Discover more from Elliott’s notebook
What we know about electability 📊 March 1, 2020
Most data back up our prior that moderate candidates have advantages over ideological ones
Welcome! I’m G. Elliott Morris, a data journalist at The Economist and blogger of polls, elections, and political science. Happy Sunday! This is my weekly email where I write about politics using data and share links to what I’ve been reading and writing.
Thoughts? Drop me a line (or just respond to this email). Like what you’re reading? Tap the ❤️ below the title and share with your friends! If you want more content, I publish subscriber-only posts 1-3x a week.
What we know about electability
Most data back up our prior that moderate candidates have advantages over extreme ones
About a month ago, I wrote for The Economist that Joe Biden looks better poised to beat Donald Trump than the other Democratic presidential candidates. (We recently turned this into a video for those of you who don’t want to read words.) I come to this conclusion by digesting a bunch of political science evidence that details how voters punish ideologically extreme candidates. They have a documented history of turning off swing voters and might juice turnout for the opposition party.
This conclusion is also substantiated by general election polling that has shown Joe Biden beating Donald Trump by margins slightly larger than those Bernie Sanders enjoys. This has been true nationally and in Midwestern battleground states.
In the weeks since I penned that take, political scientists from the University of California, Los Angeles and the Democracy Fund Voter Study Group have released a huge survey with which we can add more nuance to the debate over so-called “electability.” The researchers have been interviewing thousands of adults each week since July 2019. Last week, they released data from nearly 160,00 of these interviews spanning the latter half of last year.
We can use these surveys—which the authors are calling “Nationscape”—to break down attitudes about the Democratic primary and general election by different demographic, political and geographic groups. Here’s an outline of some findings so far:
Joe Biden has a larger margin versus Donald Trump Bernie Sanders does among all voters nationwide
Biden’s margin is also larger than Sanders’s in crucial battleground states, such as Michigan, Pennsylvania and Wisconsin
Biden does better than Sanders versus Trump among consistent Democrats, consistent Republicans, swing voters, and irregular voters (those that vote in either 2012 or 2016 but not both), but Sanders enjoys a slight advantage over Biden with non-voters
BUT: Sanders still polls well enough versus Trump to be competitive. In fact, if the Nationscape data are correct, the Vermont senator would very likely win the Electoral College if the election were held today
These data fit with our a priori beliefs about which types of candidates are most competitive (moderates), but also about Donald Trump’s popularity (it’s low) and general chances of re-election (middling at best). His net approval rating has hovered around -9 for much of the past year, which my modeling suggests is associated with a higher chance of losing the electoral college than winning it.
But the Nationscape data are not all the information we have. Notably, since I wrote my big review of what political silence says about electability, and since the UCLA/Democracy Fund conducted the Nationscape interviews (the data they’ve released so far are from July-December last year), Bernie Sanders has apparently gained ground against Donald Trump in general election matchups. The average of polls from RealClearPolitics (which is noisy and probably less reliable that a more sophisticated model, but no alternative exists... yet...) registers a 2 percentage point bounce for Sanders nationally over the past month. His prospects have improved in swing states as well.
This might lead us to believe that the Nationscape data are outdated, suggesting we should ignore them. I would disagree for three reasons:
The Nationscape surveys have much larger sample sizes
The deltas in candidate support as measured in one reliable poll avoids confounders from some weird mode, population and house effects introduced by averaging polls together.
I have re-weighted the Nationscape data to be representative of the 2016 election at the state-level, which should produce more accurate estimates of state-level attitudes than many of the polls being released by other firms
Polling and modeling issues might also make Bernie’s performance in head-to-head polls appear better than it actually is. In a study of 40,0000 Americans published last week, political scientists David Broockman and Joshua Kalla find that pollsters could be unwittingly inflating Sanders’s margin by not taking turnout into account and because young people are more likely to express preferences in a Sanders vs Trump matchup than other matchups. Essentially, they argue that Sanders is strongest with younger and disengaged voters who aren’t likely to actually turn out to vote. But many pollsters aren’t adjusting for turnout right now. Instead, most have been releasing estimates of Bernie and Biden’s strength amongej register voters instead of ones that are likely to turn out in November.
Andrew German, a political scientist and statistician, relates this to the conversation about electability by using a framework of Bayesian updating. He writes that the flaws in head-to-head polls ought to decrease their relative weight versus our priors based on the political science literature. I think this is a helpful way to think about the value of the data we’re basing our conclusions on, and I propose that we add to the equation the conclusions driven by the Nationscape polling (which are of high quality, given the above reasons).
Adding all this up, I don’t see a reason to change my original conclusions about electability. Our priors about the success of moderates versus ideologues are very strong strong, and the Nationscape data are robust and generally agree. In contrast, the contradictory evidence is full of holes.
Still, it would be wrong to end on any other note besides saying that polls indicate that Donald Trump has a slightly uphill battle for re-election regardless of who the Democrats nominate for president. The identity of the nominee is relatively unimportant compared to the impacts that the conomy and the president’s unpopularity wield over vote choice. 90-95% of partisans will vote for their candidate no matter what.
But as I’ve been telling people, it’s that 5-10% of swing voters that decide elections—especially close ones.
Posts for subscribers
Links and Other Stuff
What a whole bunch of Americans believe about a whole bunch of policies
USA Today has partnered with UCLA and the Democracy Fund Voter Study Group to highlight findings from the Nationscape polling. I recommend digging through their charts.
What I'm Reading and Working On
It’s Super Tuesday in less than 48 hours. Keep your eyes out for several pieces on the subject from me this week.
Thanks for reading!
If you want more content, I publish subscribers-only posts on Substack 1-3 times each week. Sign up today for $5/month (or $50/year) by clicking on the following button. Even if you don't want the extra posts, the funds go toward supporting the time spent writing this free, weekly letter. Your support makes this all possible!