Here's the link to YouGov's discussion of this. The methodology seems reasonably rigorous.

As Jor/dan Ell/enberg said on twitter, "All that's going on is that people don't know the official NAMES of quantities between 0 and 1."

https://twitter.com/jsellenberg/status/1504222785383587841?lang=en

It's interesting comparing the ones where people guess high vs. low.

The best takeaway (stealing from Twitter) is most people are simply innumerate. "Not most people but enough to be significant" is defined to be 20-40%.

Right, but you can pretty easily correct for that innumeracy. People just use 20% when they mean almost no one.

The big actual errors that aren't just easily explained by innumeracy. here are gun owners, obesity, over 50k income, flown, car, high school degree, voted in the 2020 election (ironically there the unadjusted numbers are right), and some smaller ones like unions, vets, white.

People were within 5% on:
* Over 100k income
* Republicans
* Married
* At least one child
* Voted 2020
* White
* Have a pet

I do not believe that 77% have read a book in the past year, would be interested in the source for that. Maybe if looking at a psalm or a familiar new testament passage counts. Mean annual expenditure on reading per capita is USD 100/year.

Estimated proportion living in NYC of 30% is up there with falt-earth "belief," a response that calls into question the idea of persistent mindset and perception of an external world. My reading of this is that people hate surveys. Maybe starting to respond out of a sense of duty but then floor sweepings come out to bring the interaction to a rapid close?

I do not believe that 77% have read a book in the past year, would be interested in the source for that. Maybe if looking at a psalm or a familiar new testament passage counts. Mean annual expenditure on reading per capita is USD 100/year.

Estimated proportion living in NYC of 30% is up there with falt-earth "belief," a response that calls into question the idea of persistent mindset and perception of an external world. My reading of this is that people hate surveys. Maybe starting to respond out of a sense of duty but then floor sweepings come out to bring the interaction to a rapid close?

I don't understand what Ellen/berg is getting at.

And he compliments the scatterplot by saying Americans can more or less rank things correctly from least to greatest. But these are averages, so yes, by the wisdom of crowds things end up kind of correct. So isn't it worth studying the ways in which the wisdom of crowds is really wrong?

People just use 20% when they mean almost no one.

I don't entirely agree with this. I bet if you asked about a non-politically salient category that is 2%, people would get it more correct. Like maybe "percent of people whose AC goes out each year" or whatever.

living in NYC of 30%

This, and the estimation that 1 in 5 people are transgender (one in five! half the population of either cis-gender!), really does lend weight to the "Americans don't understand percentages".

...interpretation.

I honestly thought the percentage of both Jews and Muslims was between 3-5%. Maybe it was back when I was more informed about US demography.

13: With the government forbidden from asking in censuses or otherwise, all percentage estimates are probably private-survey-based and problematic.

This is the flipside of people hating on 538 for saying the 2016 election odds were 70/30, they think that means Clinton is virtually certain to win, Trump's odds are no better than that a randomly chosen American lives in NYC!

What Ellen/berg is saying is that there's a function [0,1]->[0,1] and that if you apply that then people's answers are roughly accurate. You just need to remember that when people say 30% they mean something like the fraction of Americans who live in NYC, not something like 1/3.

For example, if you ask people about any question with three answers you're always going to get a total that's more than 100%.

I honestly thought the percentage of both Jews and Muslims was between 3-5%. Maybe it was back when I was more informed about US demography.

Defining who counts as "Jewish" is a famously difficult problem, so estimates probably do vary a bit. 2% has been a common one for a while now. This may not be as much the case for Muslims, but I'd be surprised if the proportion had shrunk over time.

I'm very picky on that.

Canvassing today, I was in the gentile parts of Pittsburgh. Only one door with the scroll thing.

We still haven't put ours up. Should probably get on that.

My doorway has a bit of mezuzah-shaped missing paint on our door frame from the previous owner, which amuses me because I think of a mezuzah-shaped hole as being a fairly good way to describe my own Judaism.

20: I won't count you until you do.

Well, I'll count you regardless. But I know you. Otherwise I wouldn't.

Interesting survey on Jewish adherence to practices and customs, from 2020. Nearly two-thirds own a mezuzah! I would have guessed more like 40%.

What I gather from the graph is that when people are asked to guess a number within a range, they are pulled towards the middle. Like, obviously a quarter of Americans are not Muslim or transgender, but people don't want to give very high or very low numbers. Also, people aren't thinking very hard.

25 seems right. It's striking how narrow a range all the estimates are in.

Everyone is too nice. If you want to think that think that the Democrats are about to implement Sharia law in the United States or that children are being brainwashed into being trans, which you do want to think if you want to vote for a Republican and still think of yourself as moral, you need to believe those groups are very much more common in the population than they are in real life.

I'm sure there's something to 27, but not all of these estimates relate to hot-button political topics like that and the pattern is still similar.

I already said I'd count you as Jewish. No need to push the case by being argumentative.

We can't help it.

Nearly two-thirds own a mezuzah! I would have guessed more like 40%.

The median American predicted it to be closer to 50%.

13: Same here on being wrong about the number of Muslims. Given ~1.5 billion Muslims in the world, it seems reasonable to guess that about 1% live in the US, which comes to 15 million, which is 4.5% of 330 million. It's just surprising when what seems like a reasonable guess turns out to be so far off.

92% of the population living in either California, Texas, or New York City is pretty funny though.

I am fascinated that anyone would estimate that 20% of Americans are transgender. Like, are they just shitposting when they answer these questions or are they utterly unable to reason? Where do they think all these transgender people are? (New York, along with 30% of the US population, apparently.) Come to that, if you think that 30% of Americans live in New York, why would you persistently paint New York values as crazy outliers as people so often do? Thirty percent of America may be wrong but they aren't outliers.

And if you think thirty percent of Americans live in Texas and thirty percent live in New York and thirty two percent live in California, who do you think lives in Florida? Or Chicago? Or Atlanta? I can accept that people think of the rest of the country as a howling waste with a population density of one person every couple hundred miles, but where are they putting all those Floridians?

I can connect nothing with nothing, as the poet said.

who do you think lives in Florida?

My uncle, heebie's parents, and a bunch of assholes.

I am fascinated that anyone would estimate that 20% of Americans are transgender.

That was then, now it's gotta be like 90-95 and climbing. It's invasion of the transgender body snatchers out there and I'm just sooooo tired. Must. Stay. Awake....

90-95%

Are you on the operating table?

And how can people simultaneously think that 20% of American households have incomes over a million dollars but only 65% of Americans have household degrees and 49% own a house?

Do people just think "I live in my shitty GOP suburb where there are no Muslims, no Jews, no trans people, two lesbians who are married, ultra femme and very discreet, three Black families whose kids get treated like criminals and no unions because we hate them with a passion, my mom's third cousin married a woman whose family is from Japan and I met her once at a reunion and I have never met a trans person or a Muslim...but every other part of the United States is nothing but transgender furry gays and ultra-left union halls full of Jews, Muslims and Catholics with PhDs, better stock up on gunz"? Yeah, probably.

(I grew up in a suburb like that, actually - ultra racist, ultra white, one Jewish girl in my year, two Black students against whom overtly racist remarks were made, general hatred of every good thing in life, etc.)

My uncle, heebie's parents, and that nice commenter whose baby was just in the NICU for a long time, asshole. (Not really! Thank you for your election year service!)

teo, do you have decision paralysis with the mezuzah design? Can you commission one with a local theme/materials or something, given that you've waited this long already? Might as well make it awesome...?

37 One week from today

People are bad at magnitudes, and they're bad at understanding that quantity of speech is poorly related to quantity of what's being talked about. That is compounded by the fact that this chart isn't showing medians, it's showing weighted averages. A value of 18% might be hiding a bimodal distribution where 90% of people picked 10% and 10% picked 90%.

Anyway, it's fun imagining Megacity New York (formerly known as the Boston-Atlanta Metropolitan Axis) with these demographics, where everyone is extremely rich or extremely poor and 75% of people are some combination of trans, Jewish, or packing heat--but don't try to message them, they're still using a feature phone.

I was recently walking through a well-to-do Jewish-neighborhood-adjacent area and saw a house with a mezuzah. The house next door had a prominent cross (I think in that vaguely circular Anglican style) where a mezuzah would go. Felt like a microaggression.

32.last is exactly the point. I cannot emphasize enough that they do not think that 92% of the population lives in either California, Texas, or New York City. The correct answer there is around 25%, so people would say something like 40% of people live in one of those three places, *like they always would with a number around 25%*. You can't ask people about CA, TX, and NYC individually and then add them up to get what they would answer for those three together, because they don't use percentages in a way that's additive.

If you ask them what percent of people live in Iowa they're going to say 20%, like they always do with quantities around 1%. If you ask about each of the 50 states and add the numbers up you're going to get a number between 1000% and 2000%.

Everyone over estimates Iowans because you drive through Iowa to get to the Omaha airport.

teo, do you have decision paralysis with the mezuzah design? Can you commission one with a local theme/materials or something, given that you've waited this long already? Might as well make it awesome...?

We already own two! We even had one up at our old house, then someone gave us another one as a wedding present I think. It's just one of those little things that's slipped through the cracks with all the huge changes in our lives in the past couple years. (I do like the idea of commissioning one from a local artist though.)

The formula is roughly that if the actual proportion is x percent, then people will report .6(x-5) + 50 = .6 x + 20, and conversely if they say "y percent" what they mean is that the actual percent is (5/3)(y-20). So if they say a for group 1 and b for group 2, then the rule for what they'll say for a union b is (3/5)(5/3 (a-20) + 5/3(b-20))+20 = a+b-20. Similarly for three things it's a+b+c-40. So if they say two groups are each at 30% then they'll say that the percentage in both of them is 40%.

40.1: Right. And probably my brother's mother in law.

Hassan wins Dixville Notch!

Our house, originally built for the pastor of a Lutheran church, has a plain wooden cross mounted in the garage. Christianity is not one of our household religions and I think it's time for the cross to move on. I feel strange just throwing it away, but I don't know what to do with it -- like, if we give it back to the church, I fear they'll start trying to exorcise the pride flag and so on. Does anyone want a Lutheran pastor's garage cross? I will happily ship it to you and you can put it in... your... garage?, next to the "Here I park with the passenger side door too close to the bike handlebars, I cannot do otherwise" signs or whatever is typical.

You can't ask people about CA, TX, and NYC individually and then add them up to get what they would answer for those three together, because they don't use percentages in a way that's additive.

I was literally writing a comment inviting someone to divine the computation, and then I paused to finish reading the thread and came to 46.

I think they'd be fine with you giving it back.

like, if we give it back to the church, I fear they'll start trying to exorcise the pride flag and so on

What kind of Lutheran church is it? A lot of them are pretty progressive these days.

I wonder if people would give better answers if we asked "Rank how common X group is in the US from 1 to 10, where 1 means this group does not exist the US and 10 means every American is in this group."

Hang it from the ceiling so that it lays flat and can support a marionette. Or a mobile. Who doesn't like a good Calder?

I realize that the percentage of people with incomes over 1million is but it's misleading to say 0% is reality even if it's 0.1% or 0.01%. It's not a lot of people, but they estimate that raising the tax rate on people in MA with incomes over $1,000,000 could raise 1.3 billion per year. If there were no people in that category, it would be a different story.

Also interested in two's question in 52.

In our fragmented media age, I think we've lost touch with the classic man-in-street interviews on late-night talk shows, from which you'd come away thinking "holy shit, we're doomed, how do these people even feed themselves?" These results are 102% what I would expect.

There's a 5 in there that should be a 50.

55 raises a good point, they didn't ask about any truly small quantities. Almost certainly my linear estimate is wrong if you start asking about truly tiny quantities. The right formula is going to have some kind of logarithmic behavior near zero.

Speaking of which, this is amazing.

Do they think that when they have sex with a woman, they're poking her in the pee-hole?!

(Answer: they're not thinking at all about it, obviously.)

I think this is not the progressive variety of Lutheran church, and moreover it's barely clinging to life. There are about 20 parishioners with a median age of, if I'm doing the math right, 312. I was honestly just curious to know if anyone actually wanted a cross mailed to them by a stranger. People are whimsical!

53: I think that might work, yes, or alternatively a likert of some sort.

General broadcast: if you click through at the top there's a fairly good discussion of the phenomenon, which comes down to guessing that any group that you've heard of is probably a material percentage or you wouldn't have heard of it.

40.1: HELLO. The baby is doing great and there's definitely a bunch of assholes here.

60. Chinese propaganda to make the US look bad.

I have been seeing a surprising number of Florida license plates on cars in the East Bay. This may be a function of being on the freeway more often, but I often see multiple distinct Florida-plated cars in a day. I don't remember this being the case in the past.

Chinese propaganda to make the US look bad.

As if we need the help.

Actually I've heard that the Chinese are really bad at propaganda aimed at foreigners, because the culture of the CCP is so insular that the high-up officials who approve everything have zero contact with anything resembling actual information about the world outside China.

61: Well, it sounds implausible, but then again, men poke with their pee-hole.

Also the woman asks, "Can you pee with a tampon in?" and the man says "No" - maybe he's thinking that if he had a "tampon in" it would be in his urethra.

I heard there's a shortage of tampons so I'm not going to see if one would fit in my urethra.

72: Thanks for supporting women!

I don't want the candiru to feel crowded either.

America: where the Silent Majority estimates they make up 35% of the population

It's an online poll. The percentages are off because some people tried to answer seriously, some answered 50% on everything to quickly earn some pocket money.

65: that's great news! It makes me happy.

76: The meta-analysis they linked found the phenomenon across all sorts of studies, including from back when you didn't have to pay people.

Did Queen Elizabeth never have eggs thrown at her or did she just not mind?

76 is very very cool. They give an explicit model that seems to work well (up to a constant that seems dependent on the setting?) in Eqn. 4, and it seems to work pretty well.

A highlight quote is:
"For instance, every single country in the Ipsos poll overestimated the proportion of immigrants--sometimes by a factor of eight. But after accounting for rescaling due to uncertainty, most countries' estimates were actually less than the predicted estimate for proportions of that magnitude (i.e., most green dots lie below the gray line in the online version of Fig. 2A). In other words, after controlling for domain-general psychological processing of the proportions, it appears that immigration-specific factors may be driving estimates of these populations down, not up, relative to other demographic groups of the same size. This residual error that is left after accounting for general psychophysical rescaling could be explained by immigration-specific factors--factors such as underrepresentation in the popular media and in many social networks."

Er, sorry, meant 78.

If you take into account the issue raised in 55 (which explains why my proposed formula only works for moderately sized quantities) I had worked out that the correction should be this Eqn 4 (though I hadn't posted it here because it felt like I was getting too far into the weeds). It's pretty much the only reasonable model, and it's really really cool that it seems to be quite accurate. This paper completely vindicates Jordan's claim in 2.