Tag Archives: statistics

Bush Did North Dakota

Continuing yesterday’s discussion of fake news:

Guess et al says that 46% percent of Trump voters endorsed the Pizzagate conspiracy theory. Does this mean fake news is very powerful?

We can compare this to belief in various other conspiracy theories, as measured by the 2016 Chapman University Survey Of American Fears. About 24% believe there’s a government conspiracy to cover up the truth about the moon landing, 30% about Obama’s birth certificate, and 33% about the North Dakota crash.

This last one is especially interesting because there was no unusual crash in North Dakota when the survey was written. The researchers included it as a placebo option to see if people would endorse a conspiracy theory that didn’t exist. 33% of them did.

Before we make fun of these people, consider: there’s a strong presumption that surveys don’t contain made-up questions. There was no “don’t know” option included on the poll, just various shades of “agree” or “disagree”. In order to condemn the people who “agreed” that the government was probably covering up the crash, we would have to assert that the more correct answer was “disagree”. In other words, that people should have an assumption of trusting the government, until they get some specific reason to distrust it. You can make that argument, but it’s not obvious. You could also start from the opposite assumption, where the government is guilty until proven innocent.

To put it another way, suppose I gave you the following survey:


1. Sandy Hook
2. The coronavirus
3. Obama’s birth certificate
4. The North Dakota crash

…many of us would guess he was lying about the North Dakota crash, without a second thought. And if there later turned out to be no North Dakota crash, we wouldn’t feel particularly ashamed; under the circumstances we made the right choice. If you think the government is as untrustworthy as Alex Jones, well, there you go.

I’ve previously talked about a lizardman constant of 4% on polls. That is, it’s hard to get a poll result much lower than four percent for anything, because of respondents making mistakes or trolling. If 4% of people supposedly believe something, that doesn’t mean we need to be concerned about that fraction of the population, it just means that poll has it its floor and it’s hard to conclude what the real number is.

In the same way, maybe we can posit a North Dakota constant of 33%. This is how many people believe in conspiracy theories when there’s no reason at all to believe them, not even the flimsy reasons conspiracy theories usually provide. Sometimes, if there’s a lot of evidence against them, fewer than 33% will believe in a given theory. But if it’s just “Conspiracy! True or false?” – 33% will say true.

Let’s look again at that statistic from the Guess paper – “46% of Trump voters believe”. I think their source is this poll, which finds:

Overall 38% of Americans agreed with the claim, so Trump voters (46%) were not outrageously more likely than anyone else. Other groups unrelated to ideology were about equally likely to believe it (eg 45% of Hispanics).

Like the North Dakota question, this one had no “unsure” or “what the hell are you talking about” option, forcing everyone to feign agreement or disagreement. We see that the majority of agreement is lukewarm. 75% of Trumpists and 85% of Hispanics who believe Pizzagate only “probably” rather than “definitely” believe it.

I don’t think the evidence suggests Trump voters live in an outrageously different world from the rest of us. Instead, it suggests there’s a North Dakota constant of 33% – the number of people who will believe a conspiracy theory for no particular reason. It looks like about 10 – 15% more Trump supporters than predicted believe Pizzagate, probably because it attacks Clinton, and 10 – 15% fewer Hillary supporters than predicted believe it. But these are relatively small effects, and equaled by eg whatever mysterious thing is going on with Hispanics. In any case, it all averages out to about the predicted amount.

Why is this North Dakota Constant of 33% so different from the Lizardman Constant of 4%? I don’t know. Lizardmen seem like a pretty crazy conspiracy theory, but is Hillary’s involvement with Satanic pizza parlors really that much less weird? Sure, Pizzagate is more politicized, and that might make some difference – but then how come a full 24% of Democrats believe it, six times more than Lizardman’s Constant predicts?

One part of the story is that the lizardman poll offered “don’t know”, and 7% of people chose that. If, denied that option, those people would split evenly between yes and no, that brings us up to 7%ish pro-lizardman. But that’s still nowhere near 33%.

I think this is probably a story about low-information voters. If you imagine you’ve never heard about Pizzagate, and you read the question as written, it doesn’t sound too outlandish. Some Clinton staffers’ emails contained some code words. The pedophilia and Satanic abuse are pretty out there, but post-Jeffrey Epstein we all assume somebody’s doing some kind of creepy pedophilia stuff somewhere. Maybe if you don’t know anything about this, and you don’t have the strong priors about Satanic ritual abuse that you get from studying the history of those claims in the 80s and 90s, this one seems like a toss-up. Certainly it seems like more of a toss-up than a clearly-stated assertion that reptilian aliens rule the world. If your prior is “most conspiracy theory-ish things are probably true”, this sounds like the kind of thing that could be true, whereas you might balk at the lizardman statement.

Here’s another question from the same poll:

Who believes Obama was secretly born in Kenya? Lots of people – including 28% of blacks. I’ve been told again and again that birtherism is a racist conspiracy theory and no person could possibly believe it except as a way of dog whistling white supremacy. Yet here we are with 28% of blacks supporting it – and this isn’t a small sample either! I have no idea what these people are thinking, except that 28% is pretty close to the North Dakota Constant and maybe we should just write this one off.

I conclude we probably shouldn’t draw too many conclusions from specific statements like “X% of [GROUP] supports [CONSPIRACY THEORY]”, especially if X is around 33%. It’s probably just the North Dakota Constant. Likewise, we shouldn’t interpret Pizzagate’s high polling numbers as much evidence that fake news is very convincing – though you could still make an argument that fake news plays a role in transmitting believable conspiracy theories to people who are predisposed to believe them.

Of course, there are some high-information voters who still believe these things really strongly. I think they deserve a more complete treatment, which I want to give later. I think a preliminary sketch might look like: if you start with a prior on something being true, you don’t necessarily need much evidence. The North Dakota question suggests that conspiracy theorists start with a high prior on any given conspiracy being true. What remains to be explained is why some people stick to that prior even after they get more information.