Confusion

Here's a weird result Al pointed out from yesterday's NYT race poll. They asked "Just as your best guess, about what percentage of all Americans are black: less than 10%, between 10 and 20%, between 20 and 30%, between 30 and 50%, or more than 50%?." You get these ranges:

White: 1 21 33 33 8 5
Black: 4 24 26 24 17 4

In short, blacks and whites are both massively overestimating the number of black people in the country. That's interesting. And I also think it's interesting how little racial divergence you see in the answers here. My guess would have been that thanks to residential segregation blacks tend to overestimate the proportion of African-Americans in the population while whites would underestimate it. But that's not how it goes.

Comments

Hi Matt,
If you displayed that data in a fixed width font, like Courier, it would be a lot easier to read. When you can't put data in a table, the fixed width font is your friend.

Posted by Jeff | July 17, 2008 8:40 AM

I would like to see how the results break down according to region. I grew up in the deep south, went to public schools, etc. I remember how shocked I was when I found out that black people only comprise about 10% of the US population. I had always assumed it was somewhere between 30-40.

Posted by JH | July 17, 2008 8:41 AM

White: 1 21 33 33 8 5
Black: 4 24 26 24 17 4

What are those numbers? What do they mean? Matt, are you transmitting information about troop movements to the enemies of America again? If so, I think you should hide the data more cleverly.

Posted by James Gary | July 17, 2008 8:43 AM

Bir Pirikette Sen Koy!
İnternetin Sosyallaştığı Site
http://www.piriketseverler.tr.gg
Thank You

Posted by PiriketSeverler | July 17, 2008 8:44 AM

talk about confusion! you give 5 categories and then show a double table of 6 numbers.

Posted by rickenbacker | July 17, 2008 8:47 AM

The really interesting thing about these numbers is that you average them you get a totally wrong answer. And yet the "Wisdom of Crowds" assured us that it would be more accurate than any individual answer.

Posted by John DE | July 17, 2008 8:54 AM

I believe this result obtains for most minority groups, i.e., people overestimate their percentage of the total population. I have seen similar polls overestimating the number of Jews and Asians in America. In part, these results may occur because most minority groups are overrepresented in the major media capitals of New York and Los Angeles, so they tend to be overrepresented in most of what the media produces.

Posted by y81 | July 17, 2008 8:57 AM

I used to believe in the self-correcting Wisdom of The American Electorate too, until 2004.

There's just a lot of really stupid, ill-informed people out there.

Posted by Ban Johnson | July 17, 2008 8:59 AM

Agreed, this is unreadable as presented.

The six categories are: Less than 10%, 10-20%, 20-30%, 30-50%, More than 50%, and DK/NA.

See http://graphics8.nytimes.com/packages/pdf/politics/20080716_POLL.pdf

Page 27, Question 80.

Posted by AGT | July 17, 2008 9:07 AM

Does "Blacks" include Latin American, Caribbean or African blacks?

Many americans seem to have this weird idea regarding Lat Am blacks, that they are Latin, and not Black (I would think they could be both - ditto with the Daisy Fuentes and Mike Lowell's of the world on the white scale).

Posted by JRVJ | July 17, 2008 9:11 AM

Grew up in New York, and until college, I think, I would have sworn that the U.S. was oh, maybe 15% Jewish. (Try buying a menorah in Poughkeepsie . . . ) It was a very odd realization - hey, we're a superminority!

Agree with JH - would be interesting to see regional breakdowns. I might have gone for c) 'til pretty recently, and even after exposure to actual stats, I was still misremembering it as being on the high end of b) (I'm not wrong, I'm just temporally displaced!)

The whites who picked >50% are presumably the FOABP demographic.

OT:Yahoo says "Dozens of catfish walk through South Florida neighborhood". Well then. See what gay marriage will get you?

Posted by Dan S. | July 17, 2008 9:12 AM

I think another missing piece of the puzzle is the correct answer - 12.9% according to the US census...

Posted by Andrew | July 17, 2008 9:14 AM

"
I used to believe in the self-correcting Wisdom of The American Electorate too, until 2004.
There's just a lot of really stupid, ill-informed people out there."

Ah, but what percentage?

Posted by Dan S. | July 17, 2008 9:17 AM

my guess that the shares are probably a reflection of relative media coverage.

Posted by Chris | July 17, 2008 9:23 AM

I suspect it is, in part, a function of the experience to which JH adverts: African-American populations cluster, so, when you come upon a black community, it's probably reasonably sized. I also wonder to what extent media focus on race, and the use African-Americans as the standard stand-in on race, affects the numbers.

Posted by SomeCallMeTim | July 17, 2008 9:25 AM

Dan,

27%

Posted by Brock | July 17, 2008 9:29 AM

Until I joined the military, I hadn't met too many Protestants and was shocked to find out that there are so many of 'em in the country.

Growing up on Long Island, I thought everyone was Catholic or Jewish, with the odd Jehovah's Witnesses thrown in.

I don't think I met a Baptist until I was in the Navy.

Posted by calipygian | July 17, 2008 9:40 AM

What is the real percentage of blacks? I always thought about 16%.

Are black people overrepresented on the it-leads-it-bleeds TV news? Can that be where some overestimation comes from?

Posted by ThresherK | July 17, 2008 9:59 AM

In part, these results may occur because most minority groups are overrepresented in the major media capitals of New York and Los Angeles, so they tend to be overrepresented in most of what the media produces.

Except they're not. If you look at most media (EW did a survey on this), whites are vastly overrepresented in TVs and movies. Even more so when you talk about leading, rather than supporting, roles.

Posted by Mo | July 17, 2008 10:05 AM

What are those numbers? What do they mean? Matt, are you transmitting information about troop movements to the enemies of America again?

lol

My guess would have been that thanks to residential segregation blacks tend to overestimate the proportion of African-Americans in the population while whites would underestimate it. But that's not how it goes.

On the latter point, I can see very easily why whites would over- rather than underestimate it.

Many liberal whites have diversity sugarplums dancing in their heads, leading or deluding them (us) to believe, despite significant residential segregation, that America is the great mosaic.

Simultaneously, many conservative whites, having aggressively chosen to residentially segregate, have a paranoid sense of siege -- there's all these NEGROES and we must seek shelter in the exurban cul-de-sac lest they deflower our daughters.

Maybe I'm being ungenerous.

Posted by Bart Acocella | July 17, 2008 10:09 AM

I think the real answer is: Americans suck at math. I'd bet that if you showed 1,000 Americans a picture of 25 apples, with two yellow, two green, and the rest red, and asked them to estimate the percent of green apples there were without doing the math, you'd get a distribution not far from what you see above.

But, on the other hand, one does wonder what is going through the head of a what person who honestly believes that more than 50 percent of the population is black. Does such a person live in constant fear of a Rhodesia-like rebellion? How does he explain the past 250 years of what rule? Or the typical corporate power structure with black males in the mail room and white males in the board room?

The mind boggles.

Posted by jlw | July 17, 2008 10:10 AM

And being a true American, I botched the appples example above. The numbers ought to be closer to 4 yellow, 4 green and the rest red.

Posted by jlw | July 17, 2008 10:14 AM

jlw,

The funny thing is a reader above showed you demonstrated their point. Other than BET, Univision and the WB, where the hell are minorities overrepresented in the media? Sure when you include pro sports and music it may be the case, but how can someone turn on the TV and think any group (outside of whites and Jews) are overrepresented on TV.

Posted by Mo | July 17, 2008 10:18 AM

Ask people a question they don't know or care about the answer to, give them a continuum of choices and they'll mostly pick the ones in the middle.

If you gave them the choices as
"2%, 2%-4%, 4%-6%, 6%-12%, More than 12%" you'd probably find people's answers cluster around the "4%-6%" option.

Posted by Edmund in Tokyo | July 17, 2008 10:23 AM

Depressingly, Edmund is almost certainly right. The poll should have simply asked "what percentage of the US population do you think is black", except that would have meant slightly more work for the pollsters.

But, on the other hand, one does wonder what is going through the head of a person who honestly believes that more than 50 percent of the population is black.

I suspect that this share of the population (roughly 5%) either a) are counting all non-white people as black, and live in fairly high-minority cities, b) have no idea how percentages work, c) are very, very stupid or d) all of the above.

Posted by ajay | July 17, 2008 10:43 AM

Edmund in Tokyo,

Wouldn't the people doing the survey randomize the order in which they list the options to prevent the "I'll choose the middle one" problem?

I know they don't do that in political surveys when they list "Strongly support/somewhat support/somewhat oppose/strongly oppose/no opinion" (which they almost always list in that order) but it would be easy to mix up the order in which you read the ranges off to correct for this tendency.

Posted by Pesto | July 17, 2008 10:54 AM

If you think this is bad, 18% of Americans think the Sun revolves around the earth. But we did better than the Germans or the Brits. Suck on it Copernicus.

Posted by Mo | July 17, 2008 10:58 AM

The poll question I thought was awful is:

"60) About how many of the people who live in the immediate area around your home are black — none, a few, about half or almost all?"

http://graphics8.nytimes.com/packages/pdf/politics/20080716_POLL.pdf

WFT? None, a few, about half? The big divide for whites in this country, at least in the Northeast, is where in the 1-15% blacks in the neighborhood spectrum they fall (very few whites live in >15% black neighborhoods -- would be nice to know how many). There is a huge range of 'a few' blacks in the neighborhood, and where one is in this range make quite a bit of difference.

I moved from a ~8% black urban area (in a 40% black city in a 14% black country) to a ~2% suburb recently (a one to maybe two black kids per class in school sort of place, more Chinese than Blacks). What's the right answer here? Yes, there are two black families in the immediate neighborhood of about 100-200 houses. What is that? 'None' isn't right, but I'd be hesitant to call it 'a few' as if it were 10%.

It's my perception that Americans overestimating the fractions of blacks in the population is a longstanding polling result that also shows up in free form questions, but I cannot find a citation for this. I also don't know if it is a result limited to blacks. I think people also overestimate the number of other minorities, like Jews or Chinese. Not clear if people are just bad at grasping small fractions.

And I cannot remember ever not knowing that there are around 13% blacks in the US, since I must have looked it up as a kid (

Posted by stefan | July 17, 2008 10:59 AM

And I cannot remember ever not knowing that there are around 13% blacks in the US, since I must have looked it up as a kid. Given the glaringly obvious -- from living in this country -- fact of residential segregation relying on my own perception must have seemed like a silly idea to get at this number.

Posted by stefan | July 17, 2008 11:05 AM

Ajay - 5 and 4% is the percentage who don't know.

8% of Whites and a rather staggering 17% of Blacks think Blacks are the majority.

Posted by John | July 17, 2008 11:07 AM

How would mixing up the order lead to a more valid result? If people just pick the middle choice on the page the result will be wrong in a different way. If they figure out that the choices are scrambled and pick the middle answer (percentage, not choice by position) deliberately, the result is the same. Maybe more people would pick "don't know" if they are hopelessly confused, which may be the actual correct answer, but this would be a weird way to run a poll.

Posted by skeptonomist | July 17, 2008 11:16 AM

(very few whites live in >15% black neighborhoods -- would be nice to know how many).

A great site for racial demographics in NYC.

http://130.166.124.2/ny_1.html

Posted by Jeffrey Davis | July 17, 2008 11:20 AM

Brock,

In Peoria, IL, a few years back, a man who had very publicly been disbarred got ~25% of the vote for state's attorney, so your citing of ~27% sounds about right.

Posted by Indecisive | July 17, 2008 11:30 AM

This is a well-known result in the social sciences: First, that Americans overestimate minorities by double or more. Second, that the percentages people give, when added up, exceed 100% by a lot. There have been studies, as well, assessing who makes what kinds of errors.

Posted by CSF | July 17, 2008 11:45 AM

Polls that ask these sort of questions are increasingly useless. The people being polled no longer take them seriously. I know that if I was disturbed by some pollster in the middle of dinner I would not give a damn if I was being honest.

Posted by Ben | July 17, 2008 11:49 AM

Ben:
Your skepticism about polls is well-earned. But: even with respondents like you, the findings turn out to be roughly reliable, consistent, and important. This has also been well-researched. And, people who have to put real money down on continue to invest more and more on polling.

Posted by CSF | July 17, 2008 12:15 PM

Here's a weird result Al pointed out

One prefers to be called "Frequent Comment Troll Al". But anyway...

My first thought about it is that it results from a disproportionate emphasis of blacks by the media. But I too vaguely recall having seen some surveys that show that there isn't a disproportionate emphasis on blacks in the media. So who knows.

Posted by Al | July 17, 2008 1:04 PM

I read this blog daily but have never felt compelled to comment until now. I know it's kind of an inside joke how sloppy Matt is with spelling and typos, but this takes it to another level. Summarizing what everyone else has noted, we have a post about incorrect perceptions of the percentage of African Americans in the population which does not bother to give the correct number; five categories mentioned in the text but six columns listed on the table; and a table that isn't useful to illustrate the point being discussed.

Other than that, nice attention to detail Matt.

The Poor Man calls himself the World's Laziest Blogger, so that title is taken, but I'd nominate Matt for World's Sloppiest Blogger. Or at least, sloppiest high-profile blogger that wants to be taken seriously.

Posted by Doodah | July 17, 2008 1:12 PM

I saw a poll like this about the Jewish population of the U.S. many years ago, with similarly laughable results. Most people landed nowhere near 2%.

Posted by Brittain33 | July 17, 2008 1:33 PM

Regarding my comment above about minorities being overrepresented in the media: to be honest, I was only thinking about sports (in addition to the black participants, it seems like every panel on ESPN includes a black guy) and children's programs (where there's always a suitably diverse crowd hanging out with Lizzie McGuire or Zack and Cody). In other cases, although the entertainment media may present an unrealistic picture of race relations, they don't necessarily portray a disproportionately high number of minority group members. But in any case the media may not be the correct explanation for the popular tendency to overestimate minority numbers.

Posted by y81 | July 17, 2008 1:50 PM

"The Poor Man calls himself the World's Laziest Blogger"

I firmly believe Matt's in contention and uses voice recognition software -- thus the usual "write" for "right." Only way to explain it. On the other hand, he had to use a keyboard to screw up that table, I bet (I don't think you can cut and paste by voice), so go figure.

Posted by David in NY | July 17, 2008 2:13 PM

I wrote earlier that "(very few whites live in >15% black neighborhoods -- would be nice to know how many)".

I've been looking around, and I may have to amend that to "very few economically secure whites who aren't in school live in >15% black neighborhoods." But I'm having trouble finding exactly on point references. It would be nice to know what the data look like.

Finally, one ought to be careful about the term 'neighborhood'. I've lived in one of the supposedly most integrated places in the country, Hyde Park in Chicago, and the building level residential segregation was highly significant. Some of that may be life-cycle and income effects, but it is hard to know.

Posted by stefan | July 17, 2008 2:29 PM

Does "Blacks" include Latin American, Caribbean or African blacks?
Many americans seem to have this weird idea regarding Lat Am blacks, that they are Latin, and not Black (I would think they could be both - ditto with the Daisy Fuentes and Mike Lowell's of the world on the white scale).
Posted by JRVJ

"Hispanic" is a 1969-era artificial race construct invented by two Jewish lawyers who were employees of the Nixon Administration at OMB tasked with forming up the EEO.
They also invented the artificial race of "Pacific Islanders".
The two Jews did it to limit who is "disadvantaged" as well as create Affirmative Action privileges. They didn't want 100% Mayans to claim they were "Native Americans" and access government opportunity set-asides for "real Indians". Similarly, they thought native Hawaiians were deserving - but not other Pacific Islanders like those of the Philippines, Indonesia, Australia. Though later court challenges since taxpayer money and privileged spots in college and jobs were involved added Aussie Aborigines and Maoris.

What the two Jews under Nixon created persisted because the media and identity politics practictioners of the Democratic Party loved it. They also created an instant pool of racially or ethnically "privileged" under AA people numbering in the tens of millions that liked the notion of being stakeholders and getting various bonus points in certain venues.

That it made no sense, racially or culturally, has no basis in consistancy, is besides the point.

Whites born in any Latin Country are "Hispanic", but not those born in Spain, not those whites born in Latin Countries or who moved here with less than 3 generations of "Hispanictude" - so John McCain is out, while the children of Jeb Bush are "in" as 100% Hispanics if they are looking at college or looking for a job with "diversity" bonuses in selection.. As is Mike Lovell.

Blacks were classified in a more "universal" way, by the brainstormers at OMB. Any touch of the tar brush, anywhere in the world, people can call themselves black. But many part-blood black-Portuguese, black-spanish, or black-English from a Caribbean island or Belize with no Latin heritage lucky enough to be within the "Sphere of Hispanictude" can call themselves Hispanic or white if they prefer. And many prefer to be called Latins over black.
Just as many 100% native blood Mexicans and Guatemalans prefer to be called indigenous Indians....but the OMB Jews from 1969 deny they can be that, because only those with indigenous blood from Canada or the USA can be called "Native Americans".
But Asians with "hispanic blood and culture" like in the Philippines or Guam are out of luck. They cannot say they are Hispanic just as their being pacific islands do not make them Pacific Islanders. They are screwed. No AA bonus points.

With the large amount of race-mixing in the last 40 years the two Jews who made up the EEO-1 racial and ethnic classification lists did not anticipate? With more and more Americans unwilling to describe themselves in the artificial 1969 categories? With the Human Genotype Mapping Project finishing its work soon?

Perhaps it is time to put fake races and ethnicities in the past, despite the meddling of Sandra Day O'Connor and other lawyers in robes.

Posted by chris ford | July 17, 2008 3:58 PM

"Growing up on Long Island, I thought everyone was Catholic or Jewish, with the odd Jehovah's Witnesses thrown in."

Seriously. I remember watching TV at the age of like 9 and wondering how the priest could have a wife and kids.

Posted by mad6798 | July 17, 2008 4:23 PM

This finding has been around forever -- I recall one from a couple of decades or so ago that the average American though 37% of the population was black and 15% was Jewish.

I'm sure if you added up all the estimates, they'd come out to well over 100%. People aren't very good at doing arithmetic in their heads to do reality checks on ideas.

In general, people aren't very good at thinking statistically, and those of us who do think statistically about everything aren't terribly popular.

Posted by Steve Sailer | July 17, 2008 6:12 PM

Anyone else think that chris ford was putting maybe a little too much emphasis on the Jews in that post? Just a little creepy.

Posted by AlanC9 | July 17, 2008 9:36 PM

I do research on subjective responses to numerical stimuli, among other things (my former dissertation advisor is a big name in this field). There are a number of well-known biases people mention:

-Middle category bias on discrete survey responses. This happens a lot but is systematically varies by culture and there are other kinds of response sets. For instance, Brits tend to have middle category bias while Americans tend to use the extremes more, in particular the positive ones. The really outlying examples are Koreans (middle category only) and Greeks (EXTREMES!).

-Perception of low or high probabilities are biased towards the middle in general. The bias is not symmetric and in general tends to overstate the likelihood of rare events more than it understates the likelihood of common events. This in turn leads to:

-Subadditivity, the phenomenon of having stated probabilities systematically adding up to a number greater than 100%. (Yes I know the name seems backwards, but it's not because superadditive, subadditive and additive are technical mathematics terms.)

-One big cause of this problem is the fact that the sample space (the universe of possibilities which requires things to add up to 100%) is not something most people have in mind when you're getting their estimates. It's not even clear what the sample space is in most surveys---though many survey researchers naively suppose that subjects know what *they* meant---and this can lead to some definitely odd behavior as people of different cultures respond to items. As understanding sample spaces properly took the best mathematicians a few centuries to work out and weren't really settled until the 1930s by the eminent Russian mathematician (who else?) Andrei Kolmogorov, Joe Average can perhaps be forgiven for his lapses.

-All this really has to do with the fact that our brains REALLY don't work on the very constrained mathematical space imposed by the additivity constraint of probabilities. The seemingly simple task for asking for a proportion turns out to be very, very difficult... unless you happen to know the right answer from having looked it up.

-If you renormalize subadditive probabilities, they are often much closer in proportion to accurate numbers than one might suppose, although the nonlinear biasing towards the center is still there. Alternatively, if you elicit information that is mathematically equivalent to probabilities but closer to being additive in structure, people are much more accurate.

Posted by crimfan | July 18, 2008 11:17 AM