The moderation debate fiddles with 2% while democracy’s dimensionality collapses

Presidential vote now determines 98% of House outcomes. One-dimensional partisan conflict is authoritarian-friendly territory. Could fusion voting restore competitive dimensions?

For months, analysts have been arguing over whether Democrats need to run more “moderates” in order to win more elections. As a data nerd, I find this debate fascinating. But as an alarmed citizen, I find this debate extremely frustrating.

Our democracy has an existential problem: an authoritarian rot and nothing-matters nihilism are metastasizing throughout our politics. Maybe fielding more centrist candidates wins Democrats a few more elections. Maybe it doesn’t. But either way, the argument is about one or two percentage points. And it does nothing to address the core two-party doom loop wrecking our politics.

It’s like we’re debating whether to smoke one pack a day instead of two when our democracy needs a lung transplant.

So in this piece, I want to blow the whistle on the moderation debate, call time out, and look at the bigger picture.

In a recent post, Stanford political scientist Adam Bonica argued that once you control for money and incumbency, the moderation advantage vanishes. But his broader point resonates with me: “the moderation advantage that once existed has faded... because the electoral landscape has changed.” That’s exactly right.

By my analysis (below), presidential vote share now explains 98% of House outcomes. In the Senate, it’s 91%. In 2000, roughly half of Senate races were competitive enough that candidate quality could flip them. By 2024, only 12% were.

Bonica identifies the landscape change. I want to name what that change actually is: a collapse of dimensionality. American elections have flattened from a multidimensional space into a single axis of partisan identity. Candidate quality, local factors, ideological nuance, local organizing forces—all those dimensions that used to create competition and accountability have been crushed down to near-irrelevance. In an age of hyper-partisan, hyper-nationalized voting and calcified politics, the big problem is that nothing matters.

This matters because one-dimensional politics is where authoritarians flourish. When the other party becomes the enemy, any candidate with your preferred letter, “R” or “D” – will do. Running more moderates doesn’t address this. We need new dimensions of political conflict.

In multidimensional politics, your coalition fragments when leaders threaten your core interests—farmers and autoworkers defect over tariff policy, fiscal hawks leave over reckless spending, law-and-order conservatives bail over norm violations. These cross-cutting concerns constrain authoritarian impulses.

But when politics collapses into what I’ve called the “two-party doom loop“—where there are only two sides and the political opposition is the enemy—there’s nowhere to go. You tolerate threats to your values because switching sides means empowering people who will destroy everything you hold dear. When 87% of voters believe “America will suffer permanent damage” if the other side wins, we’ve reached a level of toxic partisanship where accountability breaks down completely.

This is the “frozen state“ where any candidate with your preferred “R” or “D” will do—especially one who promises to fight the enemy harder. This is the terrain on which authoritarians thrive. It’s bad news.

But I won’t end on collapse. I promise. Stick with me to the end — I’ll offer one genuinely promising path toward higher dimensionality Fusion voting could offer candidates alternative ballot lines to signal new alliances, and begin to melt the frozen binary. It’s not a complete solution, but it’s a start.

As we’ll see at the end of the piece, fusion parties could bring more competitive elections. Stay tuned to find out how much! And yes fusion voting doesn’t fix everything. Yes, proportional representation would be even more transformative. But we have to start somewhere.

Is moving to the center really the way to win?

This debate over moderation has been burbling for a few months. In a previous post, I wrote about why I find the whole issue of “moderation” confusing — because “moderation” is a relational concept, not an ideology, and nobody can agree on how to measure it.

A quick recap of where we are: The latest salvo in this debate came from the New York Times editorial board: “The Partisans Are Wrong: Moving to the Center Is the Way to Win.” I think the ed board is actually right that a winning position would be more “angry” than traditional “moderation” (more Jared Golden, less Josh Gottheimer). As the piece argues, “It is more combative and populist. It tends to be left of center on economics and right of center on social issues.” But this is a two-dimensional approach that struggles to break out of our one-dimensional world.

During the week I’ve been working on this piece (it’s a big piece, I take my time), Adam Bonica has responded. Twice. G. Elliott Morris has also responded – both to the Times analysis, and to a separate and much sillier “moderation” analysis.

The data scientist in me finds both Bonica’s and Morris’s responses extremely convincing, and methodologically sound. But the big-picture reformer yearns for something, well, bigger-picture. So here we go.

Democratic partisans are losing faith in centrism, and the Times is on it.

Let’s start with that New York Times piece, which opens by looking at 2024’s “somewhat close” House races.

Source: New York Times. (via screenshot)

The Times then examines split-ticket districts, where one party won the presidency and another won the House seat. The commonality: “They are moderates.”

Fair point. Most would classify them as their party’s moderates.

Source: New York Times (via screenshot)

But notice the fine print: their victories came from small overperformances in razor-thin districts. Zoom out on all 2024 races, and you see the pattern much more clearly. Now we turn to my visualizations…

The big story: Presidential vote share determines House vote share with near-perfect precision. Or, more specifically, if you knew how the district voted in the presidential election, you could tell me with 98 percent accuracy how the district voted for the House. Yes, candidate quality matters, on the margins. But those margins are very very very very very tiny in today’s politics.

All those dots? They represent candidates of many different types of quality. And yet, they all hug the same line. Put another way: In today’s electoral environment, almost nothing individual candidates do matters. Only super-candidates can distinguish themselves from their party’s presidential candidate. And only barely.

Or if you like statistical terms, think of it this way: we have too little variation on our dependent variable (over-performance) to explain much. And, moreover, because the parties are much more ideologically sorted, we also have very little variation on our independent variable (ideology) to explain much either. Even the most moderate candidates today are still pretty solidly partisan — far more than they used to be.

Remember the 2000 Senate races? Republicans could rout Vermont and Rhode Island, and Democrats could dominate Nebraska and North Dakota.

Now let’s shift to the Senate and rewind to 2000. Here’s the same scatterplot. And. Well, would you look at the difference?

In most states, the Senate races are quite distinct from the Presidential election. In states above the line, Republicans over-performed. Take Jim Jeffords, a Republican in Vermont. He ran 30 to 40 points ahead of his party’s presidential candidate. But in 2001, Jeffords left the Republican Party over increasing policy disagreements. He became an independent who caucused with Democrats. Lincoln Chafee, the Republican who easily won Rhode Island in 2000, would later attempt (and fail) to win the Democratic Presidential nomination in 2016.

In states below the line, Democrats over performed. Consider Nebraska and North Dakota, where popular Democratic Senators (Ben Nelson and Kent Conrad, respectively), won easily, even as the states went decisively for Bush. In West Virginia, that’s Robert Byrd, the long-time Democratic incumbent, who won decisively (77.8 percent to 20.1 percent).

The straight black line is the overall trend line. On balance, in states where Bush did better, Republicans did better. But I want you to notice four things about this line.

First, the obvious thing: most of the dots are pretty far from the trend line. In other words, the trend line tells us almost nothing about any specific Senate race. On average, yes, Democrats do better in Gore states, Republicans do better in Bush states. But how much better depends on the candidates. A lot.

Second, the line cuts only about a 30 degree angle. While presidential partisan vote shares and Senate vote shares moved in tandem, the relationship was not one-to-one. The closer a line is to 45 degrees, the more the relationship is one-to-one.

Third, notice that at the center of the graph, the zero-zero point, the line hits slightly below and to the right. This tells us that the average Democrat Senate candidate ran slightly ahead of Al Gore (the Democratic presidential candidate that year).

Fourth, most winners sit above the trend line—they outperformed their partisan baseline. This shows candidate quality mattered. Nearly all Democratic winners overperformed their baseline, except Hillary Clinton (New York) and Jon Corzine (New Jersey). Republican winners also overperformed, except Conrad Burns’s narrow victory in Montana. This is another sign that candidate quality mattered much more then.

In short, voters in 2000 evaluated Senate candidates as individuals, not partisan avatars.

Next up: all seven 21st century presidential elections together.

And, just …wow. Would you look at how the times have been a-changin’!

First, the dots (individual Senate races) get closer to the trend line. Individual winning Senators deviate less from how their party’s president performs.

Second, the line gets closer to 45 degrees (almost perfectly by 2024), showing a clearer one-to-one relationship between presidential vote share and Senate vote share.

Third, something Democrats might find interesting: their Senate candidates ran ahead of the presidential ticket in 2004, 2008, and 2012. We see this because the trend line cuts slightly below the middle of the graph. By 2016, that advantage disappears—the trend line cuts closer to the center.

Fourth, and finally, we see more and more candidates of both parties who under-perform relative to the trend, but still win. This is another sign that candidate quality matters less.

Today, few voters consider the distinct qualities of the candidate. They are looking at the D or the R next to the candidate. Even the best candidates struggle against the crushing weight of partisan gravity.

One way to measure this transformation is to chart the extent to which presidential vote share predicts Senate vote share. We can do this by calculating the R-squared from a bivariate regression for each election year. R-squared tells us how much of the variance in Senate vote share can be explained by presidential vote share in each state - the higher the coefficient, the tighter the correlation.

In 2000, an R-squared of 0.20 meant that presidential vote share explained about 20 percent of the variance in Senate vote share across states. The remaining 80 percent of variance was attributable to candidate-specific effects and state-level factors.

Fast forward to 2024: the R-squared reached 0.90, meaning presidential vote share now explained 90 percent of the variance in Senate vote share. Less than 10 percent of the variance remained for the candidate effect or state-specific factors. Fewer states remained legitimately competitive.

Two big inter-related trends explain the collapse: partisan polarization and nationalization of politics. Party sorting means most voters now live in one coalition or the other. Fewer and fewer consider voting for the other side. Just Democratic voters and Republican voters.

And nationalization reinforces partisan voting. A Senate vote is now a vote for party control, not just an individual candidate. Political scientist Daniel J. Hopkins documented this shift in his excellent 2017 book, The Increasingly United States (which I wrote about here). The years since have only intensified it.

Take Jon Tester’s 2024 loss in Montana. Tester is a talented politician, who long defied gravity in an increasingly Republican state. He could not defy gravity in 2024. Sherrod Brown’s 2024 loss in Ohio was similar. Brown is also a talented politician. He is running again in 2026. I do not expect him to defy gravity in 2026 either.

Oh, how competitiveness has collapsed since 2000

Many commentators (myself included) have lamented the decline of two-party competitive states and House districts. For obvious reasons, lack of competitive elections is bad for democracy. Competition spurs responsiveness, innovation, accountability, dynamism. Lack of competition excuses complacency, stagnancy, corruption, and poor governance. Uncompetitive elections tolerate and breed extremism.

But how much has competition really declined? Standard competitiveness metrics have a fatal flaw. They don’t account for the collapse in candidate agency I just described. Consider a state like Rhode Island in 2000. Is that a competitive Senate seat? Lincoln Chafee, a moderate Republican, wins the Senate seat easily. Al Gore wins the presidential vote easily. In theory, Chafee could have lost. But a Chafee had held that seat for a very long time --- First John, then his son Lincoln. In 2006, a young Sheldon Whitehouse would finally win that seat.

Which leads to a bigger question: how do we even measure competitiveness? For decades, the standard has been the Cook Political Report’s Partisan Voting Index (PVI). It rates every state and district based on the candidates. Here’s the challenge: an R+10 rating means something radically different today than it did in 2000. Back then? With the right candidate, a Democrat could absolutely win an R+10 state. Today? It would take a miracle. Republican-leaning voters won’t even consider a Democrat anymore—no matter how “moderate” or compelling the candidate. Again, they are voting for which party controls Congress. It’s all partisanship and nationalization.

So how do we capture that change over time? I’ve come up with a new approach here. It’s a bit of a nerdy statistics thing, so if you just want to skip ahead and avoid the nerdy statistics thing, please do.

Nerdy statistics thing

To measure competitiveness over time, we need to know: how likely is a party to win a given seat? A toss-up means 50-50 odds. As seats become safer, one party’s odds approach certainty.

My approach uses the scatter you’ve seen in the charts. Remember that diagonal trend line showing the relationship between presidential and Senate vote margins?

The first thing to observe is the increase in how much presidential vote share explains senate vote share over time.

The key information is in how candidates deviate from that line. We call these deviations “residual standard deviations”—or just The Residual. It’s what remains unexplained after accounting for presidential performance.

Here’s why it matters: when we collect all these residuals for a given year, we can see the full range of candidate over- and under-performance. In 2000, running six points ahead of your party’s presidential candidate was pretty common. In 2024, it would be extremely rare. This tells us how many seats were truly competitive—winnable with a truly above-average candidate or other anomaly.

Below, I’ve visualized these residuals so you can compare 2000, 2012, and 2024. What you will see is a narrowing spread of how much Senate candidates over or under perform presidential candidates. The narrower the spread, the lower the residual. The residual marks the performance of a candidate one standard deviation above the median. By definition, only the top 16 percent of performers are one standard deviation above the mean.

To make this measure meaningful, I’ve broken the races into five tiers based on how unlikely it would be for a candidate to overperform their presidential nominee enough to win. The first tier contains true toss-ups—races where a slightly above-average candidate (0.25 Residuals better than the presidential nominee) could win. Lean races require a solidly above-average performance (0.5 Residuals). Pretty Safe seats need exceptional candidates running a full Residual ahead. Very Safe seats require near-heroic performances (1.5 Residuals). Foregone seats? Those need the statistically improbable—more than 1.5 Residuals of overperformance.

End of nerdy statistics thing

If that all sounds like statistics gobbledygook to you, that’s fine (you can read more about it above, or just take my word for it). But here’s the simplest explanation. I’m measuring competitiveness by asking: how much better than their presidential nominee does a candidate need to perform to win? And how likely would that performance be? I calibrate “above average” against the historical record of actual candidate performance. And I’m tethering mid-term years to the previous presidential vote.

Anyway, here’s the change over time, in five tiers of competitiveness.

The transformation over a quarter-century is just remarkable. In 2000, almost half of Senate elections were winnable for either party with only a slightly above-average candidate. Even the safest seats could become competitive with the right candidate. By 2024, toss-ups dropped to nine percent. Foregone states metastasized to two-thirds of all seats. The three middle categories also declined.

The big takeaway of this analysis: just running different candidates is not enough to make more elections competitive anymore.

And so what happens when elections are not competitive? Incumbent candidates have no real accountability. And if they want attention (and doesn’t everyone in public office?), the best way to get that attention is by being a partisan fighter. Their home-state constituents will love them for it! Heck, they can even achieve podcasting fame.

Now, let’s turn to the House, where true competition is even more unusual.

The collapse of House candidate distinctiveness and district competitiveness

House races vastly outnumber Senate contests. They tend to be lower-profile, and attract less media attention. It is even harder for individual House candidates to distinguish themselves from their party.

The same Senate trends have occurred in House elections. I’ve visualized 2016, 2020, and 2024 to show the recent progression. (I’ve omitted uncontested races from this analysis. There are a surprising number of uncontested House races. Further down the ticket, there are many many uncontested races)

2016 still shows some candidate independence. A few House candidates significantly outperform their presidential baseline. The dots scatter around the trend lines—candidates can still distinguish themselves from their parties. The R-squared is 0.84 (presidential vote share explains 84 percent of the variation)

By 2020 and 2024, any remaining independence vanishes. The dots hug the lines. The R-squared jumps to 0.97 in 2020 and 0.98 in 2024. Presidential vote explains nearly everything, leaving almost no room for candidate effects.

The House shows the same collapse in competitiveness. The foregone seats have just taken over. The genuinely competitive seats have diminished from unlikely to haystack needle.

Everything is the presidential election. With the exception of a few extraordinary candidates (Hi Marie Gluesenkamp Perez and Jared Golden!), Democrats cannot break free of their party. And even these outliers—MGP and Golden—are barely overperforming compared to past eras. Running a few points ahead of the Democratic presidential candidate is only impressive by today’s standards. Two decades ago, plenty of candidates ran well ahead of their parties.

But wait – what if we could restore dimensionality through fusion voting?

So, back to the moderation debate. I’m dubious that moderation matters, but let’s grant the most optimistic “moderation boost” scenario. It’s still only two or three percentage points. Yes, that’s not nothing. And yes, in an era of narrow margins for House/Senate control, that may be decisive.

But winning narrow margins is small thinking. It doesn’t break us out of our gridlock moment. And it definitely does not break us out of the more destructive two-party doom loop cycle I keep banging on about.

The problem is not the candidates.

The problem is the parties, and the party system.

To expand the map, Democrats would have to do a complete re-brand and elevate a whole new generation of leaders who will create a whole new set of conflicts and convey a clean break with the past. Essentially, hang a bright new ‘Under New Management” sign on the party, and then mean it. For obvious reasons, a generational baton-toss is unlikely. Older leaders like their power and are reluctant to pass it along. (See: Biden, Joe) Without transformation? Running moderates in a few districts is just playing cards in a hurricane.

Hence, I keep coming to the obvious conclusion: we need some new political parties!

Proportional representation is the end goal. But much as I may wish it, we’re not going to get proportional representation overnight. But with litigation to re-legalize fusion voting now proceeding through the courts in three states (NJ, KS, WI), it’s worth asking: what if fusion were legal everywhere? Would we have more competitive elections!

The fusion voting math is simple. Say 3% of Republicans would vote for a Democrat as a check on Trump—but won’t vote “Democratic Party.” Say another 3% of Democrats feel abandoned on working-class issues and might vote Republican—but won’t vote “Republican Party.”

Create a Moderate Party ballot line and a Working Class Party ballot line. Let candidates compete for their endorsements. Suddenly you have 6% of newly persuadable voters per race. That’s enough to restore genuine competition in some states and districts that have lapsed into lopsided partisan territory.

Fusion voting lets one candidate appear on multiple ballot lines. Think of it as multiple pathways to the same destination. A candidate endorsed by both Democrats and a Moderate Party would appear twice on the ballot. Voters choose which line to pull, but both votes count for the same person. The ballot line choice becomes a message about priorities.

Here’s where it gets interesting strategically. A Democrat might secure the Working Class Party endorsement defensively—say, by backing a $15 minimum wage—to prevent a four-point loss, while pursuing the Moderate Party offensively to bring a four-point swing. A Republican could flip the script—champion working-class economics while showing enough anti-corruption gumption against Trump to win moderate support. The parties would have to compete for these endorsements in every race.

The real competition happens in the pursuit of endorsements. Candidates must address anti-corruption and working-class priorities to win both lines. That competition is what creates new dimensionality. Suddenly elections aren’t just about Democrat versus Republican. They’re about good governance, economic populism, institutional integrity—issues that could cut across the partisan divide and force candidates to distinguish themselves beyond party labels.

These are hypotheticals, of course—but they are informed by polling data showing genuine appetite for alternatives and genuine uncertainty about which party to support given the existing two options. The precise percentages matter less than the principle: fusion parties could create cross-pressured voters and force competition on new dimensions. And competition fosters the innovation, creativity, and responsiveness democracy needs.

Below: my rough estimate of what the 2024 Senate map might have looked like if fusion voting were in play. Assume each race had potential for a three, or six, or even nine percent swing from fusion party endorsements. Simple conclusion: The Senate map becomes more competitive.

And now, the House:

The modeling translates fusion party effects into increased candidate independence from presidential results. A 3% fusion party increases competitive House races from 22 to 32 districts (7.4% of the chamber) and adds 5 Senate states to genuinely competitive or winnable territory. A 6% fusion party doubles House competitiveness to 44 districts (10.1%) while pulling 13 Senate races out of “foregone conclusion” status—leaving only 12 of 33 races effectively predetermined. A 9% fusion party nearly triples House competitiveness to 63 districts (14.5%) and transforms the Senate map entirely: only 9 of 33 races remain foregone, with 24 states where exceptional candidates could credibly compete.

(Data for both fusion modeling analyses live here)

Fusion voting doesn’t fully recreate 2000—but rolling back 4-8 years of dimensionality collapse would restore enough room for candidate quality and coalition-building to matter. We have to start somewhere.

New fusion parties could build over time. They could give disenchanted voters a new home and new power. These processes could feed back on themselves – a virtuous cycle instead of a doom loop.

Conclusion: The collapse of dimensionality and the need for more parties

Our contemporary hyper-partisan polarization crisis is a collapse of dimensionality. The political world has flattened into a single dimension: us versus them, Democrat vs Republican. And that is the world in which an authoritarian can thrive. When Democrats become the enemy for half the country, any Republican will do. Especially one who promises to fight Democrats. The same holds the other way around. And if the fighting involves some dirty tricks, so be it – they’d do the same to us! In fact, they already are!

The tightening scatterplots visualize this dimensional collapse. Watch the dots—representing individual candidates with distinct qualities—get pulled closer and closer to a single line. That’s dimensionality collapsing before your eyes. Space for candidate distinctiveness shrinking to nearly nothing.

Local factors used to matter. They were another dimension. Individual issues, personalities, local organizing, even ideologies --- these were also issues that cut across the major parties. They created some play in the joints of the political system: some responsiveness, some dynamism, some creativity. Now all of that dimensionality is gone.

In 2000, roughly half of Senate races were genuine toss-ups, and in most states, an exceptional candidate could defy their president’s party to win. By 2024, only 12% were toss-ups—while 61% became foregone conclusions. Now one factor—Democrat or Republican—explains everything, from Senate races (R² = 0.91) to House contests (R² = 0.98). One factor. One dimension. And everything is breaking.

We need more factors. We need more dimensionality. We need more parties.

New parties, whether by fusion voting or other means, are one way to bring back some of that multidimensional dynamism. They seem to me the best way. Some prefer building new factions within existing parties. But without a ballot line, I don’t understand how a faction distinguishes itself today.

And this takes us back to the problem we started with --- does moderation matter? In our current two-party system? Not really. Presidential vote explains 98% of House outcomes. Candidate quality has collapsed to statistical noise.

But could moderation matter in a different system—one where candidates can signal distinct positions through fusion ballot lines? That’s a different question with a different answer.

Right now, the only way to signal moderation is running as an Independent—getting the Democratic Party to stand down and offering voters a different label. Some ambitious politicians in the right states and districts should try this. But it’s a high-wire act few will attempt.

The fundamental problem remains: without another label on the ballot, a Democrat is a Democrat is a Democrat. The national party brand, defined by the top of the ticket, blocks out everything else. Fusion voting allows those alternative labels. We don’t need better candidates. We need more parties.

---

---