Skip to main content
ABC News
Why Clinton Might Have A Tough Time Flipping The Sanders Holdouts

We’ll be reporting from Philadelphia all week and live-blogging each night. Check out all our dispatches from the Democratic convention here.

PHILADELPHIA — Hillary Clinton is coming into her convention with a real problem. Even before WikiLeaks released thousands of Democratic National Committee emails, including some that suggested officials were actively working against Bernie Sanders, Clinton had about a third of Sanders supporters left to try to win over. The emails have exacerbated tensions with Sanders loyalists. And here’s some more bad news for the Clinton campaign about those loyalists: New data and analysis shared with FiveThirtyEight from Catalist and SurveyMonkey shows that, before the 2016 primaries, Sanders’s supporters voted less frequently than other 2016 voters, and they were less reliably Democratic than Clinton supporters.

In other words, it’s not a matter of Clinton simply coaxing Sanders supporters back into the fold — many were never in the fold to begin with. That could increase the difficulty of the task facing Clinton.

Data drawn from 14 primaries1 this year suggests that Sanders’s voters were less likely to be habitual voters than Clinton’s, Ted Cruz’s and Donald Trump’s supporters. (How do we know this? Check out the footnotes for details.2) In the 2012 general election, for instance, 88 percent of Clinton’s 2016 primary supporters cast a ballot. Same with Trump supporters. And 87 percent of Cruz supporters voted in 2012. But just 79 percent of Sanders’s voters did.

Hillary Clinton 88% 89% 78% 79%
Bernie Sanders 79 84 67 70
Ted Cruz 87 88 77 77
Donald Trump 88 89 78 79
Bernie Sanders drew more nontraditional voters

Source: SurveyMonkey/Catalist

This difference in turnout isn’t just an artifact of the overwhelming support Sanders got from young voters. Even if we take out 2016 voters who were too young to vote in the 2012 election, Clinton voters were still 5 percentage points more likely to have voted in 2012 and 9 percentage points more likely to have voted in 2014 than Sanders voters. Moreover, a higher percentage of Sanders voters had to re-register (7 percent) than Clinton voters (4 percent) for the primary, indicating that Sanders may have been bringing in more people who had let their registration lapse.3 All of this means that Sanders brought irregular voters into the process, as well as young voters, to an extent the other candidates did not. Will these people still vote without Sanders on the ballot? It’s far from guaranteed. Clinton might need get-out-the-vote programs to attract this group that Trump won’t need in order to capture Cruz primary voters who will probably vote anyway.

Perhaps even more important for Clinton is to make sure that Sanders backers who do vote cast their ballot for her. I wouldn’t bet on this for some Sanders voters. Although 56 percent of Clinton backers can be described as voting consistently Democratic in general elections,4 according to Catalist models,5 only 41 percent of Sanders backers can be. Some of the remaining 59 percent of Sanders supporters usually vote Democratic (25 percent), but about a third (35 percent) were either consistent ticket splitters, couldn’t be modeled or had usually voted Republican.6 Not surprisingly, general election polling indicates that a third of Sanders’s primary voters are not going for Clinton when third-party candidates are included. That means Clinton cannot merely count on these voters coming home to their Democratic base — because they aren’t actually part of it. Clinton is going to have to work to bring these voters aboard. (She’ll presumably get some help in Philadelphia today when Sanders gives his prime-time speech.)

Clinton can still win the election even if she doesn’t win over all these Sanders backers. I previously estimated that Clinton would gain about 1 percentage point (or 1.6 million voters) in the polls if Sanders voters currently backing Libertarian Gary Johnson or Green Party candidate Jill Stein chose between Clinton and Trump. Still, 1 percentage point can be worth a lot if the election is as close as the FiveThirtyEight models currently have it.

FiveThirtyEight: These two Bernie Sanders supporters will not vote for Clinton


  1. Arkansas, Florida, Georgia, Louisiana, Michigan, North Carolina, New Hampshire, Ohio, Oklahoma, Rhode Island, South Carolina, Tennessee, Texas and Virginia. These states were chosen because the local secretary of state had updated the voter file data in time for analysis.

  2. SurveyMonkey and Catalist partnered to match 295,000 respondents from SurveyMonkey’s national tracking poll taken from January to May with Catalist’s individual voter files from the secretaries of state in 14 states. The respondents in the SurveyMonkey poll were anonymous, so each respondent was matched to a cluster of, on average, 63 people in the Catalist file, based on ZIP code, gender, age, race and partisanship. Further, through a modified version of multilevel regression and poststratification (MRP), we have a good idea of who these people have voted for. You can read more about MRP here and here.

  3. At least some of them needed to re-register because of an address change.

  4. On the Republican side, 61 percent of Trump backers have voted consistently Republican. One can argue that Trump didn’t bring new voters into the primary process any more than Cruz or Clinton did, despite repeated claims otherwise.

  5. Based on variables including party registration and demographic and geographic information.

  6. Compare that with 24 percent of Cruz voters in the primary who fit into these categories.

Harry Enten was a senior political writer and analyst for FiveThirtyEight.