Bad times for despots: scientists can detect electoral fraud

Posted on: 30 Jun 2017

Free elections are the cornerstone of any democracy. Up until now, only through great effort could electoral fraud and manipulation be proved. Complexity researcher Peter Klimek, Medical University of Vienna and Complexity Science Hub Vienna, and colleagues from the Carlos III University of Madrid are now presenting a statistical procedure that makes it possible to detect signs of manipulation through publicly accessible electoral data. The work is published in the journal Science Advances. Findings on irregularities during the constitutional referendum in Turkey in April 2017 can be found online as well.

The more remote the election sites, the greater the irregularities:

With a newly developed statistical method, Klimek and colleagues analyzed the official election results of 21 elections in ten countries, including elections in Russia from 2007 to 2011, Venezuela from 2006 to 2013, and Uganda in 2011. Data from Austria and other established democracies were also investigated.

This large number of observations led to certain statistical regularities in electoral data, such as the distribution of electoral participation and voting behavior across different parts of the country. "If these regularities do not occur in a data set or occur only in a strongly distorted form, it’s an indication that the voters were influenced." While the elections in Austria, Finland, Canada, Spain, or France remained inconspicuous, the researchers found significant irregularities in some countries in the form of noticeable shifts of votes toward the government party. The smaller and more remote the voting stations, the stronger these displacements were. These distortions were even decisive in the presidential elections in Venezuela in 2013.

"The data clearly show that, in small and remote polling places, the influence of governmental groups on voters is a lot stronger than in anonymous urban areas, where the probability of being discovered by electoral observers is higher," says Klimek. "Our method could usher in the end of hidden authoritarian practices in formal democratic institutions," emphasizes Klimek. "In the future, electoral fraud could go from being purely suspicion and guessing to becoming a mathematically provable fact."

In this context, the authors, together with Stefan Thurner of the Complexity Science Hub Vienna, IIASA, MedUni Vienna, and Santa Fe Institute, investigated the Turkish constitutional referendum from April 16, 2017. Immediately after the election, official observers criticized the possibility of allowing unauthenticated ballot papers to be allowed in electoral identification and access authorization at the polling stations or the prohibition of opening and counting ballots. As the researchers’ data analysis shows, these practices are indeed reflected in the same irregularities observed in Russia, Venezuela, or Uganda. Compared to these countries, the effects were weaker, but the narrow outcome of the referendum made them more selective. "We identified irregularities in about 6% of the election stations whose influence was just big enough to tilt the majority situation in favor of a 'yes' to constitutional change," Klimek said.

Testing for voter rigging in small polling stations
Election forensic analysis of the Turkish Constitutional Referendum 2017