Dr. Rebecca Mercuri rebuts Dr. Shiva’s analysis of the vote in Michigan

From one of our most sophisticated experts on the subject. (Shiva has not yet replied to her, Rebecca tells me.)

We all need to focus on the other, clearer signs of theft in Michigan (e.g , the late-night vote drop, rampant misconduct by election personnel, with collusion by Dominion) and elsewhere. 

11/17/20  11:41AM EST  All Rights Reserved. Permission granted to Mark Crispin Miller to publish this content, in its entirety, to his email lists and blog.

Dear Dr. Ayyadurai —

I would like to focus on the section of your recent video <> between minutes 27:30 through 43:10. I have issues with some other parts of your analysis in the latter half-hour of the video, but we can address those later. Let’s start with just this part for now. 

This part of the video is where you are discussing P = V and P != V where P is the number of people who came in to vote and V is the number of votes recorded as cast in a particular race. Note that the race has to be 1 out of n candidates in order for P = V (in races where you can pick more than one candidate, like 3 out of 5, P would necessarily be larger than V), so let’s limit our discussion to just 1 out of n races. In actuality, P is almost never equal to V in any particular race, because it is well-understood that people skip races, EVEN ones at the top of the ticket, like President and Senate, though that tends to be more rare. We refer to this skipping as undervotes. So the proper initial formula should be P >= V which is anticipated and would be more accurate to what actually happens in voting. What should NOT be happening is P < V in 1 out of n races. My strong suggestion is that you immediately STOP using P = V and P != V in your discussions, and you should issue a correction and only use P >=V and P < V going forward (also making the distinction that V represents the tally in a single 1 out of n race).

Moving on, let’s discuss your assertions about P < V for 1 out of n races. Is it possible for this to occur using hand prepared voter verified paper ballots (HPVVPB)? Certainly. A person can (and some do) OVERVOTE, in other words, they could vote for more than one candidate in a 1 out of n race. This is less likely to occur if a voting system is used to prepare the paper ballot that the voter verifies before submitting it, because the software should block overvoting.

But overvotes can and do happen when off-precinct voting (such as mail-in or absentee) occurs with hand prepared paper ballots. In-precinct, there is the scanning step where the voter provides their ballot and it should be rejected for anomalies, including overvotes, if any are detected. The voter should be (but is not necessarily always) informed about the problem and given an opportunity to correct their ballot on a new blank one (with the overvoted ballot voided). But this checking does not typically occur with mail-in / absentee voting. In the 2020 election. This year, though, I have heard about some states giving people opportunities to “cure” their ballot — which may only apply to the signature portion on the envelopes, but could (if the ballots are not first anonymized) include detecting ballots that contain overvotes. Generally speaking, though, my assumption is that overvoted mail-in / absentee ballots will just be scanned and the races that are overvoted will just not be counted. Thus, if all things are working properly with the scanners, then P < V should never occur. Similarly, if an electronic voting system is used to prepare Voter Verified Paper Ballots, the VVPBs should never have overvotes, and if the VVPBs are used for the counting (which should be, but typically is not), then again P < V should never occur. 

This, then, leaves only the scanners and the voting machines that are electronically reading and recording the votes, where P < V could erroneously occur in the tallies. Your discussion of this seems to be focused on the “Weighted Race Feature” that Bev Harris claims to have exposed in the Diebold voting system documentation. The concept of shifting votes by percentages was well known long before Bev’s discovery. I and others had written about this possibility decades earlier. In fact, this type of vote shifting is long understood to have been used with mechanical lever machines (where tally odometers are rolled back or forward as “pre-shifting” when the machines were set up at the warehouses prior to the election). With paper ballots, this shift can also (and was believed to long be) effected through votes being “stuffed” (added) or “pulled” (subtracted) prior to the vote counting. Where such shenanigans become detectable, though, is if the balance is somehow off, such that the result exposes P < V. If you were doing this shifting electronically using percentages (as in your example), then P < V should never occur, because the amount subtracted from one candidate would counterbalance the amount added for another (it is understood that some of the subtractions could be distributed to numerous 3rd party candidates, thus making the overall result seem more plausible) and the end result would NEVER exceed P (as long as the algorithm is careful about roundings). It only would exceed P if the algorithm used a FIXED number of votes to shift or add, and the counterbalance was not calculated properly, or if there were not enough votes cast in the race on that machine to overcome the fix. 

I’m not saying that this doesn’t or couldn’t happen. What I’m saying is that vote shifting (what you are calling the Weighted Race Feature) only exposes itself if the programmers are very very stupid (which they typically are not). OR, if the fact that this IS occurring is INTENDED as a FEATURE (such as via the Weighted Race Feature) to expose to party operatives that this fix is actually going on. So, here you and I may actually agree with respect to some (but not all) of your shape analysis. There are detectable features in the statistics that can clue us in to the possibility that fraud could potentially be occurring. But, even given these statistical shapes, do we really know that INTENTIONAL fraud is actually occurring, or could it just be a software bug? 

I’m inclined to believe (and have no reason not to believe) the latter, that it is an unintentional software bug. And the evidence of this may very well be in your analysis of the straight party votes! 

But first, let’s go back to the overvotes. Say the scanners are programmed incorrectly to cast ALL votes in an overvoted race. We need to also consider the undervotes as well. Say there are 1000 voters, and in a particular race, 50 people undervoted and 100 overvoted and all of the overvotes were added into the tally, then there will be an excess in V of 50, hence P < V.  It would be easy to find the 100 overvotes, though, in the hand counting, and removing all of them from the tally will leave P > V by 150 so all would be OK.

Let’s now look at Straight Ticket Voting (STV), which has been on the decline in recent years and only 6 states currently allow it. See: <>.  As a metric of possible fraud (as you are using for your later assertions in the video), it’s not so great anymore because there’s now considerably less data. But we should consider the possibility that STV also could have been miscounted by the paper ballot scanners in those states that still use it. Back in the day of lever machines, it was fairly common for people who mostly wanted to vote for one party’s candidates to pull the STV lever and then unselect the candidates they didn’t want on that ticket, and if they wanted to, add someone else in their place. That’s perfectly legit. The problem with STV on paper ballots (and maybe part of the reason why they are going extinct) is that you can’t really do a de-selection of a particular race on a paper ballot. If you select straight party, and then select a candidate of another party in one or more races, then what SHOULD happen is that the straight party vote for that race should be subtracted and the selected candidate added. Simple enough, but it is a fairly complicated software problem, especially given the many different ballot layouts that occur in a given state. So, if we consider that all of the STV subtraction is somehow not being performed or being performed incorrectly, and if all of the votes selected outside of the STV are being added, then there could be an increase in the tallies, potentially leaving P < V if the STV overvote is not offset by undervotes.

From a programming standpoint, I think this scenario is actually likely and it could explain some problems with the vote tallies in the 6 states that are still using STV. Unfortunately, we will likely never know if the software is flawed (either deliberately or unintentionally), because all of the vendors refuse to allow independent inspection of their vote tabulating equipment, ESPECIALLY after the election, which could reveal numerous problems (including alteration of internal program code). 

That’s more than enough for now. I’m hoping you’ll be open to discussion and encourage you to write back to me with your thoughts. 

Rebecca Mercuri, Ph.D.
Twitter: @NotableMercuri

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.