By Anatoly Kurmanaev and
A close look at Bolivian election data suggests an initial analysis by the O.A.S. that raised questions of vote-rigging — and helped force out a president — was flawed.
The election was the most tightly contested in decades: Evo Morales, Bolivia’s first Indigenous president, was running for a fourth term, facing an opposition that saw him as authoritarian and unwilling to relinquish power.
As the preliminary vote count began, on Oct. 20, 2019, tensions ran high. When the tallying stopped — suddenly and without explanation — then resumed again a full day later, it showed Mr. Morales had just enough votes to eke out a victory.
Amid suspicions of fraud, protests broke out across the country, and the international community turned to the Organization of American States, which had been invited to observe the elections, for its assessment.
The organization’s statement, which cited “an inexplicable change” that “drastically modifies the fate of the election,” heightened doubts about the fairness of the vote and fueled a chain of events that changed the South American nation’s history. The opposition seized on the claim to escalate protests, gather international support, and push Mr. Morales from power with military support weeks later.
Now, a study by independent researchers, using data obtained by The New York Times from the Bolivian electoral authorities, has found that the Organization of American States’ statistical analysis was itself flawed.
The conclusion that Mr. Morales’s share of the vote jumped inexplicably in the final ballots relied on incorrect data and inappropriate statistical techniques, the researchers found.
“We took a hard look at the O.A.S.’s statistical evidence and found problems with their methods,” said Francisco Rodríguez, an economist who teaches Latin American studies at Tulane University. “Once we correct those problems, the O.A.S.’s results go away, leaving no statistical evidence of fraud.”
Mr. Rodríguez conducted the study with Dorothy Kronick, an expert on Latin American politics at the University of Pennsylvania, and Nicolás Idrobo, a doctoral student at the same university who is the co-author of a textbook on advanced statistical methods. Their study is a working paper that has not yet been peer reviewed.
To be sure, the authors said their analysis focused only on the O.A.S.’s statistical analysis of the voting results, and does not prove that the election was free and fair. In fact, there were a lot of documented problems with the vote.
In an attempt to quell the protests set off when he claimed victory, Mr. Morales called on the O.A.S. to conduct a “binding” election audit.
The resulting 100-page report, published in December, contained evidence of errors, irregularities and “a series of malicious operations” aimed at altering the results. These included hidden data servers, manipulated voting receipts and forged signatures, which the organization said made it impossible for it to validate the election’s results.
The O.A.S. found evidence of tampering with at least 38,000 votes. Mr. Morales claimed outright victory by a margin of 35,000 votes.
“There was fraud — we just don’t know where and how much,” said Calla Hummel, a Bolivia expert at the University of Miami who witnessed the election and analyzed the O.A.S.’s findings.
“The issue with the O.A.S. report is that they did it very quickly,” Dr. Hummel said. That shaped the narrative of the election before data could be properly analyzed, she said.
That initial claim by the O.A.S. is specifically what the academics are disputing in their study.
Mr. Morales’s downfall paved the way to a staunchly right-wing caretaker government, led by Jeanine Añez, which has not yet fulfilled its mandate to oversee swift new elections. The new government has persecuted the former president’s supporters, stifled dissent and worked to cement its hold on power.
Seven months after Mr. Morales’s downfall, Bolivia has no elected government and no official election date.
The O.A.S. said it stood by its statistical analysis, because it successfully detected early initial indications of fraud.
“It’s a moot point,” the organization’s head of electoral observations, Gerardo De Icaza, said in response to questions raised by the new study. “Statistics don’t prove or disprove fraud. Hard evidence like falsified statements of polls and hidden I.T. structures do. And that is what we found.”
The organization’s initial accusation came right after Bolivia’s most disputed elections since the return of democracy in the 1980s. To run for a fourth term, Mr. Morales subverted laws, staffed the electoral council with loyalists, and ignored results of a referendum that banned him from seeking re-election.
Claiming the results of the October election could not be trusted, some opposition leaders said they would paralyze the country if Mr. Morales declared victory.
For their part, Mr. Morales’s largely Indigenous supporters, fearing a return of the conservative politicians of European descent who had been the rule in the country before Mr. Morales took office in 2006, vowed to defend their political gains at all cost.
“The O.A.S., as observers, ratified the doubts that all Bolivians had and the worry that their vote has been violated,” Mr. Camacho said in a video address on Oct. 22.
As demonstrations intensified in the following weeks, Mr. Morales started to lose the support of security forces. A trickle of government defections turned into a flood.
A visibly haggard Mr. Morales went on national television to offer new elections, but by then it was too late. The same day, the military asked Mr. Morales to stand down. He fled into exile soon after.
“The O.A.S. ended up sinking any legitimacy the voting results might have had,” said Gonzalo Mendieta, a prominent Bolivian columnist.
The authors of the new study said they were unable to replicate the O.A.S.’s findings using its likely techniques. They said a sudden change in the trend appeared only when they excluded results from the manually processed, late-reporting polling booths.
This suggests that the organization used an incorrect data set to reach its conclusion, the researchers said. The difference is significant: the 1,500 excluded late-reporting booths account for the bulk of the final votes that the O.A.S. statistical analysis claims are suspicious.
Also, the academics said the organization used an inappropriate statistical method that artificially created the appearance of a break in the voting trend.
The O.A.S. consultant who conducted their statistical analysis, Professor Irfan Nooruddin of Georgetown University, said the new study misrepresented his work and was wrong. He did not provide details and did not share his methods or data with the authors of the study, despite repeated requests.
For his part, Mr. De Icaza, with the O.A.S., said that, broadly speaking, the data from Bolivia’s most recent elections was too flawed to draw any meaningful conclusions.
“You’re are doing a statistical exercise on documents that are falsified,” he said. “The question is not whether the false numbers add up. The question is whether they are false or not — and they are.”
Julie Turkewitz contributed reporting from Bogotá, Colombia.