While Hillary Clinton and Donald Trump maintain leads over their rivals in national polls, Sanders and Carson receive more support in the latest PackPoll of NCSU undergraduates. In our poll, Carson was preferred by 39% of Republicans, while US Senator Marco Rubio and Donald Trump placed second and third, respectively, at 19% and 12%. Sanders, meanwhile, was preferred by 73% of Democrats; just 20% said they would vote for Clinton.
These results are different from what the PackPoll found last Spring, when Carson and Rand Paul were both polling at 18% each. Clinton, meanwhile, had the most support at 40%. Overall support for any candidate was lower, though, in part because many students said they didn’t have a preference at that time. Students’ choices this Fall are also at odds with most recent polling of adults conducted in North Carolina by Public Policy Polling (PPP) and Elon University. Both of them reported Clinton holding a commanding lead over Sanders, and PPP said Trump was preferred over Carson.
Minds Made Up?
Even though the general elections are about a year away, and the first primaries and caucuses are months from now, polling about these races is in full swing. How accurate is it? Surprisingly, it is generally acknowledged that early polls say little about real outcomes of the elections. This is true for both predicting the winner of each Party’s nomination and the general election. An analysis from the Washington Post, those who led at this point in 2011, 2007 and 2003 were not the eventual winner except for when Obama was running largely unopposed for reelection. For this reason the Pack Poll decided to ask whether students had actually made up their mind about the candidates.
When asked if they had decided whom to vote for in the primaries, we find that, at most, only one in three students had actually made up their minds. While 33% of Democrats aid they knew who they would vote for, less than 20% of Republicans said the same thing. The smaller percentage of Republicans having decided, we believe, stems from the much larger field of candidates lacking an obvious so called, “front-runner.”
In election survey conducted in partnership with mainstream news organizations, it is a common practice to ask respondents to choose without analyzing the decisiveness of that preference. An example can be seen in a recent survey by Quinnipiac University. When not given the option to say that they were undecided, it appears that 100% of Republicans have already chosen someone to vote for, when that is clearly impossible.
The Impact of Decisiveness
The degree of students’ decisiveness affects how well we could report the candidates are doing. For Republicans who are still trying to decide or just leaning, for example, 11% pick Trump. He is supported by 21% of Republicans that have made up their minds, though, essentially doubling his support.
The opposite effect, though far less dramatic, can be seen in the case of Rubio. While 20% of those who haven’t made up their minds support him, that percentage declines to 16 among the most decisive. Support for Carson behaves more like Trump than Rubio. Carson receives 37% of Republicans’ support if they are admittedly indecisive, but 46% among those whose decisions are firm.
Of note, Jeb Bush, who has received mostly negative reviews of his debate performances lags behind these three candidates at just 9%, but only 1 person who says they are firm plans to vote for him. Bush’s support, unfortunately for him, rests almost entirely upon students who say they are open to voting for another candidate.
Among student Democrats in our Poll, Sanders was the overwhelming favorite regardless of how decisive they were, although Clinton shows room to catch up because her deficit is the smallest among those who have not made up their minds. Among those who are undecided, 31% said they could support Clinton while just 51% said the same about Sanders. Looking at it differently, though, Sanders lead appears safe because his support improves with reported decisiveness, all the way to 85% versus 13% for Clinton among those who say their mind is already made up.
One of our goals was to compare our results to news media polling that essentially forces respondents to pick a candidate. These polls convey an electorate that is all knowing and decisive, even though most voters are not that decisive at this point in time.
As we show, not only have very few students made up their minds, but also different candidates do better or worse when we include their support from students who admit to being open to voting for someone else. Carson, for example, is supported by 46% of Republicans who have up their minds, but only a fraction of student have made up their minds. Another way of calculating his “true” support is to divide the 38 decisive Carson supporters by the 449 total Republican respondents, or just 9%. Another alarming result is that even after admitting to not even leaning in favor of a particular candidate, students, when asked, will usually pick one when they aren’t encouraged to say they lack an opinion.
It is important, we believe, to recognize that early polling results can be very misleading if they are measured and reported uncritically. Worse, reporting these percentages might actually shape later polling, since uncertain voters have been found to rely on perceptions of electoral viability when choosing candidates in a primary because party identification is of little help in distinguishing one candidate within a party from the next one.
These are the results for this semester’s “Big Poll,” conducted bi-annualy since 2010 (November 5-10, 2015). The survey was administered over the internet to a random sample of 4,500 NCSU undergraduates, generating a 24% response rate for completed surveys (and 26% for partially completed). Sampling error is +/-2.93% for completed interviews and questions asked of the full sample; it is higher for sub-groups and questions asked of only portions of the full sample.
In addition to sampling error, other forms of error occur in surveys, such as confusion about question wording or the order of questions, but these are not precisely quantifiable. We do not apply post-stratification sample weights to adjust for possible demographic imbalances in our sample because our measured demographics closely approximate the known student parameters for age, year in school, and gender. Weighting the data would be unlikely to change the reported results by more than 1-2% percentage points for any question results.
Click here to see the full set of results: Toplines Fall 2015 Big Poll