So hey, something I've noticed while reading H.R. 1, section 1621.
“(g) Ballot processing and scanning requirements.—
“(1) IN GENERAL.—The State shall begin processing and scanning ballots cast by mail for tabulation at least 14 days prior to the date of the election involved.
“(2) LIMITATION.—Nothing in this subsection shall be construed to permit a State to tabulate ballots in an election before the closing of the polls on the date of the election.
Text (might take a while to load)
By requiring that ballots be scanned for tabulation before the polls close, that means the inner envelope must be opened before the polls close as well, right?
That means you either forfeit your right to correct your ballot e.g. due to errors, new information *cough*OctoberSurprise*cough*, or even coercion; or forfeit your right to a secret ballot since the state would be required to keep track of the ballot and the corresponding voter's identity in order to remove the old scanned ballot from the set of scanned ballots that are to be later tabulated.
It doesn't matter if you separate the workers that process the outer envelope and the workers that process the inner envelope, the bijection between voters and ballots must necessarily remain discoverable because the order of the envelopes must be maintained. You can't just throw the inner envelopes into a box and shake the box because then you don't know which ballot to remove in order to correct a ballot.
Even if you randomize which workers work at which tables, all it takes to reconstruct a subset of the bijection is for two people to be working on the same set of ballots, recording or memorizing the information presented (hidden fiber optic camera, anyone?), and then aligning the information with the order of the envelopes.
You're going to need a lot of people working the polls just to keep the expected nationwide number of violated ballots below unity.
My example model is for each counting precinct to randomly split the workers into two equally-large groups of processors and scanners.
Each precinct has n divided tables, where a processor and a scanner work out of sight of each other on a set of ballots, and 2p workers, with 1 <= n <= p. The workload is split into k rounds, during which each table processes s ballots and the workers are randomly selected for each table.
Procedure is as follows.
Processor:
- Processor pops and opens an outer envelope from his Work Queue.
- Processor validates the outer envelope.
- If validated, the processor pushes the inner envelope into the Scanner's Work Queue and pushes the outer envelope into the Validated Queue
- Processor then moves on to the next outer envelope in his Work Queue
- When the Processor's Work Queue is empty, the Processor pops an inner envelope from the Scanned Queue and a outer envelope from the Validated Queue, and puts the inner envelope back inside the outer envelope and seals the outer envelope.
- Processor pushes the outer envelope into the Processed Stack.
- Processor moves on to the next inner envelope in the Scanned Queue.
Scanner:
- Scanner pops and opens an inner envelope from his Work Queue.
- Scanner scans the ballot.
- Scanner puts the ballot back into the inner envelope and seals it.
- Scanner pushes the inner envelope into the Processor's Scanned Queue
- Scanner moves onto the next inner envelope from his Work Queue.
Assuming that there are exactly two conspirators in each counting precinct, for each counting precinct:
- The probability that the conspirators are placed in separate groups is 1/2.
- The probability that the conspirators are both selected for a given round, given that they are in separate groups, is ((1 choose 1)(p-1 choose n-1)/(p choose n))2 = (n/p)2
- The probability that the conspirators are both placed at the same table for a given round, given that they are selected and that they are in separate groups, is n/n2 = 1/n.
- The expected number of violated ballots for a given round is thus s*(1/2)*(n/p)2*(1/n) = (s/2)*n/p2
- Each round is independent, thus the total expectation is (k*s/2)*n/p2
And of course, each counting precinct is independent so the nationwide expected number of violated ballots for N identical counting precincts is (N*k*s/2)*n/p
2, or simply 0.5*V/p
2 where V = N*k*s*n AKA the total number of votes nationwide.
This expectation should be below unity, thus 0.5*V/p
2 < 1 <=> p > sqrt(0.5*V).
From there it follows that the minimum number of people per precinct is 2*floor(sqrt(0.5*V) + 1), which nationwide becomes a total of 2*N*floor(sqrt(0.5*V) + 1).
In the 2020 election, there were 158 383 403 (validated) votes nationwide. If we have 52 counting precincts (states + DC + overseas/territories) and if all voters voted by mail, the total number of people needed would be at minimum ~930 000 people or about 18 000 per counting precinct.
The total expectation is clearly independent from the number of tables as long as p >= n, so processing time shouldn't be an issue with enough tables. However, organizing the randomization of workers may take considerable time.
This, naturally, doesn't account for delays due to mistakes and errors that high-speed processing would inevitably cause, nor the need for observers.
Note that the number of votes is a posteriori information and thus not very useful for planning purposes. Counting precincts might use more workers than required by instead using the number of registered voters or similar.
The number of counting precincts is most likely larger in reality as well with an order of magnitude similar to that of the number of counties, so the actual number people needed nationwide might be twenty times as large or more than what the model suggests.
This also doesn't eliminate the violation of the secret ballot, the procedure only mitigates the problem.
Author
Topic: AmeriPol thread (Read 4221825 times)