Observing the Observers: The OAS in the 2019 Bolivian Elections

By Jake Johnston and David Rosnick – March 10, 2020

Executive Summary

The Organization of American States (OAS) regularly organizes observation missions tasked with monitoring elections throughout Latin America and the Caribbean. These missions, which provide assessments on the fairness, reliability and transparency of an electoral process, are generally conducted in a relatively impartial and professional way. At times, however, these missions have acted in a biased and unprofessional way and, as a result, have undermined the democratic institutions of the host country. This was the case in 2010 in Haiti, when an OAS Special Audit Mission controversially recommended changing the results of the country’s presidential elections, despite having no reasonable justification for doing so.

As this report shows, the OAS’s observation activities in Bolivia’s 2019 general elections are the latest example of a deeply problematic observation mission whose dishonest, biased, and unprofessional conduct has caused serious damage to the country’s democracy.

In a series of statements and reports released between late October and early December, the OAS has alleged that serious irregularities occurred during the vote count following Bolivia’s October 20 elections with the implication that there was deliberate manipulation of the results that changed the final outcome. In previous papers, we have shown how the very data on which the OAS based its initial allegations clearly contradicted these allegations. Jack Williams and John Curiel, two researchers at the Massachusetts Institute of Technology (MIT) Election Data and Science Lab, have replicated our findings, and concluded that “we cannot find results that would lead […] to the same conclusion as the OAS.”

This study focuses primarily on the “Final Report” of the OAS audit of the election results and shows how the authors of that report misrepresent the data and evidence found in the audit in an attempt to further bolster their claims of intentional manipulation on the part of Bolivia’s former electoral authorities. The OAS Final Report identifies many real problems with the management of the elections that should be addressed. However, despite claims to the contrary, it does not provide any evidence that those irregularities altered the outcome of the election, or were part of an actual attempt to do so.

This study is not intended to validate the election results stemming from the October 20 vote, nor does it seek to support a particular candidate or political party. Rather, our analysis is focused on the OAS’s actions and public reports on the Bolivian elections. The credibility of the OAS has tremendous implications for the next Bolivian election and for electoral processes throughout the hemisphere. A multilateral organization that is called upon to resolve an electoral dispute must be honest, impartial, and credible. This study shows that the OAS was none of these in the case of the Bolivian electoral crisis of 2019.

The unethical conduct of the OAS in Bolivia has had deeply disturbing consequences. The fraud narrative that the OAS helped promote contributed to Evo Morales, the country’s democratically elected president, fleeing the country — aboard a plane sent by Mexico — months before his term ended; other officials — including former members of Bolivia’s electoral authority — sit in jail; and an unelected de facto government has assumed power with the support of the military and has engaged in violent repression of protests, the persecution of political opponents, and a drastic change in foreign policy.¹

Background

On October 20, 2019, Bolivians went to the polls to participate in general elections. At the presidential level, the incumbent, Evo Morales, running for a controversial fourth term,² faced a fractured opposition represented by eight candidates. At the invitation of the Bolivian government, the Organization of American States (OAS) deployed an Electoral Observation Mission (EOM) to monitor the vote. At about 7:45 p.m., the country’s electoral authority ― the Tribunal Supremo Electoral (TSE) — announced preliminary, nonbinding results with 83.85 percent of votes processed by the preliminary results transmission system (TREP). Movimiento al Socialismo (MAS-ISPS) candidate Morales’s margin over his closest competitor, Comunidad Ciudadana (CC) candidate Carlos Mesa, was 7.9 percentage points — below the 10 percentage point threshold necessary for a first-round win.³ The transmission of preliminary results then stopped, though the much slower official vote count process continued uninterrupted. The TSE provided no clear public explanation as to the reason for the stoppage, or whether they intended to continue with the TREP at all (2019 was the first time the TREP was used for a presidential election, but in previous votes the TREP never processed 100 percent of votes).

Box 1. Understanding Bolivia’s Election and Results Transmission Systems

More than 7 million registered voters were able to participate in Bolivia’s October 2019 election. Across the country’s nine departments, there are 340 municipalities and 3,532 “locations,” a narrower geographic level. Across these locations, electoral authorities set up 5,131 precincts, or voting centers. Electoral authorities set up an additional 165 voting centers abroad for Bolivians living outside the country.

Each voting center contains voting tables and each table has six electoral jurors who perform a mandatory citizens’ role akin to jury duty in the US judicial system. In Bolivia, 207,322 citizens were randomly selected to be jurors and trained a month before the elections.⁴

Once voting has concluded, poll workers count individual ballots and aggregate the results into actas, or tally sheets. There are more than 34,000 tally sheets corresponding to individual voting tables within the 5,296 voting centers.

All six jurors in each voting center must sign off on the tally sheet. Representatives of political parties may also be present at the voting centers. Any person or political organization is able to monitor the vote-counting process as an observer.

There are two systems for the transmission of the results. The first is a preliminary, nonbinding count known as the Transmisión de Resultados Electorales Preliminares (TREP). For the TREP, operators in the field are provided with a smartphone and mobile application, which they use to transmit a photo of the tally sheet to a team of verifiers within SERECI (the civil registry). The field operators also input the results from the tally sheet into the mobile application. When the data inputted by the field operator and the SERECI verifier match, the tally sheet is considered verified and made available online. Due to geographic limitations and time constraints, the TREP is not intended to process 100 percent of the tally sheets.

The second vote-counting system is the official count (or Cómputo), which is legally binding under Bolivian law. The official count is more thorough and precise and takes longer. For the Cómputo, the tally sheets are physically delivered to electoral tribunals in each department and then manually scanned into the Cómputo system. At this stage, electoral officials also check associated electoral material, like voter lists, to confirm the validity of the tally sheet results. It is the only legally valid vote tallying system, and the TSE uses it to determine and announce the final election results.

Images of the tally sheets from both systems are made available online to anyone who wishes to confirm that the information on the physical tally sheets matches the information entered into the system. Periodic timestamped Excel sheets with detailed results are also made available online. Political parties can also ensure that their copy of the tally sheet matches what is published online. This makes it easy to check for inconsistencies, and for any errors to be quickly corrected. While the TREP provides a preliminary vote total, it also serves as a means of protection against fraud. With two sets of results, each inputted through separate processes, it is possible to cross-check the two systems to verify the accuracy of the results.

At 6:30 p.m. the day after the election, the TSE announced updated TREP results with 94.94 percent of the votes processed. Morales’s margin had grown to 10.15 percentage points — above the threshold for winning in the first round. Soon afterward, the EOM issued a press release denouncing an “inexplicable” change in the trend of the vote that “drastically modifies the fate of the election and generates a loss of confidence in the electoral process.” Leaders of Bolivia’s opposition, and many media outlets, interpreted this change in trend as prima facie evidence that fraud had taken place.

Protests against alleged electoral fraud spread throughout the country. Over the following days, protesters attacked departmental electoral offices, resulting in the destruction of original electoral material and the disruption of the official vote tabulation process. On October 25, the TSE announced official results showing Morales winning in the first round with a 10.56 percentage point margin — a figure consistent with the preliminary results. The opposition and those protesting alleged fraud refused to recognize these results.

The Bolivian government requested the OAS to send a team of experts to audit the results and provide clarity amid a burgeoning political crisis. Morales, whose existing mandate was supposed to last until January 22, 2020, agreed to abide by the recommendations of the OAS audit team. An agreement was signed by the Bolivian government and by OAS Secretary General Luis Almagro. The Group of Auditors (GOA), under the auspices of the OAS’s Department of Electoral Cooperation and Observation (DECO), initiated its work in Bolivia on October 31.

The mandate of the GOA was to verify the official results of the election, including by analyzing the chain of custody of electoral material and by assessing the information technology (IT) systems used in the preliminary (TREP) and official (Cómputo) vote counts.⁵

The OAS released a preliminary report on its audit on November 10,⁶ alleging significant IT security vulnerabilities that could have been used to manipulate the results and realleging prior claims of an improbable change in the trend of the vote. The preliminary report recommended a new electoral process with a new TSE. While rejecting the underlying allegations of fraud, Morales expressed his support for new elections. However, later that day Morales resigned under pressure from the high command of the country’s military, which “recommended,” during a press conference, that Morales step down. Morales received political asylum in Mexico and is currently in Argentina, which has granted him refugee status.

Under the resulting de facto government, there has been an increase in human rights violations perpetrated by state security forces, and repression of protests and activists. At the time of this writing, the de facto government has detained more than 40 electoral officials allegedly involved in electoral fraud and levied charges of sedition and terrorism against leading MAS politicians, including Morales. New elections, with a new TSE, are scheduled for May 2020; Morales has been ruled ineligible to participate as a Senate candidate. The OAS has again pledged to send an observer mission for the upcoming vote.

The OAS published the final version of the GOA’s Electoral Integrity Analysis (referred to from here on as the “Final Report”), on December 4, 2019. In an accompanying press release, the OAS stated that there was “intentional manipulation” and “serious irregularities” that made it impossible to validate the election results.⁷ The audit concludes, “There has been a series of intentional operations aimed at altering the will expressed at the polls.”

The Final Report contains five main findings:

The IT systems for both preliminary election results and the final count were flawed and allowed for manipulation.
The process to fill out the tally sheets was marred by irregularities and forgeries that benefitted one candidate.
The deficient chain of custody did not guarantee that election material was not manipulated and/or replaced.
The final 4.4 percent of tally sheets contained an elevated number of “comments” or corrections.
The trend of the final 5 percent of the vote was highly improbable.

The final two findings are considered pieces of “evidence.” The two pieces of evidence are not, in and of themselves, evidence of fraud, but rather are used by the GOA as indicators of how potential fraud may show up in the data. Additionally, the Final Report presents a list with each individual finding, which contains 12 “deliberate actions that sought to manipulate the results of the election,” 13 “grave irregularities,” and seven “errors.” At least 14 of the 25 most significant findings relate to the nonbinding TREP.

Main Findings Relating to the OAS Final Report

This study details the actions of the Bolivia EOM, DECO, and OAS Secretary General Luis Almagro, and provides an in-depth analysis of the OAS’s Final Report of the Bolivian election results.

The first section of our report focuses on the false claims made by the EOM that strongly contributed to a narrative of electoral fraud in the lead-up to Morales’s forced resignation. The EOM, on at least three different occasions, publicly claimed there was an inexplicable change in trend despite the presence of clear contradictory evidence. The preliminary report on the audit and the Final Report reiterate this false claim. OAS actions and statements played a central role in the events that led to the forced departure of President Evo Morales, and the de facto government has used OAS statements and reports to justify violations of human rights. It is important to consider the Final Report within this broader context.

The next section provides an overview of international standards and principles of international postelection audits and details postelection verification strategies. The International Foundation for Electoral Systems (IFES), warns that, while these types of audits may help reestablish confidence in the results amid allegations of fraud, they “should be used only in limited circumstances and according to clearly defined rules.” The section also discusses the only prior example of an OAS audit mission, in 2010 in Haiti, which failed to comply with basic audit principles and had a long-term damaging impact on that country’s democracy.

The remaining sections of this study focus specifically on the Final Report. We find the Final Report to be statistically flawed and lacking clear methodology or standards. Further, the Final Report fails to report the results of the GOA’s own verification exercises and buries, omits, or misrepresents facts and data that run counter to its conclusions. Far from providing a neutral and independent assessment that could provide greater clarity in the midst of a highly polarized environment, the Final Report represents an apparent attempt by the organization to justify its previous actions — including its repeated false claims about an “inexplicable” trend change after the stoppage of the TREP.

Ultimately, the responsibility for the management of the results system — both the TREP and Cómputo — lies with TSE officials and other electoral officials tasked with carrying out the elections. Certainly, the failure of the TSE to provide a clear public explanation for the decision to halt the TREP contributed to the electoral crisis and damaged the election’s credibility. Yet, the Final Report fails to provide essential contextual information that could explain the stoppage of the TREP and restore some semblance of trust in the electoral process. Whereas the Final Report could have provided clarity around these key events in the election, the report’s authors instead opted to obscure and manipulate the public record.

Despite claims to the contrary, the Final Report presents no evidence that the stoppage of the TREP or other IT vulnerabilities resulted in any “intentional manipulation” of the electoral results. The Final Report notes many real problems with the management of the elections, but fails to provide any evidence that those irregularities altered the outcome of the election, or were part of an actual attempt to do so.

Finding I: The Final Report is based on flawed statistical analyses and incorrect assumptions. The Final Report presents two pieces of “evidence” which the GOA used to determine the focus of its investigation. Both of those findings are flawed.

The GOA modifies the earlier claim made by the EOM regarding an “inexplicable” trend change following the suspension of the TREP with nearly 84 percent of the vote processed. In an earlier report from CEPR, we showed that there was no statistical basis for this claim. In the Final Report, the GOA modifies its allegation to focus on a “highly improbable” trend change over the last 5 percent of the votes. But the data presented in the Final Report directly contradicts this finding. Rather than showing a trend change in favor of Morales, the data reveal that Morales’s vote share actually decreased in the final 5 percent of the count as compared to the 5 percent counted directly before.
The authors of the Final Report cite an elevated number of “comments” or corrections affecting presidential vote totals on the tally sheets not included in the TREP. The authors do not mention that tally sheets with complex corrections are placed under “observation” during the TREP and are expected to be given greater scrutiny in the Cómputo. More than 1,000 tally sheets were not included in the TREP because of problems with the images or due to corrections. The elevated number of “comments” on tally sheets not included in the TREP is therefore normal and explainable.

Finding II: The Final Report buries or omits altogether the results of its verification exercises despite the fact that they are highly relevant. The OAS audit team’s mandate included verifying the accuracy and legitimacy of the tally sheets processed in both the TREP and Cómputo systems. Yet, the Final Report buries or omits altogether any relevant findings that run counter to its narrative and fails to properly show or explain the methodology used.

The GOA compared tally sheet images from the TREP and Cómputo systems, finding that the vote totals for each candidate matched across images in 99.8 percent of cases. This indicates that the data input into the systems match the results recorded on the tally sheets. The Final Report only mentions this finding on page 80 (of 95) and fails to explain its significance.
To check the authenticity of the tally sheets, the GOA analyzed original hard copy records for 894 tally sheets out of a sample of 2,863. There is no explanation of the methodology that was used to select the sample, nor for the choice of subsample analyzed. The Final Report states that, in 230 cases, original electoral material had been burned or was missing. But what about the cases where the OAS did find the original electoral material — did the auditors verify those tally sheets? The Final Report never mentions the results of its verification exercise even though those results could help confirm whether or not fraud or manipulation took place.
The most significant factor in the GOA’s inability to verify all the election results was the fact that 23 percent of tally sheets were burned during attacks on departmental electoral offices after the announcement of preliminary results. The Final Report includes the burning of electoral material in its findings of “intentional manipulation,” but fails to mention the context of these attacks. In particular, there is no evidence that electoral authorities or the Morales government were involved in the destruction of this material. On the contrary, all the available evidence suggests that anti-Morales protesters were responsible for the destruction. The Final Report fails to mention this and instead considers that all blame lies with a lack of security coordination between local police units and electoral offices and the inability of electoral officials to properly secure the material (which, even if true, still fails to show that there was “intentional manipulation” carried out by any of these authorities).

Finding III: The Final Report presents no evidence pointing to the manipulation of election results. Only one of the report’s findings of “intentional manipulation” or “grave irregularities” relates to actual election results. This finding is the result of a biased and methodologically unrepresentative sample of tally sheets, and does not appear to relate in any way to the problems identified with the IT systems.

From an unrepresentative sample of 4,692 tally sheets in which MAS had overwhelming majorities, the GOA conducted a handwriting analysis and identified 226 tally sheets as problematic. These sheets correspond to 86 voting centers where one person filled out parts of two or more tally sheets. The report’s authors find this to be “an intentional and systematic attempt to manipulate the results of the election.” To be clear, the Final Report does not find that the same person filled out all 226 tally sheets. No single individual is alleged to have partially filled out more than seven tally sheets. Nor does the fact that an individual filled out more than one tally sheet indicate that any election rules were violated. According to Article 49 of the Rules of the Election and the official guides issued by the TSE, it is only the signatures that are required from the electoral workers and party delegates.
The Final Report claims that all of these 226 irregularities benefitted the same candidate and that if the GOA extended its analysis beyond the original sample, more irregularities would be found. The GOA, however, only analyzed a highly biased and statistically unrepresentative sample of tally sheets with overwhelming numbers of pro-MAS votes. There is nothing to indicate that the same irregularities would be found elsewhere. The GOA only looked for irregularities in one direction. If the GOA extended their sample to areas where CC did well, they would almost certainly have encountered irregularities among those tally sheets as well.
Seventy-eight percent of the impacted voting centers are very small, with four or fewer voting tables (tally sheets correspond to these voting tables). The Final Report failed to consider sociodemographic factors in its analysis. Even if these are examples of “irregularities,” they do not inherently indicate fraud or the “deliberate” manipulation of results, as the Final Report claims. This is likely related to a well-known phenomenon: in rural areas and smaller voting centers, it is not uncommon for one person to assist with the tally sheets. This may be especially common where illiteracy rates are high or where a large percentage of the population does not speak Spanish.
Based on comparisons with previous election results and with non-“irregular” tally sheets within the same voting centers or nearby areas, it is clear that there is nothing systematically abnormal about the results that appear on these “irregular” tally sheets. Results from “irregular” tally sheets match almost exactly the results found on other tally sheets within the same voting centers. This indicates that, even if these are “irregular,” they had little or no impact on the overall results of the election. It is therefore incorrect to say one party benefitted from all of the “irregularities.”
Two-thirds of the 226 tally sheets identified were processed and published in the TREP system before the TREP interruption — indicating that these “irregular” tally sheets were not altered after or during the TREP’s interruption, the period during which the EOM and GOA have focused their suspicions of manipulation of the vote count.

Finding IV: The Final Report presents no evidence indicating that problems with the TREP, real or alleged, compromised the Cómputo or that the Cómputo was fraudulently altered. In an attempt to show that the security vulnerabilities affecting the TREP also affected the Cómputo, the preliminary report on the audit and the Final Report attempt to link the TREP and Cómputo systems by pointing to the presence of TREP images in the Cómputo. This is highly misleading as there is no evidence that these two systems were improperly linked.

More than 90 percent of the TREP images present in the Cómputo correspond to voting centers abroad and are in compliance with the established procedures of the election, as the Final Report itself notes. The remaining TREP images in the Cómputo correspond to tally sheets that were lost, damaged, or destroyed in postelection protests, which, as previously stated, the Final Report does not put into context. These tally sheets were printed, scanned, and inputted into the Cómputo system, thereby severing any unplanned technological link between the two systems as the Final Report alleges. To the extent nonforeign TREP images were used in the Cómputo, it was due to extraordinary and unforeseen circumstances. The Final Report misrepresents these facts.
The Final Report finds that the contractor responsible for the Cómputo system made direct edits in the Cómputo database, and suggests that this could be further evidence of manipulation. However, the GOA fails to note that these changes were monitored by an auditing company and were communicated to all relevant actors. The Final Report includes no analysis of what these actual changes were and if they were legitimate or not, despite having a clear record of what the changes were.

Finding V: The Final Report fails to provide clarity around the TREP stoppage and, as a result, perpetuates a false narrative of fraud. More than half of the Final Report’s findings of “intentional manipulation” and “grave irregularities” relate to the preliminary, nonbinding TREP. Yet, the report fails to take into account essential contextual information that provides a highly plausible explanation for the stoppage of the TREP that suggest that human error rather than foul play was involved. Rather than providing clarity around this key event, the Final Report uses vague language and innuendo and, in so doing, perpetuates a false narrative of fraud.

The Final Report presents contradictory information relating to the TREP. It claims there was no technically valid reason to stop the TREP, but then includes more than a dozen findings related to technical problems with the TREP that emerged before the stoppage. When the EOM pressured the TSE to restart the TREP on October 21, OAS officials were aware of many of the technical problems that the Final Report now alleges to be evidence of “intentional manipulation.” It is perplexing that the OAS would pressure the TSE to take actions that would lead to further problems that the OAS would later criticize and point to as evidence of manipulation.
The Final Report identifies the stoppage of the TREP as an act of “intentional manipulation,” but fails to reference the fact that Ethical Hacking, the audit company whose analysis the Final Report features prominently, sent a maximum alert to the TSE right before the decision was made to stop the TREP. Though it is highly probable that this maximum alert led the TSE to stop the TREP, the Final Report does not mention it.
The Final Report includes the presence of an unmonitored server (BO1) as an act of “intentional manipulation,” but fails to note that the identification of said server was the subject of the audit company’s maximum alert. The Final Report does not reference the fact that Ethical Hacking investigated the use of the BO1 server during the time in which the TREP was stopped and concluded that “there was no alteration of data.” The Final Report, which cites the Ethical Hacking audit multiple times, never mentions this.

Click here for the full, 93-page report

Notes

Long (2020).
Bolivia’s constitutional court issued a ruling allowing Evo Morales to run for a fourth term despite a constitutional referendum held in 2016 in which 51.3 percent of voters chose not to allow indefinite reelection. The TSE ratified Morales’s candidacy.
In Bolivian elections, a candidate must receive 50 percent of the vote or 40 percent of the vote and a 10 percentage point lead over the second-place candidate in order to win the election in the first round.
Guarachi and Melgarejo (2019).
The mandate of the GOA focused specifically on the vote counting process and not on the campaign or other pre-election day matters. The focus of the audit would be on “the verification of the vote count, including tally sheets, ballots and votes; the verification of the process, including matters relating to computing; the statistical and projection component, as well as the chain of custody of the ballot boxes,” the OAS noted in a press release announcing the formation of the audit team. OAS (2019a).
OAS (2019b).
The phrase “intentional manipulation” does not actually appear in the Final Report itself.

Source URL: CEPR