BREAKING REPORT: Largest Number of Cluster Votes In Wayne Co. MI Came From Psychiatric Hospital For Patients With Severe Mental Illnesses

Yesterday, data scientist Sarah Eaglesfield tweeted about some interesting facts she uncovered while looking into Wayne County, Michigan’s voter data from the November 2020 election. According to 100Percent Fed Up, her data, so far, explores cluster votes in Michigan’s Wayne County, where hundreds of affidavits have been filed sharing eye-witness accounts of voter fraud and intimidation by paid election workers and officials and Democrat activists present at the TCF Center, where absentee ballots were processed.

Eaglesfield’s tweet reads:

In order – clusters in Wayne County, Michigan with the most votes cast in #Election2020 were:

1. A psychiatric hospital (@NedStaebler is that you?)

This comment directed at Ned Staebler is in reference to his viral, videotaped, unhinged rant directed at the Republican Wayne Co. Board of Canvassers who refused to certify the county’s election last week. 

2. Apartment block (but no apartment number)

3. A convent

4. THE FOUR SEASONS Care facility (really!)

5. Homeless drop-in centre

Today, Sarah Eaglesfield shared the data she used to obtain the information she shared on Twitter yesterday. She tagged Michigan’s Democrat Secretary of State Jocelyn Benson (who insists there was no voter fraud or intimidation of poll challengers in Wayne Co), Trump attorney Sidney Powell, President Trump, Joe Biden.

RE #1: The psychiatric hospital that, according to Eaglesfield’s research, provided the largest number of cluster votes in Wayne Co.

The largest cluster of votes received in Wayne County was from the Walter P. Reuther Psychiatric Hospital, which provides treatment, care, and services to adults with severe mental illness. 97 patients at the facility either applied for a ballot or are register as absentee voters, and 78 appeared to have returned their ballot.

From the Michigan.gov website: Walter P. Reuther Psychiatric Hospital provides treatment, care and services to adults with severe mental illness.

Given that the psychiatric facility in question specifically states that it caters to those with severe mental illness it should be investigated what mental capacity these voters had and whether these ballots were potentially abused. It is my understanding that there is no capacity test for voting in Michigan, although state law allows for one to be applied.

RE #2: Apartment block ( but no numbers)

The most striking thing about the Wayne Co. dataset in this regard is that there is no normalization of data in the Mailing Address field. This is a shortcoming in the software that could be abused for fraudulent purposes.

For example, some addresses have been written using the state abbreviation, and some with the full state name. Some have the house number contained in a separate field; some have it adjoined with the street name. In one case, the zip code field has been used to note that the address is an incorrect address.

The same issue exists in the residential address field, where there is no set format for apartment or studio numbers. Apartment numbers are sometimes stored with a “#”, sometimes as “Apt”, sometimes as “Apartment”, meaning that the same apartment number in the same block could be stored differently in the system, allowing voters to receive more than one voter number.

One example of a person with the same first name, middle name, address, and year of birth, who has been assigned different voter numbers, voted twice and been counted twice, has their address stored as both:

PIO BOX 32910 FORT ST POST OFFICE, DETROIT, MI 48323 and

PO BOX 32910, DETROIT, MI 48232-0910

Due to the shortcomings with regards to address format, it is difficult to provide a thorough analysis of voter clusters. An example of the problem can be seen, for example, in apartment blocks such as The Pavillion (1 Lafayette Plaisance St) where some residents did not even provide an apartment number.

RE #3: Convent

In the case of the Felician Sisters convent in Livonia, who sadly lost 13 of their sisters due to COVID-19 in May 2020, all their deceased had already been removed from the dataset.

Washington, MI Trump rally Nov. 1, 2020.

Catholic nuns came out publicly in support of President Trump in Michigan and in other states, so it’s weird that the names of the deceased nuns were removed from the MI voter rolls, while names of dead voters across Michigan state remained on the voter rolls.

RE #4: The Four Seasons Care Facility

There are a large number of care centers catering for the elderly population within the county. Analysis found two instances where people had died in October, but their vote was still sent and counted. This isn’t problematic in itself, but there were a few care centers that stood out as suspiciouseither through having returned all their ballots from the residents on the same day (Four Seasons Care) or through having received ballots for more than one deceased person who had previously been in their care (Hope Care).

RE #5: Homeless Drop-in Center

The State’s drive to allow homeless people to vote seems to have been successful, and there were a number of votes received from homeless drop-in centers and through the Samaritans. Although these cannot be verified, the numbers are consistent with what you would expect from a raised awareness campaign.

In addition to her findings related to cluster votes, Eagelsfield also released more evidence of voter irregularities that she discovered.

Wayne County Dataset

The Wayne County Dataset was obtained from Jocelyn Benson, the Secretary of State for Michigan. It contains 613,091 lines of data, each representing a vote cast in Wayne County, including voters’ personal information (such as name, home address and year of birth). It is a partial dataset and does not contain all the voter data for Wayne County.

Forensic Analysis Questions

There are a number of questions that can be asked to help determine the integrity of the election data provided. Although one factor on its own would not point to election fraud having taken place, these questions can still help determine shortcomings in the computer systems used to process the ballots, and where human error may have occurred.

Q1. Is rejection rate in line with other years?

If the rejection rate was significantly higher than other years, this would indicate a possibility that valid ballots were being rejected. Conversely, if the rejection rate was significantly lower, it would point to potentially legal ballots being disregarded.

A1.
7,700 votes were rejected in the provided dataset, a rate of 1.256%. This is lower than the statewide 2016 Michigan rejection rate of 2.02% stated by the Election Assistance Committee in their 2016 Election Administration and Voting Survey. Although not an immediate cause for concern, given the increase in absentee ballots, rejection rates would be expected to be higher than previous elections, and ballot rejection procedures should be reviewed in the audit.

Q2. Does any voter ID appear more than once in the dataset?

As each voter ID represents a unique person, and each person is only allowed to vote once, a voter ID appearing more than once should be flagged as suspicious.

A2.

A total of 1,104 voter IDs appeared more than once in the dataset, with 4 voter IDs appearing three times, and one voter ID appearing four times.

Whilst the majority of duplicate voter IDs were processed correctly, the dataset appears to show 21 processing errors with these duplicates. 10 voters with the same voter ID who voted twice appear to have been counted twice, two of whom had changed their address, one of whom was an overseas voter. 5 voters voted twice and had neither vote counted. 6 voters had their ballot counted more than once, as detailed below.

Q3. Does any ballot ID appear more than once in the dataset?

As each ballot ID represents a unique vote, a ballot ID appearing more than once should be flagged as suspicious.

A3.

A total of six ballots were counted more than once according to the provided dataset. Of these six, four were counted twice, one was counted three times, and one was counted four times.

Q4. Were ballots received after the cut-off date rejected?

The Michigan Court of Appeals ruled that voters in Michigan must return absentee ballots to their clerk by 8 p.m. on Election Day (November 3rd) in order for their vote to count. Were any votes received after this date processed?

A4.

13 ballots are shown to have been received on 4th November, of which 11 were rejected and 2 were accepted.

It is possible that the dataset provided could have been generated using UTC, which is 5 hours ahead of Michigan Eastern Daylight Time, and this may account for the two late ballots being accepted. However, there is no exact timestamp given, and considering there is evidence that some ballots received on the 4th of November may have been processed, it is also possible that other votes were counted which arrived after 8 p.m. EDT on November 3rd, 2020.

The post BREAKING REPORT: Largest Number of Cluster Votes In Wayne Co. MI Came From Psychiatric Hospital For Patients With Severe Mental Illnesses appeared first on The Gateway Pundit.

Leave a Reply

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