Mauricio Macri’s surprisingly strong showing against Daniel Scioli in the Oct. 25 presidential election shook up Argentina’s political landscape.
Indeed, many pundits speculated that Macri would go the way of Mexico’s Andres Manuel López Obrador, claiming the election was stolen from him.
None of this happened. As Mark Jones describes in a recent post, Macri ended up with 34.3 percent of the vote, just 2.6 points behind Scioli. He’s the one with the campaign momentum going into the Nov. 22 runoff election, which will select the next president.
Nevertheless, suspicions of fraud arose after the polling stations closed, when the authorities were delayed in releasing the results.
Many people became particularly skeptical about the election’s integrity during September’s municipal elections in the northern province of Tucumán. There, the opposition claimed that more than 40 ballot boxes were torched. Those events, combined with widespread allegations of vote buying, prompted many to speculate seriously about wrongdoing. The opposition claimed that the election was fraudulent; authorities denied it. And nobody can be certain about who’s telling the truth.
It’s easy to claim fraud. It’s hard to prove it.
Unfortunately, uncovering fraudulent elections is quite difficult. How do you prove or disprove possible wrongdoing? If votes were falsified, the wrongdoers have no motive to say so; if they were not, there’s no proving a negative. Thus it is very difficult to establish a suspect election’s legitimacy or illegitimacy.
Facing this problem, social scientists have proposed different ways of distinguishing potential irregularities from voters’ true preferences. These methods try to identify perpetrators’ hidden behavior by measuring and comparing electoral behaviors that differ when there’s fraud.
For instance, in the 2004 recall referendum in Venezuela, Ricardo Hausmann and Roberto Rigobón found serious inconsistencies between the reported vote counts and two independent correlates of voters’ intention at the precinct level: first, the number of registered voters who signed the recall petition, and second, the vote intentions from an electoral poll.
Of course, each measure is imperfect. But the precincts in which the poll had the biggest errors are also those with the largest discrepancies between the number of signatures and the share of “yes” votes in the referendum.
In another example, Michael Callen and James Long found serious disparities between the results in the central government’s vote tally sheets and the numbers reported at the polling centers during the 2010 parliamentary election in Afghanistan.
Here’s how electoral forensics work.
A related approach to assessing electoral integrity uses electoral forensics — e.g., finding anomalies in a statistical feature of the vote returns that diverges from an expected theoretical distribution.
For example, when comparing a set of precincts with similar demographic and political characteristics, one expects turnout rates to follow a normal distribution. Under this assumption, turnout values far away from the mean value would suggest evidence, albeit not conclusive, of a potential irregularity. Mikhail Myagkov, Peter C. Ordeshook and Dimitri Shakin used a similar approach to flag potential irregularities in the turnout rates in Ukraine in 2004.
Perhaps the most popular electoral forensics test focuses on the digits in the vote counts and assesses how they should look in the absence of manipulation. For instance, if a candidate received 1,745 votes in a polling station, Bernd Beber and Alexandra Scacco propose focusing on the last digit — which is the “5” in this vote count — and its frequency for the same candidate’s vote counts in all polling stations. When numbers are generated at random or without any deliberate manipulation, the chance of the last digit to be any numeral between 0 and 9 is 10 percent. Therefore, if in our hypothetical example the vote counts ending in 5 appear more often than expected, there would be reasons to question the integrity of the results.
Electoral forensics has its critics.
Election forensics, however, are not free from criticism. Scholars have pointed out the necessity of reviewing the assumptions of the analysis. Addressing these concerns, Jacob Montgomery, Santiago Olivella, Joshua D. Potter and Brian F. Crisp propose a novel approach. Using what they call “informed forensics,” they combine statistic indicators, like the distribution of vote counts’ last digit, with data specific to the place where the election is held, such as ethnic fractionalization or the percent of citizens living in urban areas. The simultaneous use of forensics tests and context provides more insight when assessing an election’s integrity.
Argentina has been a trial site for catching electoral fraud
So what about Argentina? The country’s elections have given political scientists many ideas about understanding and measuring electoral irregularities.
For instance, we examine the integrity of elections in the province of Buenos Aires in the 1930s by simulating two sets of simulated elections that resemble previous elections in Buenos Aires without allegations of fraud.
Then, we manipulate one dataset based on historical accounts of electoral manipulation during that period — for example, the inflation of vote returns for the incumbent and the subtraction of votes for the challenger.
Finally, we compare the vote returns from four elections in Buenos Aires during that decade and assess the integrity of each.
More recently, Michael Alvarez, Ines Levin, Julia Pomares and Marcelo Leiras have studied the implementation of electronic voting in the province of Salta to explore how this technology has changed the perceptions of electoral integrity among voters and poll-workers.
Will the presidential runoff be a clean election?
What should we expect for the Nov. 22 runoff election? Unlike during the weeks leading up to the first round, neither candidate has so far raised doubts about the election’s integrity. But there are also other reasons to believe that the process is going to be clean.
That may, ironically enough, be the result of citizens’ distrust of the authorities. The Aug. 23 gubernatorial elections in the province of Tucuman were quite scandalous: the first court to hear the allegations of fraud declared the election invalid, but a higher court reinstated the results. That turmoil prompted the main opposition candidates to call for reform.
In response, the government modified the system significantly — for instance, adding GPS tracking systems to the trucks that carried the ballot boxes to the centers where the votes were to be counted, proving they had not been hijacked or replaced.
In addition, opposition parties also made an important adjustment in anticipation of the Oct. 25 elections. Argentina’s voting system is quite outdated. Citizens vote with slips of paper that carry the names only of a given party’s candidates, like the coupon ballots used in the 19th century United States. Each political party prints, distributes and supplies its own ballots during election day.
Given this voting “technology,” political parties often require thousands of poll-watchers to avoid ballot theft and to monitor the count in each of the polling stations. In the weeks leading to the election, the opposition parties recruited and trained an “army” of poll workers to watch over the contest. And so, as the local newspaper La Nacion notes, the elections took place without major irregularities.
Note: To invite scholars and general readers to learn more about this type of research, Political Analysis offers free access to several articles on electoral fraud and forensics until Nov. 30. The articles include a virtual issue on electoral fraud and two very recent pieces on electoral forensics and election forecasting.