Ask American partisans who’s going to win, and nearly nine in ten of them will name their own side. That’s not confidence born of polling. It’s Partisan Wishful Thinking in Polarized Times, published online Oct. 14, 2026 in Political Behavior by Seth Warner of the University of Connecticut. Political scientist John Sides covered it at Good Authority a week later, on Oct. 21, and PsyPost picked it up on Oct. 31, under a headline that says the quiet part out loud: “the math doesn’t work.” The number that should stop you: nearly 90% of voters who name a preferred candidate now expect that candidate to win, up from roughly 70% a few decades ago. Every election has exactly one winner. Something other than evidence is doing the counting. Baruch Spinoza named that something in 1677, in a single scholium buried in Part III of the Ethics, and he didn’t need a single data point to get there.

The Quick Verdict

Question	What we found
What was studied?	Whether partisan animosity explains the rise in voters’ overconfidence that their own candidate will win
Who ran it?	Seth Warner, University of Connecticut
Where and when?	Political Behavior, published online Oct. 14, 2026; covered seven days later by John Sides at Good Authority and seventeen days later by PsyPost
Data used	ANES multi-wave panel data tracking the same voters across the 2016, 2020, and 2024 elections, plus long-run ANES trend data stretching back to 1980
Headline numbers	Nearly 90% of partisans now expect their preferred candidate to win, up from roughly 70% for decades, into the high 70s in the 2000s, and the high 80s from 2012 on; a 17.6-point linear rise in in-party win expectations between 1980 and 2020
Mechanism	A one-standard-deviation rise in out-party animosity corresponded with an 8.0-8.9 point rise in victory expectations (3.3-3.7 points after controlling for candidate ratings, partisan strength, ideology, news interest, and geographic sorting); the share of voters with high partisan animosity itself rose 47 points between 1980 and 2020
Aftermath finding	Losers who’d braced for defeat ended up about 2 points less hostile toward the other party than winners; losers who’d expected to win showed no such cooling. They scored roughly 3.1 points more hostile than the losers who’d seen it coming
What Spinoza already argued	Ethics, Part III, Prop. 9, Scholium (1677): we don’t desire a thing because we judge it good; we judge it good because we desire it

The Quick Version

A new study finds that almost 90% of partisans expect their own candidate to win, up from about 70% a generation ago, and that the single best predictor of that overconfidence isn’t the polls — it’s how much a voter hates the other side. Spinoza wrote the mechanism down three and a half centuries before anyone ran a regression on it: judgment doesn’t produce desire. Desire produces judgment, and then judgment lies about where it came from.

What the Study Actually Found

Seth Warner didn’t set out to catch voters being irrational for sport. He started with a number that’s been sitting in the American National Election Study for decades, mostly unremarked: the share of voters who, asked who they think will win, name their own preferred candidate. That number used to hover around 70% for a long stretch of the twentieth century. It drifted into the high 70s during the 2000s. Then, from 2012 forward, it jumped into the high 80s and stayed there. Run the straight-line math on 1980 to 2020 and you get a 17.6-percentage-point rise in the expectation that your own side wins. That’s a trend line with no obvious reason to exist, since actual election margins haven’t gotten correspondingly less competitive.

Warner’s contribution was figuring out what’s actually driving that line, using something most single-wave surveys can’t offer: the same people, interviewed repeatedly, across the 2016, 2020, and 2024 elections. That panel structure let him test whether a voter’s hostility toward the opposing party in one interview predicted that same voter’s wishful thinking in a later one: not just whether hostile people happen to also be hopeful people, which could be any number of things, but whether the hostility measured earlier moves the hope measured afterward. It did. A one-standard-deviation increase in how coldly a voter rated the other party corresponded with an 8.0- to 8.9-percentage-point jump in the odds they’d predict their own candidate would win. Load in controls for who the voter actually likes, how strongly they identify with their party, their ideology, how much news they consume, and where they live, and the effect shrinks to 3.3 to 3.7 points. Shrinks. Doesn’t disappear.

And the raw ingredient for that effect has itself been getting more abundant. The share of voters registering high partisan animosity (real, scored, out-party hostility, not just disagreement) rose 47 percentage points between 1980 and 2020. More hatred in the system means more of the specific kind of hatred that bends judgment, which means more voters walking into Election Day certain of an outcome the polling never promised them.

The part that should worry anyone who cares about how democracies handle losing: the overconfidence doesn’t just evaporate when the result comes in. Warner found that voters who lost but had braced for it ended up about 2 points less hostile toward the winning side than the winners themselves. Defeat, expected, took some heat out of the rivalry. Voters who lost after expecting to win showed no such cooling. They came out of the election roughly 3.1 points more hostile than the voters who’d seen the loss coming. The surprise itself, not just the loss, appears to be what curdles.

What Did Spinoza Mean By This?

In the Scholium to Proposition IX of Ethics, Part III, Spinoza writes that “in no case do we strive for, wish for, long for, or desire anything, because we deem it to be good, but on the other hand we deem a thing to be good, because we strive for it, wish for it, long for it, or desire it.”

This line sits inside Spinoza’s account of conatus: the striving, built into every finite thing, to persist in its own existence. A political partisan’s conatus doesn’t show up as a tidy abstraction. It shows up as wanting your side to govern, wanting the other side’s influence reduced, wanting (desperately, often) to not be the one who was wrong about where the country is headed. Spinoza’s argument is that this wanting comes first. The judgment that your candidate will probably win isn’t an assessment arrived at and then felt about. It’s a feeling, dressed up afterward as an assessment, because the mind (Spinoza is explicit about this elsewhere in the Ethics) mistakes its own affects for perceptions of the external world.

Read against Warner’s data, the fit isn’t metaphorical. The study’s central finding is, structurally, Spinoza’s scholium with a sample size: the degree to which you want the other party to lose (animosity) predicts the degree to which you believe they will lose (expectation), and the belief presents itself to the voter holding it as a read on the world, not a readout of their own hostility. Nobody in Warner’s sample would describe their prediction as “I hate the other party, therefore I think we’ll win.” They’d point to momentum, to a friend’s yard signs, to a poll they half-remember. Spinoza would say: of course they would. The mind that’s moved by desire experiences that desire as information about the thing desired, not about itself.

Why Are Voters So Confident Their Candidate Will Win?

1. Animosity gets misread as evidence. A strong, patterned dislike of the opposing party doesn’t register to the person feeling it as hostility. It registers as a settled sense of how things are going to go, which is exactly the inversion Spinoza’s scholium describes.
2. The information environment has sorted itself to confirm it. More partisan animosity in the system means more of a voter’s daily information diet is curated by people who share, and amplify, that animosity — which gives the backward inference fresh material every day instead of correcting it.
3. Losing without expecting to costs more than losing while expecting it. Warner’s post-election numbers suggest the belief isn’t cost-free to hold. Voters who got the surprise version of defeat paid for the earlier overconfidence in extra hostility afterward, which likely feeds the animosity measured going into the next cycle.
4. The feedback loop has no natural brake. Rising animosity produces rising overconfidence; dashed overconfidence produces rising animosity in the losers. Nothing in Warner’s data points to a mechanism that resets this on its own between elections.

Why the Study and the Philosophy Rhyme

The honest version of this comparison isn’t “a 17th-century rationalist predicted a 2026 political science paper.” It’s narrower, and it holds up better for being narrower: Spinoza’s specific claim was that judgments of goodness and likelihood are generated by desire and then misattributed to the object judged. Warner wasn’t testing Spinoza. He was testing whether out-party hostility predicts a specific, falsifiable belief — who’s going to win — using repeated interviews with the same voters over time. But the shape of what he found is the shape Spinoza described: a feeling running the inference, invisibly, underneath a judgment that feels like it came from somewhere else.

It connects to the piece on Spinoza and collective nostalgia, which used the same Ethics to split passions that own a person from passions a person owns — Warner’s overconfident partisans look like the obsessively passionate half of that split, driven by an affect they aren’t examining. It rhymes with the piece on Le Bon and the conspiracy-belief study, where believers overestimated how many others shared their belief — the same mechanism, manufacturing a false consensus instead of false odds. And it sits close to the piece on Tocqueville and democratic backsliding: abstract commitments to fairness and accepting outcomes hold up fine until partisan feeling actually gets involved, at which point the feeling wins. It also connects to the piece on Festinger and devoted fans, where identity-protective reasoning bent perception of unwelcome evidence rather than an electoral outcome — the same defense, aimed at a different threat.

The Gap Spinoza Didn’t See Coming

Here’s where the comparison gets more interesting than a clean match.

Spinoza treated the desire-first mechanism as close to a fixed feature of finite minds — true of a 17th-century Amsterdam merchant guessing at his own fortunes roughly as much as true of anyone else, in any era. He had no reason to expect the mechanism’s output to trend. But Warner’s data shows exactly that: the share of voters with high out-party animosity rose 47 points in forty years, and the overconfidence riding on top of it rose with it, concentrated specifically from 2012 forward. Spinoza gave us a mechanism. He didn’t, and couldn’t, anticipate a media and social environment capable of manufacturing the raw animosity that mechanism runs on, at scale, continuously, in a way that makes the whole country’s collective wishful thinking climb decade over decade rather than hold steady.

He also offered a way out — understanding, adequate ideas, the hard-won freedom from being pushed around by affects you haven’t examined — that this study has no way to measure. Warner’s numbers describe an average effect across a huge pool of partisans. They can’t tell you which voters in that pool were doing the Spinozist work of noticing their own hostility and discounting the prediction it was generating, versus which voters simply felt certain and called it analysis. Spinoza would say that distinction is the only one that actually matters, and it’s invisible to a regression coefficient.

What This Study Doesn’t Show

Be precise about the size of this. The framing invites more confidence than the data earns in either direction.

Correlational, Even With a Panel

Tracking the same voters over time and finding that earlier animosity predicts later overconfidence is stronger evidence than a single cross-sectional survey, but it still isn’t a controlled experiment. Something else that causes both rising animosity and rising overconfidence in the same voters at the same time hasn’t been fully ruled out.

The Controlled Effect Is Much Smaller Than the Raw One

An 8.0- to 8.9-point effect shrinks to 3.3 to 3.7 points once candidate ratings, partisan strength, ideology, news habits, and geography are accounted for. Animosity matters, but it’s one ingredient among several, not the whole explanation for why nearly 90% of partisans expect to win.

This Is American Presidential Elections

The ANES data covers US presidential contests specifically. Whether the same pattern holds in parliamentary systems, multiparty elections, or lower-stakes local races with less national media saturation is untested here.

Feeling Thermometers Don’t Distinguish the Source of the Hostility

The animosity measure is a 0-100 self-reported scale toward the opposing party. It can’t separate hostility rooted in policy disagreement from hostility that’s closer to tribal identity, and those two kinds of dislike may not drive overconfidence the same way.

What Actually Follows From This

1. Treat your own certainty about an election as a symptom to investigate, not a conclusion to trust. If Spinoza and Warner are both pointing at the same mechanism, the confidence you feel about who’s going to win is at least partly a measurement of how much you dislike the other side, dressed up as a forecast.
2. Prepare for the outcome you’re not expecting, specifically because you’re not expecting it. The study’s clearest practical finding is that the surprise of losing costs more, in lasting hostility, than the loss itself. Sitting with the real possibility of defeat ahead of time is cheap insurance against that specific cost.
3. Ask Spinoza’s actual question: what do you want to be true, separate from what the evidence supports? He wasn’t arguing you can switch off desire. He was arguing you can learn to notice when a belief is downstream of it — and that noticing is the entire difference between a prediction and a wish wearing a prediction’s clothes.

Spinoza never watched a cable news panel argue about turnout models, and Seth Warner wasn’t out to vindicate a 350-year-old scholium buried in a geometric treatise on ethics. But a 2026 study built to explain why American voters have grown steadily more certain their side will win landed on the exact mechanism Spinoza described in 1677: a judgment that feels like it’s reading the world, when what it’s actually doing is reporting back on how badly you wanted a particular answer.

---

This post discusses a study published online Oct. 14, 2026 in Political Behavior (Seth Warner, University of Connecticut), covered seven days later by John Sides at Good Authority and seventeen days later by PsyPost, and Spinoza’s Ethics (1677). The study’s panel findings show that partisan animosity predicts later overconfidence about election outcomes; they do not establish that animosity is the sole cause, and the relationship is correlational rather than experimentally proven.