unpublished manuscript, Stanford University, USA.
ISSN/ISBN: Not available at this time. DOI: Not available at this time.
Abstract: In this work, we study the question of detecting electronic voting fraud based on count data alone, specifically focusing on the 2004 Venezuelan recall election. The opposition (YES, in favour of recalling sitting president Hugo Chavez), which lost the referendum, alleged that the government had committed electronic voting fraud. In this work, we use the count data to investigate two of the opposition’s claims, ultimately concluding that the claims cannot be upheld by the count data alone. The first allegation is based on the total number of ties (defined below) for the YES/NO in each particular station. In this paper we compare four or five models for the count data. We present an FDR analysis of 6507 roughly independent test statistics that shows no systematic fraud in the form of “vote-capping” and no significant differences between the dispersion of the YES votes and the NO votes. The second allegation involves Benford’s Law, which is a model for the “significant digits” of an observation. If we look at the significant digit distribution of elections generated by a “fair election model” (the null model used by the YES side as evidence of fraud), we see that Benford’s Law does not hold. Finally, we conclude with a simple scenario of fraud that is statistically undetectable based on solely the count data. Hence, our ultimate conclusion is that allegations of fraud based on count data are difficult to verify. This underscores the importance of a “paper trail” when using electronic voting machines
Bibtex:
@unpublished {,
AUTHOR = {Taylor, Jonathan},
TITLE = {Too many ties? An empirical analysis of the Venezuelan recall referendum counts},
YEAR = {2005},
URL = {https://esdata.info/pdf/Taylor-Ties.pdf},
}
Reference Type: E-Print
Subject Area(s): Voting Fraud