Election fraud in the Philippines is almost always alleged, but never really proven. We can explore an approach that can identify voting data patterns those most likely to be associated with vote padding - adding fake ballots - and find out how the sanctity of the ballot can be preserved by the thoughtful use of election data.
The use of electronic ballot counters in recent Philippine elections has allowed the accumulation of election results data, but what can we do with it? Using a simple analysis of voter turnout and votes cast, conceptualized by Klimek, Yegerov, Hanel, and Thurner from the University of Vienna(Klimek et al. 2012), we can detect election fraud in the form of ballot stuffing - adding fake ballots in favor of a particular candidate. This can be either through duplicate voters, duplicate ballots, or simply reporting contrived numbers to the higher level of canvassers.
In this article, we take a look at how it works, and apply the analysis to the most recent 2013 Philippine Senatorial Elections.
Elections in the country are always alleged to contain some form of election fraud, but these allegations are never actually proven. We can, however, use election data on a disaggregated basis to detect one type of election fraud - ballot stuffing, by just using two commonly reported numbers - the voter turnout in particular areas, and also the portion of total votes cast in favor a certain candidate.
The way to think about this is how ballot stuffing affects these two numbers. Voter turnout increases because these fake ballots are considered new voters, and the portion of total votes cast in favor of the cheating candidate will also rise.
This would go undetected on the aggregate level, but breaking these numbers up allows us to detect an unusual rise in both values that would indicate ballot stuffing in certain precincts, provinces, or cities, unless the stuffing has been done nearly uniformly across all areas.
Indeed, in recent flagrant violations of the democratic exercise of elections, such as in Russia and Uganda, the relationship can be very apparent:
These are two-dimensional histograms (or you can think of it as a ‘heatmap’) of the various disaggregations of elections in various countries. As you can see, in Uganda and Russia, there is a concentration of precincts that are ‘smeared’ towards the upper right corner, near 100% voter turnout, and 100% votes cast for the winning candidate, something which would be very unlikely save for a few ‘balwartes,’ or deliberate electoral fraud.
We can replicate this approach and apply it to the Philippines, particularly the most recent 2013 Senatorial Elections. I used data from the COMELEC and Rappler’s PHVOTE 2013, and used provinces and certain key cities as they were most disaggregated unit of analysis made possible by the available data.