The art of swaying an election is as old as democracy itself. Strategies like ballot stuffing, redistricting, voter venue switching, and temporary traffic detours have skewed regional results, and sometimes determined the winner.
While some tactics get exposed the old fashioned way -- by angry voters or investigations -- others don’t. But new research suggests some kinds of election fraud leave a trace in the voting data.
In a paper appearing today in Proceedings of the National Academy of Sciences, titled "Statistical Detection of Election Irregularities," a team led by Santa Fe Institute External Professor Stefan Thurner brought science to the problem.
"We got into this by chance, when a Russian colleague brought us the 2011 Russian Duma-election data and asked us to take a look,” says Thurner. “From the first look we were all pretty shocked, and decided to take a second look.”
Thurner, who heads the Section for Complex Systems at the Medical University of Vienna, and colleagues looked for two kinds of rigging: incremental fraud, where votes for one party are kept in the ballot box while those for the other candidates are tossed, and extreme fraud, which shows 100 percent voter turnout in a district, all voting for the same party.
The team examined data on number of eligible voters, valid votes, and votes for the winning candidate (or party) from a dozen recent elections around the world. By comparing the distributions of votes for the winning candidate against turnout numbers, they found that rigged elections show a different voting pattern than fair ones.
In fair elections, a nation's voting pattern tends to feature one cluster, showing a general trend of voter turnout and vote for the victorious party (though some nations' regional voter preferences can distort it). Rigged ones show a cluster, but with a smear of votes toward the upper right for incremental fraud. Extreme fraud has a second, smaller, completely separate cluster at the top right corner, signifying up to 100 percent turnout and votes for the winner.
Next, the team developed a model to detect how much forged or manipulated results affected the outcome, then ran through all possibilities of both fraud types playing 0 percent to 100 percent of a part in the election, and compared those to actual data to determine their prevalence.
Among the countries studied, data from recent elections in Russia and Uganda showed both the smear of incremental fraud and the second cluster of extreme fraud, with up to 64 percent of districts being affected in Russia's 2011 vote and 39 percent in 2012. Other countries’ data showed little to no such trends.
"I think it could contribute to the benefit of democracy if for every nationwide election on this planet, the raw data is made available on say a United Nations or OECD database,” says Thurner. “One could then think of a set of quality standards and checks for any election -- like the ones we presented -- or better ones.”
Read the paper in PNAS (September 24, 2012)
|All SFI News||About SFI||Follow SFI||Support SFI||SFI Home|