hello $2.90 stock image from https://www.shutterstock.com/image-vectopeople-shaking-hands-vector-handshake-two-234189016
Robert Kennedy Jr. is running for president as an independent candidate. He wants you to know he's serious, and that's he's not a spoiler. You know, that thing he called Ralph Nader in 2000? He's not that. He's going to win. And he's got the numbers to back it up, with a sleek new video to explain it all. He's got a gambit planned, to take Biden off the board with the cold and unfeeling hand of hard facts.
Disclaimer: I know that it is the job of Kennedy's campaign to make the optimistic pitch for Kennedy's chances. But since the optimistic pitch for Kennedy's chances is misleading and wrong, we should hold accountable the people who say misleading and wrong things.

Um, Nerd Alert!!!1

Kennedy's campaign recently released this video to explain the Big News, starring 'Director of Content' Jonathan Hiller. Hiller wants you to know that he's a bit of a nerd. A numbers guru. A 🤓, hold the ☝️. He just loves diving into the data and seeing where it leads, in a fair, even-handed manner. Let's explore the numbers with our new friend Jon.
i can assert with 95% confidence that this man still says 'amazeballs'

• 0:47 'we conducted a nationwide poll of over 25,000 respondents, and our margin of error was 0.6%. That makes it, by far, THE most accurate predictor of the 2024 election that we have thus far.

I could not tell you with much confidence what a 'Director of Content' actually does. But from the way he talks about statistics, I'm guessing that working with statistics is not a big part of the job description.
First of all, he tries to dazzle the audience with numbers. It is true that most polls have sample sizes between 1,000 and 5,000 people, and it is indeed true that 25,000 is a larger number than either of those numbers. But sample sizes of around 1,000 have become the standard for offering a reasonable degree of precision -- a sample of 1,000 people from America's population of ~330,000,000 results in a margin of error of +/- 3%; this means, in simplified terms, that there's a 95% chance the real value is within 3 points of the poll's result. Why not use larger samples for issues of national importance, and have lower margins of error? In the words of Columbia stats professor Andrew Gelman, 'such larger samples generally a waste of time because public opinion varies enough from day to day that it is meaningless to attempt too precise an estimate. Indeed, to do so would be like getting on a scale in the morning and measuring your weight as 173.26 pounds.' To assert that level of precision six months out from an election is even less meaningful. For this reason, high-quality pollsters tend to focus on high-quality random samples to reduce bias, and thoughtful weighting techniques to make that sample as representative of the general population as possible. In short, quality trumps quantity; a well-designed study of 1,000 participants is superior to a shoddily-designed poll of 25,000. Put a pin in this.
Second of all, and more importantly, this is not how polls work, and this is not what the word 'accurate' means. In statistics, the words 'accurate' and 'precise' have distinct definitions, and statisticians are generally extremely careful to distinguish between the two.
if you know a Director of Content, please show them this diagram before it's too late
Accuracy is 'a measure of trueness or bias, how close the average value of your results is to the true value.' while
Precision is 'measure of variability or repeatability, or, how close your results are to each other.'
A low margin of error suggests high precision, as you'd expect future surveys conducted with different samples to give very similar results. But you cannot call this, or any other poll today, 'accurate,' because accuracy depends on one thing -- the closeness of your result to the true result. You know, the election that hasn't happened yet. Maybe they should have had their Director of Statistics make this video instead.

• 1:04 'this is the head-to-head matchup between Biden and Trump. Trump with 294 and Biden with 244. Trump wins by 50 electoral votes, and if you're a political nerd like myself, you know that that is not very close.'

At this point, I"m starting to doubt not only that Mr. Hiller knows anything about statistics, but also that he's a 'political nerd' as he claims. Just as a statistician would not call a poll of an event that will only resolve in six months 'accurate,' a political nerd would not have such a boneheaded understanding of the Electoral College.
First of all, winning with 294 of the Electoral College votes would indeed be a historically close election. Winning 54.65% of the EC would result in the 9th-closest presidential election, in terms of percentage of the EC as a whole, in American history, sandwiched between Grover Cleveland's 1884 victory (which turned on just over a thousand votes in New York) and Jimmy Carter's win in 1976 (less than a two-point shift in just Ohio and Wisconsin would have flipped it to Ford).
Second of all, the EC margin does not accurately reflect the closeness of the election. You don't need to be a 'political nerd' to know this -- both 2016 and 2020 saw the victors winning 306 EVs (or 'not very close,' as Hiller calls it), but the elections came down to tiny margins in a few swing states. Trump swept the decisive states of WI, MI, and PA by less than a point each to win, while Biden's victory hinged on a 1.2-point victory in Pennsylvania.

• 1:27 'in a head-to-head matchup, Robert F Kennedy Jr. wins 367 electoral votes versus Joe Biden's 171. Now that is what we call...a landslide' (landslide video)

Why was this even polled? Issues with the methodology aside. Kennedy's campaign is using this video to tell Biden that he should drop out. Trump dropping out is extremely unlikely and Kennedy doesn't seem to be pushing for this to happen.

• 2:01 'so, to recap, Biden cannot beat Trump in a head-to-head race, and it is not close. Biden cannot beat Trump in a three-way race, and it's still not very close. Bobby wins in a landslide against Biden head-to-head. And Bobby beats Trump head-to-head. So wait, who's actually the spoiler here?

A compelling argument. I guess we're going to have to take a look at this poll.

the Oogie Boogie poll

At the bottom of that spoiler page from earlier, Kennedy provides three documents for us to peruse; two related to the poll by John Zogby Strategies, conducted between 4.13.24 and 4.21.24, and a powerpoint expounding on the poll. Let's take a look at this vaunted pinnacle of human forecasting achievement, also known as 'THE most accurate predictor of the 2024 election that we have thus far.'
I'm just going to give you a list of some of the more interesting state predictions this poll makes of a 2024 head-to-head between Trump and Biden, compared with how those states voted in 2020.
State 2024 Zogby prediction 2020 result projected swing
Utah R+2.7 R+20.5 D+17.8
Rhode Island D+4.6 D+20.28 R+16.2
Maine R+1.7 D+9.1 R+10.8
North Dakota R+18.8 R+33.4 D+14.6
Vermont D+21 D+35.4 R+14.4
Arizona R+11.6 D+0.31 R+11.9
Oregon D+3.9 D+16.1 R+12.2
Arkansas R+6.6 R+27.6 D+21
Oklahoma R+13.1 R+33.1 D+20
New York D+5.8 D+23.1 R+17.3
Virginia R+0.5 D+10.1 R+10.6
Missouri R+5.5 R+15.4 D+9.9
Indiana R+4.1 R+16.1 D+12
Now, maybe these are actually the leaked 2024 results. Maybe this is 'THE most accurate predictor of the 2024 election that we have thus far.' Maybe Traditional Democrats will power Biden to a twenty-point swing in Arkansas. But I am slightly skeptical that over a dozen states will swing by double digits in what appear to be entirely random directions. Reasonable people may ask what the hell happened here.
When you actually look at the poll itself, in the other document, an issue becomes clear from the very first page. Remember that '25,000 respondents' thing? Here's how they did it.
https://preview.redd.it/z3hbhxh6nhyc1.png?width=1146&format=png&auto=webp&s=fc931c2650217e13912a8196b6021d8ddaadc719
This was not a poll of 23,683 people, as the title suggests (which is less than 25,000 -- I guess counting isn't a skill expected from a Director of Content either); it's 50 polls, each with a sample size of around five hundred people. Remember when the campaign was dunking on those inferior polls that only had a thousand participants? Those unacceptable 3% margins of error in most regular polls? Every single mini-poll in this set falls below that mark. The lowest margin of error in this whole set comes for New York, with its n=740 for a population of 19.7 million clocking in at around 4%. This poll reminds me of the villainous Oogie Boogie from The Nightmare Before Christmas, in that it appears to be a unified and formidable presence, but when its seams are pulled, it is quickly revealed to be a sham composed of many tiny, pathetic components.
But that still doesn't explain how these numbers are so off. Look at some of these numbers -- Utah and Virginia as swing states. Missouri, Arkansas, and Oregon all close. You thought Democrats sweeping Arizona in the midterms would prevent them from a double-digit crash in the general?
This is normally where I'd look into the methodology, to see if there are any glaring errors that could result in these mistakes. But that turned out to be hard to do, as the poll only has the results. The only hint at the methodology comes from the title -- 'Interactive Survey of US Voters.' Interactive surveys usually use gamified elements ot hold the attention of the participants and convey the questions in a novel and easy-to-understand way. But obviously, when getting creative with your survey questions, you need to be extremely careful not to make design choices that will push users towards one option. And it seems like we'll never know what choices Zogby and co made.
When reading about Zogby, I quickly found that they're not the most reputable name in polling. Zogby is an old poll -- so old that Nate Silver called them the 'worst pollster in the world' in 2009, in which he criticizes their shoddy metholodogy for their online-based polls. Silver points out that, in 2008, Zogby had Obama favored in Arkansas, which other pollsters did not even consider possible (Obama lost AR in 2008 by twenty points) and had Obama close to victory in Oklahoma (he would lose it by over thirty.) Add on several misses in 2006, and it seems like Zogby's record of huge misses and secretive methods goes back a long way. If I worked for Kennedy's campaign, I might have commissioned them for this poll on their shoddiness alone, in the hopes of producing such an inexplicable poll to justify an unjustifiable campaign.

The powerpoint of desperation

In case that Hard Evidence didn't convince you, Kennedy's got one more arrow in his quiver: a good, old-fashioned slide deck laying out the case for Junior. And this thing is the cherry on the sundae. This is the document that convinced me that there is not a single person at the Junior Table who cares about the responsible use of statistics or avoiding misleading narratives. We have...
a PragerU graph -- that is, a visual that appears to convey quantitative data but, upon further inspection, contains no labeling or metrics that would allow a viewer to understand the numbers themselves. The only real takeaways are that RFK is in the Good Corner and Joe Biden is in the Bad Corner.
Junior pulls the old Truncated Axis trick, starting the Y-axis at a very high value to make the increase in this value look more pronounced than it is. Also, the year labels should be on the graph itself, rather than having to assume anything from the sources.
A table implying that Kennedy is more popular based on his website user data. This is misleading because voters know what Biden's policies are because he's been the president for the last four years. Voters don't know what Junior's policies are because few people knew who he was until recently.
https://preview.redd.it/so4sruc83ayc1.png?width=1392&format=png&auto=webp&s=05e491558d28b35b4f6863aee3be1fdf6f5f79b2
When I saw this last image, I was a little confused. A 28-point gap in the popular vote is absolutely unheard of. It would outstrip Harding's 26-point margin over Cox in 1920, the biggest popular vote landslide in a contested election ever. And a sample size of 56,000? Apparently replicated by two separate pollsters, with the same results? I was unaware that TMZ had decided to dive into the polling space. The only polls I was ever aware TMZ did were those little web polls beneath their articles. Surely he's not talking about that.
https://preview.redd.it/qcg9ns5f3ayc1.png?width=697&format=png&auto=webp&s=37670b224047a7f4dd31e99cb041861e5e866541
Oh.
It's an opt-in website poll, the example every high school statistics teacher uses on the first day to demonstrate how meaningless data can be used to mislead people. And perhaps Junior's campaign should have checked the polls again, because their guy is apparently losing to Joe.
by the transitive property Biden will win a 49-state landslide
As for the supposed corroboration on the 64-36 number on Zogby, I have no idea what they're talking about. It asserts the presence of an n=56,000 poll from Zogby conducted in May 2023, but was apparently only reported on by TMZ in 2024. I can find no evidence of this n=56,000 poll anywhere, and TMZ has a grand total of 1 article even mentioning the term 'zogby' on their site, and it's from 2006. And furthermore, if they had a huge-sample Zogby poll showing Kennedy winning in a landslide, why wouldn't they show that one off instead of the n=25,000 one they're bringing up now?
Genuinely, the most likely outcome seems to be that they just duplicated the poll because the people who made this slide deck either have no inclination or no ability to actually fact-check anything they publish.

Who cares?

I know what you're thinking. Who cares? Biden isn't going to drop out and endorse Kennedy, and this is the kind of insane moonshot thinking you expect from longshot third-party candidates. And furthermore, most of his base are the kind of disaffected voters who wouldn't necessarily vote if he weren't on the ballot. And he might take more from Trump than Biden anyway. And that's all true, but I still think it's important to point this out.
Kennedy is a scammer. He's raising the hopes of people who will inevitably be let down. He's using slickly-edited videos and official-looking documents to scam voters who don't know a lot about politics into giving him their money, their time, and their support, all to support a vengeance-driven campaign sprouted from the rotting ego of a blue-blood who has accomplished nothing of value in his seventy years on this planet and yet demands that tens of millions of people roll out the red carpet for him and acclamate him, in reverance of the one thing of note he has: his name. And I hate scammers. That's all.