Congressional Elections Modernization Act¶
The Case for STAR and Proportional STAR¶
A voting method is the rule that turns what voters want into the names that take office. Every other choice in the Congressional Elections Modernization Act (CEMA), hereafter "the Act," can be argued on its own terms -- the size of the House, the independence of the Electoral Science Office, the shape of the districts -- but none of them decides who wins. The counting rule does. And the counting rule the Act adopts is not a free-standing preference dropped into the design because it tests well in the abstract. It is the one method the rest of the architecture cannot do without.
This paper makes a single claim and builds everything else on it: the most valuable structural move the Act makes -- ending the state-administered congressional primary -- is only possible with a voting method that needs no primary to function, and STAR is that method. The properties that let STAR carry that architectural load are the same properties that make it the right choice on the merits. It handles a crowded ballot gracefully because it never asks the voter to rank a long field. It guarantees a majority-preferred winner because the runoff is built into the general election rather than outsourced to a primary. Its ballot is the one rating task tens of millions of Americans already perform without instruction. And when it fails -- every method fails somewhere -- it fails in the one way a voter can live with. None of these is a separate selling point. They are four consequences of the same fact: STAR is a method that stands on its own, and a reform that has just removed the primary needs exactly that.
The argument runs as follows. First, what the two methods do, briefly, because the rest depends on the mechanics being clear. Second, the architectural argument itself -- why removing the primary makes a no-winnowing method necessary, and why ranked methods cannot fill that space without quietly reintroducing the thing the Act eliminated. Third, the question every method must answer honestly -- compared to what? -- and the demonstration, on the actual ballots of a real election, that STAR's failures are the ones you would choose. Fourth, the ballot itself, and why a method's first duty is to be usable by the people casting it. Fifth, the multi-seat case: how Proportional STAR matches method to district magnitude. Throughout, the comparison is to the two methods that actually compete for this job -- the status quo of plurality-with-primaries, and the leading reform proposal, Proportional Ranked Choice Voting -- never to an imaginary method that fails at nothing.
A note on what the Act does, and does not, do to primaries¶
Because the whole argument rests on the elimination of the state-administered primary, one objection should be cleared at the outset, before it sits as static under everything that follows: can the Act do that to the parties?
It can, because it does not do what the objection imagines. The Act does not abolish party primaries. Parties retain the full associational right to nominate whomever they choose, by whatever internal process they prefer -- caucus, convention, party-run preference vote -- and to endorse, fund, and campaign for those nominees. What the Act ends is the state's obligation to build, administer, and pay for a private organization's nomination contest using public election machinery. The distinction is between a party's right to associate and select its standard-bearer, which is preserved entirely, and the state's choice to run that selection as a taxpayer-funded election, which is not constitutionally required and which the Act declines to continue. The Act regulates the state-administered election -- the mechanism through which public resources determine who holds public office -- not the activity of parties. Separating the two leaves associational rights intact while freeing the general election to do the work this paper describes. The fuller constitutional treatment, including the distinction from California Democratic Party v. Jones, is set out in the Constitutional Authority Technical Memorandum; for present purposes the point is narrow. The Act makes a structural decision in service of better voter-preference outcomes, and it makes that decision without touching the freedom of association.
What STAR and Proportional STAR Do¶
STAR stands for Score Then Automatic Runoff, and the name is the method. A voter scores each candidate from zero to five stars. In the first round, the scores are totaled and the two highest-scoring candidates advance. In the second round -- the automatic runoff -- each ballot counts as a single vote for whichever of those two finalists the voter scored higher, and the candidate preferred on more ballots wins. The two rounds discipline each other. The scoring round measures the breadth and intensity of support across the whole field, so a candidate cannot win on a narrow, intense faction alone. The runoff round guarantees that the winner is the candidate more voters actually prefer once the choice narrows to two, so a candidate cannot win merely by being inoffensive to everyone and the genuine choice of no one. A pure score election can elect the blandly acceptable; a pure ranked elimination can discard the broadly preferred. STAR uses each round to cover the other's weakness.
One feature of that design carries most of the weight in the sections that follow, so it deserves to be named now: the automatic runoff is a majority check built directly into the general election. It is the structural reason single-winner STAR can be trusted without a primary standing behind it. A traditional primary-then-general system uses the primary to winnow a field and the general to produce a majority between the survivors. STAR internalizes the second job -- the runoff guarantees the final winner is majority-preferred between the top two -- which is precisely what makes the primary's winnowing job dispensable rather than merely unwanted. Hold that thought; the architectural argument turns on it.
Proportional STAR extends the identical ballot to multi-seat districts. Voters score candidates from zero to five exactly as they do in a single-seat race; nothing about the voter's task changes. Seats are then filled in rounds using the Allocated Score tabulation: the highest-scoring candidate is elected, the ballots that supported that winner have their weight proportionally reduced -- spent, in effect, on the seat just filled -- and the next seat is awarded on the remaining weight, repeating until the district is full. The result is a delegation whose composition tracks the distribution of voter support, produced without party lists, without closed slates, and without asking the voter to do anything but rate the individual people on the ballot. A voter who wants to know which member represents them has a clean answer: the one you gave five stars to. Proportional STAR is the method; Allocated Score is the specific tabulation algorithm that delivers its proportional result. The detail of why this is the right multi-seat choice is held for the magnitude section below; what matters here is that the voter's experience is continuous from a one-seat race to a seven-seat one, and only the count behind the curtain changes.
Two things about the design are worth stating plainly before the argument proceeds. The method is American in origin, developed by the Equal Vote Coalition rather than adapted from a parliamentary tradition built to solve a different country's problems. And the ballot is the same whether a district elects one member or several -- one ballot action for every congressional contest a voter will ever face.
The Architectural Argument¶
Here is the center of the case. The Act eliminates the state-administered primary; under the Unified General Election Structure, every qualified candidate appears directly on the November ballot. The reasons to want this are strong on their own -- a primary is a low-turnout contest, decided by a small and unrepresentative slice of the electorate, that nonetheless sets the menu for everyone in November; it excludes independents and the unaffiliated from the consequential first round; it makes voters travel to the polls twice; it varies bewilderingly from state to state. Removing it returns time and money to the public and puts the full field in front of the full electorate at the moment turnout is highest. But the primary was also doing a quieter job, one that almost no reform proposal accounts for: it was winnowing the field. By the time the general election arrived, the primary had already cut a sprawling list of hopefuls down to a manageable handful.
That winnowing was load-bearing for the counting method, and removing it is not neutral across methods. It is precisely where ranked methods and scored methods part company.
Ranked methods depend on winnowing because their ballot and their pathologies both scale with the size of the field. A ranked ballot asks the voter to place candidates in strict order of preference. With three or four candidates this is manageable. With fifteen or twenty -- the field a general election attracts once no primary has thinned it -- it becomes a genuine imposition. The voter is asked to adjudicate the relative ordering of candidates they have barely heard of, down to a fourteenth and fifteenth preference they have no real opinion about. The ballot itself swells into the "tablecloth" format that ranked-method jurisdictions know well, a grid of candidates against rank positions whose every additional column multiplies the ways a ballot can be spoiled. And the method's failure modes intensify with the field too: the more candidates and the more elimination rounds, the more opportunities for the order-of-elimination perturbations that produce ranked voting's characteristic pathologies. Ranked counting performs tolerably when something has already pruned the field to a few serious contenders. Remove the pruning and it degrades -- in usability and in integrity at once.
This is why the democracies that run ranked elections at scale pair them with heavy pre-election winnowing, usually disciplined party nomination that controls how many candidates carry each banner. And it is the trap that the leading American reform proposal cannot escape. Proportional Ranked Choice Voting faces a fork with no clean exit. Either it keeps a primary to winnow the field -- in which case the reform has not eliminated the bottleneck at all, and one must then ask what method runs that primary. Plurality? Then spoilers and vote-splitting have simply been pushed back one round, reintroduced into the very stage that decides who reaches November. Ranked choice again? Then the field is being run through the same pathology-prone elimination engine twice, across two separate elections, doubling the exposure rather than curing it. Or, alternatively, it abolishes the primary and accepts the tablecloth -- a twenty-candidate ranked grid in a high-stakes federal general election, with the spoilage and exhaustion that come with it. There is no third door. A ranked method must either keep the winnowing it claims to oppose or absorb the field size it was never built to handle.
STAR carries no such dependency, because it never asked the voter to produce an ordering in the first place. Scoring a field of thirty candidates is a different cognitive act from ranking thirty candidates. The voter gives high scores to the candidates they favor, low or zero scores to those they oppose, middling scores to the few they have mixed feelings about, and simply leaves the rest at zero -- and leaving a candidate unscored is the natural default, not a strategic blunder that can spoil the ballot or cost the voter influence. The ballot does not grow a new column for each entrant; it adds a row, and the voter ignores the rows that do not interest them. The method's behavior does not degrade as the field grows, because its demand on the voter does not grow. A crowded STAR ballot is longer; it is not harder.
The conclusion follows directly. STAR is not merely compatible with the elimination of primaries -- it is the method that the elimination of primaries makes necessary. A reform that abolishes the winnowing step and then adopts a counting method that silently depends on winnowing has reintroduced through the back door the bottleneck it threw out the front. The Act does not make that error. It removed the primary, and it chose the one family of method that can stand in the space the primary used to occupy: a method that finds a majority on its own (the built-in runoff) and tolerates a full, unwinnowed field on its own (the scored ballot). The voting method and the elimination of the primary are two halves of a single coherent decision. That coherence -- not any isolated virtue of scored ballots -- is the heart of the case.
Compared to What? The Failures You Would Choose¶
No voting method is flawless. This is not a rhetorical concession; it is a theorem. The Gibbard-Satterthwaite result establishes that every deterministic, non-dictatorial method of choosing among three or more candidates can be manipulated by strategic voters under some circumstances. There is no method that is honest-proof, spoiler-proof, and paradox-proof all at once, and any advocate who implies otherwise about their preferred method is selling something. The serious question is therefore never "is this method perfect" -- none is -- but a comparative one: of the failures no method can fully escape, which can a democracy best live with?
That question is usually answered badly, because of a habit built into how voting methods are compared. The standard tool is the criteria-compliance table: a grid with methods down one side and desirable properties across the top, each cell a check or a cross. Such tables are useful, but they carry a hidden and corrosive assumption -- that every failure is worth one mark, that a cross is a cross. They flatten a moral hierarchy into a scorecard. And on a scorecard, a method's grave failures and its trivial ones count the same, which means a method can lose on points while being, in every way that matters to an actual voter, the safer choice. The failures are not equal, and pretending they are is the central error in much of the reform debate.
Consider what the failures actually are, ranked by what they cost the voter:
- Ballot exhaustion -- a failure of ranked elimination methods -- means a ballot that was lawfully cast stops counting before the election is decided. The voter showed up, marked their preferences, and their ballot was set aside because the candidates they ranked were all eliminated before the final round. This is disenfranchisement after the fact: a vote cast and then discarded.
- Center squeeze -- also a failure of ranked elimination -- means the candidate a majority would have preferred over every rival is eliminated early, for lack of first-choice votes, and the seat goes to someone the majority likes less. The system produces the wrong winner, not by accident of a close count, but as a structural consequence of the counting rule.
- Non-monotonicity -- endemic to ranked elimination -- means the system can run backwards: ranking a candidate higher can cause them to lose, and ranking them lower can cause them to win. A method that can punish a candidate for gaining support has lost its claim to be tracking support at all.
- Later-no-harm failure -- STAR's most-cited flaw -- means that honestly giving your second-choice candidate a decent score can, in a close contest, help that candidate edge out your first choice in the runoff. Stated plainly: the cost is that you might get your second choice instead of your first.
Set those side by side and the asymmetry is not subtle. A method that occasionally hands a voter their backup choice is not failing in the same currency as a method that throws ballots out of the count or installs the candidate the majority rejected. One is a disappointment; the others are defects of legitimacy. They do not belong in the same column of a table, and a comparison that scores them as equivalent has measured the wrong thing.
The demonstration: Alaska, on the actual ballots¶
The weight of these failures is not theoretical. It can be read off a real federal election whose every ballot is public.
In the August 2022 special election for Alaska's lone U.S. House seat, three candidates ran -- Mary Peltola, Nick Begich, and Sarah Palin -- under Instant Runoff Voting, the single-winner form of ranked choice voting. Begich received the fewest first-choice votes and was eliminated in the first round. His ballots transferred, and Peltola defeated Palin in the final round. But when the full Cast Vote Record was released, a mathematical analysis published by Jeanne Clelland in The Mathematics Enthusiast established what the official results could not show. Begich was the Condorcet winner: head-to-head, the ballots show he would have beaten Palin and he would have beaten Peltola. The candidate a majority preferred over each rival was eliminated first, for want of first-place votes -- a textbook center squeeze. At the same time, of the roughly fifty-four thousand ballots that ranked Begich first, more than eleven thousand exhausted when he was eliminated, dropping out of the count entirely. One election, two grave failures at once: the majority-preferred candidate eliminated, and eleven thousand cast ballots discarded.
Clelland then took the same ballots -- the real preferences of real Alaskans, held fixed -- and asked how the election would have resolved under STAR. The finding is that Begich, the Condorcet winner, would almost certainly have won. The reason is structural and is exactly the mechanism named earlier: under STAR the broadly preferred candidate is not eliminated for thin first-choice support, because the scoring round registers the wide second-choice support that the elimination engine threw away, and the automatic runoff then confirms the majority preference between the top two. Indeed, because Palin was the Condorcet loser -- she would have lost head-to-head to both others -- STAR's runoff structurally guarantees she could not have won under any plausible scoring. The method would have caught precisely what the elimination engine missed.
Two cautions keep this honest, and both ultimately strengthen the point. First, this is a counterfactual model, not a STAR election that was actually held: Clelland reconstructs STAR outcomes from ranked ballots under stated, conservative assumptions about how voters would assign stars, and she is careful to call the result "likely -- but not at all certain." We should be equally careful, and claim no more than she does. But the core result is robust across her assumptions: Begich's path to victory is wide, Palin's is foreclosed, and the only way Begich loses requires an implausible asymmetry in scoring behavior. Her tabulation, she notes, independently matches the published cast-vote analysis of Graham-Squire and McCune, so the underlying preference data is not in dispute. Second -- and this is the point an honest advocate must volunteer -- Condorcet failures like Alaska's are rare. Clelland cites the finding that of 182 American IRV elections between 2004 and 2022, only two failed to elect the Condorcet winner. Rare.
But rarity is the scorecard's argument, not an answer to ours. The entire point of weighing failures by severity rather than counting them by frequency is that a rare catastrophe and a common nuisance are not interchangeable. Engineering learned this lesson the hard way and built its discipline around it: the rare, severe failure is the one you design against, not the one you wave away because the odds look long. Challenger and Columbia were both, in part, failures of exactly the opposite instinct -- of treating a known flaw as acceptable because it had not yet proven fatal, letting low probability stand in for low concern until the improbable arrived. A failure mode that eliminates the majority's preferred candidate and discards eleven thousand ballots is not made acceptable by its infrequency -- and its consequences are not confined to the one seat. The 2009 Burlington mayoral election, the other of the two Condorcet failures in that dataset, was followed by the city repealing ranked choice voting outright. Alaska's special election became a national exhibit in the campaign against RCV and helped fuel a wave of state-level pre-emptive bans. And here the rarity cuts the other way from how its defenders intend. A base rate that produced two notable disasters across a couple of decades of scattered, mostly local adoption is not a base rate that stays small when the method goes national and permanent. The Fair Representation Act would put ranked elections in every congressional district in the country, every cycle, indefinitely -- and a nonzero failure rate applied to hundreds of federal contests in perpetuity does not yield two failures in twenty years; it yields a steady recurrence of them, cycle after cycle. Each one would be its own Alaska, its own Burlington: a concrete, local, nameable contest in which the candidate most voters preferred lost and lawful ballots were discarded. These would not average out into a reassuring aggregate. Each would land where it happened, on the voters who lived it, as a discrete event that the system's opponents could hold up by name -- and each would chip at public confidence in the federal legislature, the institution that can least afford to have its legitimacy questioned. Scaling a method to the whole House does not dilute its rare failures across a larger denominator; it multiplies the absolute number of legitimacy crises a national electorate must absorb, and raises each one to the federal stage. Rare failures of legitimacy do disproportionate political damage, because they hand opponents a concrete, comprehensible injustice -- "the candidate most voters preferred lost, and thousands of votes were thrown away" -- that no amount of aggregate-performance data can answer. STAR does not merely perform well on average. It structurally forecloses the specific failures that turn voters against electoral reform: it cannot exhaust a ballot, and it cannot seat the Condorcet loser. (The political history of the RCV backlash is treated more fully in the accompanying analysis of the Alaska repeal efforts; the point here is structural.)
The objections this answers along the way¶
Two of the standard objections to scored voting dissolve once the comparison is framed this way, which is why they are answered here rather than in a section of their own.
The first is the ordinal-supremacy argument: that ranked ballots are epistemically cleaner because ordering candidates is something voters can do honestly, while scoring invites strategic exaggeration. Gibbard-Satterthwaite is the reply. Strategic incentive is not a defect peculiar to scored ballots; it is a feature of every method that aggregates more than two options, ranked methods emphatically included -- they are manipulable by burial, compromising, and a documented catalog of other maneuvers. The honest framing is not "scored methods are manipulable and ranked methods are not," which is false, but "every method is manipulable, so the question is which produces good outcomes under realistic, partly-strategic voting." Held to that even standard, the ordinal-supremacy argument does not distinguish the families at all.
The second is strategic compression, or min-maxing: the worry that voters will not use the full scale, giving five stars to their favorite and zero to everyone else, collapsing the scored ballot toward a single-mark ballot. Some voters will indeed do this, and the honest concession is that they will. But the automatic runoff is the structural answer. Even if a large share of voters min-max their scores, the runoff round still resolves to "which of these two finalists did you score higher," and majority preference between the final two is recovered regardless. The scoring round rewards voters who express gradations with more influence over which two candidates advance; it does not punish those who decline, and it cannot be broken by them. A method that depended on universally honest scoring would be fragile. STAR does not, because the runoff is a floor that holds even when the scoring round is gamed.
STAR's own most serious vulnerability deserves the same honesty in return. The later-no-harm failure described above is real, and it creates a genuine strategic incentive: a voter can bury a threatening rival -- score an acceptable opponent at zero rather than honestly -- to keep that rival out of the runoff. This is the true form of the later-no-harm objection, and it is more substantial than "I fear my backup choice." But its practical bite is limited by two facts. Burial requires accurate knowledge of the competitive landscape: the voter must correctly identify which rival actually threatens their favorite, and guess wrong and the strategy misfires. And it carries its own risk: bury a candidate who would have beaten someone you like even less, and you have damaged your own interest. The pivotal-voter strategic-incentive modeling of Wolk, Quinn, and Ogren found that under STAR's two-round structure, dishonest strategies -- favorite betrayal, burial, and bullet voting alike -- are strongly disincentivized, with the residual incentive small and confined to mild "honest inflation" that preserves a voter's true preference order. The incentive exists; it should not be minimized; and it remains, measured against exhaustion and center squeeze, the failure a voter would choose.
The Ballot Americans Already Know How to Fill Out¶
A voting method's first duty is not to satisfy theorists. It is to be usable -- correctly, confidently, and without special instruction -- by every eligible voter who walks into a polling place, including the ones who did not prepare, do not follow politics closely, and will give the ballot ninety seconds of their attention. A method that performs beautifully in a textbook and trips up real voters at the margin has failed at the only test that ultimately counts. On that test, STAR has an advantage no other reform method can claim: its ballot asks voters to do something tens of millions of them already do, fluently, every week.
Americans rate things. They give restaurants and drivers and products a score out of five before they have finished their coffee. Amazon, Uber, Google, Letterboxd, Yelp, the "how did we do?" card at the bottom of a receipt -- the zero-to-five rating is among the most widely practiced quantitative judgments in American life, performed by people across every level of education and political engagement, without anyone needing to explain what three stars means relative to five. The cognitive task a STAR ballot asks for is not novel. It is the most rehearsed evaluative act in modern consumer life, transplanted to the ballot.
Now ask the comparison the other way. When was the last time any company asked you to rank its services in strict order -- to declare that the coffee was your third preference, the wifi your first, the seating your second, with no ties permitted? It does not happen, because strict ordinal ranking of more than a handful of items is a genuinely demanding task, and businesses that depend on getting honest feedback from ordinary people do not ask for it. They ask for ratings. The reform movement's ranked methods ask voters, on the highest-stakes ballot they will ever cast, to perform an evaluative act that no consumer institution has concluded is worth requesting from the same people. STAR asks for the one they perform reflexively.
And this is the same property the architectural argument turned on. The very property that lets the STAR ballot survive a crowded, unwinnowed field -- that the voter scores each candidate independently and may leave any of them unscored, rather than producing a complete strict ordering -- is the property that makes the ballot familiar and low-error. Independence of judgment is what makes rating easy and what makes it scale. The two virtues are the same virtue. And the stakes of getting this wrong are not evenly distributed. Research on ranked ballots finds that ballot-marking errors and incomplete rankings fall most heavily on precisely the voters already most likely to be underrepresented -- older voters, voters with less formal education, voters in non-English-dominant households, voters with disabilities. A ballot that is harder to complete correctly does not impose its difficulty equally; it filters hardest on the already-marginalized. Choosing the more usable ballot is therefore not merely a matter of convenience. It is a matter of which voters' intentions actually survive the act of voting.
The objection that scored ballots are too cognitively demanding -- the third of the standard objections to STAR -- has the matter exactly backwards. The demonstrated track record points the other way. The Independent Party of Oregon used scored voting in a 2020 primary with more than a million registered members eligible to participate. The Python Software Foundation elects its Steering Council by a STAR-derived method. In binding, real-world elections, the predicted usability collapse does not occur, because the task -- rate these people zero to five -- is one voters already know. The cognitive-burden objection imagines voters meeting an alien instrument. They are meeting the most familiar evaluative gesture in their daily lives, printed on a ballot.
Match the Method to the Magnitude¶
Single-seat districts and multi-seat districts pose different questions, and a serious design answers each in its own terms rather than forcing one method to do both jobs. The Act uses single-winner STAR where a district elects one member and Proportional STAR where a district elects three or more. This is not a compromise between two camps. It is method matched to the structural question the seat count is actually asking.
The clarifying point -- the one that dissolves the most common objection to STAR in multi-seat contexts -- is that proportionality is a property of district magnitude, not of the counting rule. When a district elects one member, there is exactly one majority to find, and no proportional representation is possible regardless of method; the right question is "who is the single candidate most preferred by the most voters," which is the question single-winner STAR's runoff answers. When a district elects several members, proportional representation becomes possible, and the magnitude itself opens that space: a five-seat district can seat a delegation reflecting five-fifths of its electorate's distribution. The counting rule does not create the proportionality; the magnitude does. What the counting rule determines is everything else -- how the ballot behaves, whether votes are wasted, whether the broadly acceptable are discarded, whether the method can be gamed -- the very dimensions Sections 3 and 4 examined. So the multi-seat choice is not "which method is more proportional," because at a given magnitude reasonable methods deliver comparable proportionality. The choice is "which method delivers that proportionality while also handling the field gracefully, keeping the ballot usable, and failing in the least damaging way." Asked correctly, the multi-seat question has the same answer as the single-seat one, for the same reasons.
This reframing answers the objection that single transferable vote -- the ranked multi-winner method -- is the more proven or more proportional path to a representative legislature. It is not more proportional in any way that survives scrutiny: at equal district magnitude, Proportional STAR and STV produce comparably proportional delegations, because the magnitude is doing that work. What differs is everything the previous sections weighed. STV is the multi-winner form of the same ranked-elimination engine, and it carries the same family of liabilities -- the demanding ordinal ballot that worsens as the field grows, the ballot exhaustion as preferences run out across elimination rounds, the elimination-order sensitivities -- now operating across several seats at once. Proportional STAR delivers the proportional result on the identical zero-to-five ballot voters use everywhere else, with no ballot exhaustion and no ordinal-ranking burden, and it produces that result through individuals rather than party lists: a voter is represented by the specific people they rated highly, not by a party slate negotiated above their head. The American voter's reasonable question -- "which of these is my representative?" -- has a direct answer under Proportional STAR: the one you gave five stars to.
The strongest evidence for this is not theoretical, and it comes from a body that had every reason to prefer the ranked alternative. The Los Angeles chapter of the Democratic Socialists of America conducted internal elections under single transferable vote and then deliberately moved to Proportional STAR, finding that it delivered the proportional representation they valued while dropping the features they did not -- the polarizing elimination rounds and the ranked ballot's friction -- and preserving the incentive for candidates to seek broad consensus rather than narrow factional intensity. This is a constituency predisposed toward the proportional-ranked tradition, examining both methods up close in real binding elections, and choosing the scored one on the merits. When the people most inclined to favor the ranked method switch away from it after using both, the comparison has been made by exactly the right judges.
Conclusion¶
The case for STAR and Proportional STAR is, at bottom, a single case, and it is structural rather than aesthetic. The voting method is the rule that decides who governs, and the Act's most valuable move -- ending the state-administered primary, with all the exclusion and low-turnout distortion it carried -- removed the hidden winnowing step that every counting method had been quietly relying on. A reform that abolishes that winnowing and then adopts a method which secretly depends on it has reintroduced the bottleneck it claimed to remove. Only a method that can stand in the empty space the primary left behind -- finding a majority on its own through the built-in runoff, and tolerating a full, unwinnowed field on its own through the scored ballot -- can complete the design coherently. STAR is that method.
Everything else in this paper is a consequence of that one fact rather than an independent argument for it. STAR handles a crowded ballot gracefully because it never demands an ordering. It guarantees a majority-preferred winner because the runoff is internal to the general election. When it fails, it fails by occasionally handing a voter their second choice rather than by discarding ballots or seating the candidate the majority rejected -- and the actual ballots of Alaska's 2022 election show, held fixed, that the method would have caught precisely the failure the ranked engine produced. Its ballot is the one evaluative task tens of millions of Americans already perform without instruction, which makes it both usable and equitable. And in multi-seat districts, Proportional STAR delivers proportional representation through the individuals a voter actually rated, on the identical ballot, matching the method to the magnitude. These are not four reasons standing beside each other. They are four faces of the same property: a method that stands on its own, chosen by a reform that needed exactly that.
That is the right way to read the Act's voting method. Not as a preference the design happens to express, and not as a compromise it tolerates, but as the keystone the rest of the architecture requires -- the one piece that, removed, brings the structure down, and that, set in place, lets everything else bear its load.
Works Cited¶
Clelland, Jeanne N. "Ranked Choice Voting and Condorcet Failure in the Alaska 2022 Special Election: How Might Other Voting Systems Compare?" The Mathematics Enthusiast 23, no. 3 (2026): 185-196.
Gibbard, Allan. "Manipulation of Voting Schemes: A General Result." Econometrica 41, no. 4 (1973): 587-601.
Graham-Squire, Adam, and David McCune. "An Examination of Ranked-Choice Voting in the United States, 2004-2022." Representation (2023). https://doi.org/10.1080/00344893.2023.2221689.
Satterthwaite, Mark A. "Strategy-Proofness and Arrow's Conditions: Existence and Correspondence Theorems for Voting Procedures and Social Welfare Functions." Journal of Economic Theory 10, no. 2 (1975): 187-217.
Wolk, Sara, Jameson Quinn, and Mark Ogren. "STAR Voting, Equality of Voice, and Voter Satisfaction: Considerations and Simulations." Constitutional Political Economy 34 (2023): 301-325.