Can anyone beat Cuomo? And is it Zohran?

Zohran Mamdani has made the NYC mayoral primary competitive. But ranked choice voting may shut out the very compromise choices it was supposed to facilitate.

Today’s newsletter is a guest post from Joel Wertheimer. Even if you don’t care about New York, I think you’ll like the post. It contains some nerdy diversions into game theory and ranked choice voting, a system that can introduce more chaotic outcomes in practice than you might expect on paper.

Joel is a civil rights attorney in New York and has advised candidates as a policy consultant. He worked for President Obama from 2015 to 2017 as Associate Staff Secretary. He irregularly writes about data, Tottenham Hotspur and politics at wertwhile.substack.com.

—Nate Silver

P.S. Silver Bulletin is open to guest pitches, and we pay for published posts! To see how to submit, check out the About page. What we’re looking for is posts exactly like this: subjects that are in the core Silver Bulletin wheelhouse from authors who can bring perspective or subject-matter expertise that Eli and I lack.

---

The New York City Democratic mayoral primary election is just two weeks away — and against expectations, we have a real race on our hands. The short story is that former governor Andrew Cuomo, who resigned amid a sexual harassment scandal in 2021, remains the favorite, running in the more conservative lane vacated by the incumbent Eric Adams¹. Cuomo is leading, currently averaging 37 percent of the vote in the first round of ranked choice voting, and Polymarket traders have him with about a 3-in-4 chance of winning.

But as Nate discussed in last week’s SBSQ, it has become a two-man race, with Cuomo fighting off an increasingly vigorous challenge from Zohran Mamdani, a state representative and member of the Democratic Socialists of America. Mamdani is currently in second, averaging 27 percent of the first-round vote. Cuomo’s campaign and a pro-Mamdani super PAC released dueling polls on Monday showing broadly similar contours of the race. The pro-Cuomo poll had the former Governor winning 56-44 in the final two, while the Mamdani poll showed Cuomo ahead by just two points. An average of the two polls may provide the most realistic benchmark, with Cuomo up 7 points in the final round, a surmountable deficit given the uncertainties in polling for a race like this one and what’s been something of a history of presumptive frontrunners in New York choking away their advantage.

Both Cuomo and Mamdani are polarizing candidates, however, Cuomo for his scandals and Mamdani for his progressivism. What about the others? Well, there’s plenty to pick from. Comptroller Brad Lander, City Council Speaker Adrienne Adams, former Comptroller Scott Stringer, and State Senator Zellnor Myrie² are running, in various flavors, in the center lanes ideologically relative to the New York City Democratic electorate. But if you trust Polymarket, they have less than a 1 percent combined chance of winning.

Can ranked choice voting live up to its promises?

New York City instituted ranked choice voting (RCV), also known as instant runoff voting, in its primary elections in 2021. The organization FairVote, which promotes ranked choice voting, describes RCV this way: “Voters have the option to rank candidates in order of preference: first, second, third and so forth. If your first choice doesn’t have a chance to win, your ballot counts for your next choice.” Ranked-choice voting also has its critics, with some calling it too complicated and opaque, favoring high-information voters who rank more candidates³ and understand the (as we’ll see, often complicated) strategic choices it entails.

Support for RCV arrives from a couple of different sources. One is from voters and theorists concerned about third-party “spoiler effects.” Most consequentially, in the 2000 Presidential election, had voters for Ralph Nader been allowed to rank Nader first, with the majority of them listing Al Gore as their second choice, Gore would have been able to procure the majority of votes in Florida to become the 43rd President.⁴ Indeed, in a 2018 election in Maine, where ranked choice voting is used in federal races, the centrist Democrat Jared Golden was behind in the first round against Republican Bruce Poliquin. But when left-leaning independent voters were redistributed to Golden, he won in the instant runoff, making him the U.S. Representative from Maine’s 2nd congressional district.

Ranked choice voting is often also endorsed by less partisan, good-government types who think it will better reflect voter preferences — or hope that it will put more moderates like Golden in office. The elevator pitch is simple enough — it’s a system that’s supposed to produce consensus. But let’s look at how it can play out in practice.

Imagine that in a future presidential race, you have a polarizing right-wing (say, Marjorie Taylor Greene) and left-wing (say, Ilhan Omar) candidate, and then a moderate independent (say former West Virginia Senator Joe Manchin). Polarization being what it is — even most ostensible independents lean strongly toward one party or the other — first-choice preferences might be something like Greene 40, Omar 40, Manchin 20. However, assume that all Greene voters would prefer Manchin to Omar, and likewise, all Omar voters would prefer Manchin to Greene. In a head-to-head matchup, Manchin would defeat both Omar and Greene 60-40 for a landslide win.

This makes Manchin what political scientists call the Condorcet winner: the candidate who would prevail in every head-to-head matchup in the race if there were only two candidates.

Indeed, typically around 99 percent of the time, the winner of ranked choice voting is also the Condorcet winner. But how do those 1 percent cases happen?

Well, it might look something like the Manchin example above. Under RCV, the lowest-ranking candidate in any given round of the automatic runoff is eliminated and his or her votes are redistributed until one candidate gets an outright majority. So in this example, Manchin’s 20 percent in the first round puts him on the chopping block. The election would then come down to whether more of his voters list Greene or Omar as their second choice, adding to their 40 percent from the first round.

Or for a real-world example, we can turn back to New York City, where matters are made even more complicated because voters can list up to five candidates on their ballot. Both Cuomo and Mamdani are quite polarizing, as we’ve said. And yet, one of them is overwhelmingly likely to win.

NYC nearly had a Condorcet violation in 2021, and could again in 2025

Indeed, New York City narrowly avoided a Condorcet failure in its last Democratic mayoral primary, getting very close to a situation where Kathryn Garcia, the moderate⁵ former sanitation commissioner, could have won ranked choice voting against Adams. But she was almost shut out of the final round because the liberal Maya Wiley nearly had more votes when the race was down to three candidates:

Garcia had actually trailed Wiley in the sixth round of RCV, but she picked up enough votes from supporters of Andrew Yang, with whom she had formed a strategic alliance, to leapfrog ahead in Round 7. Nate Cohn of the New York Times demonstrated that while Garcia lost to Adams by just one point, Wiley would have lost by 10 had she been in second place in the penultimate round. Wiley’s voters were much more likely to give their votes to Garcia than Garcia’s voters were to go to Wiley, as a median voter theorem model of the electorate might predict.

To give a quick state of the 2025 race with strictly non-partisan polling, let’s look at an average of ranked choice voting after eliminating everybody but Cuomo, Mamdani, Lander, Adrienne Adams, and Scott Stringer:

The newest partisan polls from Cuomo and Mamdani suggest the Emerson poll — still the most recent independent survey although it’s now nearly three weeks old — might not be far from the mark. What we have not seen is much data from pollsters on what would happen in head-to-head matchups among the various candidates.⁶

I personally reached out to each of these pollsters and asked them to run simulations to let us know how each candidate would fare head-to-head against Cuomo. Only Marist published that result (on their YouTube channel), but it was informative, if not exactly encouraging to voters who would prefer an alternative to the former governor. In Marist’s simulations, Mamdani lost to Cuomo 60-40. But despite being significantly behind Mamdani in the previous round, Lander also would have lost by the same 60-40 margin, while Adrienne Adams was down a statistically similar 62-38. This suggests the presence of an Anybody But Cuomo faction of the electorate, though not one that forms a majority.

Still, in a true head-to-head race, the whole contours of the election might be different. If it were just Cuomo and (say) Lander, Lander might position himself slightly more to the left to ensure turnout from progressives who might otherwise have preferred Mamdani. Furthermore, in real-life RCV, many voters have their ballots spoiled. Ballots can be spoiled in one of two ways: because a voter ranks five candidates that do not include one of the final two, or because she lists fewer than five choices.

And other polling suggests one of the Anybody But Cuomo alternatives might be the Condorcet winner. Politico recently published the results of an internal poll from the progressive Working Families Party that showed three head-to-head matchups between Cuomo, Adrienne Adams, Mamdani, and Lander. Cuomo beat Mamdani by 11, but Lander by just 3, and lost to Adams by 7.

We do not have much information from that poll, and one ought to take all of these internal polls with a grain of salt. Still, the Emerson poll hints at a similar dynamic. I took a look at Cuomo’s favorability numbers among Lander and Mamdani supporters to see how Lander, who trails Mamdani significantly in initial preference, might fare head-to-head against Cuomo. The results were informative: because Mamdani voters are much more likely to go to Lander than the other way around, Lander could do better than Mamdani head-to-head against Cuomo:

Should the compromise candidates coordinate?

In New York City right now, there appear to be three blocks of voters: Cuomo voters, Mamdani voters, and Anybody But Cuomo voters. Cuomo and Mamdani voters do not have much strategic thinking to do, as they are very likely to be in the final two. As such, how they should think about the campaign is not a particularly interesting exercise. Vote for your candidate, try to knock down the other guy, and pound the pavement.

For Anybody But Cuomo voters, however, the question is: is Zohran the best bet to beat Cuomo? And, if not, how does everybody pick one of the candidates to elevate them above the others?

The recent polling suggests that talk of coordination among the other candidates might be too late with Mamdani consolidating his lead in second. On the other hand, both frontrunners have an incentive to overstate Mamdani’s strength in second to shut out alternatives, and there are plenty of vulnerabilities for both frontrunners.

If you are an Anybody But Cuomo voter, then perhaps it’s actually time for some game theory. In 2021, Kathryn Garcia rode the endorsement of the New York Times to become the focal point of anti-Eric Adams coordination and nearly got over the top, especially thanks to her explicit coordination with Yang.

A focal point is a solution people default to without advance communication. Thomas Schelling, one of the fathers of modern game theory, famously asked his students at Yale if they had to meet somebody in New York City tomorrow, without knowing anything else, where would they go and when? The vast majority said Grand Central Station at noon⁷, attempting to mirror one another’s behavior.

The Anybody But Cuomo voters could do the same, although it could be hard to do so without a nudge.

One first step toward coordination could be to get more public polling that asks head-to-head questions pitting Cuomo against Adams, Lander, Mamdani, and even Stringer. I am now more agnostic as to what the polls would show given the recent polling showing Mamdani’s continued growth. But my instinct based on the data we do have is that Adrienne Adams and Brad Lander would gain nearly all of Mamdani’s voters but not the other way around, making them more likely to beat Cuomo if they were in the final two and spoiled ballots were minimized. (The lack of independent polling in this regard has been a massive disservice to New York City voters.)

A second step might be to beg the New York Times to consider endorsements. The Times has, for unclear reasons, opted out of the business of endorsements in its home city, precisely the place where its endorsements have the most value. The highly educated liberal readership of the New York Times and the New York City Democratic primary electorate have substantial overlap, and guidance would likely be useful to the readership in finding a focal point.

Finally, if any of the other candidates are themselves, deep down, Anybody But Cuomo voters, they could act strategically with endorsements at the end. With early voting set to begin on Saturday, there isn’t much time left. But as we’ve seen, RCV isn’t a magic bullet, and it may produce another Democratic mayoral nominee that many voters find hard to stomach.

1. Adams is now running as an independent.

2. For full disclosure: I worked for Andrew Cuomo briefly in 2017 and was quite vocal regarding my views on him before he resigned. I support Zellnor Myrie in this race and have donated to both him and Adrienne Adams.

3. High-information voters may be more likely to learn about all of the candidates, use all the ranking spots available to them, and know about the possible strategic considerations to ensure they are voting for at least one candidate likely to be in the final two. Voter turnout is higher among more educated voters to begin with, and one concern is that a voting system that asks voters to learn even more about candidates exacerbates the burden required to participate.

4. Although Nader voters did not fall as neatly along a left-right spectrum as one might guess: he was also a choice for voters fed up with the establishment in general.

5. Moderate by New York City standards, that is.

6. In 2021, pollsters ran those simulations, showing voters that Kathryn Garcia had a chance to beat Eric Adams but Maya Wiley likely did not. Ultimately, as we saw above, that is exactly what happened.

7. As Nate points out in his book, this was partly because the New Haven Line of the Metro-North Railroad originates from Grand Central. Had Schelling taught at Princeton, Penn Station (where trains to New Jersey originate) might be a more natural choice instead, the focal point that his students assumed others would also pick.