National Popular Vote

Electing the President by nationwide popular vote, so that every ballot carries the same weight wherever it is cast.

AI evaluation · not yet reviewed by a human

This evaluation was produced and sourced by an AI model; a human review is still pending. Figures and conclusions may still change. The review log is at the foot of the page.How review works →

The President would be chosen by whoever receives the most votes across all fifty states and the District of Columbia, replacing the Electoral College and the state winner-take-all rules built on it. If no candidate reaches 50 percent, the top two meet in a runoff — the arrangement used for the presidency in most democracies that elect one directly. This touches the presidency only: the Senate weights small states by design and the House inherits part of that through apportionment, and neither is affected. Two routes exist, a constitutional amendment or activation of the interstate compact, which stands at 222 of the 270 electoral votes it needs after Virginia joined in April 2026. This evaluation assesses the change itself rather than either route, and looks twenty years ahead.

Balance

Much better for the future · 0.93 previous scale

Balance on the previous scale. The Bilanz 2.0 simulation is not yet available for this evaluation. The category comes from the share of the debate on the pro side (r).

For 93 · 93 % Against 7.4 · 7 %

Size class: medium Scale of this evaluation: Normalised Impact — unitless, calibrated to this topic. For comparison: one point here is worth roughly 200 million euro per year. One vote is valued at 1,000 euro in this evaluation, the same figure used across democracy topics, derived from what an American election cycle costs per ballot cast and raised by a factor for the legitimacy a ballot carries beyond its administration. How we score →

Arguments for

Arguments against

7 arguments evaluated · Scoring v1.3 Δ absolute +85.6

Arguments — For

5 arguments · top 3 shown

Every ballot carries the same weight

57of 100

One electoral vote stands for 194,000 people in Wyoming and over 700,000 in Texas, Florida and California. The shortfall for an under-weighted voter is a fraction of a ballot rather than a whole one, but it falls on roughly half the electorate.

Value 9 · Equal vote weightImpact 7Plausibility 9

▸ Show reasoning & sources ▾ Hide reasoning & sources

What is at stake is whose preference decides who governs. A citizen in a populous state casts a ballot that is worth a fraction of one cast a few hundred miles away, for no reason connected to that citizen at all. Equal suffrage is the mechanism by which every other political right is defended, which is why it sits at the constitutional core rather than among ordinary institutional questions. The same weight is assigned here as to the arguments about disputed and manipulated votes below, deliberately: a ballot diluted by the allocation formula and a ballot lost to a recount dispute are the same injury to the same good, and putting them on one scale is what allows the comparison to be about size rather than about which value the reader prefers. The value is at the highest level short of life and health.

Without the reform, the weight of a ballot depends on where the voter lives, and the spread is measurable rather than rhetorical. One electoral vote stands for about 194,000 people in Wyoming and for more than 700,000 in Texas, Florida and California, where even distribution would put each elector at about 623,000 ^[12]. Wyoming holds 0.18 percent of the population and 0.56 percent of the electoral votes ^[12]. Measured against turnout instead of population, a Wyoming ballot carries a weight of 2.97 and a Florida ballot 0.78, with Louisiana landing exactly on the national average ^[13]. It matters what that bounds: the largest individual shortfall is about a fifth of a vote, not half and not a whole one, and most under-weighted voters sit far closer to parity than the Wyoming-to-California headline suggests. Two mechanisms produce the gap: one hundred of the 538 electors come from the Senate half of the formula and are handed out two per state regardless of population, and state winner-take-all then discards the minority vote inside each state. With the reform both disappear, because the total is simply the count. Against 158 million ballots cast in 2024 ^[7], with about half the electorate under-weighted at a population-weighted mean shortfall near seven percent, roughly 5.5 million vote-equivalents are restored at each presidential election. This argument counts only that restoration and not the loss to voters who currently hold more than one ballot's worth of influence, on the ground that no one is entitled to that. The Impact is the largest here because the shortfall is small per person and reaches half the country.

There is nothing to estimate about whether the inequality exists. It follows from the allocation formula written into Article II and the Twelfth Amendment, and anyone can compute it from census figures and elector counts. That places this argument in a different class from every other one here: it rests on arithmetic rather than on evidence about how people will behave. What is uncertain is only the size, and only mildly: the share of the electorate that is under-weighted and the typical size of the shortfall depend on how the state boundaries fall against population, which moves slowly with each census. The band runs from 1.0 to 4.5 and even its floor exceeds the whole opposing side several times over. One dependency deserves naming: the decision to count only the restored side, and not the mirrored loss, is a judgement rather than a measurement. The plausibility is as high as it gets for a claim of this kind, and it is capped short of certainty only because that judgement sits underneath it.

evidence basis: Mechanism · P ceiling 9.5

Campaigns and policy would address the whole country

20of 100

Of 262 campaign events in 2024, 246 were held in seven states. The same incentive shapes what gets promised, not only where: ethanol for Iowa, steel for Pennsylvania and Ohio, Cuba policy for Florida.

Value 9 · Equal political voiceImpact 5Plausibility 4.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is whose interests the presidency is oriented toward — who gets asked for their vote and whose concerns end up in federal policy. It reaches voters in states that have been safe for a generation, where someone can go an entire adult life without a presidential campaign competing for their support, and it reaches every taxpayer through the sectoral commitments the current geography rewards. The value is at the constitutional core, because equal standing in the choice of a president is the same good as equal weight in counting it.

Without the reform, a candidate has no reason to spend a day, or a policy commitment, anywhere the result is already known. The travel side is counted: of 262 general-election events in 2024, 246 were held in the seven closely divided states, which hold a little under a fifth of the population, while 43 states and the District of Columbia shared sixteen between them ^[2]^[3]. Turnout follows the same contour, with Wisconsin at 75 percent and Hawaii at 49 ^[8], though attention is one cause among several. The policy side runs on the same incentive and matters more: carrying a pivotal state is worth more than gaining a point nationally, which produces a recognisable pattern of commitments — ethanol tied to Iowa, trade and steel positions tied to Pennsylvania and Ohio, Cuba policy tied to Florida. These shape federal energy, trade and foreign policy for decades. With a national count the geographic premium disappears from both at once, which is why they are counted as one argument rather than two. Turnout component: two points across roughly 125 million eligible citizens outside the battlegrounds, about 2.5 million ballots per election. Policy component: 100 million to 1 billion per year of distortion, central 400 million. The Impact is moderate and its band is wide for an honest reason, since the policy that would have existed instead is unobservable.

The two halves of this argument are evidenced very differently, and the plausibility is set by the weaker one. The concentration of campaign events is counted, not estimated: FairVote tracked every event of the 2024 general-election campaign ^[3]. The step from attention to turnout is a projection, since battleground states differ from others in registration rules and local political culture as well as in presidential attention, and the assumed two points is a round number rather than a derived one. The policy half is weaker still. The mechanism is standard in political economy and the examples are real, but how much federal policy is actually driven by presidential electoral geography rather than by the same interests working through Congress is contested — ethanol has a farm-state Senate coalition quite apart from Iowa's place in the calendar. Nor is it clear that a national median target produces better policy rather than differently distorted policy. Low to middling plausibility: the concentration is beyond dispute, everything the argument builds on it is inference.

evidence basis: Projection · P ceiling 6

Close elections become far less fragile

12of 100

In 2000 the presidency turned on 537 votes in Florida while the national margin was about 540,000 — a thousand times larger. A national count makes the decisive margin the national one, which is almost always wide enough to survive a recount.

Value 9 · Robustness of the countImpact 2.5Plausibility 5.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

What is at stake is whether the losing side accepts the count. A recount fight is the moment a democracy is most exposed, and 2000 showed how much institutional damage a few hundred contested ballots can do when everything hangs on them. Everyone is affected at once rather than any particular group. The value is at the constitutional core, alongside the arguments about vote weight, because acceptance of the result and equality of the result are the same good seen from two sides.

Without the reform, decisiveness is concentrated into whichever state happens to be pivotal, so a minute margin there settles the whole country. In 2000 that margin was 537 votes in Florida ^[14], against a national margin near 540,000 — a factor of about a thousand. Pivotal states landing within a few thousand votes has happened repeatedly, in 2000, 2016 and 2020; the national margin has fallen below half a percentage point three times since 1960. With the reform, the decisive number is the national one, so an election of the same closeness produces far less to litigate. This argument is netted rather than split: the objection that a national count nationalises recounts belongs to the same consequence stream and is deducted here rather than booked separately. What is deducted is the tail case, which is real — when the national margin genuinely is minute, all fifty states are contested at once under fifty recount regimes with no federal mechanism to resolve it ^[5], which is rarer than today's disputes and worse when it arrives. The net removes roughly 2 million vote-equivalents per election from dispute exposure. The Impact is moderate and follows from arithmetic about margins rather than from any behavioural assumption.

The core comparison is documented and not seriously contested: the 2000 margins are a matter of record ^[14], and that pivotal-state margins are smaller than national margins follows from the fact that a state is a subset of the country. What is estimated is the frequency and size of disputes on both sides of the netting, and there the numbers are constructions. Nobody has run a national recount, so the tail case is sized by analogy rather than observation, and the frequency rests on a handful of historical margins, a small sample for a tail event. Running against the argument, most of what makes a national tail case dangerous could be fixed in advance by uniform standards, which would also shrink the deduction and make the net larger. Middling plausibility: the direction is arithmetic, the size is not.

evidence basis: Precedent · P ceiling 6

A shorter handover

2.2of 100

Eleven weeks pass between election day and inauguration because electors once had to travel. With a national count the winner is known in days, and the Twentieth Amendment already shortened this gap once for the same reason.

Value 6 · Administrative qualityImpact 0.8Plausibility 5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The stream is the capacity of a government to act, and the risk carried by a long period in which authority is formally held by someone who has lost it. This is institutional functioning rather than anyone's rights, which places it at the same level as the other arguments about how well the machinery works. The value is middling.

Without the reform, the calendar is set by machinery rather than by need: electors meet in mid-December, Congress counts the certificates on 6 January, and the new President is sworn in on 20 January, about eleven weeks after the vote. That schedule was built when electors travelled by horse. It has been shortened once already — the Twentieth Amendment moved inauguration from 4 March to 20 January in 1933, for exactly this reason — which establishes both that the gap is adjustable and that shortening it is considered worthwhile. With a national count the winner is known within days and the remaining steps are the only thing holding the date. What a shorter handover buys is a smaller window in which a defeated administration holds full power while its successor can neither appoint nor decide, which costs in deferred decisions and carries risk of its own. Valued at 150 million per year, which is a placeholder rather than a measurement. The Impact is small.

The mechanical claim is solid: the elector machinery is what the calendar is built around, and removing it removes the reason for most of the delay. The historical precedent is firm, since the gap has already been shortened once by amendment. Beyond that the argument is weakly supported. No study puts a figure on what an interregnum costs, the estimate here is a placeholder, and a shorter transition has a real cost of its own: incoming administrations use those weeks to staff a government, and compressing them could produce a less prepared executive rather than a faster one. Nor does the reform automatically shorten anything — the dates would have to be changed deliberately. Low to middling plausibility: the opportunity is clear, the benefit is asserted.

evidence basis: Plausibility · P ceiling 5

The winner would be the candidate most people chose

2.1of 100

Five of 47 presidents took office having received fewer votes than their opponent, two of them since 2000. A national count makes that outcome impossible by construction.

Value 9 · Equal vote weightImpact 0.4Plausibility 5.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The consequence is that a country is governed by someone the smaller number of its voters chose. What is damaged is not a particular person's ballot but the connection between the count and the outcome, which is the thing that makes losing an election acceptable to the side that lost. The weight is the same constitutional core as the other arguments here; the difference lies entirely in how many ballots the effect touches, which is what the balance is meant to express.

Without the reform, the two counts can diverge, and they have: five of 47 presidents entered office without the most votes ^[1], twice in the last seven elections. With the reform the divergence is impossible, because the national count is the outcome. The size of the delta is the margin that would have had to be overturned for the result to match the plurality, which was about 540,000 votes in 2000 and 2.9 million in 2016. How often this recurs depends on how close national margins stay and how the swing states sit relative to the country, which currently points to something near a one-in-five chance per election. That gives an expected value near 340,000 vote-equivalents per election, 85,000 a year. The Impact is small because the event is occasional and the margins involved are a fraction of a percent of the electorate, even though each occurrence is highly visible.

The historical record is settled and the mechanism is definitional: a national count cannot contradict the national count. What is projected is the forward frequency, and that is genuinely uncertain in both directions. Divergence became more likely as the swing states drifted away from the national average, which argues for a higher rate than the long-run five-in-47. Against that, 2024 saw the popular and electoral winner coincide with a 1.5 point national margin ^[9], and the alignment could persist for a decade. A range from one-in-twenty to one-in-three would all be defensible. Middling plausibility: that the reform eliminates the divergence is certain, how much divergence there would have been is not.

evidence basis: Projection · P ceiling 6

Arguments — Against

2 arguments

A second round, decided by fewer voters

6.2of 100

A runoff below 50 percent removes the minority-president problem entirely. It also triggers often: five of the last nine elections produced a leader under 50 percent. In those years the presidency is settled by a round some first-round voters do not return for.

Value 9 · Equal vote weightImpact 1.3Plausibility 5.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

The consequence is that the person who governs is chosen by a smaller group than turned out to choose. It is a milder injury than being prevented from voting, because nobody is stopped and everyone was present for the round that set the choice, but it is the same good: whose preferences decide. It falls hardest on voters for whom turning out is costly — shift workers, people without transport, those who needed a mail ballot — since they are the ones least able to do it twice. The value is at the constitutional core, and the low score reflects the size of the effect rather than any discount on what it touches.

Without the reform, one round decides everything and every voter who turns out is present for the decision. With a runoff below 50 percent, that holds only when someone clears the threshold — and in the United States that is the exception rather than the rule. Minor candidates routinely pull the leader under half: of the nine elections from 1992 to 2024, five ended with a leader below 50 percent, including 2024 at 49.8 percent ^[9]. So roughly every second election would go to a second round, and the second round is the one that counts. What is lost is the voters who do not come back. The right comparison is France, where the two rounds are three weeks apart and both feature nationally known candidates: in 2017 turnout fell from 77.8 to 74.6 percent ^[11], about three points, far less than American primary runoffs suggest. On roughly 155 million ballots that is 5 to 9 million fewer voters in the deciding round, occurring about half the time, and discounted because those voters did have their say in the first round and determined who was left. The Impact is small: the round they miss is real, but they were present for the round that shaped it.

Two of the three steps are solid and the third is not. That the threshold would bite often is arithmetic on the historical record, not a forecast ^[9]. That drop-off occurs between rounds is observed wherever two-round systems run. What is projected is how large it would be in an American presidential runoff, and there is no domestic case to transfer from: American runoff experience comes from primaries and southern local races, where drop-off is severe and the comparison misleading, because those second rounds involve low-salience contests. France is the better analogue and points to a small effect, but France votes on Sundays, has compulsory-adjacent civic norms and a different media environment. The discount applied for first-round participation is a judgement rather than a measured quantity. Middling plausibility: the frequency is certain, the size is a transfer from one foreign case.

evidence basis: Projection · P ceiling 6

Two rounds cost more than one

1.2of 100

Elections cost money under any system, and the coordination this reform adds is minor. The real item is paying for a second nationwide election in roughly half of all cycles.

Value 5 · Administrative costImpact 0.6Plausibility 4.5

▸ Show reasoning & sources ▾ Hide reasoning & sources

This stream is public money, raised from taxpayers and unavailable elsewhere once spent. It touches no one's rights directly and its weight here is the ordinary weight for public expenditure. It is deliberately not raised because the spending happens in a constitutionally sensitive area: an expensive reform is not thereby a more serious harm. The value is moderate.

Without the reform, one national election is held every four years and each state certifies its own result. With it, two things change, and only one of them is expensive. The coordination the reform requires — a common definition of a valid ballot, a shared certification timetable, a national canvass — is real but small against what is already spent, because states would continue to run and count their own elections; assumed at 10 million per year. The second round is the real cost: a full nationwide election, with polling places, staff and ballots, in about half of all cycles given how often the 50 percent threshold would bite ^[9]. A presidential election costs the public purse on the order of a billion dollars, which amortises to roughly 100 million a year. Money spent here is unavailable for poll worker recruitment and equipment replacement, which is the more relevant constraint than the sum. The Impact is small: this is an administrative programme of ordinary size against a change of constitutional scale.

That two elections cost more than one is not in doubt, and the frequency is arithmetic on the historical record rather than a forecast ^[9]. What is uncertain is the price tag. No agency has costed a national runoff, American experience with runoffs comes from primaries and local races at a different scale, and the billion-dollar figure is an order-of-magnitude estimate rather than an accounting one. The band spans more than a factor of three. Working against the objection, a second round reuses everything already built for the first, so the marginal cost is well below the first round's full cost. Low plausibility on the size, high on the fact that the cost exists and recurs.

evidence basis: Plausibility · P ceiling 5

Summary

The problem is that a ballot's weight depends on where it is cast: one electoral vote stands for about 194,000 people in Wyoming and over 700,000 in Texas, Florida and California [12], which against turnout works out to 2.97 for a Wyoming ballot and 0.78 for a Florida one [13]. The shortfall for any individual under-weighted voter is a fraction of a ballot rather than a whole one, but it falls on roughly half the electorate and it is systematic. The reform solves it completely rather than partially, and the claim rests on the allocation formula itself rather than on any prediction about behaviour — a class of certainty none of the objections can match. Three further gains follow: campaigns and federal policy would stop being aimed at a handful of pivotal states that hosted 246 of 262 events in 2024 [2], no president could take office having lost the count, and close elections would become far less fragile, since in 2000 the presidency turned on 537 votes in Florida [14] while the national margin was about a thousand times larger. Against this stand two real costs, both attached to the runoff rather than to the popular vote itself: the 50 percent threshold would trigger a second round in roughly half of American elections [15], that round is settled by a slightly smaller electorate, and it has to be paid for. Two arguments commonly made do not survive inspection and are set out with reasons rather than dropped silently: manipulation does not become more attractive under a national count but less, because winner-take-all concentrates the payoff into a pivotal state where a few hundred ballots can decide everything; and small states are not protected by the present system but ignored by it. The hinge is not a number but a rule: this evaluation counts the restoration of under-weighted ballots and does not count the loss to voters who currently hold more than one ballot's worth of influence. A reader who thinks excess weight is an entitlement worth protecting should read this balance as close rather than lopsided.

Outlook — effect over time

Much better for the future · 0.93 previous scale

Sources

National Popular Vote: Agreement Among the States to Elect the President by National Popular Vote. nationalpopularvote.com
National Popular Vote: Almost All (94%) of 2024 Presidential Campaign was Concentrated on 7 States. nationalpopularvote.com
FairVote: Presidential campaign events in 2024. fairvote.org
NPR: Virginia ups the National Popular Vote Compact to 222 votes. npr.org
Ballotpedia: National Popular Vote Interstate Compact — arguments for and against. ballotpedia.org
Rock the Vote: The National Popular Vote Interstate Compact — an explainer. rockthevote.org
U.S. Election Assistance Commission: Election Administration and Voting Survey 2024. eac.gov
The National Vote: Fact Sheet. thenationalvote.com
Wikipedia: 2024 United States presidential election. en.wikipedia.org
Passarelli & Bergman, Political Studies Review: Runoff Comebacks in Comparative Perspective: Two-Round Presidential Election Systems. iris.uniroma1.it
Wikipedia: Two-round system. en.wikipedia.org
USAFacts: Representation in the Electoral College: How do states compare?. usafacts.org
The Conversation: Whose votes count the least in the Electoral College?. theconversation.com
The Saturday Evening Post: How Much Does Your Vote Weigh?. saturdayeveningpost.com
The Almanac of American Politics: The 49 Percent Nation. thealmanacofamericanpolitics.com

Last reviewed by Claude Opus 5 · August 25, 2026 · 1× AI, not yet reviewed by a human

August 25, 2026AI reviewClaude Opus 5First evaluation
Taken from the archiv/usa-source-2026-08/ delivery (revision draft); derived fields recomputed.

Evaluations are produced with AI support and reviewed on a schedule for new developments; human passes are marked separately.How we review →