I see the point as that small average effects of observed moderation are perfectly consistent with 1) issue positions mattering a ton to individual voters, 2) large effects of moderation on some issues, and 3) zero or negative effects on others
I don’t think the absolute value of the point estimates can be taken to the bank, the relative value is the main message. Unclear how you’d aggregate across the estimates to get a total - eg if you added up all the positive point estimates together you’d get ~30 pp for Ds, clearly overestimate
Long live Hinich 1977!
Yes, I wouldn't read too much into absolute magnitudes. Main point of the paper is the argument, and the data helps illustrate it.
About to be on the road for a couple weeks but will reach out when I get back!
More generally, the idea here is to make an argument from first principles about what we should expect from aggregate data given what we know about voters & use survey data to try to show those theoretical dynamics are at play.
As I flagged, I wouldn't reach much into the candidate-level results this early - the Cuban result is just interesting.
Yes I do believe voters care about issue positions and that that is why decades of research find that moderate candidates do better.
The core finding that moderation improves vote shares isn't survey-based, it's consistent across decades of research including in recent elections.
P.S. Bonus fun fact — while it’s very early and our data isn’t ideal for assessing 2028 electability, @mcuban.bsky.social ’s performance in the trial heats we did is literally off the charts.
It also means moderation on specific issues—particularly issues where a party is out of step with the public—may yield bigger electoral gains.
These gains will still be small, though—meaning moderation across *many* issues is likely needed for parties to make a meaningful dent.
So what does this mean?
Studies finding small average returns to moderation provide a lower bound on how many voters vote on issues. E.g., in our simulations, every simulated voter is voting on issues, yet the average effect of moving to the middle can be small, zero, or negative
This matters because some argue that cross-pressured survey responses are just measurement error—that these people are really moderates
But if so, they should reward moderation across the board
They don't. They only reward it on issues where they personally disagree w the party
We see this at the individual level too. Voters who agree with a party's standard position on an issue don't reward & may actually penalize candidates who abandon that position.
Even voters who look "moderate" overall behave this way on issues where they hold a party's position.
This is not intuitive if you think voters are at the elite center on every issue / always prefer a position between the parties. But it's exactly what you'd predict if voters are cross-pressured and care about proximity on specific issues.
More broadly: on issues where a party's position already has majority support, moving to the elite middle yielded negative effects for Republicans (-2.4pp, p<0.001) and a negative (though not significant) effect for Democrats.
We also find moderation can backfire
When Republicans moved to the middle on transgender sports or K-12 teacher accountability—issues where the GOP position is already popular—they lost support. Same story w/ Dems and Social Security
Moving to the middle ≠ moving toward voters
Thank you @mattyglesias.bsky.social for writing up some of these results in this morning's @nytimes.com ! I agree with his take-aways: www.nytimes.com/2026/03/16/o...
It is notable that these are the issues where each party’s donors are also most out of step with the electorate (Democratic donors too left-wing on social issues; Republican donors too right-wing on economics), per some work I did w @nmalhotra.bsky.social academic.oup.com/poq/article-...
This pattern lines up with where each party is "out of step." The electoral payoff to moving to the center appears biggest on the issues where each party has drifted most from public opinion.
For Republicans, the biggest gains were on economic issues like healthcare and minimum wage.
See Appendix for policy item wording.
For Democrats, the biggest gains from moving to the middle came on social/cultural issues like affirmative action (+4.5pp) and LGBTQ issues—where the party's standard positions are furthest from the median voter.
To illustrate this, we ran a conjoint experiment (N=6,000 voter-file-matched registered voters) with hypothetical 2028 presidential candidates.
Each candidate had randomized positions on 3 of 29 policy issues—either their party's standard position or an "elite middle" alternative.
This suggests moderation is *more important* than many argue: a candidate/party will need to moderate on *lots* of issues where its positions aren’t popular in order to win more votes.
The small average effect of moving to the middle thus isn't evidence voters ignore issues. It's *what you'd expect* if voters care about issues but their preferences don't all sit neatly between the two parties.
Small aggregate effects are consistent with lots of issue voting.
The same can lead the net effects of moderation on an issue to be small: cross-pressured voters who liked you on that issue but preferred the other party on others may flip against you — even if you win over slightly more voters who now prefer you overall
This is due to multidimensionality.
Moving toward the other party on one issue can cause cross-pressured voters—who liked you on that issue but preferred the other party on others—to flip against you. You win some voters but lose others.
Here's why those matter
If a party's position on an issue is already popular, moving toward the other party means *not* moving towards most voters - & might mean moving *away* from most voters - on that issue
"moderation" on that issue won't win many or should actually cost votes
We think this interpretation is wrong, for two reasons:
1. "moving to the middle" is usually measured as moving toward the other party -- not toward the median voter. These aren't the same thing.
2. voters aren't one-dimensional. They're cross-pressured across issues.
The standard story: empirical research finds that candidates whose positions move toward the other party's gain fewer votes than classic theories expect.
This has led scholars and pundits to argue voters mostly don't vote on issues -- they vote on party loyalty, identity, vibes.
New short paper w @jkalla.bsky.social !
Candidates gain from moderation, but less than many theories expect.
Many conclude voters must not care about issues.
This is wrong. Small *average* effects mask large effects on specific issues & are consistent with widespread issue-based voting 🧵