Polling in presidential primary campaigns is notoriously fluid. In an effort to gauge movement among voters as they might change their mind and as candidates pop in and out of the race, pollsters will often ask voters about both their first-choice candidate and their backup. This question has been particularly popular this cycle, given how crowded the field is, and can be useful for understanding the overall level of support a candidate has, including who benefits if one candidate loses support or whether voters are grouping candidates into “lanes.”
Take, for instance, Sen. Kamala Harris’s exit from the race last Tuesday. Attention soon turned to who her supporters might back: One Morning Consult poll found that 22 percent of Harris’s supporters listed former Vice President Joe Biden as their second choice, while 21 percent put Sen. Elizabeth Warren and another 14 percent put Sen. Bernie Sanders.
FiveThirtyEight Politics Podcast: Harris Drops Out
But just how accurate are voters’ second-choice picks? Do they really tell us who voters would support if their first-choice candidate dropped out? This is a hard question to answer, not only because voters’ first-choice picks aren’t set, but also because any test for accuracy would require re-interviewing respondents and comparing their second-choice answers over time.
For the last three months, I’ve worked with Iowa State University and Civiqs to try and answer these questions by interviewing the same group of Iowans to see how voters’ minds are changing in the lead-up to the caucuses. Iowans’ second-choice candidates could be particularly important because candidates who receive support from less than 15 percent of the caucus attendees are not considered a viable candidate.1 This means a candidate’s backers must either shift their support to a candidate who cleared the 15 percent threshold or go home — so someone’s second choice might be especially revealing as to who the voter might end up supporting.
But before we dive into what voters’ second-choice preferences tell us, it is worth noting just how many Iowa Democrats are still changing their minds. Of the 306 likely caucusgoers who supported a candidate in September and responded again in October, nearly a quarter (23 percent) changed who they were supporting. Slightly more respondents were consistent from October to November — 20 percent of 226 respondents supported a different candidate. This is hardly evidence that most voters have made up their minds. In total, 30 percent of the 210 likely caucusgoers who responded in all three waves of our surveys named the same candidate as their first choice all three times.
So perhaps unsurprisingly, we found even less consistency in second-choice picks. Slightly less than half (44 percent) of likely caucusgoers who listed both first and second choices in September and October selected the same second-choice candidate, although over half (52 percent) of people who responded to the October and November waves selected the same second-choice candidate both months. Similar to the frequency with which respondents stuck to their first choice pick, almost a third (30 percent) of Iowans who responded to each wave also named the same second-choice candidate in all three surveys.
Overall, Mayor Pete Buttigieg of South Bend, Indiana, surged from fourth to first and Biden fell from second to fourth over the last three months, according to our survey. But interestingly, these changes weren’t primarily driven by voters throwing their support behind the candidate who had been their second choice the previous month. In fact, we found that less than half (45 percent) of the respondents who changed their minds between September and October switched to the candidate they initially listed as their second choice.
It is possible, though, that voters’ second-choice picks are becoming better predictors of how they might switch their allegiances. We found that, of the voters who changed their minds between October and November, 55 percent listed their October second-choice pick as their top candidate in November. Keep in mind though that since about 20 percent of respondents shifted candidate preferences between survey waves, and only half of those respondents landed on their second-choice candidate, there is still about 10 percent of the electorate that changes its minds in ways not predicted by their first and second choices.
But with this many candidates in the race, the fact that voters’ second-choice preferences can account for about half of the movement means we see that these responses are at least somewhat useful. If people changed their minds by randomly selecting a candidate from the field, they would do much worse than a coin flip. That said, treating these answers as if they accurately capture who voters might support if their candidate drops out provides a false sense of certainty about how the race is going to evolve, especially given the closeness of the polls in Iowa. At this point, there’s still just a lot of uncertainty about who Iowans will end up supporting in February.