The Off-Year Election Variants of Mad Poll Disease
And other observations about Tuesday’s elections.
Note for new subscribers: Welcome! You can read more about my background and this publication here. I also hope you’ll take the time to check out previous posts.
Note for everyone: Over the last few months, I’ve concentrated on covering important topics deeply, most recently the Christian nationalist hack of the House of Representatives. But that has been at the expense of Weekend Reading’s original purpose: providing regular insights on politics and our ongoing crises. So, with this edition, I’ll be resuming that approach. Going forward, I aim to publish more frequently, with shorter takes on multiple topics. I’ll also be working on longer single-topic pieces, which I’ll publish as they are ready.
In this post:
An update on the discourse around what I call “Mad Poll Disease.”
How to – and how not to – think about presidential elections in the Trump era.
Why Democrats’ recent victories aren’t just about abortion.
Why American elections (even the “good” ones) don’t always deliver democratic results.
Check out my appearance on The Ezra Klein Show, which just came out today. I was honored to be featured, and I hope you listen.
The Off-Year and Special Election Variants of Mad Poll Disease
This has been a week! Last Sunday’s Weekend Reading post, “Mad Poll Disease Redux,” was picked up widely, and was seen as something of an antidote to the truly bleak results of the New York Times/Siena survey released the same day. Then, last Tuesday’s elections resulted in major wins for Democrats and progressive ballot measures. Those results were seen by many as not only rebutting the Times’ and others' grim horse race results, but as auguring an eventual Biden re-election.
I’m grateful for all the links and positive feedback. However, I need to underscore that my purpose in writing about Mad Poll Disease was not to reassure people that Biden will be reelected. (As regular readers know, I’m hardly a reflexive cheerleader for Democrats’ electoral prospects or strategic choices.) Rather, I sought to show why polling today cannot tell us anything we don’t already know about the outcome in the presidential race.
Here’s what we do know:
The Electoral College will be decided by margins that were too narrow to be accurately seen on Election Day 2016 or 2020 (let alone more than a year away) in Arizona, Georgia, Michigan, Nevada, Pennsylvania and Wisconsin.
Who prevails will depend on whether the majorities who turned out to reject Trump in 2020, and his MAGA candidates in 2018 and 2022, turn out again in those key states.
Therefore, it is more important to pay attention to what needs to be done to defeat Trump and MAGA, because that won’t happen on its own. That means we can’t afford to coast on overconfidence based on Tuesday’s election results, just as we can’t afford to give into despair based on the latest poll numbers. Every election within the margin of error is within the margin of our efforts to shape the outcome.
The Times was quick to warn readers that although “Tuesday Was Great for Democrats. It Doesn’t Change the Outlook for 2024.” I couldn't agree more! But that’s because neither the horse race surveys nor Tuesday’s election returns can tell us what we need to know about the “outlook for 2024” – whether the anti-MAGA majority in the Electoral College battleground states will turn out a year from now.
That said, the special elections are a powerful counterargument to the Times/Siena poll’s literal results, which suggest Biden is headed to the most lopsided Democratic defeat since Dukakis in 1988.
In other words, the elections last Tuesday remind us that the only evidence that Biden is headed to a blow-out defeat is the Times’ and others' horse race polling. But, again, Tuesday’s election results should not make us more sanguine about 2024 than we had reason to be before the Times’ survey either.
Did the Polls Really Get It Right?
After Tuesday, many commentators seemed eager to suggest that Democrats shouldn’t take these victories to be good news about 2024. After all, the polls had accurately predicted these results – and polls are also predicting trouble for Biden in 2024!
It’s a familiar pattern. After every election, pollsters and political analysts will find a way to validate their own pre-election work, whether or not that work actually gave voters useful information. In general, analysts make two arguments about why fielding and covering polls is important: that they can predict the outcome, and that they can tell us what voters care about.
Therein lies a neat trick. If polls get the point estimates wrong, they fall back to saying that polls are still useful because they tell us what mattered to voters in this election. And if the polls fail to tell us in advance what people would base their votes on, they fall back to saying the polls are accurate because the point estimates came close. (And remember, when you read how close this or that pollster came to the final result, that bragging depends on comparing the polls taken closest to the election; they never report how accurate their polling was any earlier. In other words, they’re asking you to trust their polling a year out because their polling the week of the election was accurate.)
In the 50+ midterms that took place in the era before everyone began fetishizing polling and poli-sci savviness, a blatantly criminal act by one party – such as, say, trying to overturn the previous election – would have dominated coverage of the next election where that party was vying for control of the House of Representatives. That’s what happened with Watergate in 1974. But in the 2022 elections, the mainstream media’s lodestar was poli-sci research about thermostatic midterms and polling that made it seem like voters didn’t want to hear any more about Trump and January 6th.
The coverage leading up to the 2022 midterms took it as a given that a Red Wave was coming. As I’ve written before, this narrative choice had consequences that arguably could have shifted the balance of power in the House. When the Red Wave failed to materialize as predicted, it should have triggered more intense soul-searching about why that happened. Instead, it was treated as an interesting surprise, and an opportunity to circle the wagons around the inherent validity of pre-election horse race polling.
In post-mortem analyses, the real measure of how well the polls did should be how well they informed election coverage ahead of time. It’s easy to look back and crow about the point estimates being close. It’s apparently much harder to look forward and allow the data to challenge pre-existing narratives. If the polls were really so good at predicting Tuesday’s results, where were all the breathless headlines about why Republicans should be panicking or why Democrats should be excited? If Tuesday’s election results were so unsurprising, why didn’t the people who boosted the Times/Siena survey on Sunday pre-but the results they “knew” would be coming two days later?
How to Think About Elections Since 2016
Tuesday’s election returns confirmed that the anti-MAGA majority is still the most important, and least discussed, fact about American politics today.
To a very great extent, our federal elections beginning in 2016 have, more than anything else, been about one choice: Trump/MAGA, or not. This is one of those rare times in our history when several successive elections have been choices between two profound and sharply different visions of America's future. Since the Civil War, there have been two such times: at the end of the 19th Century, when successive elections pitted the populist vision of William Jennings Bryan against the corporatist vision of William McKinley, and again in the 1930’s when the contest was over the New Deal Order.
With that in mind, it’s crucial to approach the 2024 election beginning with the recognition that about 178 million Americans have cast at least one ballot on the “Trump/MAGA or not” question. (That’s about three quarters of those eligible to vote.)
Let’s unpack that statistic – beginning with the 2016 election, which, in terms of turnout, seemed pretty consistent with previous elections. Clinton won the popular vote by 2.1 points. About 110 million people who voted in both 2016 and 2020 are still on the rolls. While some of those who supported Clinton then have switched to Trump, and some former Trump supporters switched to Biden, there’s more evidence than not that this switching has been a wash.1
Since 2016, turnout has literally been off the charts. As this chart shows clearly, turnout rates in the last three elections have been ahistorical. Before 2016, except during world wars and when women and then 18 year olds were given the vote, turnout rates stayed within a narrow 3 point range, often for decades at a time. But since 2016, an additional 68 million Americans have cast at least one ballot. Nonetheless, the media spends most of its attention on the vanishingly few numbers of voters who might change their minds when the real story is whether this group of voters turn out out or not.
These voters favor Biden and Democrats by about a dozen points nationally. (I shorthand this group as “post-2016 voters,” since they were either not registered until after 2016, or failed to turn out that year.) In “The Emerging Anti-MAGA Majority,” I showed that in states that consistently vote for Democrats or for Republicans, changes in turnout are unlikely to determine the winner, since post-2016 voters are net Democratic in blue states and net Republican in red states. But in swing states, while 2016 voters lean Republican, post-2016 voters lean Democrat.
Thus, the razor-thin outcomes in those swing states are determined by how many new anti-MAGA voters show up in November – as was the case in 2020, according to the VoteCast post-election survey. In the five states that Biden won by such narrow margins, he won post-2016 voters by 8 to 16 points.2
… But This Isn’t How “Experts” Think About Presidential Elections
I know it sounds insultingly obvious to say that Biden’s victory depends on turning out his supporters (or Trump’s opponents). That’s the point.
The media’s approach is to all but ignore what we know about voters based on the choices they’ve already made, and instead divide the electorate into categories defined by very broad brush demographics, most frequently by race/ethnicity and by education. From there, endless analyses of survey crosstabs make it seem like voters from certain demographic groups are changing their minds en masse. In reality, most of these apparent “trends” are just reporters narrating statistical noise and relying on unexamined ecological fallacies.3
No campaign manager would ever develop a campaign plan built around questions like, “How do I get more non-college voters to support Biden?” In the real world, campaign plans are built around questions like, “How do I get the people who have voted for Biden before to stick with him, and how do I get them to vote again?” As remarkable as it sounds, every vote counts equally on Election Day. You don’t get bonus points for coveted demographic groups like “non-college voters.”
That all of this isn’t painfully obvious to everyone is because there is almost no recognition or interest in the nature of how the electorate has changed since 2016. The proportion of the marquee demographics in the electorate didn’t change much after 2016; what changed was the proportion of the electorate voting for the Democrat. The Biden campaign doesn’t hope that the turnout fairy brings more college voters. The Biden campaign wishes for more Biden voters.
Once you make this mindset shift, you start dividing up those who have voted for Biden according to why they voted for him – again, regardless of their demographics. If one of the reasons people came out to vote for Biden is that they fear losing the right to an abortion, you make abortion a salient issue to keep them in the fold and turning out. The logic in the media conversation runs in reverse – that college-educated voters turn out more reliably and care a lot about abortion rights, so you should talk about abortion.
Speaking of which…
Abortion Isn’t Just Abortion
Especially after Tuesday, the idea that elections are now “about abortion” is becoming a new piece of conventional wisdom. I think there’s no question that Dobbs and state abortion bans have been significantly motivating voters. But it is essential to see that it’s about much more than one narrowly defined issue. It’s about a broader story, which includes all of these parts:
Reproductive freedom - The discrete issue of whether there should be draconian limits on women’s reproductive freedom.
The bundle of MAGA issues - The Dobbs decision, and Republicans’ subsequent doubling down on abortion restrictions, have made it much clearer to voters that the wildly unpopular Republican agenda is not just empty rhetoric for “the base,” but a clear statement of what Republicans will do if they have power. And abortion isn’t the only plot point in this story – as Greg Sargent pointed out, Democrats also had success in local elections pushing back against MAGA “culture war” items like banning books and bullying transgender kids.
Who they are - MAGA Republicans’ post-Dobbs actions, including unanimously elevating Mike Johnson to be Speaker of the House, demonstrates their commitment to governing based on a worldview that is wildly out of step with most Americans.
Their determination - Even if abortion rights are protected, MAGA Republicans won’t give up on their agenda, any more than they have given up on the belief that the 2020 election was stolen. These are politicians who are so determined to realize their agenda that they are willing to storm the Capitol.
Thus, while many voters have known for a long time that Republicans want to ban abortion, few seriously believed Roe v. Wade would ever be overturned. Or that, given the opportunity, Republicans really meant what they said. That view was consistently reinforced by the mainstream media, which, for example, routinely dismissed activists’ warnings during Supreme Court nominations.
Even if Republicans somehow manage to discipline themselves and avoid talking about abortion for the next 10 months, this isn’t the kind of story that is easily changed or forgotten. We have an anti-MAGA majority because Americans have learned the hard way what MAGA’s story is really about: taking away our freedoms.
The Elections that Prove American Elections Are Broken
American elections are supposed to grant “the consent of the governed,” as the Declaration of Independence put it, to those who win them. But somehow, our elections have become synonymous with democracy itself, rather than a tool we use to practice democracy. No matter how broken our tools get – and no matter how much we complain about the obvious disrepair – we keep using them, and we keep declaring the results to be the finest possible workmanship.
On Tuesday, Ohio voters chose to add the right to an abortion to their state’s constitution, defying their state legislature’s attempts to ban it. This was a victory for freedom. But it wasn’t an unqualified “victory for democracy,” as many might put it. It was a victory for popular democracy that reveals how routinely our election system produces undemocratic results.
Throughout the states with Republican trifectas, further restrictions on abortion are extremely unpopular, while raising the minimum wage and expanding Medicaid is extremely popular. That’s why voters in Ohio, and other states where this is allowed, have been overruling their state legislatures. For instance:
Protecting Abortion - Ohio voters approved a constitutional amendment to protect abortion in 2023 after the Dobbs decision reinstated an extreme 6-week abortion ban that had passed in 2019.
Medicaid Expansion - Voters chose to overrule their states’ decisions to reject Medicaid expansion in Idaho, 2018; Missouri, 2020; Nebraska, 2018; Oklahoma, 2020; and South Dakota, 2022.
Minimum Wage Increases - Voters passed minimum wage increases via ballot initiative in Alaska, 2015; Arizona, 2016; Arkansas, 2018; Florida, 2020; Missouri, 2018; Nebraska, 2022; South Dakota, 2014.
If the outcome of state legislative elections truly represented the consent of the governed, we wouldn’t be seeing so many instances of a majority voting for something that directly contradicts the laws enacted by their state representatives. And keep in mind that even if people in every MAGA-governed state feel this poorly represented, we don’t see the evidence of it unless that state allows people to push back via ballot initiatives (which many don’t).
Of course, initiatives on health care or minimum wage could be written off as a truly representative legislative body being out of step on just one or two isolated issues. But Michigan’s recent experience shows us how democratic elections can be used to legitimize imposing a comprehensive anti-democratic regime on the citizens of the state.
As the following table makes clear, in nearly every election from 2006 through 2020, Democrats won a majority of the votes for the state house in Michigan. But losing in 2010 swept into power a MAGA trifecta that redrew state legislative lines to guarantee MAGA majorities for the rest of the decade, contrary to what the majority of Michigan voters chose.
Once in power, those gerrymandered majorities enacted everything from union-killing “right to work” laws and the notorious emergency manager bill which led to the Flint water crisis, to barbaric “rape Insurance” laws. They passed six restrictions on abortion, delayed opting into ACA’s Medicaid expansion until 2014, and made voting more difficult, especially for Black voters.4
However, in 2018 Michigan used the initiative process to pass several democracy-bolstering measures, including Proposition 2, which dictates that both congressional and legislative districts will be drawn via an independent citizen commission. The result was not only a complete reversal – a Democratic trifecta – but the repeal of many of the most damaging policies of the MAGA years.
Voters in similar states, like Wisconsin, aren’t so lucky. They are still living daily under an anti-democratic regime that they cannot realistically hold accountable in the same way the voters of Michigan have. While Michigan voters could vote to redraw the lines more democratically, Wisconsin voters can’t insist on fair lines, because in Wisconsin there is no legal provision for voters to bring initiatives or referendums to the ballot.
This trend is also in line with a broader pattern I documented a few weeks ago – the white Christian nationalist movement hacking our democracy by exploiting its advantage in low turnout primaries in safe Republican districts to implement a regressive agenda that most Americans don’t want. With enough interest group power concentrated in one place, it can be frighteningly easy to “hack” legislative districts to subvert the will of most voters.
Thanks for reading Weekend Reading ! Subscribe for free to receive new posts and support my work.
In the recent Times/Siena survey, only 3 percent of 2020 definite Biden voters and <1 percent of 2020 definite Trump voters say they will definitely vote for the other candidate. That’s not the best sign for Biden, but it’s not catastrophic.
If it seems a contradiction to see polling results cited as evidence in a piece that is critical of polling, my point is that horse race polling is not reliable with respect to very small margins – but these margins are huge. For example, Biden won post-2016 voters by 8 points in Wisconsin, which he won by 0.6 points; he won post-2016 voters by 16 points in both Arizona, which he won by 0.3 points, and Georgia, which he won by 0.2 points.
An ecological fallacy is believing that what is true for a group is true for the individuals in a group. In this context, the problem arises when, for example, it’s stated that based on two surveys, Latino voters are trending away from Biden. We have no way of knowing the extent to which that reflects individual Latino voters who had supported Biden now supporting Trump, or the extent to which the original sample randomly included more Biden supporters than the second sample (which would mean that no individuals had switched sides). I expand on this idea more here.
The devastating consequences of a decade of anti-democratic rule in Michigan are best understood in comparison with two other Great Lakes states: Wisconsin (where a similar story played out) and Minnesota (where that decade looked entirely different). While in 2010 the three states were similar in many respects, by 2020 gerrymander-enforced GOP majorities had made MI and WI significantly less free, democratic, prosperous, and safe relative to MN. I wrote about that story here.