# State Ballot Position Crosswalk [[How a Candidate Gets on the Ballot]] establishes **who** is printed. This map covers **where** — the order in which qualified candidates and parties appear, who decides it, and on what basis. It is the fiftieth-state companion to [[Recount Trigger Crosswalk]], and it exists because of that map's most unexpected finding: **Georgia's SB 3EX now conditions a recount *duty* on where an office is printed on the ballot.** Ballot order has been known to affect votes since the 1950s. In at least one state it now affects procedure. The queue item that commissioned this map predicted that a state assigning first position to the governing party has written a small permanent advantage into its ballot design, and that roughly a third of the states do it. The first half is nearly right and wrong in an instructive way; the second half is unanswerable as posed, and the reason it is unanswerable is a finding. **No state gives first position to the governing party.** The party-order states give it to **whichever party won a *named prior statewide race*** — governor in Florida, Georgia, New York, Pennsylvania and Texas; **secretary of state** in Michigan and Indiana; the last presidential *or* gubernatorial race in Wisconsin. That is a different rule with a different failure mode: the position is held by the **last winner**, not the current one, so a party that takes the governorship does not inherit the top line until the following cycle, and a party that has lost power keeps it until then. The advantage is a **ratchet on a delay**, not a perquisite of office. **And the count cannot be given.** Ballotpedia's own category counts — 13 states plus DC by random draw, 8 party-based, 8 alphabetical, 4 precinct rotation, 1 incumbents-first, 1 filing-order, and **15 "multiple methods"** — partition the fifty states exactly once, which means the residual bucket of 15 absorbs states that plainly belong in the substantive columns too. Georgia, Pennsylvania, Texas and Wisconsin each order parties by a prior statewide vote *and* sit in the multiple-methods group. So "8 party-based states" undercounts party-ordered states, and no published fifty-state ballot-order number is comparable to any other. This is the same defect [[Recount Trigger Crosswalk]] found in the recount tables, arrived at independently. The third finding is the one worth the page. **The measured harm and the statutory design point in opposite directions.** The best-identified estimates — Ho and Imai on California's alphabet lottery, 80 statewide races, 1978–2002 — find **no detectable primacy effect for major-party candidates in general elections** (D +0.05 pp, R −0.06 pp, both inside standard error) and roughly **+2 points of party vote in primaries** (D +1.89, R +2.16), rising to **+3.44** for nonpartisan contests and **+6.59** for Libertarians in a primary. General-election *party column* order is precisely the thing the party-order statutes control, and it is the one context where the effect is near zero. The contexts where the effect is large — primaries, nonpartisan and low-salience races — are mostly ordered by lot or alphabet, because there is no party to rank. **The states that assign order assign it where it does not matter much, and randomize it where it does.** > [!note] How to read this crosswalk > `■` marks a part · `▸` marks a design, tier or column · `⚡` marks what sets the order · `◇` marks who decides it · ` marks what it costs the loser · `⚖` marks the legal status of a challenge · `⊘` marks a duty or remedy that does not exist · `✱` marks the structural point of a block · `⚑` marks something that **changed for November 2026** · `◈` marks **a count that is a definitional artifact rather than a fact**, explained in the Notes · `†` marks **a figure taken from the primary source against a tracker that disagrees** · `⊙` marks **an inference from ballot layout rather than from statutory text**, flagged as such · `[bracketed]` notes are dated caveats. Everything inside the fence is a **statute, a design or a state** — **no current officeholder is named anywhere on this map**, by design, so it stays true as seats turn over on 3 November 2026. ``` STATE BALLOT POSITION CROSSWALK — who gets the top line, who decides, and the one state where position became procedure (as of 10 October 2026) │ Legend: ▸ design / column ⚡ what sets the order ◇ who decides $ cost to loser │ ⚖ legal status of a challenge ⊘ does not exist ✱ structural point │ ⚑ changed for November 2026 ◈ count is a definitional artifact │ † primary source over tracker ⊙ inference from layout, not text │ │ THE ONE-LINE VERSION: THE STATES THAT ASSIGN BALLOT ORDER ASSIGN IT IN THE ONE │ CONTEXT WHERE THE MEASURED EFFECT IS ~ZERO (major-party general elections), AND │ RANDOMIZE IT IN THE CONTEXTS WHERE THE EFFECT IS +2 TO +6 POINTS (primaries, │ nonpartisan races). Read Part D before drawing any conclusion from Part A. │ ■ PART A — THE SIX DESIGNS (and why they do not add to fifty) │ ├─▸ 1. PARTY-BASED ORDER ........................ 8 states ◈ see Note 1 │ ⚡ the party's vote in a NAMED PRIOR STATEWIDE RACE, descending │ ◇ nobody, each cycle — the arithmetic of the last election decides │ ✱ NOT "the governing party." The top line belongs to the LAST WINNER of a │ specified race, so it lags the actual transfer of power by one cycle. │ FL · GA · NY · PA · TX ....... last GOVERNOR's race │ MI · IN ..................... last SECRETARY OF STATE's race │ WI ......................... last PRESIDENTIAL *or* GUBERNATORIAL race │ MO ......................... the party that WON the last governor's race │ [MD and DE order by majority/principal-minority party rather than by a vote │ count; Ballotpedia's 8-state count does not include them. See Note 1.] │ ├─▸ 2. RANDOM DRAW / LOT ........................ 13 states + DC ◈ │ ⚡ a drawing — bingo ball, coin, envelope, or a lot drawn by the state board │ ◇ the secretary of state, the state board, or a county board │ ✱ THE ONLY DESIGN THAT ASSIGNS NO ADVANTAGE ON PURPOSE. It does not remove │ the advantage — somebody still prints first — it removes the PATTERN, │ so the gain is noise across cycles rather than a standing asset. │ AR (county board) · KY · NH · SD · VA · WA · WV · DC │ RI · OK ................ parties by lot, then independents │ ├─▸ 3. ALPHABETICAL ............................. 8 states ◈ │ ⚡ the candidate's surname │ ◇ nobody — and that is the problem: the advantage is INHERITED │ ✱ A FIXED ALPHABET IS THE ONLY DESIGN IN WHICH THE ADVANTAGE IS HEREDITARY. │ AL · HI · LA · ME · MS · NV · VT ...... straight alphabetical │ TN ................................... primary ballots only │ ├─▸ 4. RANDOMIZED ALPHABET ...................... CA · NM · OR · UT │ ⚡ a letter drawn by lot, then the alphabet rotated from it │ ◇ the secretary of state │ ✱ THIS IS WHY WE KNOW ANY OF THIS. California's lottery is a randomized │ natural experiment, and every credible primacy estimate in Part D comes │ out of it. The state that neutralised the effect is the state that │ measured it. │ ├─▸ 5. ROTATION ................................. ~6 states ◈ │ ⚡ the order CHANGES between precincts, districts or ballot styles │ ◇ the statute, mechanically │ ✱ THE ONLY DESIGN THAT CANCELS THE EFFECT RATHER THAN ASSIGNING OR HIDING │ IT. Every candidate is first somewhere, so the primacy gain nets out │ inside the race instead of across cycles. It is also the most expensive │ to administer, which is why it is the rarest. │ OH (by precinct) · MI · NE · WY · KS (by geographic division) · MT │ ND (alternated) · ID (rotated; random alphabet under 100k registrants) │ AK (rotated, except state representative by random draw) │ MN (president by formula, other offices rotated) │ └─▸ 6. THE SINGLETONS ........................... 2 states MA ...... ⚡ INCUMBENTS FIRST, then parties, then everyone else ✱ the only state that writes the incumbency advantage into the ballot itself, as a rule about the officeholder rather than about their party. Nobody litigates it. IL ...... ⚡ ORDER OF FILING, with a lottery among simultaneous filers ✱ converts ballot position into a queueing contest decided before any voter is involved, at the clerk's counter. │ ├─▸ RESIDUAL. "MULTIPLE METHODS" ................ 15 states ◈ THIS IS THE DEFECT │ ✱ Ballotpedia's seven categories partition the 50 states EXACTLY ONCE, so │ 15 states sit in a bucket that overlaps every substantive column above. │ GA · PA · TX · WI each order parties by a prior statewide vote AND │ appear here. Therefore "8 party-based states" UNDERCOUNTS the │ party-ordered states, and NO fifty-state ballot-order count in │ circulation is comparable to any other. Do not add these columns. │ ■ PART B — GEORGIA: WHERE POSITION BECAME PROCEDURE ⚑ NEW FOR NOVEMBER 2026 │ ├─ ⚑ SB 3EX — signed 25 June 2026 (special session). House 94–79 (R 93–3, │ D 1–76); Senate concurred 36–16 (R 32 yes; D 4 yes, 16 no). │ ├─▸ WHAT IT DOES │ ⚡ An AUTOMATIC FULL HAND RECOUNT of every ballot cast statewide, ordered │ by the secretary of state, REPLACING the machine recount counties │ otherwise run. The hand count becomes the official result. │ ⚡ TWO CONDITIONS, BOTH OF WHICH MUST HOLD: │ (1) the certified margin is within 0.5%, AND │ (2) the office "appears FIRST OR SECOND on the ballot" │ ◇ the secretary of state orders it; the state reimburses counties that │ finish within 17 days of the order │ ▸ EIGHT COVERED OFFICES: Governor · Lieutenant Governor · Secretary of │ State · Attorney General · State School Superintendent · Commissioner │ of Insurance · Commissioner of Agriculture · Commissioner of Labor │ ├─▸ ✱ THE STRUCTURAL POINT — CONDITION (2) IS A WORKLOAD CAP WRITTEN AS A RULE │ ABOUT TYPOGRAPHY, AND IT EXISTS NOWHERE ELSE IN AMERICAN ELECTION LAW. │ ⊙ In a Georgia gubernatorial cycle, Governor and Lieutenant Governor are │ the offices printed first and second — so of the eight offices NAMED, │ the position condition reaches TWO. The other six are covered in text │ and unreachable in that cycle's layout. [⊙ INFERENCE FROM BALLOT LAYOUT, │ NOT FROM BILL TEXT. The statute's own language is the position test; the │ count of offices it actually reaches in a given year follows from how │ that year's ballot is ordered. Verify against the certified 2026 ballot │ before relying on the number.] │ ✱ TWO RACES AT AN IDENTICAL MARGIN NOW GET DIFFERENT PROCEDURE DEPENDING │ ON WHERE THEY WERE PRINTED. The 0.5% margin is the same; the duty is not. │ ✱ AND IT COMPOUNDS, because Georgia is ALSO a party-order state (Part A.1): │ the party column is ordered by the LAST governor's race, and now the │ office order down the ballot gates a recount duty. Two separate │ typographic variables, each set by something other than this election. │ └─ ⚑ The same bill delayed the 2024 ban on QR/bar-code ballot tabulation from 1 July 2026 to 2028 and ordered a committee to recommend equipment standards. [Logged here because it rode the same vehicle; it is not a ballot-position provision.] │ ■ PART C — WHAT ORDER IS ACTUALLY WORTH (the primacy effect, measured) │ ├─▸ SOURCE: Ho & Imai on the CALIFORNIA ALPHABET LOTTERY — 80 statewide races │ (36 general, 44 primary), 473 candidates, 1978–2002. Randomized order, so │ these are causal estimates rather than correlations. † │ ├─▸ GENERAL ELECTIONS — percentage points of total vote, gain from first position │ Democrats ................. +0.05 (SE 0.46) ⊘ NO DETECTABLE EFFECT │ Republicans ............... −0.06 (SE 0.53) ⊘ NO DETECTABLE EFFECT │ Green ..................... +0.56 │ Reform .................... +0.25 │ Libertarian ............... +0.23 │ American Independent ...... +0.16 │ NONPARTISAN ............... +1.95 (SE 0.30) ◀ largest general effect │ ├─▸ PRIMARY ELECTIONS — percentage points of PARTY vote │ Republicans ............... +2.16 (SE 0.46) │ Democrats ................. +1.89 (SE 0.32) │ NONPARTISAN ............... +3.44 (SE 0.78) │ Green ..................... +3.15 │ Peace and Freedom ......... +6.31 │ Libertarian ............... +6.59 ◀ ~50% of baseline vote share │ [Median primary effect in 1998 and 2000: roughly 1.6 points.] │ ├─▸ OUTCOME IMPACT │ ⚡ Ballot order "might have changed the winner in as many as 12 PERCENT of │ all primary races examined" — significant positive effects in 7 of 59. │ † THE 9.5-POINT FIGURE IS DEAD. Krosnick, Miller & Tichy's much-cited │ 9.5 pp estimate for the 2000 presidential race shrinks to about 1 point │ and loses significance once district partisanship and multiple testing │ are accounted for. Any source still quoting 9.5 is quoting a reanalysed │ number. │ └─▸ ✱ THE SHAPE OF IT: THE EFFECT GROWS AS INFORMATION FALLS. It is ~zero where voters have a party label and an incumbency cue, and largest in primaries, nonpartisan contests and low-salience down-ballot races. Which means the party-order statutes of Part A.1 govern the LOW-EFFECT context, and the lotteries and alphabets govern the HIGH-EFFECT one. The design is misaligned with the harm in both directions at once. │ ■ PART D — WHY NOBODY HAS WON A BALLOT-ORDER CASE (and what changed in 2026) │ ├─▸ THE WALL: Jacobson v. Florida Secretary of State, No. 19-14552 │ (11th Cir., 3 Sept. 2020) │ ⚡ Florida gives first position to the party that won the last governor's │ race. The district court declared it unconstitutional and enjoined the │ secretary of state and all 67 county supervisors. │ ⚖ VACATED AND REMANDED WITH INSTRUCTIONS TO DISMISS FOR LACK OF │ JURISDICTION — on two independent grounds, NEITHER on the merits: │ ⊘ NO ARTICLE III STANDING. No plaintiff proved injury in fact; and any │ injury was neither traceable to the secretary of state nor │ redressable against her, because it is the COUNTY SUPERVISORS — │ not sued — who actually place candidates on the ballot. │ ⊘ NONJUSTICIABLE POLITICAL QUESTION, on Rucho v. Common Cause. │ ✱ THIS IS A REMEDY PROBLEM, NOT AN EVIDENCE PROBLEM. The harm in Part C │ is real, small and spread across millions of voters, which is exactly │ the profile that defeats particularised injury. The official who WRITES │ the rule is not the official who PRINTS the ballot, so even a plaintiff │ with an injury sues the wrong defendant. │ ├─▸ ⚑ THE DOOR: Bost v. Illinois State Board of Elections, No. 24-568 │ (decided 14 January 2026, 7–2, Roberts; Jackson and Sotomayor dissenting) │ ⚡ Held: POLITICAL CANDIDATES HAVE ARTICLE III STANDING to challenge the │ election rules governing their own races, before voting or counting │ begins — a candidate has a "concrete and particularized interest" in │ the rules under which votes for them will be counted. │ ⚖ STANDING ONLY. The Court decided nothing about the underlying Illinois │ mail-ballot grace period, which it left for separate consideration. │ ✱ Jacobson's plaintiffs were VOTERS AND ORGANIZATIONS. Bost addresses the │ plaintiff class Jacobson's reasoning never reached. Whether a CANDIDATE │ disadvantaged by a ballot-order statute can now get to the merits in the │ Eleventh Circuit is UNTESTED — but the standing ground Jacobson rested │ on is narrower today than it was in 2020. │ ⊘ Nothing in Bost touches the political-question holding, which is the │ second and independent lock on the door. │ └─▸ $ WHAT IT COSTS THE DISADVANTAGED PARTY TO DO NOTHING: in a party-order state, the top line for a full cycle, worth ~0 points in a major-party general election and up to ~2 points in that state's primaries — where the party-order rule does not apply. The asset is worth least exactly where it is assigned. │ ■ PART E — WHAT TO WATCH │ ├─ ⚑ NORTH CAROLINA ran its 2026 drawing on 19 DECEMBER 2025 under │ N.C.G.S. § 163-165.6(c): a bingo ball drew the letter "F" and a coin flip │ landed tails, giving REVERSE alphabetical order from F — F, E, D, C, B, A, │ Z, Y, X, W, V, U, T, S, R, Q, P, O, N, M, L, K, J, I, H, G — for every │ 2026 primary and general contest except municipal races, which run strict │ alphabetical from A. The drawing was held the day candidate filing closed. │ ├─ ⊘ THERE IS NO FEDERAL BALLOT-ORDER RULE. HAVA sets requirements for voting │ systems, provisional ballots and registration; it says nothing about the │ order in which candidates are printed. Order is wholly a state question. │ └─ ⊙ THE OPEN QUESTION THIS MAP DOES NOT ANSWER: whether any state has ever measured the primacy effect inside its OWN design. California can, because it randomizes. A fixed-alphabet or party-order state cannot — there is no variation to measure against — so the states with the strongest claim to be assigning an advantage are the states structurally incapable of proving they are not. ``` ## Notes - **Note 1 — the counts are definitional artifacts, marked `◈`.** The category counts in Part A (13 + DC random draw · 8 party-based · 8 alphabetical · 4 precinct rotation · 1 incumbents-first · 1 filing-order · 15 multiple methods) are Ballotpedia's own, current as of that page's August 2025 revision, and they partition the fifty states exactly once. Reading the underlying state-by-state table instead produces materially larger groups — around 15 states ordering parties by a prior statewide result, and around 10 rotating — because the residual "multiple methods" bucket overlaps every substantive column. **Georgia, Pennsylvania, Texas and Wisconsin appear in both the party-based list and the multiple-methods list.** Treat every number in Part A as a count of *how one tracker classified a state*, never as a count of states. This is the same defect [[Recount Trigger Crosswalk]] documented in the fifty-state recount tables, found independently here. - **Note 2 — Maryland and Delaware are party-order states that the 8-state count excludes.** Maryland lists the majority party first, then the principal minority party, then parties by registration; Delaware has a fixed party order. Neither keys to a vote count, which appears to be why the tracker's party-based category omits them — and which is itself a distinction worth keeping, because a registration-based or fixed order does not lag a transfer of power the way a prior-vote rule does. - **Note 3 — the `⊙` marker is an inference, not a statute.** Part B's claim that SB 3EX's position condition reaches two of its eight named offices in a gubernatorial cycle follows from how a Georgia statewide ballot is ordered, not from the bill's text. The statute's operative language is the first-or-second test. Anyone relying on the office count should check it against the certified 2026 ballot layout. - **Note 4 — Georgia's existing recount right is unaffected.** SB 3EX's automatic hand recount sits alongside, not in place of, Georgia's long-standing right for a candidate to *request* a recount at a margin of 0.5% or less, which applies to any office and carries no position condition. The new provision adds a *duty* at the top of the ballot; it removes nothing. See [[Recount Trigger Crosswalk]]. - **Note 5 — the 9.5-point figure marked `†` in Part C is superseded.** Krosnick, Miller and Tichy's widely quoted 9.5 pp estimate for Bush in the 2000 presidential race is reduced to roughly 1 point, not statistically significant, once district partisanship and multiple testing are accounted for. It still circulates in advocacy material and in litigation filings. - **Note 6 — no current officeholder is named on this map.** The only person named anywhere on the page is the signer of a dated bill. Every other cell is a statute, a design or a state, so the map does not decay on 3 November 2026. [[Brian Kemp]] signed SB 3EX on 25 June 2026. - **Note 7 — what this map deliberately leaves open.** Whether *Bost* actually reopens ballot-order litigation is untested and cannot be resolved from the opinion: *Bost* is a standing holding, and *Jacobson* rested on standing **and** on the political question doctrine as independent grounds. A post-*Bost* candidate plaintiff clears the first lock and walks into the second. Nothing in the record says how the Eleventh Circuit would treat it. ## See also - [[Recount Trigger Crosswalk]] - [[How a Candidate Gets on the Ballot]] - [[Ballot Order Effect]] - [[Jacobson v. Florida Secretary of State]] - [[Bost v. Illinois State Board of Elections]] - [[Post-Election Audit Crosswalk]] - [[Ballot Marking Validity Crosswalk]] - [[How a Ballot Is Counted]] - [[How a Midterm Election Is Certified]] - [[State Election Administration Tree]] - [[Help America Vote Act]] - [[Political Question Doctrine]] - [[Rucho v. Common Cause]] - [[Foundations Hub]] - [[2026 Midterms Hub]]