[Lea esta página en Español]

Analysis of Voting Data from the Recent Venezuela Referendum

Edward W. Felten, Department of Computer Science, Princeton University
Aviel D. Rubin, Department of Computer Science, Johns Hopkins University
Adam Stubblefield, Department of Computer Science, Johns Hopkins University


After the August 15 referendum in Venezuela on whether or not to recall president Chávez, opposition groups examined the polling data and made accusations of fraud due to statistical anomalies in the reported election results that they claim could not have occurred if the election were run fairly. However, our analysis of the same data, based on simulations, did not detect any statistical anomalies that would indicate obvious fraud in the election.

We emphasize that a lack of statistical evidence does not imply the absence of fraud. Rather, it rules out certain classes of fraud. In any case, the fraud that is alleged is not the type that we would expect a cheating government to employ. In particular, we believe that the forms of election fraud that are most likely to succeed, such as voting machines silently switching some fraction of Yes votes to No votes inside the computer, would not produce observable statistical anomalies.

Electronic voting is more susceptible to widespread fraud than less automated mechanisms. The fact that the opposition is highly suspicious of the outcome is due, in part, to the choice of electronic voting machines in a simple Yes/No election. While we did not find any statistical evidence for the claims of caps on the machines or other specific accusations of fraud, we are concerned that wide scale unobservable fraud is much easier to realize in electronic voting machines than in, for example, precinct based paper systems.