The New York Times’ “Moderation Advantage” Is a Statistical Illusion

After accounting for money and incumbency the supposed electoral bonus for moderate candidates vanishes entirely.

The Democratic establishment is building its entire comeback strategy on a premise: moderate candidates win. The New York Times just gave this conventional wisdom its most prominent endorsement, in an editorial arguing moderates consistently outperform progressives.

The New York Times editorial, however, rests its case on one simple metric: PAC endorsements.

In my last piece, I explained why the electoral gains from a moderation-centric strategy are tapped out and proposed an alternative: a Clean vs. Corrupt anti-corruption platform that can actually win back voters.

In this post, I’m going to show why the evidence used to support the claim that moderates perform better is a statistical illusion.

The small electoral bump the Times attributes to moderation isn’t a feature of ideology at all. It’s a phantom effect created by ignoring the two most powerful forces in American elections: fundraising and incumbency. When we apply even the most basic statistical controls, the “moderation advantage” disappears completely.

This isn’t a complex academic debate. It’s a simple case of looking at the right numbers. Let’s walk through it.

The Flaw in the Simple Story

The Times’s analysis rests on a straightforward correlation: candidates they define as “moderate” (based on PAC endorsements) tend to do slightly better than those they don’t. But correlation, as the saying goes, is not causation.

People who carry EpiPens have higher rates of severe allergic reactions. But the EpiPen doesn’t cause allergies; people carry EpiPens because they have allergies. The Times made the same mistake: they saw candidates funded by centrist PACs winning more often and assumed moderation caused the wins, without checking whether these “moderates” were more likely to be well-funded incumbents.

They were. The candidates the Times labels “moderate” are significantly more likely to be well-funded incumbents. The problem becomes clear once you see who they’re actually comparing.

The Problem: Bad Comparisons

The Times compared “Moderates” (candidates funded by centrist PACs) against “Non-Moderates” (everyone else). But “everyone else” mixes two completely different groups:

1. Non-Moderates: Candidates who actually ran competitive campaigns but were not endorsed by moderate PACs.

2. “No PAC Funding” Candidates: Candidates who raised little to no money from any PACs. These are often non-competitive candidates running in unwinnable districts with little to no resources.

Unsurprisingly, these unfunded candidates underperform by 1.9 percentage points. But they’re not underperforming because they’re too progressive. They’re underperforming because they were never competitive to begin with. By construction, the Times’ non-moderate group is weighed down by candidates who were never going to win.

This contamination problem isn’t unique to the Times’ framing. We can prove it by flipping their analysis: what if we compared candidates backed by progressive PACs against everyone else?

Using the same methodology, progressives outperform Harris by +1.4 points while non-progressives underperform by -0.4 points. This is the mirror image of the Times’ result showing that moderates outperform by +1.4 points while non-moderates underperform by -0.6 points. Why? Because now it’s the non-progressive group (which includes moderates) that gets dragged down by unfunded candidates.

This is the Times’ first error: building their comparison on a contaminated sample. We can correct for this by filtering out all non-competitive candidates while still using their preferred measure of ideology. Let’s see what happens when we only compare PAC-funded moderates against PAC-funded non-moderates.

Step 1: The Surface-Level Correlation

What happens when we only compare funded candidates? After filtering out “No PAC Funding” candidates, we’re left with 335 Democratic candidates. When we run the Times’ simple, unadjusted analysis on this data, we still find a small moderation advantage of +0.99 percentage points.

Step 2: The Vanishing Effect

Now for the moment of truth. What happens when we adjust for money and incumbency?

The effect vanishes.

Moderation advantage after accounting for money/incumbency: +0.24 percentage points (a difference that is statistically and substantively indistinguishable from zero).

Once you compare like with like—candidates with similar funding and similar incumbent status—the ‘moderation advantage’ disappears.

The Real Methodological Problem

The Times editorial anticipates objections to its analysis by dismissing academic research as overly complex and disconnected from voter perceptions:

“[S]ome progressive analysts and professors say it has little effect. The debate involves statistical complexities that are difficult for most people to follow. Yet there is a simple way to see the weakness of the argument that moderation is unimportant.

The researchers making that claim rely on data about candidate ideology that can have little connection to voter perceptions… The data ends up being so messy as to produce bizarre results.”

This preemptive dismissal is telling. The Times needs to discredit academic research because that research has reached an inconvenient consensus: the electoral benefits of moderation have largely disappeared. An influential 2018 study claiming large advantage for moderates was corrected by its own authors, who now conclude the evidence “is far weaker than we previously thought, and should not be relied on.” More recent studies consistently find that moderation’s effect on vote shares is either small or statistically indistinguishable from zero, with party brand and turnout dynamics now dominating electoral outcomes.

Rather than engage with the evolving evidence, the editorial dismisses it as too “complex” and turns to what it presents as a simpler measure: PAC endorsements. But the editorial’s “simple” approach—lumping well-funded progressive candidates together with unfunded candidates running in unwinnable districts—is precisely the kind of “messy data” problem it warns about.

The editorial claims academic measures are disconnected from voter perceptions. Yet when we use data entirely based on voter perceptions (the Cooperative Election Study), we find no moderation advantage. In fact, measuring candidate ideology based on how voters perceive candidates showed the smallest effect size of any measure tested.

The issue isn’t methodological complexity. It’s that rigorous analysis—whether using various academic measures, voter perception data, or even the Times’ own PAC endorsements—consistently fails to show the large moderation advantage the editorial describes.

The idea that moderation helps candidates win isn’t extraordinary. My own research has documented real electoral benefits from moderation in past eras. But even ordinary claims require some evidence.

If the data showed a sizable moderation advantage today, I would be the first to report it. The issue here is that the evidence doesn’t support the claim. The uncomfortable truth is that the moderation advantage that once existed has faded, not because the theory was implausible, but because the electoral landscape has changed.

Moderation Will Not Save The Democratic Party

The Times based an entire political strategy on a statistical artifact. The small advantage they found was never about ideology. It was simply incumbency and fundraising in disguise. Once you account for those basic factors, there is zero statistical evidence in their own data that moderate candidates perform better. The phantom effect they discovered is just that: a statistical shadow cast by money and incumbency.

Even if we set aside the above analysis and accept the Times’ inflated estimate at face value—an effect we know to be artificially large—it would not have changed the outcome in a single competitive seat in 2024. The strategy has already been fully implemented. There are no more gains to be had. That would be true even if a moderate advantage existed, which it doesn’t.

This matters because the Democratic establishment is going all in on a strategy built on a mirage.

The real question isn’t whether to run moderates or progressives in swing districts. Democrats already run moderates in nearly every competitive seat—and they’re still losing. The question is what those candidates stand for.

Voters aren’t looking for ideological triangulation. They’re looking for someone who will fight a system they believe is fundamentally rigged. This is why I’ve argued for reframing the central conflict as Clean vs. Corrupt, an anti-corruption platform that directly confronts the institutional rot and elite capture that fuels voter rage across the political spectrum.

This doesn’t mean moderation itself is the problem. But it’s not the answer, either. If you’re looking for candidates who can win back the House, don’t just look for moderates who split the difference on policy. Look for moderates who can be credible anti-corruption reformers, candidates who can channel voter fury at a broken system into a concrete agenda for change.

To their credit, the Times engaged constructively with this analysis. After I replicated their work and identified a labeling error that had effectively doubled the apparent size of their effect, they promptly issued a correction and engaged in substantive dialogue about the data. I appreciate that willingness to engage, and I believe we share the same goal: understanding what actually works in modern elections.

That shared goal is precisely why the core problem can’t be dismissed. Even using their corrected numbers and their preferred measure of ideology, the “moderation advantage” disappears completely once you control for fundraising and incumbency. The Democratic Party is building its comeback strategy on this premise. If the premise is wrong, the strategy will fail—and they can’t afford to get this wrong.

The stakes are too high for wishful thinking. The data tells a different story than their editorial suggests, and Democrats need to hear it.

---

Data and Code Availability

All data and code used to generate the results in this analysis are publicly available and can be replicated. The repository includes:

• Complete replication code for all analyses and figures

• Cleaned datasets including candidate ideology scores, vote shares, fundraising data, and incumbency status

• Documentation explaining data sources and variable construction

The repository can be accessed at: https://github.com/abonica/nyt_moderates_analysis

This transparency allows readers to verify the findings, explore alternative specifications, or build upon this work.