A Replication of My Work.

Mr. Brian Amos, Ph.D. candidate at the University of Florida was dedicated enough to replicate some of my work and acknowledge that he gets the same results I reported.

He does have a few disagreements with my approach. For example, what he describes as a nitpick, I would respond with: That’s a feature, not a bug! My choice of limiting an analysis to the precincts with more than 500 votes cast results in what he considers an overemphasis on the effect I’m am concerned with. This is absolutely true. That particular analysis was designed to draw out that effect and make it more apparent. The vote share data is very noisy and impacted by many different factors. The trend is real, but is easily missed in the inherent noise of the larger dataset.

Wichita 2014 Election Results
Wichita 2014 Election Results

Mr. Ames wonders if some other, correlated factor such as the voter registration numbers, would display a similar trend in the cumulative chart. He shows this is true for the share of Republicans in this particular data set. But this is not a universally correlated trait across the different states where such trends have been found, and it was not enough in Sedgwick County Kansas to account for the difference in vote share.

I discuss this factor at more length in my recently published paper “Audits of Paper Records to Verify Electronic Voting Machine Tabulated Results” in the Summer 2016 issue of The Kansas Journal of Law and Public Policy. The graph displayed above is from that paper, illustrating that although there is an upswing the cumulative graph for share of Republicans, it is much smaller than the upward surge of the vote share for various republican candidates in 2014.

His parting comment “While the charts may be explainable through vote fraud, there are other, perfectly innocuous explanations that can be put forward, as well.” is true. Yes, there are other possible and innocuous explanations. Statistical analysis only illuminates correlations and other relationships. Further investigation is needed to determine cause. Just because the trend is a predicted sign of election fraud does not mean election fraud occurred.

The only way to tell if our machine tabulated vote count is accurate or undermined is to conduct a proper audit. That’s never been done here in Sedgwick County. I’ve requested access to do this as a voter and been denied. I filed the proper paperwork in a timely manner asking for a recount of those records after the 2014 election and was denied. I’ve sued for access as an academic researcher and been denied.

Why should I trust a vote count that our officials will not allow to be publicly verified? Why should anyone?

Another Analysis of 2016 Democratic Primary

This is a solid analysis. I say this without having vetted their data collection, I’m assuming they did that part right. If so, the conclusion is obvious. They authors confine all analysis to the appendix, so you can read the paper without having to understand any math.

Are we witnessing a dishonest election?

They found Sanders won 51% to 49% in places that had a paper trail. They found Clinton wins 65% to 35% in places that don’t. That’s amazing! Yes, those are different states. Yes, they looked at a different possible causes They tested for that difference while accounting for the % whites and the ‘blueness’ of the state. No, they didn’t find anything sufficient to explain that difference.

You don’t have to be a statistician to understand that’s a huge difference in proportion. It helps to be a statistician to understand the tests they ran checking other explanations and the resulting output. They are running appropriate tests and the output is unequivocal. Which they stated. I concur.

“As such, as a whole, these data suggest that election fraud is occurring in the 2016 Democratic Party Presidential Primary election. This fraud has overwhelmingly benefited Secretary Clinton at the expense of Senator Sanders.”

Redacted tonight makes this article their lead story.

BTW, I absolutely loved their fake commercial for “Shut your f***ing tweethole” at the 15 min mark.

Authors response to criticisms

My work, some of my graphs and my previous post, are included in the appendix of the response article. Lots of interesting graphs there too.

An Open Letter to Bernie Sanders

Dear Bernie,

If you want to win the presidency and elect a revolutionary congress, you must find a way to force accurate counts of votes across the country. There is no reason to believe that machine generated vote counts are accurate when they are not checked for accuracy. This is particularly difficult in places like South Carolina and parts of Kansas, where no paper trail exists to even attempt a public recount. Or Arizona where manual hand counting of ballots is not permitted.

I live in Kansas. I’m a professional statistician and an ASQ Certified Quality Engineer. I find certain patterns in election results quite disturbing. Graphs of Oklahoma primary results are below. Both exhibit a common and concerning pattern: as the number of votes cast in a precinct increases, so does the vote share for the candidate favored by the Washington establishment. This pattern is NOT due to random chance nor do voter demographics explain it. In the fall, the Republican candidates across the board can be expected to show such a pattern wherever machine counting of votes is combined with poor to non-existent auditing of those results. The pattern is consistent with election rigging.

Citizens like myself have had little success in forcing our officials to show the paper trails so we can have confidence in their reported results. I’ve been trying for more than three years to get access to the paper records that would allow me to assess how accurate our computer tabulated official vote counts are. After my latest legal setback, it will be another year before I might get permission. In the meantime, we will be having another election on non-transparent voting machines.

You, as a candidate, have the right to demand manual recounts. Well, in some places anyway. If you were to do so, irrefutable evidence of problems with vote counts will emerge in some of those places. If and only if your supporters can find and correct those problems can your revolution win at the ballot box.

In states that have paper trails, I suggest you start asking for manual recounts of the paper ballots and Voter Verified Paper Audit Trails (VVPAT) where you can. Whether you won or lost the contest doesn’t matter. The point is to evaluate the size and number of discrepancies and check for bias. Laws vary from state to state. Typically there is a short window of time to request recounts. Many jurisdictions will balk and try to keep you from doing so by various legal maneuvers. But there will be many opportunities through the primary season. You have supporters that can be trained and provide labor hours when needed. A 100% manual recount isn’t necessary. A random sample of precincts is sufficient.

If you recount and find discrepancies, you might receive additional delegates. More importantly, if you were to demand recounts, it would highlight the fact that in many states, those machine counts are never audited or verified with the original paper records. Most citizens are shocked to discover that their vote counting process is not verified, or in some places, verifiable. I know I was when I first discovered this truth about Sedgwick County Kansas in 2012.

Thank you

Beth Clarkson


The Charts below show the cumulative share of the vote each candidate acquires as the size of the precincts increase. This model clearly shows that as the size of the precinct increases Clinton and Rubio gain a larger share of the votes while Sanders, Trump and Cruz lose votes. This is NOT a random fluke, this is a consistent pattern with machine counted votes. While in OK, this trend was not enough to change who won the election, it may have had an impact on the number of delegates each received.

2016 Oklahoma Republican Presidential  Primary
2016 Oklahoma Republican Presidential Primary

2016 Oklahoma Democratic Presidential Primary
2016 Oklahoma Democratic Primary

eta: Link to data at OK.gov
eta2: I’ve updated the charts with labels for the Dem candidates and added Kasich to the Rep chart.

After February Hearing, What’s Next?

What’s Next?
The loss at the hearing last week and the length of time to have an appeal (approx. year) means I need not worry about my lawsuit for a while. I can concentrate on other things. I have an offer to write a book. I have a day job I need to give more focused attention to. I’m working an editorial on voting system recommendations for Sedgwick County.

I am extremely disappointed that I won’t be able to verify results of 2014 before the 2016 election, which will be run on the same equipment. Our election commissioner has decided to postpone purchasing a new system that would provide a complete paper record of all ballots until 2017.

Sedgwick County
So…what can we do to improve the situation in 2016? I have a few ideas, but I would need help to accomplish any of them. The hearing has shown me just how many other people care about this issue. It’s nice to know that there are others who want to see transparency in our election results. With help, here are some things that we could accomplish.

“Paper Please” Promotion
This isn’t something I can do; promotion of any sort is outside my bailiwick. I’m a shy nerd. I don’t know what would be required to get the word out to all voters in Sedgwick County that they have the right to a paper ballot and that their vote has a better chance of being counted properly if they mark a paper ballot. But this isn’t something that needs my special skills either. The buttons and bumper stickers that people brought to my hearing were awesome and unexpected. Thank you.

Exit Poll the 2016 election
This requires volunteers and not much else. Well, tables and chairs and paper ballots, stuff like that. I would need to do some research on what is allowed and what isn’t when conducting an exit poll, then train volunteers to do the job and schedule shifts for the polling place(s).
The basic idea is to staff a relatively large polling station with volunteers to pass out paper ballots and put them into a box to be counted at the end of the day. They would report their results. I would get copies of the total tapes for those voting stations and compare the two. We wouldn’t have 100% coverage, but if we can manage more than 80% I would be delighted. Even 20% would provide a reasonable check on the accuracy of the reported totals. If anyone is interested in volunteering to do this next November, let me know. With help, we can organize such a poll.

I even found and joined a facebook group for this kind of activity: Citizen Exit Poll

Recount the 2016 election
In Sedgwick County, only a candidate may request a recount. The candidate may specify the precincts to be recounted and the method of the recount. The candidates themselves are not allowed to be in the room where the votes are being recounted by the election board, but they are allowed to specify a representative who is able to observe the ballots to assure that they are counted properly.

In order to do a solid audit through recounts, I need as many candidates as I can get to request a recount and appoint me as their representative. I will also need funding to cover the costs; recounts aren’t free. Finally, I doubt I will manage to get a recount done to my satisfaction without a judge ordering them to do it. That means I’ll need legal assistance before I’ll be granted access. While I have pro bono lawyer on my current case, I can’t expect him to add on another for free. Nor can I donate as much of my time as is required without compensation, so I’ll be soliciting donations to fund our time as well.

I know how to select a representative sample in a low cost way, but thus far the election office have refused to even give me an estimate of the costs of doing such a recount. The cost for a hand count of the paper ballots in 2014 was IIRC around $10,000. Recounting RTAL ballots will be harder, even more time consuming, and there are an order of magnitude more votes to be counted that way. I expect it to be costly. I estimate the cost will be between $20,000 and $100,000. If it happens. I give a higher probability to the outcome that my ability to conduct a recount in 2016 will be thwarted in some way.

A good critical look at the mail in ballots and provisional ballots wouldn’t be a bad idea either. But again, this requires more time of all the officials involved, which in turn means they will charge me more for doing this work. So, I guess it’s time for me to find out if I can overcome all my mental blocks to asking people for money and raise sufficient funds.

Other places
The three activities outlined above can be modified to other jurisdictions. Promoting “Paper Please” won’t be needed in places with a solid paper trail already. Recounts are expensive and the laws vary about who and when they can be requested, so that’s a step requiring a strong commitment.

If people are interested in pursuing recounts, I can help them understand what steps would be needed to verify the output of an election using a sample of paper records – i.e. I can minimize the costs of the sample for the greatest return on information gained. But the most likely outcome is a lot of legal hassles and grudging limited access for those who can get a judge’s permission to see the records. One guy was allowed to photograph ballots but not touch them. He found discrepancies ranging from 3% to over 5%.

I leave it to every individual voter reading this to decide what they personally are willing to do to make sure their vote is counted accurately. Voting with a paper ballot is an option nearly everywhere and requires only that you ask for it when you vote.

Activists, I would recommend running an independent exit poll on election day. Setting one up locally is not all that expensive with volunteers. If that’s something that interests you, let me know. I can help with the details. Then you can compare your results with the reported results and see what you think of the comparison. If they match up well, then congratulations. You have no reason to suspect the count is inaccurate.

If they don’t match up, it’s more evidence in support of the hypothesis that we have a serious problem with the foundation of our democracy. Once you have established that the foundation has rotted away, you have to replace it with a solid one. Hopefully, without bringing the entire edifice down upon you while the replacement is being put in.

2014 Wichita Election Results

Wichita 2014 Election Results
This is the final graph from my presentation at the Lawrence Symposium yesterday. My friend Eileen encouraged me to make a post about it as it’s one she feels quite illustrative of the reason for concern.

During the Q&A after my talk, I sorta kinda said I didn’t think the Governor legitimately won the popular vote in 2014. It occurred to me later that probably wasn’t a very ‘politically astute’ thing to say. I hope the Governor does not personal offense. Such thoughts didn’t occur to me at the time because I was talking about how I read the data shown in this chart. I get engrossed in the data analysis and don’t think about anything else.

This chart shows each of the four statewide races, Governor, Senator, Secretary of State, and Attorney General graphed in a cumulative summation analysis. In addition, the cumulative summation analysis of Wichita registered republican voters is provided to give a comparison.

In the cumulative summation models, as the models accumulates more data (think moving left to right across the graph) we expect it to converge to the average quickly and show decreasing differences randomly distributed around the mean. When we see a pattern of the model increasing or decreasing, it indicates that the sequencing of the summation is not random, but has a relationship with the share of the vote achieved.

In the graph above, the % registered republican voters shows a pattern that is not visually distinguishable from random. The four races all show a clearly non-random pattern with striking similarities. All show the pattern of increasing republican share of the vote with increasing votes cast.

These are not independent datasets. The four races all show a similar pattern because they are highly correlated. The pattern itself is the troubling thing.