lulu Fries’dat & Anselmo Sampietro
in collaboration with Fritz Scheuren
& Election Justice USA
An Electoral System in Crisis, is a 39-page independent in-depth examination of the accuracy and security of U.S. electronic voting equipment. This research has been invited for publication in the Journal of the International Association of Official Statistics (IAOS). Due to the unusual time constraints of the primary election cycle, and the right of the public to have access to this information, the authors are taking the unusual step of publishing ahead of time online. The full report is now available online at the website of the co-author’s documentary; and will be posted in a number of locations including the forum of The American Association for Public Opinion Research, and the forum of Social Research Methods. Below is a summary of our findings. We encourage everyone to download and read the full report.
Having confidence in our elections is central to our faith in our government, and all the decisions that we make collectively as a nation. But are the candidates who win the ones that we actually vote for? A large and growing body of research provides convincing evidence that U.S. electronic voting equipment in many parts of the country may not be counting votes accurately. This could be due to malfunctioning computer equipment that in 43 states is over a decade old, and long past its natural life. However, in many cases, the data points toward fraud as a likely explanation.
The documentation consists of statistical graphs analyzing data from five presidential cycles, as well as off-year races from across the country. The data illustrates that there are unusually large discrepancies between small precinct and large precinct election returns, and noticeable differences between hand-counted and machine-counted precinct results.
In our research we examined the election results of the 2016 presidential primaries, and found irregularities in the overwhelming majority of the twenty-one states that we analyzed. The data indicates, in particular, that the totals reported on the Democratic side in the race between Hillary Clinton and Bernie Sanders may not be correct. In state after state, independent examination by two separate analysts found suspect statistical patterns giving Clinton inflated percentages, that in all likelihood, are not fully based on actual votes, and showing Sanders with what appear to be artificially depressed totals.
The difference between the reported totals, and our best estimate of the actual vote totals, varies considerably from state to state. However, these differences are significant—sometimes more than 10%—and could change the outcome of the 2016 Democratic presidential primary. We found irregularities in the 2016 Republican presidential primary as well, and while concerning, we do not believe they are large enough to change the outcome of that race. It is important to note that the fact that a candidate benefits from irregularities does not imply that a candidate is responsible for them.
Fritz Scheuren, a member of the statistics faculty at George Washington University, and a former president of the American Statistical Association, has been a collaborator in this research. Examining the data from the study, Scheuren said, “As a statistician, I find the results of the 2016 primary voting unusual. In fact, I found the patterns unexpected [and possibly even] suspicious. There is a greater degree of smoothness in the outcomes than the roughness that is typical in raw/real data.”
The underlying analysis for this research was presented in an article by Beth Clarkson in the Royal Statistical Society journal, Significance. This method was also covered in a number of other scientific papers and blogs, some of which were weakened by incorrect assumptions and flawed analysis. This research is the most comprehensive and academically rigorous treatment of the material that has been published to date, using blind replication of the data by more than one statistician and verification by respected members of the academic statistical community.
If voting-machine results were inaccurate on a regular basis, there would be some evidence of it. One indicator would be that votes counted by machines would give different results than votes counted by hand. In fact, this is now being seen in elections all over the country.
In the 2016 Democratic primary in Kings County, New York (Brooklyn,) a group of provisional ballots were hand-counted by a group of volunteers. Comparing the hand-counts with the machine-counts, there is a noticeable difference (Figure 1). With one exception, in every single assembly district that we examined, Hillary Clinton performed better when the votes were counted by machine. In the lone instance where Clinton did better by hand, it was only by half a percent. In every other assembly district, Bernie Sanders did better when the votes were counted by hand. This is a small sample of the overall ballots cast, but the consistency of the results makes a convincing case that something is amiss.
Comparisons have been made previously between voting results in hand-counted precincts and machine-counted precincts. When there have been discrepancies, they have been passed off as the result of demographics. This is a reasonable concern, since it is possible that voters of a particular political perspective could tend to choose a particular type of voting equipment. However, in this instance, because the two sample sets (hand-counted and machine-counted) are from identical precincts, with voters participating in the same election on the same day—there is no demographic variable to take into account.
In this case, the provisional ballots are from voters who were not able to vote by regular ballot. In the New York 2016 primary, over 120,000 voters were purged from the rolls in Brooklyn alone, and a large number of voters also had their voter registration changed without their knowledge or intent. It is this subset that, when counted by hand, shows a consistently higher percentage for Sanders. There are two possible explanations for this. One is that the machines are counting the votes differently. The other is that the voters who were forced to use provisional ballots were targeted Sanders voters. Possibly, both of these factors are at work. Either way, the data indicates the footprint of manipulation in the election, and calls into question the validity of the reported results.
The paper uses a variety of techniques, including bar charts, scatterplots, and a graph called the “cumulative precinct vote tally chart” to illustrate the large distance between the vote totals currently being reported and statistical norms. It compares election results over time – since the year 2000, by geography and demographics, and also responds in detail to previous criticisms of this work. A broad range of plausible explanations for the statistical irregularities are explored.
In cases where statistical discrepancies exist, our recommendation is that election results be examined, and if found to be inaccurate, decertified. Where paper ballots are available, it would be informative to count them by hand in a secure process that is open to the public, invites media scrutiny, and has strong chain-of-custody protocols.
About the Team
lulu Fries’dat is an Edward R. Murrow award-winning journalist and documentary filmmaker. She received a Best Documentary award for her first feature-length documentary (producer/director) Holler Back — [not] Voting in an American Town, a film that explores systemic issues in our elections that discourage voter participation. Clips are available for viewing here. Her network news experience includes editing assignments for CBS Evening News, Nightline, Sunday Morning, The Today Show, and Good Morning America. She produced and edited profiles of Democratic candidates for MSNBC, and has done long-format documentary work with NBC News and CNBC. She was on the editing team of Gideon’s Army, an Emmy-nominated documentary that follows the personal stories of public defenders in the Deep South. Her full bio is available here. Follow her on twitter @shugahworks.
Anselmo Sampietro holds a Master of Statistics degree from the University of Bologna, Italy and has also studied at the University of Technology Sydney, in Australia and the University of Warsaw, in Poland. He currently leads a team of data analysts for a company based in London, UK. He collaborated (through InnoCentive) with General Fusion, a Canadian start-up that is developing clean and reliable nuclear fusion reactors, to build a statistical model predicting plasma performance. He has specialized in the use of statistical analysis software R, the analytic tool used throughout this article.
Fritz Scheuren is a Senior Fellow and Vice President at NORC in the Center for Excellence in Survey Research. Scheuren has an unparalleled record of work on complex substantive tasks related to sampling and to the analysis of data from government agencies and private sector institutions. Scheuren also serves on the Statistics Faculty at The George Washington University. At GWU, he created a successful survey sampling certificate program which he still teaches. Most of his over 450 applied and theoretical papers, presentations, monographs, and books are on sampling aspects of data collection primarily in a survey context. Scheuren served at the 100th President of the American Statistical Association (ASA) and chaired the ASA Sections on Survey Research Methods. He is a Fellow of the American Statistical Association and the American Association for the Advancement of Science (AAAS.)
Election Justice USA – the paper was assisted in its research and development by Election Justice USA, a national non-partisan coalition of seasoned election integrity experts, statisticians, attorneys, journalists and activists, whose mission is to make sure each American’s right to vote can be exercised without issue in accurate and honest elections.