The First Repair: Change who gets in, and what they may do once there
Start with the maps, because nothing else works until they are neutral. Politicians should not draw the districts in which they run. If a district is drawn seventy-thirty, November is a formality no matter how the primary is designed; reforming the ballot while leaving the map alone is rearranging furniture. So take the pen away from everyone. Maps should be computed by a public formula from the Census’s geography and population counts and nothing else: districts of equal population, each one connected, splitting the fewest counties, then the fewest towns, then the fewest census tracts, and among the maps that do all that, the most compact.16
Neutrality here is not a matter of trusting good people. It is engineered: the formula never receives an election result, a party registration, an incumbent’s address, a poll, or a candidate’s name, and it receives no demographic data beyond head counts except where federal voting-rights law requires it. The code is public, the inputs are public, and the same code runs in every state, so anyone can run it and check the answer. Iowa has come closest for forty-five years, with nonpartisan staff forbidden to use political data; this takes that rule the rest of the way, so there is no one left to capture. The one place a human hand still touches the map, the voting-rights exception, is published, explained, and reviewable in court. Two honest limits. A neutral process is not a proportional outcome: where one party’s voters crowd into cities, a compact map will tilt against them, and no one will have chosen it. And the formula promises nothing about competitiveness; that is measured separately, on the scorecard, where it belongs.
Then put every candidate on one primary ballot. No separate party primaries. Everyone runs, every eligible voter participates regardless of party, and the two candidates with the most support go to November.
But how you mark that ballot decides whether the reform works at all, and I want to be careful here, because I changed my mind about it.
The obvious design is the one most places use: pick one name. It fails in a way that matters. When four candidates on one side split their supporters and two on the other side split only two ways, the side with more choices loses. This happened in California in 2012, when four Democrats in a Democratic-leaning district split nearly half the vote among themselves and two Republicans took both November slots.17 Once that risk exists, parties stop leaving it to chance: they clear fields, move money, and sometimes promote the opponent they would rather face. The party still picks your finalists. It simply does it earlier and less visibly.
So rather than argue about plumbing, let me state what any such system must accomplish, and then give my answer.
- 1.Everyone votes in the election that actually decides the seat. In a district drawn safe for one party, the primary is the election. Excluding most citizens from it cannot be defended.
- 2.Supporting one candidate must never harm another candidate you would also accept. This is the test that matters most and the one pick-one fails. If backing a moderate can cost you your favorite, voters will calculate and parties will manipulate, and you arrive back where you began.
- 3.The winner must hold real majority support, not a plurality of a divided field.
- 4.An ordinary voter must be able to understand it while standing in line, and it must be countable on the equipment we already own and checkable by hand.
My answer is the simplest design I can find that passes all four: check every candidate you would accept, and the two with the most support go to November. No ranking, no rounds of elimination, no penalty for generosity. St. Louis has done it this way since 2021.
Why not ranked choice, which is what most reformers now advocate? Because of one election. Alaska, 2022. Three finalists remained: Mary Peltola, Sarah Palin, and Nick Begich. On the ballots Alaskans actually cast, Begich would have beaten both of the others in a straight two-way contest. Ranked choice eliminated him first, because fewer voters had written him in the top slot, and the seat went to a candidate a majority had ranked below him.
The theorists have a name for this and it is not a plain one. Here is the plain one: the middle gets squeezed out. The candidate who is broadly acceptable but nobody’s passion is precisely the candidate ranked choice discards, and precisely the candidate this pamphlet exists to elect. Under an approval ballot with two finalists it cannot happen: Begich is approved by Palin’s voters and by many of Peltola’s, he advances, and the runoff produces the answer the ballots already contained.
I will not oversell my own answer, either. St. Louis is now the only city in America using approval voting; North Dakota outlawed both approval and ranked choice in 2025, which ended Fargo’s experiment.18 And St. Louis has seen a repeal attempt led by members of its Board of Aldermen, on the argument that approval voting can dilute the strength of a united minority: a group voting as one for a single candidate can be outvoted by a larger group approving several. That is a serious objection and I do not think it has been well answered.19 Ranked choice’s own record is worse: in 2024 five states rejected adopting it, Missouri banned it outright, and Alaska kept its system by 743 votes after a recount; seventeen states have now banned the method.20 Neither is proven.
So it is the best answer I can come up with, and it has never been tried above the level of a city. I am very open to a better one from people who study elections for a living, on one condition: that it passes those four tests more completely than this does. The tests are the proposal. The mechanism is my answer to them, and I would rather be corrected now than be wrong at scale later.
Index funds only, while in office. Members of Congress and their spouses and children may hold broad-market funds, Treasury bonds, cash, and a home. Nothing they can move with a vote, a hearing, or a phone call. Full and fast disclosure. And the rules against evasion, through trusts, shell companies, relatives, arranged trades, written first, not added later, because “later” is where every ethics rule in American history has gone to die.
Any bill with real support from both parties gets a recorded vote. If a measure clears a meaningful threshold of bipartisan sponsorship, it goes to committee and then to the floor within a fixed window. Leadership loses the power to kill a popular bill by simply never scheduling it. Members lose the ability to hide.
If Congress misses the budget deadline, two things should happen. First, funding continues automatically at existing levels: no shutdown, no hostages, no federal workers wondering about their mortgage. That is ordinary legislation, and a bipartisan bill for it already exists. Second, the responsible chamber should face the voters at once: after a mandatory conference and protected votes on the same final package, every member of a chamber that still fails to pass it stands for a special election within ninety days. I must be honest about the legal weight of that second half. House and Senate terms are fixed in the Constitution, so the immediate election requires a constitutional amendment, not a statute.21 A narrow, defined catastrophe can postpone the election briefly; it cannot cancel it. And that pause belongs to this consequence alone: the emergency rules for borrowing, which I will come to, run on a separate standard and a separate clock, so one broad declaration can never suspend both.
Everyone gets representatives who answer to voters rather than to donors and party leadership. The right has spent a decade complaining about party machines choosing candidates; this ends it. The left has spent a decade complaining about legalized corruption; this ends that. There is less genuine ideological conflict in this repair than in anything else I propose, which is why I would do it first.
SOURCES AND NOTES FOR THIS SECTION
- 16.Iowa Code ch. 42, sec. 42.4: nonpartisan legislative staff have drawn Iowa’s congressional and legislative maps since 1980 under population-equality, contiguity, fewest-split, and compactness standards, with incumbent addresses, party registration, prior election results, and demographic data other than population counts forbidden except as federal law requires. Verified September 10, 2026. On outcomes: Jowei Chen and Jonathan Rodden, “Unintentional Gerrymandering: Political Geography and Electoral Bias in Legislatures,” Quarterly Journal of Political Science 8, no. 3 (2013): 239-269, find that neutral, compact districts produce a pro-Republican seat bias where Democratic voters are concentrated in cities, which is why the text claims a neutral process and not a neutral result. Removing politicians from the process has been associated with more competitive seats: a Brennan Center analysis of the 2024 cycle found independent commissions drew about 19 percent of districts but 41 percent of toss-ups. That figure concerns commissions, the design this plank replaced on September 10, 2026, and is cited for the direction, not for the formula’s outcomes. The 15.0-versus-5.7-percent comparison formerly in the source record was retired on August 7, 2026, after the underlying study could not be located. One state’s commission (Michigan) had thirteen districts corrected by a federal court in 2023 over its use of race; the formula confines that judgment to one published, reviewable exception.
- 17.California Secretary of State, Statement of Vote, June 5, 2012 primary, 31st Congressional District: Gary Miller (26.7 percent) and Bob Dutton (24.8 percent) advanced under the top-two rule while four Democratic candidates split roughly 48 percent of the vote in a district Democrats were favored to carry that November. Verified August 7, 2026.
- 18.City of St. Louis Board of Election Commissioners, March 4, 2025 Primary Election Final Official Results; North Dakota H.B. 1297 (2025), prohibiting approval and ranked-choice voting statewide.
- 19.The minority-vote-dilution objection to approval voting, raised in the St. Louis repeal effort, is treated in the implementation volume as serious and not yet fully answered.
- 20.Verified August 6, 2026, against Ballotpedia’s 2024 results and 2025 legislative reports: adoption measures failed in Arizona, Colorado, Idaho, Nevada, and Oregon; Missouri approved a prohibition; Alaska’s repeal failed by 743 votes after a December 2024 recount; seventeen states had banned ranked-choice voting by mid-2025. On the 2022 Alaska special election, Graham-Squire and McCune, “Ranked Choice Voting and the Center Squeeze in the Alaska 2022 Special Election” (arXiv:2303.00108), confirm from the cast-vote record that Begich would have won both head-to-head contests.
- 21.U.S. Constitution, art. I, sec. 2 (two-year House terms) and amend. XVII (six-year Senate terms). The immediate-election consequence therefore requires a constitutional amendment; automatic continuing funding is ordinary legislation with an existing bipartisan model.