San Diego County, November 3, 2026: Which Contests the 1% Manual Tally Checks

CitizensOversight (2026-11-20) RayLutz

This Page: copswiki.org/Common/M2072

More Info: ElectionAudits

San Diego County, November 3, 2026: Which Contests the 1% Manual Tally Checks

Summary

California requires each county to hand count a random sample of its ballots after every election. This is the 1% manual tally. San Diego County drew its sample in two public drawings. A Citizens' Oversight observer recorded the batches drawn.

We matched the batches from the first drawing to the county's own published report of votes by batch. The sample checks the statewide and countywide contests well. It checks many local contests barely or not at all. Six contests have no ballots in the sample. Twenty-one more have fewer than ten.

The county already publishes the vote totals of every batch it scanned. So the totals of the sampled batches are already public.

The data

The county's report is the "Unofficial Semi Statement of Votes Cast by Batch," in two Excel files on sdvote.com. Both were last saved on November 7, 2026. They list each batch on each scanner, with the votes for every choice in every contest. Together they cover 5,580 batches of mail ballots and 1,258 batches from vote centers, and 766,723 ballots.

The drawings were held on November 4 and November 9. The list of batches drawn comes from the county's own sheets at the drawings.

  • November 4: 13 vote center batches and 57 mail batches.
  • November 9: 20 mail batches, from mail scanned after the first drawing.

Each batch is named by a number such as 4006-182. The first digit is 4 for mail and 6 for vote centers. The next three digits name the scanner. The number after the dash is the batch on that scanner. So 4006-182 is mail scanner HiPro 06, batch 182. Every batch from the November 4 drawing appears in the county's report.

The November 9 batches are not in the report, which was saved before that drawing. The second drawing started at batch 706 of scanner HiPro 01, and the report ends at batch 705 of that scanner. So the report holds the batches the first drawing chose from. This analysis covers the 70 batches of the first drawing.

How large the first sample is

The 70 batches hold 8,030 ballots. That is 1.05% of the ballots in the report. Vote centers gave 1,304 of them and mail gave 6,726.

The 57 mail batches are 1.02% of the 5,580 mail batches. The 13 vote center batches are 1.03% of the 1,258 vote center batches. This holds if all 1,258 vote center batches were in the drawing on November 4. The county's report to the Secretary of State will give its own count.

Ballots sampled in each contest

The election has 220 contests. Every ballot in the county has the statewide and countywide contests, so each has about 8,000 sampled ballots. Local contests appear on fewer ballots, and their samples are much smaller.

Sampled ballots in the contest Contests
None 6
1 to 9 21
10 to 99 58
100 or more 135

The six contests with no sampled ballots are:

  • City of San Marcos council district 2, with 6,836 ballots in the report.
  • City of Solana Beach council district 4, with 1,431.
  • City of Imperial Beach council district 1, with 941.
  • City of Solana Beach council district 2, with 875.
  • Permanent Road Division 9B, Harris Annex, Measure Q, with 56.
  • Canebrake County Water District board, with 18.

Two contests have a single sampled ballot: Solana Beach Measure C, with 4,631 ballots in the report, and Poway council district 3, with 3,789.

California Elections Code section 15360 requires one more batch for each contest left out of the 1% sample. Only that contest is counted in the added batch. The six contests above would each need one. We do not yet know whether the county added batches for them. The batches from the second drawing may cover some of them.

A contest with one sampled ballot is not left out, so the code adds nothing for it. But one ballot cannot check how the machines counted a contest. Citizens' Oversight holds that every contest needs a real sample, from at least one whole precinct or batch.

Why mail batches sample local contests poorly

Mail ballots are scanned in the order they arrive. A mail batch holds ballots from all over the county. A local contest is spread across hundreds of batches, a few ballots in each. San Marcos council district 2 appears in 212 mail and vote center batches.

A vote center batch holds ballots cast at one place. It can hold many ballots of one local contest. One vote center batch holds 270 ballots of San Marcos council district 2.

So adding one vote center batch for each thin contest would check it far better than more mail batches. The attached table of thin contests gives, for each one, the batch in the report with the most of its ballots.

Small cells and the secrecy of the ballot

Batch totals are sometimes withheld on the grounds that a contest with fewer than ten voters in a batch could reveal how they voted. This section measures how often that occurs.

A cell here is one contest in one batch. A small cell has fewer than ten ballots. A cell is unanimous when every ballot in it chose the same candidate or side, with no blank or overvoted contest.

Sampled batches All batches in San Diego's report
Cells 5,003 478,806
Small cells 1,027 99,729
Small and unanimous 447 40,394

Small cells are common. 56 of the 70 sampled batches have at least one. But San Diego already publishes the totals of every batch, with about 100,000 small cells. The totals of the sampled batches are a small part of that public record.

A small cell reveals a vote only to someone who knows whose ballot is in that batch. Mail batches are formed by the order of scanning, not by voter.

Attached files

Next steps

  • Repeat this analysis for the 20 batches of the second drawing, once the county publishes a later batch report.
  • Compare the hand counts with these machine totals, once the county publishes the results of the manual tally.
  • Ask the county whether it added batches for the six contests left out.

Notes on method

The batch reports were flattened into one table of scanner, batch, contest, choice and votes. For each contest, the ballots in a batch are its votes divided by the number of choices allowed. The reports do not state that number. It was taken as the largest number that divides the total of every batch row. Seventeen contests allow more than one choice, from two to seven. The scripts are attached.

The status of all our 2026 letters and requests is on Auditing the 2026 General Election with AuditEngine.

View MD Edit MD Attach