Democratic Primary Gets Rare Glimpse at Sane Polling

If you’re a Democrat and like options for your nominee, you’ve been in luck this election. Over 25 candidates have decided to run.

But just because your candidate runs doesn’t mean they get to participate in the debates. Only 20 candidates made it to the first debate, and only around ten have made it in recent debates. That candidate number will only drop further for the November elections.

The main reason candidates miss the debates — and then choose to drop out — is because they aren’t polling well. The Democratic Party itself has been controlling this primary debate access using these polls. The rationale for using polls for debate access is that candidates who poll well will be stronger candidates for the party and more likely to win in the general election.

But are those polling numbers meaningful when there are so many candidates?

With traditional polls, respondents are asked to choose only one candidate-our current plurality voting method. This choose-one voting method is notorious for vote splitting. That is, when candidates who have overlapping ideas run, their votes can divide between their shared supporters. When we see the person in first getting under 30%, this is typically why. This division can present itself in sometimes unpredictable ways that can even change the order on the leaderboard.

So what’s to be done about this vote splitting? The first suggestion is often to further eliminate candidates. But when there’s so much vote splitting, it can be difficult to tell who makes sense to eliminate, particularly before candidates are able to effectively get their message out. That means we risk eliminating strong candidates, and we miss hearing their ideas.

Vote splitting isn’t the only issue either. Respondents to these choose-one polls may not pick a candidate even if they like that candidate best. If they feel their favorite candidate has little chance of winning, there’s a fear of wasting one’s intended vote. This abandonment of preferred candidates can lead to a downward cycle of continuing neglect. The candidate gets marginalized from low polls and those low polls in turn cause further low polling from this wasted vote fear. These candidates seldom have a chance and miss out on opportunities to gather new support.

It’s clear the current way we measure candidate support through this choose-one approach is flawed. Even worse, that flawed approach is used for a high-stakes decision — who gets to debate — which often determines who stays in the race (and who is ultimately elected).

So what do we do?

Fortunately, there’s another voting method that’s especially easy called approval voting. Approval voting lets voters choose all the candidates they approve of, most votes wins.

This approval voting approach addresses vote splitting, and it always lets you vote your honest favorite regardless of that candidate’s perceived viability. Vote splitting is addressed by letting you support all the candidates you find similarly appealing. And approval voting lets you support your long-shot favorite because you can always hedge your bets with one of the frontrunners at the same time. You keep your say in the election’s outcome while also showing your favorite candidate that you like their ideas.

Interestingly, The Economist recently ran a poll with YouGov that approximated this approval voting method. The pollsters asked respondents all the candidates they were considering voting for. The question looked like this:

“Which candidate or candidates are you considering voting for in the Democratic Presidential primary or caucus in your state in 2020? (Select all that apply)”

As expected, this substantially addressed the vote splitting. The leader even changed entirely. Other candidates shifted up and down the leaderboard, and other candidates saw their support spike.

When poll respondents could choose multiple candidates, Biden went from a small lead to a slight trail (a virtual tie) with Warren and added more separation from Sanders. To be clear, there was a switch on the leaderboard with this sample. We know this because the same respondents were asked under both a choose-one approach and under the option where they could choose all the candidates they wanted.

One interesting candidate here was Andrew Yang. He both benefited and suffered in this approval-voting-like approach. Yang’s big benefit was that he jumped from 3% to 12% with Castro and Klobuchar keeping his company. The downside for Yang was that his competition also got a boost — and their boost was larger. Booker and O’Rourke, for instance, shot past him into the 20’s under this alternative approach.

Harris and Buttigieg had a small separation from the end of the back under the choose-one method. But under an approval-like approach, they created further separation from the rear and pushed near 30%, just a 10% hop away from Sanders.

Then there were the candidates who saw very little benefit, even under this fairer handed approach. Gabbard, Bennet, Blasio, Williamson, Steyer, and the rest of the bunch found themselves stuck with very little support.

This view under an approval-like approach is clearly a different picture than what we’ve seen. Given the severe limitations of our current choose-one approach, this approval-like approach is plainly a more accurate reflection of candidate support.

There is a caveat, however, with this poll. This poll mimics the characteristics of approval voting and makes a good proxy measure, but it’s not quite the same. That’s because the polling question prompt doesn’t explicitly ask respondents to answer under the condition that their selections for multiple candidates would determine the nomination outcome. Rather, the question is asking them all the candidates they’re considering voting for under the present choose-one voting method.

That implication can result in a couple issues. One is, under actual approval voting, they may vote for more or fewer candidates. For an example of voting for fewer, consider a situation where Sanders is even less competitive from third place. Someone who doesn’t like Sanders and is considering mainly Biden and Warren might decide to pick only one of the two candidates under approval voting. On the other hand, if Sanders were more competitive, the voter may be more inclined to pick both Biden and Warren to hedge against Sanders winning.

There are other cases where actual approval voting might result in more candidate votes than through this approval-like poll. For example, candidates may not have been considered due to fear of a wasted vote. Yang (and others in a similar spot) may not have been considered at all by some voters due to fear of throwing their vote away in a choose-one system. So technically, by a respondent explicitly imagining our current choose-one method, they didn’t consider voting for someone like Yang. But under an explicit approval voting prompt, they may have supported a candidate like Yang.

What can we say about this approval-like poll?

This poll is a nice look into the window of approval voting. It’s a glimpse of the kind of information we could be gathering. At The Center for Election Science, we did this kind of information gathering in 2016 and explicitly compared different voting methods. As you’d expect, it too offered a drastically different picture. We’ll do this again in 2020 where we can directly test approval voting alongside other voting methods.

We fail our ourselves and our elections when we continue to look at candidate support through the distorted choose-one voting method. We need to do better. Approval voting does that job, and it does it in a way that’s easy to understand. The stakes are too high to continue with the status quo.

Originally published at https://www.electionscience.org on October 8, 2019.