TOO LONG FOR REDDIT............. blah blah

It's fatal for the statistics, at least. It's only possible to [reproduce](https://twitter.com/Master0fNull/status/1296344422024728577) their graphs and tables when you make these errors. Once corrected, the purported anomalies vanish. Like, *poof*. It was known that the analysis of the official count was severely flawed, but it wasn't clear exactly what he did to mess it up. I looked into it early this year and quickly determined that his ordering of the tally sheets didn't match the public data, but I couldn't figure out the specific mistake he'd made. I tried to get at it by taking one of his scatter plots, selecting a few of the most extreme outliers on the periphery, then matching them to specific tally sheets and checking them against the public data. I saw that the dots weren't where they were supposed to be, so I tested a few different errors that came to mind, but I could never crack it. I always knew it was some stupid, lazy mistake, just like the rest of what he did, but I couldn't put my finger on what it was. Once he released his data and code, someone else solved it in, like, ten minutes. You wouldn't believe the effort I put into unwinding some of his work. God, back in January or February, I tried to use optimization solvers to reverse-engineer a split point from his analysis. Why? Because he wouldn't share his data or code! (Also, I'm dumb and have too much free time.)

Also, I think it's necessary to explain some context around this statistics debate. The statistical analysis from the OAS's [final report](https://www.oas.org/fpdb/press/Audit-Report-EN-vFINAL.pdf) published last December claimed there were anomalies in the progression of the vote count for both the TREP (preliminary count) and the Cómputo (official count). The initial TREP claims were [challenged](https://cepr.net/press-release/oas-should-retract-its-press-release-on-bolivian-election-cepr-co-director-says/) literally the day after they were [first made](https://www.oas.org/en/media_center/press_release.asp?sCodigo=E-085/19), then [challenged](https://www.cepr.net/what-is-happening-in-bolivia-cepr-s-recent-work/) again a few weeks later, but the later versions of these accusations could not be attacked head-on because there was a hole in the public data which made it impossible to determine the timing of tally sheet verifications during a critical 23-hour pause in the TREP. In essence, the vote totals on the tally sheets verified during this time could be predicted based on earlier tally sheets, but the order in which they were verified was unknown. The OAS claimed there was a statistical discontinuity in the progression of the vote count *during the 23-hour pause*, which meant their analysis was order-dependent and thus impossible to directly refute based on the incomplete public data. Researchers including myself sought the non-public timestamps many times from the OAS and their statistician, but they ignored all of these requests because they already had the implicit trust of mainstream opinion-makers who never once considered asking them to show their work. When they were [challenged](https://archive.vn/GtQNT) again back in February, this time in the Washington Post (uh-oh!), the Post allowed the OAS's statistician to publish an evasive [non-response](https://archive.is/ZfrX3) in which he merely reproduced the same defective graphs he'd contributed to the OAS report, even though his critics had been [unsuccessfully](https://twitter.com/JohnCuriel14/status/1233056756474904576) [requesting](https://twitter.com/Master0fNull/status/1234927869698224132) the data and code behind them. What's more, this statistician's response was so lackluster that, in a still-unclear arrangement, the OAS effectively swapped him out for a former World Bank statistician who could more capably [bullshit](https://twitter.com/BoliviaOEA/status/1256294524269314049) the public. (Your tax dollars at work, folks!) Then, a set of high-profile researchers who'd been intrigued by the Washington Post dust-up looked into the issue. They were stonewalled by the OAS and their statistician like everyone else, but they managed to go around them and get the NY Times to obtain the non-public data straight from Bolivian electoral authorities, which allowed the TREP claims to be refuted directly. This led to the [study](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3621475) summarized in the [NY Times](https://www.nytimes.com/2020/06/07/world/americas/bolivia-election-evo-morales.html) piece, which refuted the OAS's entire TREP analysis and discredited the work enough that it seems to have forced them, two months later, to at last [disclose](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/SGOFSC) their data and code, which their statistician had declined to share. Along with this data and code, the statistician published a [response](https://drive.google.com/file/d/1PV4kGx_MtUuubP2RXe6nY1KyaamtYTVh/view) to the paper from the Times piece, but this was instantly [refuted](https://twitter.com/frrodriguezc/status/1296801773844344833) and the release of his data and code exposed even more errors such as this now-infamous sorting one. It was well-known among those of us who checked the OAS's claims that their statistician was incompetent and his analysis was [trash](https://twitter.com/futurebum/status/1233819774406316032), but you needed to have the data and code to definitively refute his work and these were extremely difficult to obtain. I did what I could, emailing anyone I thought might have it, making feeble requests to Bolivian electoral authorities, pressing influential Bolivian journalists to obtain it for me, and even contacting my representatives in Congress, but I struck out time and again. Thank God it finally came out. So, to sum up: That entire section of the OAS report is dead. There will be a response to aspects of the refutations, I'm sure, but the space in which the other side has to argue has become quite confined, and the arguments they make won't be vindications of the specifics of the OAS's analysis but rather attempts to pick at the edges of their critics' models.

I figure you probably don't care about all of the above, but I think it's critical to understand the context in which all this occurred. The OAS has been lying to the public this whole time and continues to lie. In fact, according to a CEPR researcher, when he privately reached out to a high-level OAS official shortly after their erroneous October 21st press release, they acknowledged that they were wrong but simply did not care. They thought they could ignore their critics forever and get away with it, but eventually CEPR managed to launder their criticisms via a third-party who got into the Washington Post, which forced the OAS out of its defensive shell and eventually resulted in the complete dismantling of the statistical analysis. At this point, it's become clear that the OAS will do anything to preserve its own reputation. For example, after that rather cautious NY Times piece in June, the OAS had a meltdown and put out a bizarre [rant](https://www.oas.org/en/media_center/press_release.asp?sCodigo=E-064/20) in which they made reckless attacks against the Times, calling them enablers of Hitler and Stalin and so on. The OAS isn't credible on this issue, so it's extremely naive to trust their interpretations of other data which is also non-public. What got the ball rolling on the statistics was the fact that there was public data that was sufficiently detailed to cast doubt on their claims, and this eventually led to inquiries that resulted in the release of the non-public data and the full refutation of the OAS's statistical analysis. But does the public have access to the system logs or the source code for the electoral software? No. What about the voter rolls and supplemental electoral material? No. The logs from the chain-of-custody system? No. None of this is available to the public, and in fact Bolivian electoral authorities abruptly disabled the public [website](https://computo.oep.org.bo/) in January, possibly in response to an email of mine, so there isn't even a full, authoritative source for the [electoral data](https://archive.org/download/election_bolivia_2019) anymore.

Despite the fact that you can't get this non-public data, there are still ways to attack the other parts of their report. In fact, CEPR - you know, the people who were right about all the statistics from day one - put out a whole [report](https://www.cepr.net/report/observing-the-observers-the-oas-in-the-2019-bolivian-elections/) refuting all of the OAS's claims back in March. In the specific case of the computer system, there was the public website as well as multiple detailed reports from the contractors who were hired by the electoral authority to operate and oversee the system. I used all those sources and many more back in December to construct a [explanation](https://www.reddit.com/r/BOLIVIA/comments/e983ts/why_the_trep_was_suspended/) of what happened with the system, but I couldn't get any traction for the same reason it took forever for the OAS's statistics to be publicly discredited. The truth will come out in time, I hope, but for now the government the OAS ushered into power is doing all it can to railroad former officials who've been held in pre-trial detention since last November. Their investigators blab to the media the second they learn any new detail, often before they can even analyze it, and after almost a year of investigation they haven't come out with anything solid. In the past few months, it's become evident that their case isn't as strong as they'd like and they've become desperate. They leaned on the president of the electoral authority - someone hand-picked by Añez who a decade earlier covertly [organized](https://www.reddit.com/r/BOLIVIA/comments/iadssv/tse_president_salvador_romero_covertly_organized/) a media campaign against the MAS - after he appeared to drag his feet on joining their formal fraud complaint. When he finally did join it, the report he provided spoke of possible authors of possible crimes and he didn't point fingers, although he was later successfully pressured into 'fixing' this. They keep saying they're almost done with it, that it's 90% finished, but that's been their story for months now. There's even the bizarre situation where Carlos Mesa, the former and future president of Bolivia as well as the runner-up of last year's election, got so upset about the lack of progress in the investigation that he initiated a parallel one that the Bolivian press dubbed *Fraud Case #2* and which appears to be retracing in the exact same footsteps as *Fraud Case #1*. It's an absolute farce. God only knows what they're cooking up now. I figure it'll come out in the media right before the October 18th elections, since this is all a bullshit political game, so for now there's not much to do but wait.

If you want to be ahead of the discourse, you should ignore all the accusations except for the OAS's handwriting analysis and whatever real evidence the criminal investigation turns up. I'm sure you won't trust me on this, but I know the fraud complaints backwards and forwards, as I spent a while researching them, and I'm right. The OAS knows the handwriting analysis is all they have, which is why they actually released it and even [translated](https://www.oas.org/en/spa/deco/Report-Bolivia-2019/Anexo-de-Pericia-Caligrafica_Redacted_English.pdf) it to English, but withheld every other internal technical report. I have to say, too, that the value of this handwriting analysis is often overstated. In the part of it where they analyzed signatures, the OAS only found one worksheet - that is, not the official tally sheet but a kind of practice sheet - which had forged signatures. That's it. One. The rest of their signature analysis is hot air, which I tried to explain [here](https://twitter.com/futurebum/status/1236469638063587329). It all comes down to the 226 tally sheets where the juror names were written out by someone who also wrote out the names on other tally sheets in the same polling place. I have a deeper analysis of those, though, which shows that much of it is, again, hot air and a few are outright errors, albeit for a complex mix of reasons that I really ought to explain somewhere.

After looking at all this in excruciating detail, my conclusion was that there was definitely fraud, but it occurred at the polling place level and was limited to MAS strongholds where their genuine level of support wasn't substantially lower than the reported one. I think that, most likely, this fraud did not alter the result of the election and I figure that ultimately the evidence for fraud will rest on bland statistical analyses that point to improbably high turnout and vote share in the Tropic of Cochabamba and in rural Potosí. It'll be difficult to estimate this fraud, but someone who has tried is Walter Mebane of the University of Michigan, who put out a range of [counterfactuals](http://www-personal.umich.edu/~wmebane/Bolivia2019.pdf) in which the purported fraud was undone based on his own statistical model. However, in order to alter the final result, these counterfactuals would require the main opposition party to get a higher-than-expected percentage of the vote in the MAS's areas of deepest support. What's more, the corrections he makes are huge and geographically incoherent. Still, his was perhaps the only analysis that tried to cut through all the noise and it's telling that the fraud he points to only amounts to, like, a fraction of a percentage point.

Most Bolivians deeply misunderstand the fraud claims. Even the high-level debates between academics are full of bad information to the point where it gets kind of embarrassing. There's a lot of people who still believe that, like, every third vote for Morales was fraudulent or some nonsense like that. I talk to these people all the time and it's so exhausting. They tell me that they're Bolivian and they lived through the fraud, so how dare I contradict them. I ask them for details, if there was an incident they witnessed or heard about, and they invariably send me viral videos they saw on a social media site, point to dumb errors on the election website that even the OAS dismissed, link me terrible studies like the one from Édgar Villegas which don't merit mention in the serious fraud discourse, cite claims the OAS quietly abandoned last year, etc. I've spent the past, like, nine months watching this all unfold from the sidelines, and it's been unbearable. I sometimes bark on Twitter about all the mistakes that get made, but their press is a closed system and won't even retract clear errors. A lot of Bolivians see the criticisms and think it's all a plot by foreign leftists to trick them, but the reason these criticisms have to come from overseas is because their national media doesn't check anything! They do little in the way of real investigation and generally limit themselves to copying, at times incorrectly, from other, mostly official, sources. You'll hear that same criticism from the right, too. It isn't just me. Everyone there knows that the press stinks. There needs to be a serious retrospective about the media's role in what happened last year, both the domestic media and overseas. CNN presenter Fernando del Rincón, last seen attempting to goad Arturo Murillo into killing blockaders, was one of the worst offenders. And it wrecks me when I see people praise Ximena Galarza as a serious journalist. God, she even got a pat on the head from the US State Department. How gross. Every useful contribution these two ever made occurred when their interviewee coughed up a revealing detail which they didn't understand and failed to follow up on. I think they all wanted Morales gone and opened up a firehose of bullshit on the country to make it happen. He shouldn't have run for a fourth term, especially after the referendum, but it's the responsibility of journalists to be careful and honest with the public, especially in such a volatile political situation.

Sorry, this is far too long. Who knows, maybe someone will get something out of it.

tldr; The OAS's statistics are bad and most of the rest of their report is bad, too. There are domestic criminal investigations that the OAS is hoping will bail them out, but these have stretched on and on without making any formal public accusations, and it's unclear what, if anything, they've found to support the fraud claims.