Articles and Analysis


Exploring the Internals of the CNN/WMUR/UNH Poll

If polls are political crack, than I've been happily snorting the latest from the CNN/WMUR New Hampshire primary polls conducted the University of New Hampshire (also known as the "Granite Poll") for the last 24 hours or so. I've become a fan of this survey because of the "internals" included in their questionnaire and the very helpful set of crosstabs (for the Democratic and Republican samples) released as always by University of New Hampshire Survey Center.

Our original posts on the two surveys highlighted the same trial heat numbers that virtually everyone else is focusing on: Clinton's lead has increased since July, while support for Mitt Romney has fallen off so that he now runs neck-and-neck with Rudy Giuliani.

However, to elaborate on a point I tried to make a week ago, the trial heat numbers in such polls (yes, the ones we track here at Pollster) may be the least useful at this stage, particularly if you interpret them as a prediction of what voters will do several months from now. Surveys are good at telling us what voters think and who they currently prefer, but those preferences are always subject to change. The latest survey of Democrats from UNH/CNN/WMUR helps put some of those preferences in perspective. Here a few highlights:

1) A preference is still not a final decision -- Just before asking the likely Democratic primary voter who they would vote for "if the election were held today," the UNH pollsters ask them how close they are to making a final decision.


So even though 91% can name a candidate they would support, only 17% are "definitely decided," 28% are just "leaning to someone" and more than half (55%) "still trying to decide." Of course, voters may ultimately decide to support their current preference, but by their own report most have not yet made that final choice.

I wrote the last two paragraphs before the Republican results had been released, but those numbers show even more evidence of uncertainty. Although 84% of Republicans express a candidate preference, 66% say they are "still trying to decide" while only 13% are definitely decided and 21% are leaning to a candidate.

2) A Big Shift on "Can Win"- One of the biggest and most noteworthy shifts on the Democratic poll involves the question about which candidate "has the best chance of beating the Republican nominee in the general election next November?" Although. Hillary Clinton's share of the vote has increased four points (from 39% to 43%) since June, we see a 17-point increase (from 37% to 54%) in assessments that she has best chance to win in November:


These results are consistent with similar findings from the most recent NBC/Wall Street Journal national survey of Democrats, which show the percentage choosing Clinton as the Democrat with the "best chance to defeat the Republican candidate" growing from 39% to 54% from April to September. Also, the Pew Research Center shows 53% of Democrats now naming Clinton as the candidate with the "the best chance of winning the general election against a Republican" (although they had not asked the question previously).

I noticed something interesting buried in the cross-tabs. Most of the shift on this measure in the UNH polls has occurred among the roughly 60% of likely Democratic primary voters that are less than "extremely interested" in the primary. That suggests a common pattern: Less attentive voters reacting mostly to the dominant campaign news story, "horserace" coverage relentlessly portraying Clinton as front runner consolidating her lead.

One thing to keep in mind though. Electability ratings are not always a great predictor of, well...of electability. The 2004 National Annenberg Election Study (NAES) tracked electability ratings of each of the Democratic candidates (how likely respondents considered each candidate to beat George Bush in November) among likely Democratic primary voters from October 2003 though February 2004. In a paper presented to the American Political Science Association annual meeting in September 2004), Annenberg's Kate Kenski reported:

Dean's electability and viability ratings were higher than those of Kerry until the Iowa caucuses. After the Iowa caucuses, Kerry's electability ratings surpassed those of Dean. Impressions of Dean's electability against Bush reached a high point of almost 53% around November 19. Two months later, on the evening of the Iowa caucuses, this perception had decreased eight points.

3) A McCain Resurgence? Support for John McCain among likely Republican primary voters, which had fallen sharply in New Hampshire as elsewhere, has increased to 17% on this survey from a low of 12% in July. McCain's support had hovered around 30% in UNH polls conducted during 2006 and early 2007.

An outlier? Perhaps, but dig deeper and the survey yields evidence of some underlying strengths for McCain in New Hampshire that may provide a foundation for a future resurgence there. His favorable rating now (63% favorable, 24% unfavorable) is slightly but not significantly better than it was in February (59% favorable, 27% unfavorable) when he had 28% of the vote and ran a point ahead of Giuliani. And McCain now leads the Republican field (with 32%) on the question of which candidate has the "right experience to be president."

It bears repeating: Eight years ago, John McCain defeated George Bush by a huge margin in New Hampshire (49% to 30%) among these same voters. Voters have a way of falling back to past preferences.

4) Favorable Ratings - Talk to campaign pollsters about the value of the trial heat results and most will tell you a similar story: Vote preference is usually the last thing to change. If you want to see evidence of the campaigning and paid advertising that candidates do, look to the movement in their favorable ratings. The table below shows the most vivid evidence of the real progress that the candidates are making in New Hampshire, starting with the Democrats:


And the Republicans:



Chris S.:

The wording on the electability question is kind of confusing:

"Which Democratic candidate has the best chance of beating the Republican nominee in the general election next November?"

Might some voters not be interpreting that question the way it was intended? Specifically, if a voter thinks that Clinton has a 90% chance of winning the nomination, then they might think "There's a greater probability that Clinton will beat the GOP nominee in the general election (than there is that Obama would do the same) because of the greater probability that Clinton will be *in* the general election."?

In other words, the probability that Clinton is elected to be the 44th president is the product of the probability that she wins the nomination times the probability that, if nominated, she wins the GE. In this poll question, the respondant is supposed to compare the probabilities of just the second piece of that equation (probability of winning the GE **if nominated**). But isn't the wording a little ambiguous on that point?



"If polls are political crack, than I've been happily snorting the latest"

Not to get too snarky Mr. Blumenthal, but clearly your experience with crack is lacking - it would be a disaster to snort, you would very much prefer to smoke it, I assure you.


Mark Lindeman:

Yeah, the image of Mark snorting crack just about did me in.

Chris, I think that's a good point. I doubt it matters much, but it would be interesting to have someone ask which candidate "would do best" in the general, or some other language that didn't lend itself to the ambiguity you describe.



"It bears repeating: Eight years ago, John McCain defeated George Bush by a huge margin in New Hampshire (49% to 30%) among these same voters. Voters have a way of falling back to past preferences."

The is one huge difference this time though that mitigates against a McCain resurgence - the undeclared voters split more towards the Democrats this year while they broke overwhelmingly towards McCain in 2000. McCain beat Bill Bradley far more than he beat Bush.

In this poll, the breakdown in registration is:

24% - Registered Dem
25% - Undeclared but voting Dem
19% - Undeclared but voting GOP
33% - Registered GOP

In the last 2000 UNH tracking poll it was:

26% - Registered Dem
13% - Undeclared but voting Dem
23% - Undeclared but voting GOP
38% - Registered GOP

This year, the undeclared voters are more likely to vote in the Dem primary by a 56%-44% margin while in 2000, the undeclared voted in the GOP primary by a 63%-37% margin and provided McCain with his decisive margin of victory. While he did go up in this poll, the overall composition of the electorate provides McCain with a much lower ceiling and it is difficult to see with the experience of the past 8 years and McCain's leading role in surge cheerleading, that he is going to gain the support of undeclared voters who are more interested in the Dem field this year.


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.