (with Daggett's support falling) and the fact that no matter how hard pollsters appear to push, Corzine does not seem to rise beyond 42%.

So while the empirical evidence says this race is still too close to call, my hunch is that Christie will emerge the narrow victor.



While I wouldn't criticize a "hunch," I would also point out that one factor that you never mentioned in this commentary is the "New Jersey effect." It's pretty well known historically that in NJ Democratic candidates on a statewide basis seem to always do significantly better than the pollsters give them credit for. That was true for Lautenberg, it was true for Corzine in 2005, it was true for both Gore & Kerry and it was certainly true for Obama. And going back awhile, I recall that it was even true for Toricelli.

GOP candidates always get their hopes up in NJ in the days leading up to the election and then get wiped out on election day. I don't know if it's because NJ voters are simply embarrassed to tell pollsters they like Democrats, or that the Northern NJ and Philly suburb Democratic machine is very effective in getting out those votes.

If it's the latter, we've already seen it in the double-digit lead for Corzine in the early voting reported by at least 2, and maybe, 3, pollsters. You mention that but don't seem to see it being significant. As a candidate, if I've "banked" a 12-point lead -- statewide -- on the first 10-12% of the vote counted, I'd be hard pressed to see how I'd lose the election. Remember, this banked vote doesn't simply reflect the early returns from an expectedly strong region, these would reflect a very widespread universe of voters across the whole state.



Should be a fun one to watch :)

Wish I had some insight to offer but I'm too busy getting the popcorn going.



nelcon, part of what you describe may be what some people call the "what you see is what you get" rule. In a strong Republican or strong Democratic state, undecided voters tend to break overwhelmingly in favor of the majority party at the last minute. So when a poll shows a 42-42% split, the minority party candidate will get 42% and the majority party candidate will get undecideds. The idea is that a minority party that has not convinced a majority of voters to support her by the day of the election then she is not going to convince them, and they simply default to the other candidate.

That happened last year in Wyoming where polls showed Lummis and Trauner in a 43-42 dead heat the week before the election, yet Lummis wound up winning the election 53-43 because the 10% undecided voters all broke for her.


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.