PhD Thesis, Universitat Konstanz.
ISSN/ISBN: Not available at this time. DOI: Not available at this time.
Abstract: Election forensics is an approach to assess the legitimacy of an election using sta- tistical tools to detect election fraud. The dissertation focuses on the forensics of digit tests, for which applications and the impact is constantly growing. Digit tests compare a theoretical distribution of a specific digit with the empirical frequencies of that digit in vote counts of election results. A significant deviation indicates anoma- lies or fraud. Despite their popularity it remains uncertain how digit tests work, what kind of election fraud they can and cannot detect and if they are trustworthy as fraud indicators. The precision in detection and the accurateness of accusations is important, as they are sensitive issues that can affect the democratic process. The dissertation contributes to the accurateness of fraud detection through solving most of the riddle surrounding digit tests. In the first paper we investigate presidential elections in France, Finland and Rus- sia, which differ strongly in their perceived legitimacy, and evaluate the validity of the Benford’s Law test, the last digit test as well as their linkage. In particular, we consider specific fraud mechanisms that each digit test can (and cannot) capture and cross validate the results in Russia with extreme fraud indicator. We conclude that significant digit tests do not necessarily indicate manipulation. In the second paper I distinguish between different levels of fraud coordination and actions that capture relevant fraudulent activities. The mechanisms are calibrated with the 2008 Canadian election results, that are considered to be fraud-free and then used to simulate artificially manipulated election data. This gives a unique setting to estimate the sensitivity and specificity of the Benford’s Law test. The findings narrow the fraudulent activities that are detectable by Benford’s Law to very specific manipulation settings. In the third paper I use a laboratory experiment to investigate how humans manipulate election returns by hand. The analysis of the experimental data can capture “fingerprints” of manipulation that are often not visible to the eye and improves our knowledge about fraudulent activities. Additionally, I investigate if the applied strategies are affected by the manipulation setting and if both, setting and strategies, affect last digit tests’ detectability. I find many different fraud strategies, which in combination with the manipulation extent in deed impact the detectability.
Bibtex:
@phdThesis{,
AUTHOR = {Verena Mack},
TITLE = {The Fingerprints of Fraud: An In-depth Study of Election Forensics with Digit Tests},
SCHOOL = {Universitat Konstanz},
YEAR = {2016},
TYPE = {Thesis ({Ph.D.})},
URL = {https://kops.uni-konstanz.de/handle/123456789/41741},
}
Reference Type: Thesis
Subject Area(s): Voting Fraud