Guilty Until Proven Innocent: The Diego Escobari Approach to Bolivia’s Elections

March 26, 2020

Guilty until proven innocent. That appears to be the electoral fraud approach favored by Walter D. Valdivia and Diego Escobari in their recent piece in Project Syndicate. It is also the exact approach the Organization of American States (OAS) had a responsibility to avoid in their audit of the October 2019 Bolivian election. As we write in our response to the OAS, citing the International Foundation for Electoral Systems:

[A]n audit that results from a losing candidate’s allegations of fraud means “that the election commission is starting from a position of weakness … Essentially, the electoral process is considered guilty until proven innocent, often undermining public confidence in the electoral system.” [emphasis added.]

Let us preface this with some very important context that Valdivia and Escobari (V&E) left out of their piece. On October 21, 2019 — the day after the election, and as the official counting of the votes had only begun — the OAS put out a press release asserting that the Tribunal Supremo Electoral (TSE)

presented [preliminary count] data with an inexplicable change in trend that drastically modifies the fate of the election and generates a loss of confidence in the electoral process. [emphasis added.]

The OAS went so far as to presume that a second round in the presidential election would be coming, despite the fact that the official count was less than two-thirds complete.

With its irresponsible press release, the OAS itself actively “generated a loss of confidence in the electoral process.” The OAS statement emboldened a political opposition that had openly questioned the validity of the election even prior to the first vote being cast. Protests spread throughout the country:

The OAS observers confirmed that the violence forced the interruption of the [official count] in six departments: La Paz, Cochabamba, Chuquisaca, Potosí, Oruro and Beni. In Potosí, Pando and Tarija, the infrastructure of the Departmental Electoral Tribunal was completely burned, as were the facilities of the Civic Registry Service in Potosí and Chuquisaca.

Not until November 10 did the OAS even attempt to provide evidence to support their claim of a “drastic” and “inexplicable” trend change. One prominent statistician described their preliminary statistical analysis as “a joke.” In the meantime, our own analysis of the electoral data clearly showed that the fate of the election was foreseeable well before the “inexplicable change” revealed itself in the preliminary count. In so far as the trend did change, the swing toward Morales’s first-round victory was entirely predictable.

Absent an “inexplicable change” which “drastically modifies the fate of the election,” the whole OAS audit falls apart. It is the logical starting point for any analysis of the election. In this context, the significance of the OAS’s initial statement the day after the election cannot be understated.

That also means that the question of whether or not the OAS claim is supportable is bound to be contentious. So what do V&E have to say?

John Curiel and Jack R. Williams (C&W) claimed in The Washington Post to have “found no reason to suspect fraud.”

Given its narrow scope, the C&W study cannot, in fact, dispel doubts about the election’s fraud. Nonetheless, the researchers’ widely publicized blanket statement detonated in Bolivia like a cluster bomb …

Put aside the gross hyperbole. V&E misrepresent the headline — not even written by C&W. The full sentence reads “Our research found no reason to suspect fraud.” [emphasis added.] For their part, C&W emphasize the “narrow scope” of that research, writing:

We do not evaluate whether these irregularities point to deliberate interference — or reflect the problems of an underfunded system with poorly trained election officials. Instead, we comment on the statistical evidence. [emphasis added.]

So yes, neither the report by C&W nor their piece in the Monkey Cage proves there was no fraud. C&W did not set out to prove there was no fraud — merely to investigate whether the data supported a specific claim by the OAS. There is no amount of analysis that can suffice to prove there was no fraud. Neither expanding the scope of the analysis, nor changing any underlying assumptions, can possibly overcome this impossible hurdle. And C&W nowhere imply that they have.

However, C&W do raise doubts about the OAS allegations of fraud to the extent that they are conditioned on the supposedly “inexplicable change.” V&E simply attempt to shift the burden of proof. But the burden of proof absolutely must lie with those alleging fraud. Otherwise, mere allegation — impossible to disprove —will on its own suffice to invalidate an election. This is clearly intolerable.

Even if sufficient, credible, evidence for fraud is found, that is not sufficient by itself to overturn an election. If evidence against a particular person is established, that person should face legal consequences. But even an election with established fraud must not be overturned unless the legally cast ballots would point to a different outcome. In this light, there are two questions: is there credible evidence of fraud, and what impact might it have had on the election?

Despite its obvious importance to the overall fraud narrative, V&E criticize C&W for focusing “on the vote-reporting blackout on election night ― a single event that had little to do with structural fraud.”

This is especially comical as Escobari (along with his coauthor Gary Hoover) have an entire paper based exclusively on the premise that the “blackout” permitted them to “estimate the size of voting fraud” by comparing results before and after the interruption. V&E even cite that paper, claiming that they “found statistically significant evidence of fraud.”

V&E move the goalposts. The overriding concern regarding the election is whether or not the legitimate votes sufficiently favored Morales to warrant a first-round victory. Because the official results showed Morales winning by more than one vote, the size of the alleged fraud matters much more than the statistical significance.

However, even in that paper, Escobari and Hoover (E&H) in no way found evidence of fraud. “Fraud” in their paper is a simple catch-all for any difference between the pre- and post-“blackout” vote that their models fail to explain. Specifically, the model that E&H favor fails to account for any differences between votes counted before and after the interruption — except for the fact of the timing. That is, they do not even attempt in that model to explain why the late votes favored Morales more heavily than the early votes.

Once they do account for geographic differences (municipality and precinct controls) they find that these are sufficient to entirely explain how the late break for Morales put him over the 10 percent margin required to claim victory in the first round. This confirms — not contradicts — C&W. Their results affirm what we have argued since October 22: that the change in trend was entirely predictable.

V&E also cite the work of Rómulo Chumacero as “robust” support for their argument. We are still in the process of attempting to reproduce those results, but several of his summary tables appear to show that geography explained the late votes that delivered Morales’s first-round victory. Unlike the OAS, Chumacero has so far cooperated in responding to substantive questions about his methodology critical for reproducing his results.

Nevertheless, V&E diminish findings that geography accounts for the change in trend:

[L]ack of evidence of trend-change after the TREP blackout is not evidence of an absence of fraud. If fraud was already baked into the 84% of votes reported before the interruption, then C&W’s results suggesting no shift in trend when reporting resumed are incapable of shedding light on the existence of fraud before the blackout.

This is pure burden-shifting. C&W properly address OAS allegations about the shift in trend. But there is no evidence of fraud “baked into the 84%” either. The OAS provides none; V&E provide none. There is no analysis that may be performed to satisfy V&E in this regard; by their own logic, lack of evidence is not evidence of an absence. It is up to V&E or the OAS to provide proof.

Still, if we take V&E seriously here, then we must also account for the fact that evidence is lacking that there was no fraud in favor of the opposition baked into the 84 percent. We may just as easily suggest that Morales’s victory was close only because the opposition committed fraud. Simply making the suggestion does not make it credible. Surely, E&H would not accept this suggestion to be a valid critique of their work. Why is V&E’s argument of C&W any more valid? In any case, the entirety of E&H’s analysis is conditioned on the idea that the 84 percent is clean. Absent that assumption, E&H’s analysis offers nothing.

V&E further call out C&W for comparing late-counted precincts to similarly sized precincts counted early. Yet they merely suggest that this could be a problem “[i]f electoral fraud is easier to engineer in small precincts.” Again, there lies the unstated assumption that such fraud would be in support of Morales. The presumption of guilt persists.

Nevertheless, the OAS levered the suspicion of fraud — suspicion it helped manufacture and promote — into a demonstrably unbalanced audit of the election. The OAS restricted much of the focus of its audit exclusively on MAS-heavy geographic areas with no consideration to the possibility of irregularities occurring elsewhere. Having (inevitably) found irregularities, the OAS pointed to the fact that these areas heavily favored MAS as evidence that the irregularities themselves benefited MAS.

This would be circular logic if the last step held at all. Consider that such a presumption of benefit — even if the OAS had audited more broadly — would actually reward the opposition for spoiling tally sheets where Morales performed best. More importantly, the OAS provides no specific evidence that the legitimate votes cast on the 226 tally sheets it identified as “irregular” should look less favorable to Morales than the official results indicated — let alone to a degree sufficient to invalidate the results. In fact, what we discovered was that the results reported on these 226 tally sheets look almost exactly the same as the non-irregular tally sheets from the same voting centers.

Again, V&E start from a supposition of guilt, stating, “the fact that Morales was even on the ballot at all last October indicated that fraud was afoot.” V&E may not like that the “Constitutional Court caved in” by allowing Morales to run in the 2019 election. It may even suffice to make opponents vigilant in hopes of securing the integrity of the election, but it cannot possibly indicate fraud.

They continue, writing:

In addition, the preponderance of evidence that has since emerged points to rampant electoral fraud. An audit by the Organization of American States (OAS) revealed major irregularities and manipulation, including falsification of poll officials’ signatures, altered tally sheets and databases, and a broken chain of custody. Most damning of all, the transmission of voting data was redirected to two unauthorized hidden servers.

We discuss these concerns in our report, but neither V&E nor the OAS provide any credible evidence that these irregularities in any way favored Morales, let alone were part of a coordinated attempt to manipulate the electoral results. According to the public reports of both private contractors involved in the election, the TSE actually believed at the time that fraud was being committed against Morales.

Even the “broken chain of custody” concerns raised by the OAS included opposition protesters burning electoral materials. The OAS used this to cast doubt on the election results saying this hindered their ability to verify the results. That may be true, but it is the height of chutzpah to accuse electoral authorities of manipulating the election by pointing to opposition destruction of electoral materials. V&E go even further and actually point to this destruction as evidence of fraud.

There is no credible evidence that Morales failed to win by the necessary margin based exclusively on legitimate votes. V&E have no basis for criticizing C&W — who make no claims unsupported by their research. V&E don’t even disagree substantially with the findings of C&W. Rather, V&E must accurately report what others argue and quit pretending that their inability to model the election results justifies their cries of fraud.

Ultimately, the most important thing is whether or not the actions taken by the OAS were justified. By denying the clearest explanation for the change in trend, using that ignorance to conduct a completely biased audit of the election, and then misrepresenting the report’s findings as proof of fraud, the record is clear: the OAS’s actions have been consistently irresponsible.

Support Cepr

APOYAR A CEPR

If you value CEPR's work, support us by making a financial contribution.

Si valora el trabajo de CEPR, apóyenos haciendo una contribución financiera.

Donate Apóyanos

Keep up with our latest news