Manual Tally Oversight Procedure
Citizens Oversight (2016-11-07) Ray Lutz
This Page: http://www.copswiki.org/Common/ManualTallyOversight
More Info: Election Integrity
, Election Team
, Snapshot Protocol
, Snapshot Protocol
The 1% manual tally in San Diego: (click image for full size)
This procedure is concerning oversight of the Manual Tally itself.
Prior to this procedure, please see Random Draw Oversight
for overseeing the random draw which is just prior to this phase.
The manual tally audit differs from state to state. In California it's called the "1% manual tally" and is specified by Election Code section 15360. California requires that all races have at least one precinct which is tallied, and all races are to be tallied in 1% of the precincts. They can also use a two-step process where they tally by precinct and then tally by batch. However, we find a general problem with the batch method, because they do not produce a report of all batches prior to the selection process but instead make those reports after the selection. This breaks our oversight protocol since it would be easy to then "unfix" any batches that were previously hacked once the insiders know what the selected batches are.
The manual tally process can take many days and most is not too important to oversee since they know this is a public process. The issue is not what happens in the room during tallying, but what happens before and after. The step before the Manual Tally process the random draw, and the related procedure is the Random Draw Oversight
procedure. Although this will seem boring and uneventful, it could potentially detect a central tabulator hack. We have learned a lot about what to look for. Mainly, it is at the beginning and the end.
During the Manual Tally process itself, you should note the following:
- GETTING THE BALLOTS:First thing we want to know is how the ballots were secured prior to getting into the manual tally room. Did staff have an opportunity to pre-count the ballots in the precincts so they would match the computer reports? Are the boxes "sealed" before coming into the room? Try to watch election staff access the ballots from storage.
- What we really DO NOT want to see is excessive handling of the ballots prior to getting into the manual tally room. A bad audit process is one where elections staff has to go into the ballot boxes and rifle through them for a week with 40 employees looking for the ballots to be counted, probably using a computer report to help them find them. (This is actually how they did it in San Diego in the 2016 Primary making us wonder what they are trying to hide, given that Bernie Sanders won over Hillary Clinton by 58% to 42% in the primary, while the VBM ballots, which received all this additional handling, we noticed Hillary Clinton beat Bernie Sanders by the same (nearly exact) margin, 58% to 42%. Definitely, this is a red flag!)
- TALLYING:Although this looks easy, manual tallying is pretty difficult for humans to do accurately. Studies show the error rate of the read and tally method is about 1% to 2%. So definitely, you must not be disruptive at all in the room.
- Change of Method -- If the registrar changes from PRECINCT to BATCH or vice versa in the middle of the process, we need to know about this as it is a serious red flag.
- What method are they using?
- Read and Tally -- In this method, there are usually one (or two) readers and two talliers. The readers read off the races on each ballot, usually processing each ballot in full (all races) before they go on to another ballot. The two tallier work independently but in synchronized fashion, thereby checked their progress as they go. This method is faster and has a lower error rate than the sort and stack method.
- Sort and Stack -- In this method, the ballots are processed one race at a time, and they are sorted by vote. So in one pile might be "Yes" and other one "No". If there are multiple candidates, then there maybe many stacks. Once sorted, the stacks are counted. This sounds faster and less error prone than the Read and Tally method but it is slower and more error prone. So this method should not be used as the primary method.
- No knowledge of correct results: Talliers should NOT know the numbers in the race. Sometimes we see the supervisor walking around with the computer report and hinting that they should re-tally until they can match the report.
- I have been told that in Chicago City election district (which is different from Cook County, the rest of the county) they stop tallying when they get to the desired tally as the readers continue to read. If they do this, then truly our election system is corrupt! Please let us know about this and try to document it as much as you can.
- Variances -- The big thing we are looking for here is large variances of maybe a few votes to perhaps 10 votes per precinct that can't be explained as misfeeds, voter mis-marking, etc.
- Re-scanning -- What we really DO NOT want to see is RESCANNING the Precinct (or batch) and then seeing that there was no discrepancy in the rescan, and then ignoring the problem. If the rescan of the ballots comes up with a different answer they have only proved there is potentially a serious problem
- No variances -- Another red flag is no variances at all. We do expect some minor variances due to misfeeds, double feeds, voter intent mistakes, etc.
- Actual results -- What were the counts in the races that were tallied? Reports that we have seen DO NOT include this critical information, only that "there were no variances." We need those total numbers so we can compare with the Snapshot Data File to see if the "correct results" have been modified from what they initially said they were going to be. Therefore, please DOCUMENT THE FINAL RESULTS OF EACH OF THE RACES TALLIED because they usually do not provide this information!
- FOR EXAMPLE: Assume the precinct being tallied was fixed by a compromised employee, flipping ten votes in favor of the desired candidate and away from the other. If this were done in say 1000 precincts, that would be a flip of 10,000 votes or a change in votes of 20,000 (since they are flipped). This can change and election! So assume we do get the Snapshot data file of all precincts before the Random Draw. Then this data file will include the fraudulently changed results. The insiders learn of the precincts selected for the manual tally audit, and they change those results back so they will match the ballots (and then make changes elsewhere to account for the 10 flipped votes). The produce a new report for the manual tally teams that they know will match. This is tallied, and they find no variance. Unless you get the vote counts in their report, you can't detect the hack. you need to get the vote counts AND compare with the Snap Shot data file. Only then can you detect the hack and corner the fraudsters.