By Anatoly Kurmanaev and Ethan Singer
Venezuela’s electoral body announced early this week that the country’s president, Nicolás Maduro, comfortably won another six years in office, beating his main opponent by 7 percentage points in a vote that was marred by widespread irregularities.
But partial election results, provided to The New York Times by a group of researchers associated with Venezuela’s main opposition alliance, supply new evidence that calls the official result into question.
Their figures suggest that an opposition candidate, a retired diplomat named Edmundo González, actually beat Maduro by more than 30 percentage points. The researchers’ estimate of the result — 66% to 31% — is similar to the result obtained by an independent exit poll conducted on Election Day across the country.
It was not possible for the Times to independently verify the underlying tallies, which the researchers say were collected from paper receipts produced by about 1,000 voting machines, about 3% of the country’s total. By Wednesday, Venezuela’s government-controlled election authority had still not released detailed results, despite growing international pressure.
But several independent survey and election analysts reviewed the researchers’ approach and said that, based on the tallies shared by the researchers, the estimates appeared credible. And given the partial tallies, the Times was able to broadly replicate the researchers’ estimates of the result within 2 percentage points.
The announcement of Maduro’s victory triggered deadly protests on Monday and led several Latin American countries to suspend or downgrade diplomatic relations with Venezuela, plunging the polarized country into a new period of flux.
The Times analysis shows that the election tallies provided by the researchers are not compatible with a victory by Maduro, by any margin.
What the researchers did
There are about 30,000 voting machines in Venezuela. After polls close, each machine prints a vote tally, and volunteers observing the vote on behalf of political parties have a legal right to a copy.
To estimate the overall vote count, the research group, called AltaVista, created a random sample of 1,500 voting machines across the country, about 5% of the total. They designed the sample to proportionally represent the overall geography and partisanship of the country.
After the polls closed, AltaVista began collecting scanned copies of the tallies collected by opposition volunteers. Intimidation by government supporters, organizational lapses and patchy cellphone coverage prevented the researchers from obtaining all the tallies in the sample, but they collected and verified data for more than two-thirds of those precincts.
AltaVista made the data, including images of the printed receipts, public Tuesday night.
The vote totals in these precincts show González, the opposition candidate, would be projected to defeat Maduro by a margin of 66% to 31%, the researchers said. Even in the ruling party’s strongholds in previous elections, the data shows that voters shifted away from Maduro.
Reasons to believe the researchers’ data
A similar approach would have yielded accurate estimates in previous elections. Because historical election data in Venezuela is public, we can see that the precincts sampled in this election have been representative of the nationwide vote in the past. In three prior elections, the weighted results from these precincts were within two points of the final national result, a Times analysis found.
The only significant difference is with the government-declared results of Sunday’s election.
Venezuelan elections under Maduro have long ceased to be considered free or fair, beset by voter suppression and other irregularities. But the validity of the actual voting tallies has never before been cast into such doubt in a competitive national election.
The researchers’ results are consistent with an independent exit poll. An exit poll by Edison Research conducted on Election Day found overwhelming support for González. Its poll, conducted across 100 polling stations and based on interviews with nearly 7,000 voters, found González leading, 65% to 31%.
Independent survey and election analysts said AltaVista’s methodology and techniques appeared sound. Dorothy Kronick, an expert on Venezuelan electoral data at the University of California, Berkeley, independently assessed the validity of the opposition researchers’ sample at the Times’ request. She arrived at a similar conclusion.
Kronick found the sample slightly favored the opposition-leaning areas, but not nearly enough to explain the tremendous difference between the numbers claimed by the government and the opposition. “The numbers are irreconcilable,” she said. “One of them is wrong.”
Adam Berinsky, an expert in survey methodology at Massachusetts Institute of Technology, said that the researchers’ approach seemed reasonable and that, if anything, challenges the researchers faced while collecting data, including witness suppression, may lead to an underestimate of opposition support.
Different statistical techniques would not affect the overall conclusion. The researchers’ data suggests a strikingly different result from the one claimed by the Venezuelan government.
If the claims were closer, smaller methodological choices — how to select which precincts to validate or how to weight data from those precincts — could be decisive. But that is not the case here. This kind of analysis may not be conclusive enough to adjudicate a two-point difference, but it can offer clarity on a 30-point difference.
What’s still uncertain about the data
The researchers are aligned with the opposition party. The activists behind the project have a history of opposing Maduro’s government, and supporting opposition causes, which makes them a vested party in the country’s political standoff. But they have not held political office, and they have designed their methodology in collaboration with independent academics in the U.S. and Brazil.
There may be some biases in the data we don’t fully understand. The researchers’ sample of 1,500 precincts was drawn to assure representativeness in both partisanship and geography. But it’s possible that the selection is unrepresentative in some way the researchers did not detect.
It’s also possible that the precincts for which the researchers were not able to find voting receipts were meaningfully different from the ones where they were successful. (That would be unlikely, however, because the researchers selected the precincts randomly and say they were able to find voting records in most places where they sought them.)
Historical comparisons are challenging in a country with a history of election problems. Ensuring a sample with a representative number of Maduro supporters and opposition supporters is challenging without trusted, detailed historical election data. The migration of millions of Venezuelans in recent years makes any demographic estimates particularly challenging.
Comments