November 3, 2019 📊 Where are the swing voters?
Plus, who's the most electable Democrat? And (the pitfalls of) some early 2020 prediction models.
Welcome! I’m G. Elliott Morris, data journalist for 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? Share with your friends and tap the ❤️ below the title! It helps my newsletter rank higher in Substack’s curation.
Dear Reader,
This week’s main read: The key puzzle for 2020 will be whether the swing states have swung back toward Democrats since Trump’s surprise victories there in 2016. I check in with some survey data to suggest an answer to the question. There are more swing voters than you think.
Plus, I’ve got links to a study about the most “electable” 2020 Democratic candidate and a look at some early 2020 predictions based on electoral “fundamentals” (and a note on why they might not be reliable).
Thanks all for reading my weekly email. Please consider sharing online and/or forwarding to a friend. The more readers, the merrier! If you’re shy, the best way you can support my newsletter is to press the heart button below the title (this makes it rank higher in Substack’s curation). If you’d like to read more of my blogging I publish subscriber-only content 1-3x a week on this platform. Click the button below to learn more!
—Elliott
This Week's Main Read
Where are the swing voters?
Everywhere (except there’s not many in Alabama)
In the latest data from the Fox News Poll, Joe Biden leads President Donald Trump in a hypothetical 2020 general election by 12 points. This is roughly where he has been for the past 4 months. Also of note: Donald Trump is held to ~40% of the voting public in matchups against all major 2020 Democrats (though I don’t know exactly why they included Hillary Clinton in the poll), something I’ve pointed out in this newsletter before. I noted as much on Twitter:


If these numbers hold until next November, voters will deliver a remarkable defeat to the president. But that’s a BIG “if”. Polls at this point in the election cycle (precisely 365 days out today!) tend to have little predictive power, though they have gotten better in recent years.
But here’s what I really want to know: If voters prefer Biden to Trump by 10 percentage points more than they preferred Clinton to Trump in 2016 (she won by 2 points, he’s up by 12), where are all the swing votes? More importantly, are they in the states that Democrats need to win the electoral college in 2020?
I booted up fresh downloads of the Census Bureau’s 2017 American Community Survey (a survey of Americans’ demographics), the 2018 Cooperative Congressional Election Study (a massive 60,000-person survey of Americans) and the Census Bureau’s 2018 Current Population Survey (which asks more than 100,000 people whether they voted) to find an answer to these questions..
My analysis was relatively straightforward, but bear please forgive me as I delve into a basic summary of the technical details. You can read this document for a similar explainer if you want more hand-holding.
First, I used the ACS to estimate how many people of each demographic group (white, black, female, college-educated or not, etc, and all the interactions in between) lived in each state. Then I used the CPS to estimate how likely Americans in each of those demographic groups are to be registered to vote. I did so using a statistical technique called multi-level regression and post-stratification. (Again, you can read about that here.) Essentially, for each “type” of demographic—white, female, age 18-30, living in Iowa without a college degree, etc—we know how likely they are to vote.
Then, using the CCES, I counted any voter who (a) voted for Donald Trump and disapproves of him now and (b) any voter who DID NOT vote for Trump but approves of him now as swing voters. Anyone else who reported voted in 2016 but has consistent opinions was counted as a non-swing voter—in other words, it’s assumed they’ll vote for/against Trump if the approve/disapprove of him. This might not be an entirely correct assumption, but it’s as close as we’re going to get. I used MRP to know how many voters in each demographic group were likely to be these so-called “swing” voters.
Then, I just counted up (a) the number of voters and (b) the number of swing voters in each state. Dividing this the total number of projected voters, per the CPS, we get the percentage of all voters that are “swing” voters for every state.
That ends up looking like this:

But please note that there are confidence intervals around the estimates. They look like this:

I think this analysis is interesting, but honestly, the takeaways are kind of a mess. First, it’s clear that enough voters are “up for grabs” to make the electoral college competitive for 2020. Six percent of registered voters in Iowa have changed their mind about approving/disapprove of the president. So have 5 % of voters in Michigan, 4% in Wisconsin and 4% in Pennsylvania. Thats more than enough to tip the electoral college next year, presuming those changing their minds have gone to Democrats (which isn’t a sure bet). We perhaps already knew this given Biden’s 12-point lead in Fox’s polling.
What I see instead of proof of President Trump’s imminent demise are some interesting states to cherry-pick. First, I don’t think the North Dakota and Idaho estimates are that great (see: their confidence intervals), so I’m not really paying attention to them. The relatively swingy voters in Arizona and Iowa are notable in their possible status as swing states next year. The Deep South is also interesting and notable for its inelasticity, which we should expect given the high population of African Americans and non-college-educated whites, two groups noted for their devotion to Democrats and Republicans, respectively. But this cuts both ways, as the estimates for GA and VA surprise me.
I haven’t yet decided whether I’m done playing around with this analysis. There could be some things to fine-tune. But let this serve as a rough guide: there are plenty of swing voters in the electorate, and they could absolutely decide 2020.
And here are some selected links to the work I read and wrote last week:
Posts for subscribers:
November 1: How liberal are Democratic primary voters? More liberal than they are moderate or conservative, but just barely.
Political Data
Cory McCartan (The Monkey Cage): “Who’s the most electable Democrat? It might be Warren or Buttigieg, not Biden.”
In the figure below, you can see how each candidate’s popular vote would translate into electoral votes, according to the model. Candidates with an electoral college advantage win more electoral votes for the same national popular vote totals, because the demographic groups that dominate key swing states prefer that candidate. You can see which candidate gets more electoral college bang for each vote by whose lines are highest for each popular vote percentage.
Every Democratic candidate would need to win more than 50 percent of the popular vote to win the electoral college. But exactly how much varies by candidate: Warren makes it to 270 electoral votes first, with about 50.7 percent of the popular vote, followed by Buttigieg, then Kamala D. Harris, and finally Joe Biden and Bernie Sanders. These estimates are not perfect; though Warren wins first most of the time, she isn’t guaranteed to do so.
Nathan Kalmoe: According to the fundamentals, “2020 looks like a tossup election”:


For what it’s worth, the lack of a correlation between the economy and presidential vote share in elections since 2008 suggests to me that we should pay attention to the left panel of the graph Nathan posted. It shows that Trump is basically in the same place as he was in 2016 as far as predicted national vote share goes.
Perry Bacon Jr. (FiveThirtyEight): “What We Learned From The First House Vote On Impeachment”
Simply put, it’s a good bet that not much will change no matter what happens in the hearings. That’s both because Americans’ views on the president are very partisan (basically Republicans almost universally support him but a majority of country does not), and because many of the most damning details about President Trump and his administration’s dealings with Ukraine have probably already come out in the last month. The resolution on Thursday passed 232-196, with two Democrats and no Republicans breaking ranks, and it’s just hard to imagine many members switching sides.
All in all, I think Thursday’s vote is a pretty good representation of what we can expect from the House impeachment process: Party unity on both sides resulting in Trump’s impeachment.

David Byler (The Washington Post): “Baghdadi’s death won’t save Trump from his political problems”
In the months leading up to bin Laden’s death in early May 2011, Obama’s approval rating was hovering in the mid-to-high 40 percent range. As news of bin Laden’s death broke, Obama’s approval rating jumped into the 50s, stayed there for about a month and then returned to normal as the story faded from public consciousness. Bin Laden’s death was a net positive for Obama in the long run, in that it gave him a big foreign policy achievement to point to in his upcoming reelection campaign. But the bounce didn’t last.
Some voters may price Trump’s record on the Islamic State into their overall estimation of his competency, but this won’t be a game changer for him. After the dust settles, he’ll still be stuck with his same set of problems — including, and most importantly, himself.
Robert Gebeloff (The New York Times): “The Border Between Red and Blue America”
If 2016 is an indication, the battle lines are clear for 2020. Hillary Clinton dominated the inner-ring suburbs, and Donald J. Trump was dominant in the outer ring.
This is as true within states that are deep red or deep blue as for the nation as a whole. In Alabama, for example, Mrs. Clinton dominated in densely populated neighborhoods. But only one-quarter of the state’s voters live in such neighborhoods, and Mr. Trump won the state easily.
In true battleground states like Michigan, the pattern held, though the voting population was more evenly distributed by neighborhood density. Yet the fact that about one-fifth of Michigan votes came from rural tracts helped Mr. Trump pull off his upset victory.
In the end, the most important takeaway from this work is that the suburbs can’t be depicted with a broad brush, and that when we analyze them for demographic or political purposes, we should be aware of the polarization within.


Other Data
From me for The Economist: “An NBA controversy sparks social-media manipulation”
Nearly a quarter of the accounts tweeting about the NBA in the days after the incident had a higher than 50% chance of being bots or trolls. Many of these “suspect” accounts were created soon after Mr Morey’s tweet. Indeed, in the following week, the number of new suspect accounts increased by 400% over the previous seven days. In that earlier week, about 47% of the newly-created accounts in our sample were suspected of malicious activity by the machine-learning algorithm. After the tweet, this proportion rose to 60%.

Political Science, Survey Research, and Other Nerdy Things
Erika Franklin Fowler et. al.: “Political Advertising Online and Offline”

This graph shows that candidates for political office still spend way, way, way more money on TV than they do on Facebook.
Jamie L. Carson et. al.: “Nationalization and the Incumbency Advantage”
Legislative scholars have investigated both the growth in the incumbency advantage since the early 1970s and its decline in recent decades, but there are several unanswered questions about this phenomenon. In this paper, we examine the incumbency advantage across a much wider swath of history to better understand its connection with changing levels of electoral nationalization. Based on an analysis of U.S. House elections extending back to the antebellum era, we find that the incumbency advantage fluctuates in predictable ways over time with changes in nationalization, which can be a product of both institutional and political conditions. We also demonstrate that the increased influence of local forces in congressional elections may not be strictly necessary nor sufficient for the existence of an incumbency advantage.
What I'm Reading and Working On
As you could infer from the main piece in this week’s newsletter, I’m working on some analyses using MRP to smooth out national polling and distill estimates to the state and demographic level. I can’t say more right now, but get excited. This will be great. I’ve been too busy and too sick to want to pick up a book.
Something Fun
I have had a hard time staying away from candy this halloween season. I’ve done a pretty good job, except that several of y’all roasted me on Twitter for admitting my love for Almond Joys. So I enjoyed this article in the Post on Halloween, candy and adulthood.
Thanks for reading!
Thanks for reading. I’ll be back in your inbox next Sunday. In the meantime, follow me online or reach out via email. I’d love to hear from you!
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!