The big news out of Iowa is that Ted Cruz has surged into first place in the GOP nominating contest in that state, displacing Donald Trump. But is it true? Within hours of that result being reported in the media, another poll came out suggesting that Trump had regained his Iowa lead. In the first poll sponsored by Monmouth University Cruz was up by 5 points (after earlier polls had shown him surging into a statistical tie). But in the latest poll, sponsored by CNN/ORC, Trump was back in first place with a commanding 13-point lead over Cruz.
There are a lot of myths about presidential campaign polling. However, at least two important observations might be made in this case. First, the difference in these polling results is probably far too large to be accounted for by the sheer size or composition of the sample. In both cases, pollsters surveyed Republican respondents who said they were “likely” to participate in the Iowa caucuses in 2016. Defining “likely” might seem pretty subjective, and even speculative, and to a certain extent it is. However, there are pretty well-established ground rules among pollsters for defining such voters. Mainly it comes down to whether the survey respondent participated in the Iowa caucuses four years ago and their voting pattern overall. Pollsters rely on self-attestation, of course, so it’s always possible that some voters are misrepresenting their past voting behavior. However, there is no reason to believe that the rate of misrepresentation would vary so sharply from poll to poll.
Second and possibly more revealing is the time frame of the two polls. The CNN/ORC poll was conducted from November 28 through December 6. The Monmouth poll was conducted from December 3 through December 6. Note that this second time frame overlaps with the first, but does not include several earlier days of polling. It is quite possible that the Monmouth poll captured a significant shift in the polling that occurred in the late period. Would it be enough to change a +13 Trump advantage into a +5 Cruz advantage? Probably not, but given the margin of error in these polls, the swing may be smaller than it appears. The important point is that polls contain trends that are captured within the context of the polling process itself. Ideally, one would want to break out that portion of the CNN/ORC survey that was conducted on December 3-6, to see if it suggests a distinct surge in the polling toward Cruz, and matches, at least, in part, the pro-Cruz tilt recorded in the second poll. If so, the difference in the two polls may not be as sharp as the final reported outcome seems to suggest.
It could be that the two pollsters also have distinct ways of sampling voters. For example, one might rely more heavily on landlines than cellphones, which might create a skew. The pollsters might sample in different Iowa countries, with different allegiances, which could create another skew. In addition, the wording of the surveys questions – for example, on basic voting preference — might be sufficiently different that someone who leans toward one candidate is reported as simply supporting that candidate in one poll, while in the second poll, they are considered “undecided.” One would have to drill down into the polling practices of each firm to unearth these differences.
There is some evidence of a basic skew. If you compare CNN/ORC polls with other polls in the past several months, both in Iowa and nationally, it appears that CNN/ORC tends to report slightly higher numbers for Trump. For example, the three latest national polls all have Trump leading the field. The polls were conducted over roughly the same periods. However, IBD/TIPP and Quinnipiac have him leading the second-place finisher by 10 and 12 points, respectively. The CNN/ORC poll has him leading by 20 points. That’s a substantial difference.
A word of caution, however. Polling at this stage of the race is highly volatile, and voter candidate allegiances are still relatively weak (though they appear to be much stronger for Trump). It could well be that developments during the December 3-6 period, either in Iowa or nationally, created a surge of support for Cruz. A spate of campaign ads might have appeared during those particular polling days, or voters were reacting to comments the candidates made about the San Bernardino attack or some other event during that period.
In all likelihood some combination of all of these factors has played a role in the extraordinary difference reported in the most recent polling. Overall, only two conclusions seem certain. First, there is a growing interest in Crux’s candidacy, especially among evangelical voters, but also among GOP voters generally. Second, no matter who or what comes his way, Trump always seems to rebound from apparent setbacks to retain his leading position.