Election Data Analyzer

“Perhaps we have lived with this miracle too long to be properly appreciative. Freedom is a fragile thing and is never more than one generation away from extinction. It is not ours by inheritance; it must be fought for and defended constantly by each generation, for it comes only once to a people. Those who have known freedom and then lost it have never known it again.” - Ronald Reagan

Hey folks, I have been quiet here but very busy preparing EDAv5.0 for release. This will be a major update and what is exciting about it is that the new features are directly a result of feedback from canvassers for the most part. There will be a new dashboard tab that allows you to visualize all of your pivot table data. I added a methodology to extract a statistically significant RANDOM sample of registrants from the overall population and generate a walk list. There will also be a template form that you can use built in to generate a pdf walk list that can then be printed out. Of course, these updates take 5x longer than planned so that explains the lack of case study examples posted here. I know that is why most of you tune in. Those of you with USEIP can expect a release phone call anounced in the near future.

For those of you not participating in the latest anonymous poll above, here is what people want. 77% Yes, 4% No, 19% Not sure on 57 Votes. So the Yes have it clearly and this is what will be coming next in v4.3. Thanks for the feedback.

For those of you who may be new to the channel due to the recent interview above, please read the FAQ that I have pinned. Thanks.

I am in the process of beta testing a new feature for the next release of EDA. Statistics are a complex topic so if there are folks out there who know this topic well, let me know what you think of the direction I am going here with a comment or a direct message. This is how it would work. The canvasser decides what the overall population (registration address in this case) are of interest based on your own criteria (entire county, city, subset etc). Then you put together your canvass questions for each address. You canvass the address and ask the questions. The output of that is binary: it either checks out as OK or there is a problem you discovered with the registration data. The confidence interval is measuring if you found a problem or not at the address. So let's say you want to use a 99% confidence level with a 5% interval on 50,000 registrants, the sample size is 657. This dashboard calculates that for you (continued)

This is an EDA analysis using the Pivot Filters tab with exports that I then used to manually generate this dashboard for the county with the most registrants in my state. I have been curious about the registration "surges" that happen leading up to an election. I am focusing on the Sep - Nov surges in the bottom chart on the left. The pie graph and first table below it show the total registration baseline in November 2020. The three tables below that show the distribution during the surge periods and the difference between the baseline and the surge period. See anything interesting? I do, several items. Leave some comments. The next posts we will dig deeper on this same topic using EDA to query other demographics of the "surge" registrants. How old are they? Are they clustered in certain precincts? What address types are being used? It might be interesting. Stay tuned. You can do this at home as well.

Taking a break from registration role case studies, this is the voter history role for a single county (which is a separate database in my state). We need to reverse trends and make in person voting the dominant trend again. It looks like sometime in 2004, when many of us were "asleep at the wheel" on the election integrity topic, our system was compromised. The surge in the absolute number of mail in votes can be partially explained (I suppose) by population growth in this high growth state. But looking at this, the in person votes were hovering around 5000 going back to 1990-2004. Then the mail in surge starts. From a technical perspective, voter history roles tend to be VERY large files that a spreadsheet solution is not optimal for analyzing. I also wanted to present this data type to gauge interest in potentially analyzing more of it in future posts.

To wrap this up, here is the break out of the Party of those 60 people in the registration role...and now the punch line....58 of the 60 who changed there party have a status of "Failed to Vote". So I am to believe 60 people changed their party affiliation but 58 of them forgot to vote? By the way, 60 party affiliation changes is 10x more the number of changes on any given day going back to 1944. What happened on 11/3/2020 to cause this? I can make no sense of this right now but it might be a good of example of where you would need to call your clerk for an explanation of this oddity.

I then inspected the pivot tables looking for anomalies and the only item that popped out to me was the Party Affiliation Date. As you can see above, nobody registered to vote on 11/3/2020, but, there were 60 people who apparently asked for or themselves changed their Party Affiliation on 11/3/2020.....but there is more....

To wrap up this case study with a few posts, here is what I did if you want to try this at home. Again, I did this for one county and working with a larger data set would probably yield more insights but for this purpose, it is good enough. The goal was to map the most populous retirement communities in my county. I took the original unique address data pivot table and in a new excel file, using find and replace, removed the APT and UNIT references using a simple find and replace wildcard search. I then replaced all of the address with that modified address and in a blank EDA file, ran the analysis again. This will allow EDA to roll up the number of residents at a unique address. Looking at group home of 10+ residents (I ignored the rest), there were 32 of those with exactly 900 residents. I then inspected 32 address on a google map by using the EDA feature which exports the address data. I found no issues with the address themselves they are indeed all group homes.....so far no problems....

I am going to be posting the final part of this case study looking at 85+ year olds living in retirement communities soon. Looking ahead, if you have ideas on what types of case studies would be most helpful to examine in your work with EDA, poste them in the comments. I will take those on next. I want to be most helpful to canvassers....thanks.

Before I continue the case study, I realize this analysis should be extended over all counties for this age cohort in this state to be ale to draw any firm conclusions. I took the 3817 registrants in a county over 85 yr old and put those records through EDA by itself. What can we learn from the above outputs? There are 40 people on the roles greater than 100 including one who is 110 years old. Is that realistic? About 12% of the mailed ballots were not voted by these people (returned, undeliverable etc). What happens to those ballots? In the last name length table you can see there are people using there initials instead of the actual last name to register. Not legal. From what I see in the data, there are 6 people using initials for registration name. This is the first time I have encountered total name lengths exceeding 30 characters. This is a small data set so extrapolating the results as a mental exercise might be interesting. The address analysis will be next.

Let's get back to case studies now that EDAv4.0 has been released. In "honor" of the non-sense going on in Racine, WI, I looked at elderly in my county. These are 85+ on the reg roles by reg year. The reg dates range from 1944-2021. 18% have registered to vote between 2016-2021. Some of that may be families moving there loved ones into state and updating the registration data. [edited] The next level I will look at is to extract those address that are common to large groups of these registrants and inspect them on a map and confirm it is a legitimate group home. Then, we can further dissect the data for that subset. I will rerun the subset data through EDA and see what emerges. This data was extracted from EDA using the Pivot Data Filters tab on YOB for those of you wanting to "try this at home".

Folks, As of today, EDA is now being used in 47 states. We have 100s of patriots collaborating on the USEIP EDA team plus hundreds more users helping them with data management, analytics and canvassing. This is incredible and all was accomplished grass roots style in about 4 months! There are even international supporters. All I can say is thanks for the support and more importantly, ask yourself how you can get involved using whatever talents you have. Just jump in and do something. MS, ND and VT.....ready to get involved?

TOGETHER WE WILL MAKE AMERICA GREAT AGAIN! Subscribe: @Dan_Scavino_Official