Prepared for presentation at the 2015 Annual Conference of the European Political Science Association, Vienna, Austria, June 25–27.
ISSN/ISBN: Not available at this time. DOI: Not available at this time.
Abstract: Many statistical methods that have been proposed to use low-level election vote counts to detect election frauds have a hard time distinguishing distortions in vote counts that stem from voters’ strategic behavior from distortions that originate in election frauds. Even a finite mixture model implementation of the recently proposed positive empirical approach of Klimek, Yegorov, Hanel and Thurner (2012) has this limitation. We use latent variable models for polling station voting data and postelection complaint data from the 2005 and 2009 German Bundestag elections to show that the types of multimodality emphasized in the Klimek et al. (2012) model are valid symptoms of the localized incidents that trigger the complaints. The postelection complaints in Germany certainly do not represent the full variety of election frauds that occur in various election systems. They reflect a mix of simple administrative failures, concerns about voter privacy, worries about ballot access and disagreements with the way Germany’s mixed system operates—a banal “fraud” latent variable at best. But the fact that some methods that can be shown to discriminate strategies from frauds—even if imperfectly—is an important advance for election forensics.
Bibtex:
@inproceedings{,
AUTHOR={Mebane, Walter R. and Klaver, Joseph},
TITLE={Election Forensics: Strategies versus Election Frauds in Germany},
BOOKTITLE={Proceedings of 2015 Annual Conference of the European Political Science Association},
ADDRESS={Vienna, Austria},
MONTH={June},
YEAR={2015},
URL={http://www-personal.umich.edu/~wmebane/epsa15.pdf},
}
Reference Type: Conference Paper
Subject Area(s): Political Science, Voting Fraud