Why Not Ranked Choice Voting?¶

A Position Paper of the Congressional Elections Modernization Act (CEMA)¶

Prepared by Albert Ramos for The American Policy Architecture Institute

Among people who want to reform American elections, the diagnosis is close to consensus. Winner-take-all, single-member districts distort representation, suppress competition, and entrench a two-party system that leaves millions of voters effectively unrepresented. The result on the wall rarely tracks how people actually voted. The Congressional Elections Modernization Act (CEMA) accepts that diagnosis in full, and it shares the goal most reformers are reaching for: a House whose composition reflects the electorate that chose it, filled by candidates who had to earn broad support rather than survive a narrow contest. The disagreement this paper takes up is not whether to fix the status quo. It is which mechanism does the fixing.

That disagreement is narrower than it first appears, and locating it precisely is the whole task. At any given district size, the competing proportional methods deliver comparable proportionality -- the magnitude does that work, not the counting rule. So the real question is everything else: how the ballot behaves in the voter's hands, whether votes are discarded before they are counted, whether the count can be gamed, and whether the method can stand on its own or needs another institution propping it up. On those dimensions the methods diverge sharply, and the divergence traces to a single piece of machinery.

This paper makes its case against one defendant: the sequential-elimination engine. The most prominent ranked reforms in America -- single-winner Ranked Choice Voting and its multi-winner proportional extension -- both run on that one engine, and the engine is the source of the trouble in both. It misbehaves in single-winner races, and bolting proportional representation on top of it does not cure the misbehavior; the proportional method inherits every flaw of the single-winner one and adds a new failure of its own. The recurring pathologies catalogued in the pages that follow -- center squeeze, ballot exhaustion, non-monotonicity, surplus-transfer arbitrariness, the swelling ballot -- are not a list of separate complaints. They are what one mechanism does, seen from several angles. CEMA rejects both the single-winner and the proportional ranked methods for the same reason, because the defect is common to both, and it adopts a method built on a different engine entirely: scoring rather than eliminating.

The comparison is not to a strawman. Two live federal bills carry the ranked approach into Congress right now, one for each form. The Ranked Choice Voting Act (H.R. 6589 / S. 3425, 119th Congress) would mandate single-winner ranked choice voting for every House and Senate seat. The Fair Representation Act (H.R. 4632, 119th Congress) would mandate the multi-winner proportional form for the House -- the only live federal proportional representation bill, carried by its lead sponsor in five consecutive Congresses since 2017. CEMA, the Ranked Choice Voting Act, and the Fair Representation Act all ground themselves in the same constitutional authority, the Elections Clause of Article I, Section 4. So this is not a quarrel about whether Congress may set the manner of congressional elections. Everyone here agrees it may. It is a quarrel about which manner it should choose -- and that makes the comparison clean, because the methods can be judged on their merits with the constitutional question set aside.

The case that follows is not that ranked choice voting is worse than what we have now. It is plainly better than winner-take-all plurality; that much the reform consensus has right. The case is narrower, and it turns on the engine. A method built on sequential elimination carries structural costs that a method built on scoring does not, and those costs fall hardest exactly where a proportional, primary-free system most needs the method to be strong. The argument is in the machinery, and that is where it begins.

What These Methods Are¶

Before the critique, the mechanics -- and the vocabulary, which is genuinely tangled. Terms are defined here as they arise, in bold at first use, and collected in a glossary at the end.

Ranked Choice Voting (RCV) is the American umbrella term for any system that asks voters to rank candidates in order of preference on a single ballot. It names the ballot format, not a specific counting rule, which is the first source of confusion: two quite different tabulation methods both wear the "ranked choice" label depending on how many seats are being filled.

When one seat is at stake, the method is Instant Runoff Voting (IRV) -- the single-winner form. Voters rank the candidates; if no one holds a majority of first preferences, the candidate with the fewest first preferences is eliminated and those ballots transfer to their next-ranked choice; the process repeats until someone has a majority of the ballots still in play. This is the version most Americans have actually used, in Alaska, Maine, New York City, and dozens of municipalities, and it is the version the Ranked Choice Voting Act (H.R. 6589) would install in every congressional race.

When multiple seats are at stake, the method is the Single Transferable Vote (STV), which American reform organizations have rebranded Proportional Ranked Choice Voting (Proportional RCV) to connect it to the single-winner version voters may already recognize. Voters rank candidates the same way. A quota -- the share of the vote that guarantees a seat -- is calculated; any candidate who reaches it is elected, and the surplus beyond the quota transfers to those ballots' next preferences at a fractional value. When no one meets the quota, the lowest candidate is eliminated and their ballots transfer onward. Electing, transferring surpluses, and eliminating continue until every seat is filled. This is the method the Fair Representation Act would mandate for the House. This paper uses "Proportional RCV" as the advocates and the bill do, reserving "STV" for the technical and international record where that is the precise term.

Here is the fact that unifies the two and the reason a single paper can take on both: they are the same engine. STV reduces to IRV when there is only one seat to fill -- the identical fewest-first-preferences elimination rule, minus the surplus transfers that only matter when seats can be over-filled. Ireland, which uses the method for everything, does not even distinguish them in law: its constitution describes both its single-winner presidential elections and its multi-winner parliamentary elections as "proportional representation by means of the single transferable vote." The ballot is the same, the elimination rule is the same; only the number of winners changes. Whatever the elimination engine does to a single-winner race, it brings with it into the multi-winner one.

A Note on History: The Weight of a Century¶

A brief origin story for each method clarifies a comparison that the marketing usually obscures.

The transferable-vote idea came first. Thomas Wright Hill proposed it in 1819; Carl Andrae put a version into use in Denmark in 1855; and the British barrister Thomas Hare developed the canonical multi-winner form in 1857. STV entered sustained national governance in Ireland and Malta in 1921, and the Australian Senate adopted it in 1948. By any reasonable measure the multi-winner method has roughly a century of continuous use behind it.

Single-winner IRV came after its multi-winner parent, not before. William Robert Ware -- the founder of the architecture schools at MIT and Columbia -- observed around 1871 that Hare's multi-winner system could be applied to a single-seat race, and the method was first used in government in Western Australia in 1908, then for the Australian House of Representatives from 1918. This inverts the intuition most Americans carry. The popular framing treats single-winner RCV as the original and the proportional form as the exotic extension; mechanically and historically it is the reverse. The proportional method is the parent. IRV is the special case someone noticed could be carved out of it. When advocates describe Proportional RCV as the natural next step beyond single-winner RCV, they have the genealogy backwards: the "next step" is where the method started.

STAR Voting is the new arrival. It was invented at the Equal Vote Conference at the University of Oregon in October 2014, when reformers from the rival ranked and rated camps converged on a hybrid: score the candidates, then hold an automatic runoff between the top two. Its proportional form, Proportional STAR -- the method CEMA adopts for multi-member districts -- is newer still, settled by the Equal Vote Coalition's multi-year research committee around 2020. So the comparison is between methods with a century of history and methods with roughly a decade. That gap is real, and it cuts both ways.

The long operational history of the ranked methods is genuinely valuable, and this paper will not wave it away. A century of use means the failure modes are known rather than speculative. We are not guessing how the elimination engine behaves under stress; we have watched it. But that same record carries two liabilities that the "tested for a hundred years" framing leaves out. First, the way we know the engine's pathologies are real and recurring -- rather than teething troubles of a young reform -- is precisely that a century of use has surfaced them again and again; time has been the diagnostic, and the diagnosis is not flattering. Second, that long history coincided with remarkably little methodological competition, especially in the United States. For most of the century the ranked ballot was effectively the only alternative to plurality on the table, so its persistence reflects path dependence and first-mover advocacy as much as demonstrated superiority. It was adopted, and re-adopted, in an era that had no tools to evaluate it against alternatives that had not yet been invented.

STAR's short record is, by contrast, a real limitation, and this paper concedes it plainly rather than spinning it. But the thinness is the leading edge of something, not simply a deficit -- and what it is the edge of is a question this paper returns to at the end, once the structural case has been made. For now it is enough to mark the asymmetry: the older methods were implemented on intuition and advocacy, before the analytical tools to characterize a voting method existed; the newer one was designed with those tools in hand. That does not settle the comparison. It does mean the comparison need no longer be settled by waiting a century to see what breaks.

How the Engine Fails in Single-Winner Races¶

The case against the ranked methods starts where the engine is simplest to see: the single-winner race, the form the Ranked Choice Voting Act would impose on every congressional seat. The failures below are not growing pains of a young reform or artifacts of bad implementation. They are what the eliminate-the-weakest rule does, and a century of use is how we know it.

Center Squeeze¶

The marquee failure -- the one in lights, the one that has driven a wave of legislation against the reform -- is the center squeeze. A broadly acceptable candidate, one who would beat every rival in a head-to-head contest, can be eliminated early precisely because that broad acceptability is spread across voters' second and third preferences rather than concentrated as first choices. The engine cannot see breadth. It sees only first-preference counts, and in each round it cuts the candidate holding the fewest -- even when that candidate is the one most voters could live with. Over time the effect compounds beyond any single bad result: knowing the center is the dangerous place to stand, broadly appealing candidates are discouraged from running at all, and the method pulls toward polarization by design.

This is not a textbook hypothetical. In the 2022 Alaska special election for the U.S. House, Nick Begich -- the candidate who, the ballot data showed, would have defeated either opponent in a direct race, the Condorcet winner -- was eliminated in the first round for holding the fewest first preferences, handing the seat to an outcome a majority would have rejected head-to-head. The 2009 Burlington, Vermont mayoral race ran the same way: the Condorcet winner, preferred by a majority over each rival, finished out of the money because first-preference support sat elsewhere. These are the documented signatures of the elimination engine in single-winner use, and they are exactly what the social-choice literature predicted for it decades before Alaska or Burlington.

Ballot Exhaustion: The Quiet Problem¶

Center squeeze is the famous failure. The more pervasive one is quieter, less discussed by advocates, and in its cumulative effect arguably the larger problem: ballot exhaustion. A ranked ballot exhausts when every candidate it ranked has been eliminated; at that point it exits the count and influences nothing further. The "majority" the eventual winner is said to hold is then computed only from the ballots still active at the end -- not from the ballots actually cast. A winner can claim a majority while a substantial share of voters have already dropped out of the tally entirely.

This is not a rare edge case. Burnett and Kogan's study of four American IRV elections found inactive-ballot rates between 9.6 percent and 27.1 percent -- in the worst case, more than a quarter of voters whose ballots were spent before the decisive round. The reason advocates rarely foreground exhaustion is that it does not produce a dramatic villain the way center squeeze does; it produces a slow leak. Ballots quietly fall out of the count, the reported majority is computed off a shrinking base, and nothing in the result announces that it happened. The problem is structural to elimination: a voter realistically ranks only a handful of candidates, so the more candidates in the race, the larger the share each voter leaves unranked, and the more ballots run dry before the seats are settled. Exhaustion is the failure that scales with the size of the field -- a fact that becomes decisive when the method is moved into multi-member districts, as the next section shows.

Non-Monotonicity¶

The third failure is the strangest. The elimination engine fails monotonicity: the property that ranking a candidate higher should never cause them to lose, and ranking them lower should never cause them to win. Under IRV it can. By changing which candidate is eliminated in an early round, a bloc of voters moving their favorite up can reshuffle the sequence so that the favorite is eliminated where they previously survived. A system that can respond backwards to the very preferences it is counting has a coherence problem at its core, and it is theoretically the gravest charge against the engine.

Honesty requires scoping it correctly: in practice it appears to be rare. Studies of real and simulated elections put monotonicity failures in the low single-digit-to-teens percentage of competitive multi-candidate races, not the majority. It is named here not because it strikes often but because it strikes at all -- a method whose result can move in the opposite direction from the voters who caused the movement is doing something a voter cannot be asked to reason about, and it is a property of the elimination rule itself, not of any particular implementation.

The Count Cannot Be Done Locally¶

The last single-winner failure is administrative, and it is rarely mentioned at all. Single-winner IRV is not precinct-summable. Under plurality, each precinct counts its own ballots, reports a tally, and the jurisdiction adds the numbers up; the count is local, transparent, and auditable at the level where the votes were cast. The elimination engine breaks that. Because you cannot know which candidate has the fewest first preferences -- and therefore who is eliminated in round one -- without first aggregating every ballot in the jurisdiction, the actual tabulation has to be done centrally, on the assembled record of how every individual ballot was marked. The rounds are sequential and jurisdiction-wide; a precinct cannot resolve them on its own.

This is a genuine cost in transparency and administrability that has nothing to do with center squeeze or exhaustion -- a separate strike against the same engine, in its simplest one-seat form. It matters here because it sets up a clean contrast: single-winner STAR, as the affirmative case will show, keeps the local count that the elimination engine destroys. The engine surrenders precinct summability the moment it is adopted, even for a single seat.

Four failures, then, in the single-winner case alone -- center squeeze, exhaustion, non-monotonicity, and the loss of local counting -- and the Ranked Choice Voting Act would write all four into every congressional election. Each is a direct consequence of resolving an election by sequential elimination. None is fixed by adding seats. The next section shows what happens when the same engine is asked to do proportional work.

What Proportional STAR Does¶

Before turning the engine loose on multiple seats, the alternative needs to be on the table, because the proportional critique is sharpest when the reader can see what the other method does in the same situation.

Proportional STAR is the proportional application of STAR voting to multi-member districts -- the same scoring the method uses for single-winner races, paired with a proportional counting rule. The ballot is identical to the single-winner one: the voter scores each candidate from zero to five stars, scoring as many or as few as they wish, and leaving a candidate unscored simply counts as zero. Nothing about the voter's task changes between a one-seat race and a seven-seat one -- the same rating gesture, the same ballot, only more names in the column.

Seats are filled in rounds. Each seat costs one quota of ballot weight -- in this setting the Hare quota, the total number of voters divided by the number of seats. In each round the highest-scoring candidate is elected. Then one quota's worth of ballot weight is deducted from that winner's supporters, strongest first: voters are sorted by the score they gave the winner, from highest to lowest, and weight is consumed from the top of that list downward until one full quota has been spent. Voters who scored the winner highest are spent first; if the quota boundary falls inside a tier of voters who all gave the same score, that tier shares the remaining cost equally; voters below the boundary keep their full weight untouched. The next seat is then decided on the weight that remains, and the process repeats until every seat is filled. The result is a delegation whose composition tracks the distribution of support across the district, produced without party lists and without closed slates. A voter who asks which member represents them has a direct answer: the one they scored highest among those elected.

That reweighting rule -- the specific procedure for deducting weight after each win -- is the Allocated Score algorithm. The distinction matters for precision in what follows: Proportional STAR is the method, the thing a voter experiences and a designer selects; Allocated Score is the tabulation algorithm, the deterministic formula that delivers the proportional result behind the curtain. Claims about how the method behaves -- that it is monotonic, that it resists center squeeze, that it never exhausts a ballot -- are properties of Proportional STAR. The weight-deduction formula itself is Allocated Score. The two are not interchangeable, and keeping them distinct is what lets the comparison stay exact.

One feature carries most of the argument, so it is worth naming at the outset: Proportional STAR's demand on the voter does not grow as the candidate field grows. Scoring forty candidates is the same kind of act as scoring four -- rate the ones you have a view on, ignore the rest. There is no ordering to construct, no rank that must be left blank at a cost, no ballot that sprouts a new column with every entrant. That single property is the hinge on which the rest of this paper turns, because it is exactly the property the ranked ballot lacks.

The Engine Under Proportionality: Inherited and Amplified¶

Here is the move at the center of the paper. Proportional RCV is the elimination engine with quotas and surplus transfers bolted on. Adding those parts does not repair the single-winner defects -- it carries them into multi-member races intact, and then adds one more failure that only appears when seats are multiplied. Proportionality does not cure the engine. It inherits it.

The single-winner failures all survive the transition. Center squeeze does not vanish in a multi-seat district; it relocates. A broadly acceptable candidate can still be eliminated for thin first-preference support before their breadth is counted, and the contest for the final seat in any district -- once the early seats are filled and the field thins -- collapses into a single-winner IRV race with the full single-winner vulnerability restored. Ballot exhaustion does not vanish either; it gets worse, for reasons the next paragraph makes central. Non-monotonicity persists, because the elimination rule that produces it is still running. The loss of local counting persists and deepens: Proportional RCV requires central tabulation for the same reason single-winner IRV does, compounded by the surplus transfers. Every structural cost catalogued in the single-winner section reappears here. None was solved by adding seats.

Surplus-transfer arbitrariness is inherited and amplified. This one is specific to the multi-winner case, where it has room to operate. When a candidate exceeds the quota, the surplus must be redistributed -- and the rule chosen for doing so is a genuine choice with no uniquely correct answer. Random selection, the Gregory method, Meek's method, Warren's method: competing, respectable implementations that can hand the same set of ballots to different winners. The Fair Representation Act specifies a fractional Gregory-style transfer truncated to four decimal places, a defensible choice, but a choice nonetheless -- a hidden parameter, set by drafters, outcome-determinative in close contests and invisible to essentially every voter. The result depends not only on how people voted but on which surplus-transfer rule the statute happened to pick.

And then the new failure, the one that only the multi-winner case reveals: field-size fragility. Recall the property established back in the single-winner section -- the elimination engine degrades as the candidate field grows. More candidates mean more elimination rounds, more preferences a voter must produce and a ballot must hold, more rounds in which a broadly liked candidate can be cut before their support is counted, and more ballots that run dry before the seats are filled. In a single-winner race the field is naturally limited. But multi-member districts produce large fields by their nature -- that is what multi-member means, more seats drawing more candidates. So the method is at its structurally weakest precisely where it is meant to be used. The ranked proportional ballot has a narrow workable band: enough candidates to make a proportional result possible, few enough that voters can actually rank them and the ballot stays legible. Push past the top of that band and exhaustion climbs, the ballot swells, and the count strains. The method needs the field held inside the band.

The international record confirms that the winnowing is load-bearing, not optional. Ireland sustains STV through disciplined party nomination -- parties carefully limit how many candidates they run per constituency -- which works, but concentrates gatekeeping in party organizations, the opposite of the candidate-centered promise Proportional RCV advocates make in America, and a mechanism with no analogue in nonpartisan elections. Australia shows what happens when the field is not disciplined: its STV Senate saw ballot papers swell until the 2013 election averaged more than thirty party columns per state, the notorious oversized ballot driven by group-voting-ticket harvesting. Australia reformed hard in 2016, abolishing group voting tickets, and the gaming abated -- but the reforms could not abolish the underlying pressure. The 2025 Senate still fielded hundreds of candidates for a few dozen vacancies, and voters are instructed to number a handful of boxes rather than rank the field. The ballot itself encodes the concession that the full field is too large to order.

What Field-Size Fragility Requires, and Why CEMA Cannot Supply It¶

A method that needs the field held inside a narrow band needs some institution upstream of the general election to do the holding -- to decide how many candidates reach the voter and trim the rest. In the American system that institution is the primary. The Fair Representation Act's own machinery presumes it: the bill's field-management provisions, sorting how many candidates a party may nominate and how many a nonpartisan contest must advance, only cohere if a winnowing election exists upstream. Strip the primary away and nothing guarantees the band; the ballot is free to grow until it breaks.

This is the precise point at which the ranked proportional method becomes incompatible with CEMA specifically -- not because the Fair Representation Act drafted it poorly, but because the primary is the institution CEMA does away with. CEMA consolidates congressional selection into a single Unified General Election and retires the state-administered partisan primary, because the harms primaries carry are the harms CEMA exists to end: chronically low and unrepresentative turnout that lets a small slice of the electorate set the November menu, the exclusion of independents from the consequential first round, the doubled trip to the polls, and the public subsidy of a private nomination contest. A reform that retires the primary to cure those harms cannot then adopt a counting method that quietly needs the primary to function. It would be reintroducing through the back door the institution it removed through the front. The fuller case for why a primary-free congressional system is not only possible but preferable is made in the companion paper, No Primaries? No Problem!; what matters here is the structural lock. The field-size fragility of the elimination engine is one inherited defect among several -- but it is the one that, in a primary-free system, is disqualifying rather than merely costly.

Proportional STAR carries no such dependency, because the property that creates it is absent. Its demand on the voter does not grow with the field; a score ballot of forty candidates is the same task as a score ballot of four. There is no ordering to construct, no rank left blank at a cost, no exhaustion when preferences run out. The ballot tolerates a large field gracefully, so no upstream winnowing is needed, so the primary can be retired with nothing left depending on it. The large field that breaks the elimination engine is, to a score ballot, merely a longer ballot.

The Strongest Case for Ranked Choice Voting¶

A position paper earns its conclusion only by meeting the best version of the opposing argument, not a caricature. Ranked choice voting has real strengths, and several are strong enough that CEMA adopts the same goals.

First, it is candidate-centered rather than party-centered. Voters choose among individuals, not party lists -- a critical feature in the American context, where a large share of elections are nonpartisan and closed party-list systems are politically untenable. This is a genuine advantage over European-style list proportional representation, and advocates are right to insist on it. It is also one Proportional STAR shares in full: it, too, is candidate-centered and list-free.

Second, in the proportional form, it delivers proportional outcomes. Any faction commanding a quota's worth of support can elect a representative. This is the core promise of proportional representation, and the method keeps it -- but, as established earlier, the proportionality comes from the district magnitude, not from the ranked ballot. At the same magnitude, Proportional STAR delivers a comparably proportional result. Proportionality is not the dimension on which these methods differ.

Third, the ranked methods have a track record in functioning democracies that no score-based method can yet match. Ireland and Malta have governed under STV since 1921; the Australian Senate since 1948. These states have held peaceful transfers of power under broadly representative parliaments for a century. This is not a trivial argument, and this paper meets it head-on rather than waving it away: it is the single strongest empirical point in the ranked methods' favor, and the discussion of tradeoffs below concedes it plainly.

Fourth, ranked ballots capture real preference information -- ordering candidates is more expressive than a single categorical choice, and it lets a voter name fallbacks behind a true first choice. This too is real, and it is a strength Proportional STAR matches and arguably exceeds, since a score ballot captures not only order but intensity.

These strengths are genuine, and CEMA grants them. But notice what they are strengths of. Candidate-centeredness belongs to any list-free design. Proportional outcomes belong to district magnitude. The century-long record belongs to STV's history. None of these is a property of the sequential-elimination engine that does the actual counting. Strip away what the ranked methods share with their alternatives, and one question remains: does the elimination engine at their core carry structural costs that a different engine avoids? The single-winner and proportional sections have answered yes. The affirmative case is what the alternative engine looks like.

The Affirmative Case: Scoring Instead of Eliminating¶

The positive case for STAR is not the list of the elimination engine's costs reversed. It is a different engine with a different defining property, and the advantages are what that one property looks like from several angles. STAR and Proportional STAR do not eliminate. They score -- each candidate judged on their own merits, independently of the others -- and from that single difference everything follows.

The defining property is the one named at the start: the voter's task does not grow as the field grows. Because each candidate is scored independently rather than placed in an order, there is no elimination sequence to misbehave. So the method is monotonic -- raising a candidate's score can never cost them a seat, because there is no elimination order for the higher score to perturb. There is no center squeeze -- breadth of support registers directly as score rather than waiting in lower rankings to be counted, so the consensus candidate the engine would cut is the candidate the score rewards; a contender rated three stars by 80 percent of voters outscores one rated five by a narrow 30 percent and zero by the rest. There is no exhaustion -- a score ballot expresses a view on every candidate at once and stays fully operative through every allocation round; weight is spent as preferred candidates win, but the ballot never falls out of the count. And the proportional reweighting is deterministic -- the Allocated Score algorithm spends one quota per seat by a fixed rule, so the same ballots yield the same winners every time, with none of the surplus-transfer arbitrariness that competing implementations introduce. These are not four separate improvements. They are four consequences of scoring rather than eliminating, which is the same as saying the method does not buckle when the field is large.

That single property is what lets Proportional STAR do the structural work CEMA requires. A reform that holds one unwinnowed general election needs a proportional method that can stand in a crowded field without an upstream institution thinning it first. The elimination engine cannot, for every reason the preceding sections gave. Scoring can, because the large field that breaks elimination is, to a score ballot, merely a longer ballot.

The Ballot a Voter Already Knows How to Mark¶

One affirmative argument stands on its own rather than as a consequence of field-tolerance, and it is among the strongest the method has, so it deserves its own footing: the STAR ballot asks voters to do something most of them already do fluently.

The zero-to-five rating is among the most rehearsed quantitative judgments in modern American life. People rate drivers, restaurants, products, and businesses on five-star scales constantly, across every level of education and civic engagement, without anyone explaining what three stars means relative to five. A STAR ballot transplants that familiar gesture onto the ballot unchanged -- and it is the same ballot for single-winner Senate races and multi-winner House races, so a voter learns one action for every congressional contest they will ever face. Set against that the task a ranked ballot demands in a large field: place twenty candidates in strict order down to a fourteenth and fifteenth choice the voter has no real opinion about, on a grid where every added candidate is another row and another way to spoil the ballot.

This is not a matter of convenience, and it is where the equity stakes are highest, because the errors a complex ballot produces are not distributed evenly. Research on ranked ballots -- Pettigrew and Radley's multi-state analysis, Neely and McDaniel's study of San Francisco, Cormack's New York City work -- converges on a consistent finding: ballot-marking errors and incomplete rankings fall most heavily on the voters already most likely to be underrepresented, including older voters, lower-income voters, voters in non-English-dominant households, and voters of color. Overvoting -- the error most likely to void a ballot outright -- rises with the complexity of the grid. A harder ballot does not impose its difficulty equally; it filters hardest on the already marginalized, and the larger the field the harder it filters. Choosing the more usable ballot is therefore not a usability footnote. It is a question of whose expressed intent survives the act of voting. And it is the same property -- independence of judgment, candidate by candidate -- that makes the ballot both easy to mark and immune to the field-size collapse the elimination engine suffers. The two virtues are one virtue.

The Local Count, Recovered¶

The single-winner section closed on an administrative loss: the elimination engine cannot be counted at the precinct. Scoring recovers it, at least where it can. Single-winner STAR is precinct-summable. A precinct can total the scores for each candidate and report its own head-to-head runoff tally locally; the jurisdiction adds the numbers up. Same ballot, same local-count-and-aggregate transparency as an ordinary plurality election -- and a clean advantage over single-winner IRV, which surrenders local counting the moment it is adopted. For the Senate and single-seat House races, CEMA keeps the count where the votes are cast.

Honesty requires the other half of the ledger, and it appears in the tradeoffs below: in the multi-winner case, Proportional STAR is not summable either, because ballot weights change after each seat is filled. So the scorecard is precise rather than sweeping. Single-winner: STAR summable, IRV not -- a clean win. Multi-winner: neither summable -- a wash, and CEMA claims no advantage it does not have. That asymmetry is more defensible than a blanket "STAR is summable" would be, and it maps exactly onto the engine thesis: summability is a property of how you count, and elimination destroys it even in the one-seat case, while scoring preserves it until the proportional reweighting -- not the scoring -- takes it away.

Accepted Tradeoffs¶

No method escapes every flaw -- Gibbard's theorem guarantees every reasonable system can be strategically manipulated under some conditions -- and honesty requires meeting Proportional STAR's costs with the same directness applied to the ranked methods'.

Later-no-harm failure. Under STAR, honestly scoring a second choice can help that candidate defeat your first in the automatic runoff. STV satisfies later-no-harm; STAR does not. But this is not a free-standing defect -- it is the exact price of solving center squeeze, and the two cannot both be bought. A method satisfies later-no-harm only by ignoring lower preferences in early rounds, which is precisely the mechanism that produces center squeeze. One can protect the consensus candidate or guarantee later-no-harm, not both. CEMA accepts the later-no-harm failure deliberately, because the center squeeze it buys out is empirically the more damaging of the two.

Strategic burial. A voter can score an acceptable rival at zero to keep them out of the runoff. The incentive is real, but its practical bite is limited: burial requires accurate knowledge of who actually threatens your favorite, and it backfires if the buried candidate would have beaten someone you like even less. The pivotal-voter modeling of Wolk, Quinn, and Ogren found dishonest strategies under STAR strongly disincentivized, with the residual confined to mild honest inflation that preserves the voter's true order.

Centralized tabulation in multi-winner races. As noted just above, single-winner STAR is precinct-summable but Proportional STAR is not, because ballot weights change after each seat is filled, requiring central aggregation of individual ballot records. This is the same requirement STV imposes -- CEMA claims no tabulation advantage in the multi-winner case, and says so.

Reweighting legibility. A voter whose ballot weight drops after a round may not immediately grasp why, even though the Allocated Score formula is public, deterministic, and fully auditable. STV's fractional surplus transfers are arguably even less transparent, but that does not make reweighting self-evident. CEMA addresses this through the Electoral Science Office's public-education mandate and an auditable-tabulation requirement, but voter comprehension will take sustained institutional work, and that is a real cost.

No national track record. The largest objection is not a flaw in the method but a gap in its resume: no nation has elected a parliament by Proportional STAR, while Ireland has run STV since 1921. This is the strongest point in the ranked methods' favor, and it is conceded plainly. Two things bound it. First, the ballot is not novel even if the count is: the zero-to-five scale is globally ubiquitous, and STAR has run in binding elections -- the Independent Party of Oregon's 2020 statewide primary with more than a million eligible participants, the Python Software Foundation's Steering Council, and, most tellingly, the Los Angeles chapter of the Democratic Socialists of America, which ran both STV and Proportional STAR for its leadership elections and deliberately moved to the latter, finding it delivered the proportionality they valued while dropping the polarizing elimination rounds. A constituency predisposed toward the ranked tradition, having run both, chose the scored one. Second, and more fundamentally, the choice no longer has to be settled by waiting a century to see what happens. STV entered national governance with no tools to predict its behavior; Proportional STAR enters a field that can characterize a method's properties -- through satisfaction-efficiency simulation, incentive-distribution analysis, and formal criterion testing -- before a single ballot is cast. Operational experience and structural soundness are both legitimate criteria. The ranked methods lead on the first. This paper's argument is that the second is knowable in advance, and that it points the other way.

The Advocacy Landscape: A Note to Fellow Reformers¶

The strongest argument for the ranked methods is not structural at all -- it is that they can actually pass, and no competing method has anything like their political infrastructure. FairVote has built, over decades of campaigns, a capacity no rival possesses; the Fair Representation Act has sponsors and a five-Congress history; and public familiarity with single-winner RCV gives the proportional form a ready-made on-ramp. On immediate political viability the ranked methods are plainly further along, and this paper does not pretend otherwise.

But the history section left a thread hanging, and this is where it is paid off. The thinness of STAR's record was called the leading edge of something. Here is what it is the edge of -- and the argument runs, for once, toward the reform community rather than against it.

Consider the position the ranked-method movement is now actually in. As of early 2026, nineteen states have enacted laws prohibiting ranked choice voting, most of them preemptively, in states that had never used it -- a defensive wave touched off in no small part by the visible failures the century of use surfaced, the Alaskas and the Burlingtons. And the drafting of those bans matters enormously. Many were written against the ranked ballot itself, not against IRV tabulation specifically: statutes barring any "ranked-choice voting or instant runoff voting method" in any election. A ban written that way forecloses the entire ranked family in one stroke -- single-winner IRV and multi-winner STV alike. The Fair Representation Act, arriving in such a state, would hit a wall that predates it.

That leaves ranked-method advocates three doors, and two of them are bad. They can litigate -- argue a ban violates a state constitution -- which is slow, expensive, and a long shot. They can wait for the political winds to turn a statehouse, which surrenders the clock to the other side. Or they can notice the door the bans left open. Those statutes target the ranked ballot. STAR uses a score ballot. As currently drafted, most of these prohibitions do not reach a score method at all -- though candor requires the caveat that nothing stops a legislature from extending a ban once a score method became a live threat, and some future statute surely will. An advocate sitting on a defeated RCV campaign in a banned state could set down the ranked ballot, pick up the score ballot, and get a fresh start, carrying forward every hard-won lesson of RCV organizing without the legal deadweight.

But the deeper lesson is not "switch methods and push STAR through the legislature instead." That invites the same preemptive reflex; the bans show incumbents will foreclose a ballot type they have never even used. The lesson is about sequence, and it is a lesson the reform community has had to learn the hard way. Voting reform is unlike most reform, because the thing being reformed is the mechanism that selects the people whose approval the reform requires. The incumbents have both the motive and the power to slam the door first. So the durable path does not aim at government first. It establishes the method where collective choice already happens at scale and no statehouse's permission is needed: student governments in high schools and on campuses, professional and member organizations, businesses, party-internal contests. Build a local, visible, satisfied track record -- a place to point to and say "it works, people like it, here is the data" -- and only then approach government, arriving with proof of concept rather than a theoretical pitch.

This is exactly the path STAR's record already traces. Astoria, Oregon adopted it for municipal use; the Independent Party of Oregon ran it for a statewide primary; the Los Angeles DSA and the Python Software Foundation chose it for internal elections; in several of these settings it displaced plurality, approval, or a ranked method that had been tried first. The thinness the history section flagged is not weakness. It is the near edge of the bottom-up strategy that voting reform, uniquely, requires -- the slower route that does not hand incumbents an easy preemption. The ranked-choice movement pioneered that patient, build-it-locally organizing in America. The argument here is that the same playbook, run with a score ballot, threads the very wall that has lately been built against the ranked one.

One uncomfortable observation follows, and it belongs in a paper addressed partly to neutral readers rather than only to allies. That foreclosure -- the sense that Proportional RCV simply is the proportional method rather than a proportional method -- has left a real gap in the evidence. No published peer-reviewed study systematically compares STV against score-based proportional alternatives on common criteria. Ogren's 2024 incentive analysis comes closest, including STV in a multi-winner comparison, but it does not evaluate Proportional STAR directly. The comparative claim that the ranked method deserves the movement's reform energy over the structural alternatives has never actually been run head-to-head; the literature is largely a conversation among proponents. This does not make the ranked method the wrong choice. It means something narrower and more uncomfortable: the choice has been treated as settled before the comparison was run. This paper does not ask the reader to take its verdict on faith. It asks that the comparison be run at all.

Conclusion¶

Ranked choice voting is a genuine reform with genuine strengths, and this paper has tried to credit them honestly. Its advocates are right that proportional representation beats winner-take-all, right that a candidate-centered method suits American elections, and right that the ranked methods have governed democracies for a century. The Fair Representation Act is carefully drafted, and its designers saw the ranked ballot's central constraint clearly and built to respect it. The Ranked Choice Voting Act would, in the single-winner case, improve on the plurality system it replaces. None of that is in dispute.

What this paper disputes is structural, and it reduces to one defendant. The objections raised across these pages -- center squeeze, ballot exhaustion, non-monotonicity, the loss of local counting, surplus-transfer arbitrariness, the swelling ballot -- are not a catalog of separate defects to be weighed one against another. They are what a single mechanism looks like from several angles: the sequential-elimination engine. That engine misbehaves in single-winner races, and the Ranked Choice Voting Act would write the misbehavior into every congressional seat. Bolting proportionality on top of it, as the Fair Representation Act does, cures none of those defects -- it inherits every one and adds field-size fragility, the failure that bites hardest exactly where multi-member districts put the method to work. And field-size fragility is the one inherited defect that a primary-free reform cannot absorb, because escaping it requires the very winnowing institution -- the primary -- that CEMA exists to retire. A reform built to retire the primary cannot adopt a counting method that secretly needs it.

Proportional STAR needs no such thing, because it runs on a different engine. It does not eliminate; it scores. Its demand on the voter does not grow with the field, so it stands in a crowded general election without an upstream contest thinning the field first -- which is what makes a single unified general election possible at all. The same property that makes its ballot the most familiar gesture in American civic life makes it immune to the field-size collapse the elimination engine suffers. The two virtues are one virtue, and it is the property the ranked ballot structurally lacks.

So the question is not whether ranked choice voting beats plurality. It does, decisively. The question is whether it is the best method available now that the alternatives can be characterized in advance rather than discovered in office, and now that the ranked ballot itself has run into a wall of preemptive bans the score ballot was not built to trip over. On the structure, on the evidence, and on the politics of getting a durable reform adopted at all, the answer points the same direction. That answer has not been seriously contested, because the comparison was treated as closed before it was opened. This paper is the argument for opening it.

Glossary¶

Allocated Score -- The deterministic tabulation algorithm CEMA uses to count Proportional STAR elections. Fills seats in rounds: elect the highest-scoring candidate, then deduct one Hare quota of ballot weight from that winner's supporters, strongest first, and decide the next seat on the remaining weight. The algorithm, not the method (see Proportional STAR).

Ballot exhaustion -- Under a ranked method, the condition in which every candidate a voter ranked has been eliminated, so the ballot exits the count and influences nothing further. Causes reported "majorities" to be computed off a shrinking base of still-active ballots. Worsens as the candidate field grows.

Center squeeze -- The failure mode in which a broadly acceptable candidate -- one who would win any head-to-head contest -- is eliminated early under a ranked method because their support is spread across lower preferences rather than concentrated as first choices.

Condorcet winner -- The candidate who would defeat every other candidate in a direct one-on-one race. The elimination engine can fail to elect the Condorcet winner (as in Alaska 2022).

Field-size fragility -- The property of the sequential-elimination engine whereby its performance degrades as the candidate field grows. Decisive in multi-member districts, which produce large fields by nature.

Hare quota -- The share of ballots that secures a seat, computed as total voters divided by number of seats. Used by Allocated Score.

Instant Runoff Voting (IRV) -- The single-winner form of ranked choice voting: eliminate the candidate with the fewest first preferences, transfer their ballots to next choices, repeat until one candidate holds a majority of active ballots.

Monotonicity -- The criterion that ranking or scoring a candidate higher must never cause them to lose, and lower must never cause them to win. The elimination engine fails it; scoring methods satisfy it.

Precinct summability -- The property that lets each precinct count its own ballots and report a tally the jurisdiction simply adds up. Plurality and single-winner STAR have it; the elimination engine (IRV and STV) does not.

Proportional RCV (Proportional Ranked Choice Voting) -- The American advocacy name for the multi-winner Single Transferable Vote; the method the Fair Representation Act would mandate for the House.

Proportional STAR -- The proportional, multi-winner application of STAR voting, counted by the Allocated Score algorithm. The method CEMA adopts for multi-member House districts. The method, not the algorithm (see Allocated Score).

Ranked Choice Voting (RCV) -- The American umbrella term for any system using a ranked ballot; names the ballot format, not a specific counting rule.

Single Transferable Vote (STV) -- The multi-winner ranked method: quota-based election plus surplus transfer plus sequential elimination. Reduces to IRV when only one seat is filled.

STAR Voting -- "Score Then Automatic Runoff." A single-winner method: voters score candidates zero to five, the two highest-scoring advance, and the one preferred by more voters wins the automatic runoff.

Surplus-transfer arbitrariness -- Under STV, the dependence of the outcome on which surplus-transfer rule (Gregory, Meek, Warren, random) is chosen, since competing rules can elect different winners from identical ballots.

Works Cited¶

Australian Electoral Commission. "2025 Federal Election: Nominations Overview." Media Release, April 2025. https://www.aec.gov.au/media/2025/04-12.htm.

Burnett, Craig M., and Vladimir Kogan. "Ballot (and Voter) 'Exhaustion' Under Instant Runoff Voting: An Examination of Four Ranked-Choice Elections." Electoral Studies 37 (2015): 41-49.

Cormack, Lindsey. "More Choices, More Problems: Ballot Error and Voter Engagement with Ranked Choice Voting in New York City." Working Paper, 2023.

Equal Vote Coalition. "Proportional Representation." https://www.equal.vote/pr. Accessed June 2026.

Equal Vote Coalition. "STAR Voting Origins." https://www.equal.vote/origins. Accessed June 2026.

Graham-Squire, Adam T., and David McCune. "An Examination of Ranked-Choice Voting in the United States, 2004-2022." Representation 61 (2023): 1-19.

Graham-Squire, Adam, and David McCune. "A Mathematical Analysis of the 2022 Alaska Special Election for US House." arXiv:2209.04764, 2022.

Green, Antony. "Party and Voter Reactions to the Senate's Electoral Changes." Antony Green's Election Blog, 2025. https://antonygreen.com.au/fed2025-party-and-voter-reactions-to-the-senates-electoral-changes/.

Miller, Nicholas R. "Closeness Matters: Monotonicity Failure in IRV Elections with Three Candidates." Public Choice 173, no. 1-2 (October 2017): 91-108.

National Conference of State Legislatures. "Ranked Choice Voting: State Legislation and Prohibitions." Accessed June 2026. https://www.ncsl.org/elections-and-campaigns/ranked-choice-voting.

Neely, Francis, and Jason McDaniel. "Overvoting and the Equality of Voice Under Instant-Runoff Voting in San Francisco." California Journal of Politics and Policy 7, no. 4 (2015): 1-27.

Ogren, Marcus. "Voting Methods Shape Electoral Incentives: Candidate Incentive Distributions." Working Paper, 2024.

Pettigrew, Stephen, and Dylan Radley. "Ballot Marking Errors in Ranked Choice Voting." Working Paper, 2024.

Ramos, Albert E. Congressional Elections Modernization Act, Rev 6.0. The American Policy Architecture Institute, 2026.

United States Congress. Fair Representation Act. H.R. 4632, 119th Congress, 1st Session. Introduced July 2025.

United States Congress. Ranked Choice Voting Act. H.R. 6589 / S. 3425, 119th Congress, 1st Session. 2025.

Wolk, Sara, Jameson Quinn, and Marcus Ogren. "STAR Voting, Equality of Voice, and Voter Satisfaction: Considerations for Voting Method Reform." Constitutional Political Economy 34, no. 3 (2023): 310-334.

Revision history available in the raw file.

📥 Download this document (opens on GitHub -- click the ⬇ download button)