2019 Bolivian general election

The pause in results transmission for 24 hours, which took Morales from a tight race with Mesa to an outright win, was challenged by people in Bolivia and other countries, who questioned the legitimacy of the results. Protesters and opposition politicians called for a second round to be held despite Morales' lead, as did the governments of Brazil, Argentina, Colombia, the United States, and the European Union.[3] Support for the results of the election came from the governments of Mexico, Nicaragua, Venezuela, Cuba, Palestine, the Non-Aligned Movement and the new President-elect of Argentina.[64][61][65][66]

The day after the election the vice-president of the TSE Antonio Costas resigned, citing his disagreement with the decision to stop transmitting results.[67] The president of the Santa Cruz Electoral Tribunal Sandra Kettels also resigned on 30 October.[68]

After an updated vote tally was announced on Friday 25 October, including previously annulled ballots in Beni, the United Nations announced that it supported an audit of the process and results, to be carried out by the OAS.[69] Responding to concerns about vote tampering and violent protests, Morales asked the Organization of American States (OAS) to conduct an audit of the vote count.[70] Morales said he would call for a second-round runoff vote with Mesa if the OAS audit found evidence of fraud.[71][72]

On 27 October, Morales declared that a coup d'état was in the making against his government, saying that political rivals were planning to stage a coup the following week.[73] On 6 November, the opposition published a report stating there had been electoral fraud, including cases where MAS allegedly obtained more votes than the number of registered voters.[74]

On 10 November, the Organization of American States Electoral Observation Mission in Bolivia published a preliminary report of the audit conducted during the elections. The report found significant irregularities overseen by the Electoral Commission,[4] including widespread data manipulation and altered and forged records.[2] adding that it was statistically unlikely that Morales had secured the 10-percentage-point margin of victory needed to win outright, saying that election should be annulled after it had found "clear manipulations" of the voting system that called into question Morales' win and that "The manipulations to the computer systems are of such magnitude that they must be deeply investigated by the Bolivian State to get to the bottom of and assign responsibility in this serious case.[75] The OAS recommended new elections and appointment of a new elections commission.[2][58]

Within hours, Morales announced that fresh elections would take place.[76][77] By late afternoon of that day,[78] Morales and his vice president, Álvaro García Linera, resigned from office after losing support from the police, the military, and former political allies.[11] Adriana Salvatierra Arriaza, the president of the Bolivian Senate, was next in the line of succession,[78] but she too resigned from office on 10 November.[79]

On 5 November, Professor Walter R. Mebane at the University of Michigan used his own "eforensics" model to detect and predict the level of fraud that occurred during the election.[80][81] He estimated that there were between 20,450 and 24,664 fraudulent votes which were subdivided into votes that were abstentions (no votes) that were then transferred to MAS and votes that were initially for other parties but later changed to MAS. With this level of fraud, he initially determined that it would not have been enough to change the results of the elections (Morales would have had a margin of 10.16-10.27%, depending on assumptions) although on 13 November Mebane said that feedback from colleagues led him to believe that "best formula" for the model led to a new reallocation which indicated that Morales would have had a lead of 9.9% over Mesa, requiring a runoff election.[81]

On 8 November 2019, Ethical Hacking, the tech security company hired by the TSE (under Morales) to audit the elections, stated that there were multiple irregularities and violations of procedure and that "our function as an auditor security company is to declare everything that was found, and much of what was found supports the conclusion that the electoral process be declared null and void".[82] In their official report, one source for the OAS, they stated "We cannot attest to the integrity of the electoral results because the entire process is null and void due to the number of alterations to the TREP source code, the number of accesses and manual modifications with the maximum privileges to the databases being created during the electoral process and the inconsistencies in the software that arose in the TREP and Computo."[83][84][85]

On 12 November, the OAS's preliminary conclusions were contradicted by a separate analysis by the Center for Economic and Policy Research (CEPR), a left-wing policy think-tank based in Washington.[57] The CEPR report said the OAS "provided no evidence to support these statements suggesting that the quick count could be wrong" and postulated that the irregularities they perceived were instead merely the result of normal geographic voting patterns, noting that "later-reporting areas are often politically and demographically different from earlier ones".[51] The CEPR report argued that due to Morales' voter base being in more rural regions, the later-arriving results from peripheral areas were more likely to be in his favor.[86]

On 5 December, the full 95-page OAS report[5][6] was released along with 500 pages of corroborating details as appendices. The audit involved the work of 36 specialists and auditors of 18 nationalities including electoral lawyers, statisticians, computer experts, specialists in documents, calligraphy, chain of custody and electoral organization. The findings included that an outside user who controlled a Linux AMI appliance with "root privileges" — conferring the ability to alter results — accessed the official vote-counting server during the counting and that in a sample of 4,692 returns from polling stations around the country, 226 showed multiple signatures by the same person for different voting booths, a violation of electoral law. On those returns, 91 per cent of votes went to MAS, approximately double the rate recorded elsewhere.[7][87] The identity of this user was later claimed to be Sergio Martínez, who subsequently fled the country.[88]

On 21 December, the Technical Mission of Electoral Experts sent by the European Union published a 67-page report made similar observations and conclusions to that of the OAS. They noted that "there were minutes with an unusually high number of null votes, blank votes and a hundred percent participation of voters in a series of polling stations" and highlighted the general failure of the TSE to declare these irregularities.[8][9]

On 27 February 2020, a further CEPR statistical analysis was published via the Washington Post. The work was carried out by Jack Williams and John Curiel, MIT researchers working as independent contractors for CEPR.[89][90] The researchers stated that "there is not any statistical evidence of fraud that we can find — the trends in the preliminary count, the lack of any big jump in support for Morales after the halt, and the size of Morales's margin all appear legitimate. All in all, the OAS's statistical analysis and conclusions would appear deeply flawed" and that "it is highly likely that Morales surpassed the 10-percentage-point margin in the first round" as originally presented.[90]

After the Washington Post article, Bolivian government officials wrote to MIT about the report. The MIT Associate Provost for International Activities responded stating "this study was conducted independently of MIT... it should be referred to as a CEPR study... we do not endorse or otherwise offer an opinion on the findings".[91] Bolivian newspaper, Página Siete, notes that one of the authors of the report, Jack Williams, had previously signed a letter to the US Congress to oppose the "military coup" in Bolivia and which supported the previous CEPR study.[92] The OAS reiterated their criticisms of the original CEPR report and issued a statement to say that "the mentioned article contains multiple falsehoods, inaccuracies and omissions."[93] Bolivian Minister for Foreign Affairs Karen Longaric called the study "lacking in scientific and academic value".[94] It has also been noted, by political scientist and electoral analyst Rodrigo Salazar Elena, that, except for a few details, the two linked CEPR studies are replicas of the same analysis and that lack of statistical knowledge led commentators to be guided by the prestige of MIT and Washington Post and take the conclusions of the CEPR for granted.[95]

On 10 March 2020, Irfan Nooruddin, Professor in the School of Foreign Service at Georgetown University and author of Elections in Hard Times: Building Stronger Democracies in the 21st Century, wrote a Washington Post article to defend the analysis he performed as the head of the OAS statistical study included in their audit.[96][95] In it, he criticises the 27 February CEPR report by questioning the plausibility of their extrapolation, as well as their assumption that there was no discontinuity in the data beyond the point where the preliminary count was halted. Nooruddin states that at the point where 95% of votes were counted, Morales's vote share began to rise more quickly than it had previously, which is consistent across all six departments reporting at that point. These findings, he says, are consistent with the rest of the findings in the OAS report. He also notes that they are consistent with a separate analysis conducted by Diego Escobari, Associate Professor at University of Texas Rio Grande Valley, and Gary A. Hoover, Head of Economics at University of Oklahoma.[97] CEPR also said that the results of this study were in error.[98]

On 12 March 2020, Professor Rodrigo Salazar Elena, researcher at the Latin American Faculty of Social Sciences in Mexico, wrote an article in Voz y Voto magazine in which he compares and discusses the claims and evidence shown in the OAS and two CEPR studies.[95] He defends the OAS audit by stating that the "duly justified" statistical analysis rests on the "continuity assumption": even with different voting groups, change in vote trend should not exhibit large discontinuities around a single point in time. He states that in order to rebut the OAS analysis and account for the increase in Morales's vote share, it would be necessary to identify a feature distinguishing voters on either side of the threshold. He does not dispute CEPR's method, but notes that it rests on the assumption that voting patterns are geographically contiguous "despite the fact that they are different in terms of reporting the votes to TREP". He offers two potential objections to this assumption. First, he says that geographic contiguity is less plausible than the "continuity assumption" made by the OAS. Secondly, he says that the patterns of stations voting before and after the TREP cutoff are not due to chance.

In June 2020, the New York Times reported on a study by independent researchers Francisco Rodríguez (Tulane University), Dorothy Kronick and Nicolás Idrobo (University of Pennsylvania) which said that the OAS's statistical analysis was flawed, and that the OAS likely used a dataset that incorrectly excluded 1,500 late-reporting voting stations. With these stations included, there was no sudden change in the voting trend, contrary to the OAS's finding. The authors also stated that the OAS used a statistical method that improperly created the appearance of a break in the voting trend where none existed.[99] The researchers' study was published in the Journal of Politics in 2022.[100] Reached for comment by the New York Times, Irfan Nooruddin, who conducted the OAS's statistical analysis, said that Rodríguez and colleagues' study was wrong and did not accurately represent his work. Nooruddin later responded in a comment that the OAS audit included the 1,511 polling stations in all the analyses conducted and that changing the statistical method to the one advocated by the researchers doesn't change the finding.[101] Nooruddin published software to replicate the results found in the OAS audit.[102] The authors note that they "do not assess the integrity of the election overall.. the OAS presented many qualitative indicators of electoral malpractice".[103] The head of electoral observations for the OAS, Gerardo De Icaza, called the conclusions of the study "a moot point", saying that proving or disproving electoral fraud with statistics alone is impossible, a sentiment echoed by Nooruddin in his own article.[99][96] Calla Hummel, a Bolivia expert at the University of Miami also commenting in the NYT article stated: "There was fraud — we just don't know where and how much".[99]

In August 2020, after Nooruddin published the dataset he used for his audit to a Harvard University digital repository,[104] CEPR reported that they had found a "fatal flaw" in the data which "negat(ed) the OAS's claims that fraud affected the results". They noted that timestamps for the tally sheets recorded in the dataset were formatted as alphanumeric strings, rather than in a purely numeric format: therefore, when they were sorted using this variable, they would have been sorted alphabetically, rather than chronologically, so that tally sheets which had been timestamped with a time of 1:00 pm would be earlier in the order than those timestamped at 1:01 am on the same date, despite the latter having been timestamped nearly twelve hours earlier. The CEPR's David Rosnick argued that "the OAS had no real-world chronology of Bolivia's vote count, even though it made accusations that there was a change in the trend of the votes over time that suggested fraud".[105][106]

On 25 August 2020, Nooruddin acknowledged the timestamp sorting error identified by the CEPR resulted in "figures where the x-axis is generated using using this variable were incorrect," but stated that the mistake "does not affect any of the results or conclusions reported in the original OAS report."[107] Nooruddin updated a comment explaining the statistical analysis correcting the timestamp sorting mistake and updating the statistical estimator to a local linear regression as argued for by Idrobo, Kronick and Rodríguez.[108]

In October 2020, the Bolivian government presented the results of a police investigation into electoral fraud during the election, alleging that former minister of the presidency Juan Ramón Quintana set up a "war room" to plan electoral fraud together with a number of members of the electoral bodies. The investigation also alleged that a number of foreign individuals, some linked to Mexico's Labor Party, an ally of the governing National Regeneration Movement, were involved in the meeting .[109][110]

The same month, the Bolivian prosecutors office also released a report corroborating 16 pieces of evidence indicating willful manipulation of the election results. This included redirection of server traffic to a network outside the control of the TSE, falsified tally sheets, burned voter index lists, poor chain of custody not guaranteeing that the material had not been tampered with and modification of data from a number of polling stations. These incidents and more led to the OAS concluding that they could not endorse the results of the elections.[111]

In December 2020, CEPR released another response to the concerns raised by the OAS, writing that "when within-locality variation is taken into account, the election results stand up to scrutiny".[112]