Questions and Responses, 2004

Archive record — page as published on ballotintegrity.org, October 2004.

This page held a question-and-answer document the 2004 campaign wrote for volunteers who expected to be asked hard questions about its proposal. Its interest today is almost entirely practical: stripped of the advocacy framing, it is a compact explanation of how precinct hand counting actually works, written by people who had looked into it in some detail.

What has been removed

The original document was addressed directly to a volunteer preparing to meet a legislator, and framed each item accordingly — anticipating what an official or their staff might raise, and closing with encouragement and a request to report back. Those framing sentences are not reproduced, along with the contact details that accompanied them. What follows is the substance, in the past tense, as a record of the campaign's own answers.

How the proposal was to work in different counties

The first and most practical question was how any of this could happen without changes that were already impossible so close to an election. The campaign's answer distinguished between two situations.

In counties that already voted on paper — hand-marked optical-scan ballots or punch cards — nothing new needed to be printed. The federal contests were already on paper; the change was procedural, in that those contests would be counted by hand at the precinct rather than only fed through a scanner, with the machine count retained for comparison. The detailed version of that procedure is on the optical-scan procedures page.

In counties using direct-recording electronic machines or mechanical lever machines, there was no paper record to count, so the proposal required a separate short paper ballot carrying only the two or three federal contests. That is the harder version, and the campaign did not pretend otherwise; it argued that a single-sheet ballot with three contests was a small printing job compared with a full ballot.

Combining paper and electronic voting in one election

The obvious objection is that mixing methods within one election is unworkable or unprecedented. The document answered it with a specific example it returned to repeatedly: a municipal election held in Pompano Beach, Florida in March 2004, in which paper and electronic ballots were used together in the same election. The campaign cited it as a demonstration that the administrative problem had a known solution rather than as a general endorsement of how that election was run.

How long a hand count takes

The question that decided the argument in practice was time. The campaign's estimate was that counting two or three federal contests in most polling places would take under four hours after the close of polls — a figure that depends heavily on precinct size and on how the count is organised, and which it presented as an estimate rather than a measurement.

The underlying arithmetic is worth understanding, because it is what makes the estimate plausible in principle. Hand counting scales with the number of contests, not simply with the number of ballots: counting one contest across a stack of ballots is a sorting operation, and a team can process a precinct-sized stack in a manageable period. It is counting every contest on a long ballot — often dozens of races, judicial retentions and referendum questions — that turns a hand count into an all-night undertaking. Restricting the proposal to two or three federal races was not a rhetorical narrowing; it was the entire feasibility case.

Jurisdictions that already hand counted

The document also pointed out that precinct hand counting was not hypothetical: it was ordinary practice in parts of Texas and Idaho, among other places. The descriptions it gave of that practice are the most useful thing on the page, because they show what the procedure requires beyond willing hands.

Two features stand out. The first is the sequestration of counters: once counting begins, the counting team stays in the room until it is complete, so that the process cannot be interrupted and resumed with a different composition. The second is the presence of observers throughout, watching the whole count rather than sampling it. Together these turn a hand count from an unverifiable clerical exercise into an auditable procedure — and they also explain why hand counting is labour-intensive in a way that is easy to underestimate, since it commits both counters and observers for the full duration.

How the measure related to legislation then pending

Finally, the document addressed a question about a voting-systems bill then pending in the U.S. House of Representatives in the 108th Congress (H.R. 2239), which proposed a voter-verified paper record requirement. The campaign's position was that the two were not alternatives: the pending bill was a durable statutory requirement that would take effect over time, whereas its own measure was a temporary step intended for a single election that was weeks away. It also restated, as the proposal itself did, that the measure was not a position for or against electronic voting in general — the point made on the plan page.

In context: hand counting today

Full hand counting of all contests remains rare in the United States and is generally confined to very small jurisdictions, for the reason the arithmetic above implies: the labour required rises with ballot length, and American ballots are long. Where hand counting is used at scale in current practice, it is as an auditing method — examining a sample of paper ballots to check a machine-produced result — rather than as the primary count. The National Conference of State Legislatures maintains current summaries of audit and count procedures by state.

The empirical questions the 2004 document could only estimate — how accurate hand counts are relative to scanners, how long they take, how error rates vary with fatigue and team size — have since been studied directly. The MIT Election Data and Science Lab publishes research on counting methods and their error characteristics, and the Election Assistance Commission's national survey data records how jurisdictions actually count. Anyone interested in the substance of this page should read that current evidence rather than relying on a 2004 estimate.