The Constitutional Court of Austria issued an order today to rerun the election after the right-wing populist Freedom Party filed a legal challenge alleging “failures and irregularities” and claiming a “system failure” in the counting of mail-in ballots.” The Freedom Party’s Nobert Hofer lost the May 22 election to the left-win Green Party’s Alexander Van Der Bellen by 30,863 votes – about .69 percent of the valid votes counted, according to official government reports.
Ken Kollman, the CPS director and the Frederick G.L. Huetwell Professor of Political Science, and his colleagues used novel statistical techniques to analyze election returns for evidence of fraudulent behavior.
Led by the path-breaking research of Walter Mebane in the field of election forensics, the team has developed the Election Forensics Toolkit, a website prototype that runs computer programs to analyze election returns that are reported at the precinct, polling station, or district levels.
Electoral manipulation can happen in the collection, counting, and reporting of vote totals, turnout percentages and numbers, and valid vote totals versus total votes cast. When humans manipulate election returns, those returns show traces of that manipulation in the patterns of digits, and in the correlations across reported numbers, that would not occur in a naturally occurring process (i.e., in an election without any human manipulation), Kollman says. Similar techniques are used by statisticians and investigators when they try to detect evidence of financial fraud.
The team uses multiple methods, and when egregious fraud occurs, multiple tests show evidence consistent with fraud. If only one method shows evidence, the claims of potential fraud are naturally weaker and indicate that such results can occur by chance or sometimes by the strategic behavior of voters. With geocoded election data the team can use maps to pinpoint where likely fraud occurred.
The team’s work was featured today in Washington Post’s The Monkey Cage blog, after analyzing Austria’s election and finding no evidence of fraud. The team also analyzed Turkey’s 2015 parliament elections and found evidence consistent with fraud in certain areas.
The team includes Mebane, Kollman, and Allen Hicken aresearch associate professor for CPS, and Kirill Kalinin, a graduate student in the U-M Department of Political Science. The CPS is part of the U-M Institute for Social Research. In 2015, they received a grant from USAID (in collaboration with David Backer at the University of Maryland, a PhD from Michigan who worked in ISR) to create the website and to test for the effectiveness of the methods on data from Albania, Bangladesh, South Africa, Afghanistan, Uganda, Kenya, Libya, and Cambodia. Their goal is to give a rendering of the integrity of a given election; more specifically, they hope to be able to analyze data from countries around the world and give results indicative of the likelihood of human manipulation of election results.