# Council incumbency and the one-third hypothesis

Reviewed November 6, 2026. Results are the November 5, 4:17 p.m. Pacific,
Preliminary Results 4 tabulation, not certified outcomes. This supplement does
not add ballots or change the underlying snapshot.

## Download and reproduce

`election-results.zip` is the complete bundle: this supplement plus the original
seven-contest source files in `official/`. The previous November 4 and November 5 supplement ZIPs remain
unchanged. `manifest.json` records SHA-256 and byte counts for every payload.
The November 4 comparison inputs, including their raw reports, are in `previous/`.
No private correspondence, voter identities or ballot-level records are included.

- `incumbency.json`: reviewed roster source, definitions, first-round and final
  group totals, per-candidate values, immediate transfers and limitations.
- `candidates.csv`: every candidate in the two council contests, including write-ins.
- JSON vote values use integer units of 1/10,000 vote. CSV uses vote values.
- `official/analysis.json`: the original normalized seven-contest snapshot.
- `official/README.md`: detailed source/report definitions and known limitations.

From the project repository, run `python3 ingest/elections/build_results.py`, then
`npx tsx ingest/elections/build_incumbency.ts`. The page and exporter use the same
calculation module, `src/lib/election-results/incumbency.ts`. The test suite checks
against the source tallies, cross-group conservation and the complete ZIP manifest.

## What is being measured

Incumbent means holding the same council seat before this election, independently
verified against the City's current elected officials list, reviewed November 6:
https://www.portland.gov/auditor/elections/elected-city-officials

D3: Steve Novick, Tiffany Koyama Lane, Angelita Morillo.
D4: Olivia Clark, Mitch Green, Eric Zimmerman.
All remaining candidates and the uncertified write-in entry are in `others`.
First choices are votes counted for candidates in round 1, not necessarily every
literal rank-1 mark (skipped/invalid ranks are handled under the counting rules).
The denominator excludes undervotes and initially inactive ballots. It is not
turnout or the registered electorate. A combined incumbent share describes votes
in this contest; it does not estimate the causal effect of holding office.

Threshold = floor(initial valid vote value / 4) + 1. Under City Code
2.08.030(C)(2)(b)(3), a threshold-reaching candidate is assigned the quota in all
future rounds. Three equal retained quotas have identical one-third shares if
only their combined retained total is used as denominator. That is arithmetic,
not evidence of equal popularity. Other candidates still hold vote value when
counting stops. Threshold crossings occur in different rounds; their totals
cannot be summed as unique voter support. Transfer value can appear more than
once as it moves between candidates. Remaining candidates can also fill remaining
seats without reaching quota under C(2)(a); that exception was not needed here.
Rules: https://www.portland.gov/code/2/08/030

Flows name the immediate sender's pile. A challenger's pile can contain value
previously transferred by an incumbent. Do not label these original first-choice
cohorts or unique voters. Only transfers entering a reported round are included.
Incumbent held = incumbent first + other-to-incumbent - incumbent-to-other
- incumbent-to-inactive - incumbent-to-rounding. Within-group transfers cancel.
Whole ledger: incumbent held + other held + later inactive + rounding = initial.

## Ballot rankings still missing

No November 2026 CVR download was found on the County results and statistics pages
reviewed November 6. The County says statistics will follow certification Nov. 30;
that statement is not a specific promise of a CVR release date.
https://multco.us/info/november-3-2026-election-results-multnomah-county
https://multco.us/info/turnout-and-statistics-november-2026-general-election

The County previously published anonymous CVRs for 2024:
https://multco.us/info/turnout-and-statistics-november-2024-general-election

Needed for a stronger test: an anonymous CVR for both council districts, all
contributing counties, matched to a dated tabulation, with rank/overvote/undervote
codes and the tabulator configuration. This would permit candidate rank 1-6
counts, any-rank support, pairwise preferences under explicit missing-rank rules,
and overlap (zero, one, two or three eventual threshold-reaching candidates on
each ballot). Any-rank shares overlap and need not sum to 100%. None of those
statistics can be recovered by adding aggregate transfer values. “Fringe” is not
a numerical classification available from these results.

The complete ZIP also contains `official/changes.json`, a candidate-level
comparison with the previous November 4 afternoon tabulation.