Summary: debunking "2000 Mules"
The movie doesn't show what viewers think it showed -- it's dishonest, improbable, and most importantly, unsubstantiated.
Last May, partisan podcaster Dinesh D’Souza published a film called “2000 Mules” that makes a new allegation about 2020 election fraud, that there existed a massive “ballot harvesting” operation. This is a practice illegal in most states of people influencing an election by collecting and delivering ballots from only one side. The movie was based on a reports by “True The Vote”, a partisan activist group. They claim to use geolocation (GPS) data to track people who have visited ballot dropboxes multiple times. There has been a lot of fact-checking of this movie. In this post, I summarize the most important bits.
The biggest struggle is that the movie doesn’t show what viewers think they saw. The movie doesn’t show anybody stuffing an unusual or illegal number of ballots into a dropbox. It doesn’t show anybody visiting dropboxes more than once. It doesn’t show any GPS data. Many viewers believe they saw such things, but this is because the movie misrepresents what they showed.
Yes, I saw the movie. Supporters shrug off critics, claiming they didn’t even watch the movie. I watched it. I have screenshots below, with timestamps. Yes, it’s true their unbeliever relatives/neighbors probably haven’t watched the film, but fact-checkers like me have.
This post can easily be debunked — simply cite where in the movie somebody stuffs an illegal number of ballots, visits a dropbox twice, or where geolocation data is shown. I can’t cite where it doesn’t happen, of course, but supporters can easily cite where it does. This has never been done.
Said more simply, the movie’s claims are unsubstantiated. Regardless if the claims are actually true, we shouldn’t believe them without evidence, without proof. No such evidence is shown in the movie, and they haven’t released their raw data for us to look at outside the movie.
Republicans have legitimate concerns about the 2020 election, whether hasty pandemic rules favored one candidate, or whether these rules allowed more fraud. Reasonable people can disagree on such matters. But they require evidence to backup claims.
Baseless claims of election fraud harms our county. A number of Republicans have pointed this out out in a recent letter. True patriots either show the evidence backing such claims, or if they don’t have evidence, refrain from making the claims that hurt our Republic.
Thus, substantiating these claims should be our highest priority. If you believe they are true, you should demand D’Souza and True the Vote release the raw data so that they can be proven true. If you believe they are false, you should demand they release the raw data so we can all see they are proven false. No matter what you believe, you should demand release of the data, and criticize D’Souza and True the Vote for failing to do so.
D’Souza’s deliberate deceptions
The movie doesn’t actually show what viewers thought they saw. This is due to the misrepresentations.
A good example is this segment (at 0:48:05). D’Souza’s voice over claims “What you are seeing is a crime. These are fraudulent votes.”
You are not seeing evidence of a crime or fraudulent votes here.
This man is named Mark Andrews. The Georgia election commission investigated this him. They tracked down his license plate number (blurred in the movie, but visible in the original video). They found:
he dropped off 5 votes
he has 5 eligible voters in his household
ballots for all 5 of those voters were dropped off on this date [[see Correction #1 below]]
it’s perfectly legal for somebody to drop off all the ballots for their household in this manner
In short, this video isn’t evidence of a crime. Mark Andrews is now suing D’Souza for defamation.
D’Souza may have other evidence. He may have video and GPS evidence of this same man dropping ballots in other dropboxes, or at this dropbox at other dates. Thus, we can’t prove the man is innocent from what we see in the movie. We just know that this video contains nothing suggesting a crime, despite D’Souza’s deceptive voice over.
At 0:30:07, they show a man putting two ballots into the dropbox. The voice over describes it as “In this case, he drops a few on the ground, picks them up, stuffs them into the box”. But in the video, he’s only placing two in the box, not stuffing a bunch as implied. He drops only a single one to the ground, not a “few” as they explicitly claim. A lot of dropbox videos show somebody dropping off two ballots, because in Georgia it’s legal for only one member of a married couple to drop off both ballots.
These examples show the fundamental problem of the film: they claim one thing while showing another, and viewers leave the film remembering what was claimed rather than what was shown.
The movie contains a lot of “artistic recreations” and “dramatizations” that are deceptive.
To be fair, not all are deceptive. For example, D’Souza gets criticized for stock images of Moscow to represent cities. In this case, nobody is tricked into believing something wrong.
But it’s not clear where “artistic recreation” ends and “evidence” begins. Real documentaries clearly label their artistic recreations for this reason. It’s obvious from social media that many people couldn’t tell the difference between reality and recreations.
The movie shows maps with little red dots representing GPS data. These are not real data. For example, at timestamp 24:59 in the movie, it shows this picture. This is not a picture of data they have, for one thing it’s mall in New Jersey, where they don’t claim to have data. Instead, it’s a stock photo onto which an artist has simply drawn little red dots. It’s dramatization, showing the sorts of data they claim to have, rather than actual data.
Moreover, GPS isn’t his accurate. At best, a phone gets to about 15 feet, though normally it’s closer to 100 feet. These dots imply accuracy down to inches.
To repeat: nowhere in the film do they show any GPS/geolocation data. It’s all just artistic renderings.
The most obvious case of them lying is when the movie points out an unusual high number of ballots dropped in one box, many times the average, saying “one box in Gwinnett County that had a chain-of-custody document with 1,962 ballots noted on this”. It showed the following image (timestamp 0:38:28). This is not the original document, but an artistic recreation.
The actual document looks different.
Virtually no viewer of the film would’ve understood this was a recreation and not the original document.
In many cases, they are deliberately misleading rather than obviously lying.
If you read social media, you’ll see that many viewers of the film think they helped solve a murder with the GPS data. They didn’t — but neither do they explicitly claim this.
This confusion comes from a conversation about a murder “ebbing on cold case status” and handing over GPS data to the FBI. The case wasn’t “cold” so much as already “solved”.
It’s a neat trick. Fact-checkers debunk the claim they solved a murder, and then D’Souza responds by pointing out (truthfully) that they never explicitly said they solved a murder, and thus all fact-checking of the film can’t be trusted.
But they did lie about it being a “cold case”. And they did lie by implication, even if not explicitly. Indeed, when you look at the fact-checkers, they debunk D’Souza’s “suggestion” that he solved a murder rather than his “claim”.
This incident shows that the fact-checkers are indeed trustworthy — and D’Souza isn’t.
Geolocation doesn’t work that way
I’m going to disagree with other fact-checkers here and say that the GPS algorithms in the movie are merely improbable rather than impossible. As other fact-checkers point out, phone geolocation data isn’t accurate to prove mules — by itself. It can’t place people within a foot of a dropbox. But it’s certainly accurate enough to place somebody within 100 feet of multiple dropboxes at specific times, which can be correlated with the surveillance videos. The two sources of data combined could certainly be used to find mules, which is why it’s so notable that the movie fails to do so. It identifies not a single mule using this technique.
The movie suggests that because geolocation was accurate enough for Jan 6 prosecutions, then it’s accurate enough for tracking mules. This isn’t true. We can look at the details to see why. In this prosecution, Google claims its geolocation accuracy is less than 100 feet.
We can draw a map of the Capitol with a 100ft circle to show that this is sufficient to catch people within the building:
We can likewise draw 100ft circles around dropboxes to prove that it’s insufficient to locate a mule near a dropbox. This is shown below, the circle is centered on the precise location of the dropbox in Sandy Spring (outside Atlanta).
Thus, phone geolocation is accurate enough to locate rioters in the Capitol on Jan6, but not accurate enough to locate mules.
But there’s more to the story: data brokers.
In the Jan6 prosecutions, prosecutors got the data from Google with a warrant. Data brokers do not have access to Google’s data.
Instead, data brokers get their data from CSLI (cell-site location information) and third-party apps/libraries running on the phone.
Neither of these sources is sufficient to give the data that True the Vote claims.
CSLI gets its data from cell sites, not the phone. It’s way worse, struggling to get locations within 1000 feet, sometimes locations are a mile away. Even with improving the data with interpolation, extrapolation, and other algorithms, it’s still far too inaccurate to deliver anything close to what True the Vote claims.
Third-party app/library data only covers about 5% of mobile phones. It’s as accurate as anything else on the phone (such as Google Maps), but it only gets a tiny percentage of the population. That means any estimates True the Vote makes would have to be increased 20 times, from 2000 mules to 40,000 mules. With their relaxed algorithm, that suggests closer to half a million mules and that virtually all ballots dropped in dropboxes were muled.
So we know two things:
mobile phone GPS data isn’t accurate enough to prove mules by itself, without confirmation from other data (like surveillance videos)
data from brokers isn’t close to even that
A naive use of this data will catch delivery drivers, inter-library lending programs, people dropping off lawn signs, and the like. They do nothing in the film to make me trust they did any more than the naive algorithm, and a lot to make me believe they are covering up flaws.
Regardless of any of this, though, the data speaks for itself. if they claim the GPS data shows mules alone, then simply publish it.
Conclusion
The primary message you should take from this posting is that the movie is unsubstantiated. The evidence viewers think they saw inside movie wasn’t what they thought. Outside the movie, none of the terabytes of video and GPS evidence they claim to have has been made available. Partisans are willing to believe without evidence, but the rest of us won’t.
The secondary messages are that the claims are incredibly improbable and that D’Souza and True the Vote are deliberately dishonest. We aren’t going to believe such a vast conspiracy without evidence anyway, but these problems give us even more reason to disbelieve.
If the claims are true, if there were such a ballot harvesting operation, then this is far too important to coverup. Partisans should be demanding D’Souza and True the Vote release their raw data so the rest of us can see.
If the claims are not true, then patriots should be condemning D’Souza and True the Vote for sowing distrust in our democracy, harming the country.
Either way, whether you believe or don’t believe, you should be demanding that D’Souza and True the Vote release their data — and that it’s not believable until they do.
Sources
True the Vote gave part of their data to the Georgia government, who found nothing, nothing suspicious, no probable cause to believe this happened.
I’ve done my own fact-checking of the film, but my efforts are influenced by those who’ve put a lot more work into this than I have, such as @AngryFleas and @PootDibou. The latter has done a lot of work collecting as much data from the film as possible, such as independent copies of the videos seen in the film, or original documents (such as the original “chain of custody form” shown above). This post is just the major points, they’ve done a thorough job debunking a lot of more minor points.
Corrections
Minor typos have been corrected.
Correction #1 (itself corrected)
Above, I repeated the claims from the investigator that all 5 ballots were dropped off at the same ballot box on the same data (“this drop box on this date”). D’Souza pushed back on this, claiming there’s no way to tell they were in the same dropbox. I have independently verified that all 5 votes were returned on the same day as the video (October 6, 2020), but I have not yet been able to verify they were placed into the same dropbox. I’m still researching the matter.
D’Souza may be correct that the public information does not show which dropbox was used. I’m working from public records, as is D’Souza, I suspect. However, public records don’t contain all the information. I’m looking to verify whether non-public government records might have this information.
Great work here, Rob, as usual.
FWIW geolocation data provided by Google, and perhaps other sources, often includes “GPS” derived geolocations. Google reports those locations as accurate to 2 to 16 meters, i.e., 6.2 to 50 feet. However visits to an area are determined by indentifying geolocations inside a geofence. That fence is almost always larger than the area of interest to ensure that all potential visitors are counted, at the expense of falsely counting some non visitors. For the specific case of True the Vote, a geofence with a radius of 100 feet was used.