“How segregated is your city?”: A story of why every graph, no matter how clear it seems to be, needs a caption to anchor the reader in some numbers

Aleks points me to this article showing some pretty maps by Eric Fisher showing where people of different ethnicity live within several metro areas within the U.S. The idea is simple but effective; in the words of Cliff Kuang:

Fisher used a straight forward method borrowed from Rankin: Using U.S. Census data from 2000, he created a map where one dot equals 25 people. The dots are then color-coded based on race: White is pink; Black is blue; Hispanic is orange, and Asian is green.

The results for various cities are fascinating: Just like every city is different, every city is integrated (or segregated) in different ways.

New York is shown below.

No, San Francisco is not “very, very white”

But I worry that these maps are difficult for non-experts to read. For example, Kuang writes the following::

San Francisco proper is very, very white.

This is an understandable mistake coming from someone who, I assume, has never lived in the Bay Area. But what’s amazing is that Kuang made the above howler after looking at the color-coded map of the city!

For those who haven’t lived in S.F., here are the statistics:

The city of San Francisco is 45% non-Hispanic white, 14% Hispanic, 7% black, and 31% Asian (with the remaining 3% being Native American, Pacific Islander, or reporting multiple races).

“Very, very white,” it ain’t.

I’m not trying to pick on Kuang here–I’m sure it’s not easy to write on deadline. My point is that even a clean graph like Fisher’s–a graph that I love–can still easily be misread. I remember this when I was learning how to present graphs in a talk. It always helps to point to one of the points or lines and explain exactly what it is.

And now, here’s the (amazing) graph of the New York area:

NewYorkB.jpg

18 thoughts on ““How segregated is your city?”: A story of why every graph, no matter how clear it seems to be, needs a caption to anchor the reader in some numbers

  1. I wonder if when he says "San Francisco proper is really, really white," he really means "the white part of San Francisco proper is really, really white."

  2. I just emailed Kuang, the author of the Fastcompany article, with a link to this blog. He said he corrected his article, and that he had mistakenly thought that "San Francisco" was just the pink blotch at the upper part of the peninsula. (There's no scale on those maps, which is probably part of the problem). Actually that pink blotch is just the Marina District, which, at about 3 square miles, is roughly 1/15 of the area of the city (and even less of the population). That long empty rectangle coming in from the west is Golden Gate Park, so those splotchy green-and-pink areas above and below it are very much in the CIty of San Francisco.

    So perhaps the story is that every map needs a scale, even more than every map needs a caption.

    But: I just checked the Fastcompany story and it still has the "very, very white" line, so either Kuang's correction hasn't been posted yet or there's a holdup somewhere.

  3. Kuang "corrected" the article to "some parts of San Francisco are very, very white", which is closer to being true. The part of the city that Kuang thought was the whole city is the Marina District, which is 85% white. That's pretty white, alright, but I'm not sure it merits the "very, very" claim. It doesn't seem to me that it does. I don't know why Kuang is so wedded to that claim.

  4. With regards to the New York map – there is a very dense blue part in the lower middle of the map, towards the middle of the blue there is a very dense pink part.

    Geographically, what is that pink part and why do white people congregate there?

  5. It would really make it quicker for viewers to grasp these graphs if dot colors weren't randomly assigned. It's too hard to decode a map in which ethnicities are arbitrarily assigned. It's too hard to remember that Asians are green, Hispanics orange, blacks blue, whites red, and so forth.

    Instead, American culture has well-known stereotypical colors for each ethnicity: black for blacks, and so forth. Assuming you want a white background to your map, that leaves the question of what to do with whites. I usually use blue, because blue eyes are only found in whites.

  6. I agree with Steve that the colors could be better chosen, and I said so on Fisher's blog. But I don't like blue for whites either. A stereotypical, offensive-to-some-I'm sure mapping would be something like black for blacks, white for whites, brown for hispanics, yellow for Asians…except what would you use for a background color? Also, you really want all of the colors to have equal visual weight (which the current ones don't either) and yellow tends to be overwhelmed by other colors.

    Pink for "white" people makes sense, since we are kinda pink. I would try pink for whites, reddish brown for hispanics, black for blacks, and maybe orange for asians, and adjust the intensities to try to give them about equal visual weight.

  7. In response to Steve, I'm perfectly comfortable with the color scheme, though I do think that the yellow-orange for hispanics and the yellow-green for asians is a bit to close.

    The colors used in Fischer's maps (which are really derived from the colors in Rankin's map of Chicago, though the magenta is a bit redder, the blue a bit greener, and the orange a bit yellower. The complaint I would have would be the lack of a color key, something that Rankin provided.

    To me, I had no problem because I saw Rankin's map before I saw the set of Fischer's maps that had been generated on the same basic mold, albeit with a few tweaks and some of the details left out.

  8. Yair and I did white (that is, open circles on a white background) for whites, black for blacks, red for Hispanics, and green for others.

    My favorite of Fisher's maps are those on the original (largest) scale, but for those, I think he should make the dots smaller.

    Also, I'm glad that Kuang corrected his news article, but I'd be even gladder had he credited me with finding the mistake!

  9. Definitely agree about scale. It is worthwhile to look at the original resolution in it's full 3009×3009 glory, and futher I agree that the dots are too large at that scale. However, that is somewhat of a platform quirk. Fischer is using Flickr, so the smaller resolutions are simply risized version of the 3009×3009 original. So, smaller dots would give the smaller maps much less contrast, especially in cities other than NYC and patterns would be less apparent to the eye.

    As for my non-expert view on color coding, it would seem to me that colors with a strong hue would be more effective in this particular format than black and white, as the dots give a sense of concentration rather than merely identity. If you have a white background, white dots with black rims doesn't easily convey the information in the same way that a red or a green dot would.

    Furthermore, the visual weight of the dots depends in part on how they are colored and black and white dots don't carry the same weight as colored dots visually, giving the viewer a somewhat lopsided view of the relative weight of the categories.

    That is not to say that they don't have merit. It seems to me that white and black combined with a set of hued colors would be useful if we consider those tagged as white and black to be categories of different relevance than the hued colored categories, but if one were to hypothetically recolor Fischer's maps (actually probably wouldn't be hard to do…I may try it if I have time) with a black color for one category of another (probably for blacks), it would likely set the category apart from the rest in a way that, if unintended, would harm the presentation of the map.

    At this point I'm going to stop because (a) you have better things to do than read the armchair philosophy on the color-coding of maps by a random commenter on this site and (b) I would like to test some of my assertions for myself should I find the time.

  10. Andrew, you want credit for finding a mistake that thousands of other people also noticed right away? I think that's somewhat excessive. I have to think that anyone who lives in the Bay Area and read his article would know it was a mistake.

    Hey, how about if he credits _me_ for pointing out the mistake; I'm the one who emailed him, after all.

  11. Though I generally consider it to be bad form to post links to one's own blog posts one comment threads as responses I'm going to link to my most recent post as it's (1) tailored to this comment thread, (2) very long, and (3) contains visual elements, making it unsuitable content for a commment.

    As well, I was wondering if Andrew would be able to point me to the work he did with Yair [Ghitza, I presume] that he referenced in his above comment.

  12. Phil:

    I agree with you. There's no reason for him to credit me in particular; it would be fine for him to credit you. From a blogging perspective, though, I think he should credit somebody for the correction.

    You are quantitatively minded and live near San Francisco, so it is unremarkable to you that the city of S.F. is not anything close to "very, very white."

    To others, though, the notion of a "very, very white" San Francisco makes sense. After all, we know that San Francisco is in California, it's liberal, it's full of gay people. Liberals are all white, right? "Stuff white people like," and all that. But, no, that's not the case at all. San Francisco has a mix of races.

    It all comes back (of course) to Rich State, Poor State: People have simplified and false conceptions of political geography, and they can inadvertently distort what facts come to them to fit these templates.

    I'm not, of course, saying that Kuang was trying to mislead–any more than I think that Tucker Carlson was purposely mistaken when he made his notorious statement about rich voters. Rather, my guess is that Kuang and Carlson are both journalists who made innocent mistakes.

    But, in their innocence, these mistakes reveal confused world-views, and I think it's worth correcting these mistakes (by bringing them face-to-face with data) and acknowledging the corrections. It's much easier to move forward, I think, when you acknowledge a mistake. As note above, I think it would be fine for Kuang to credit you, or me, or anyone else. The main point is to own the error (which, for reasons above, I doubt is merely a simple misreading of a map).

    Meng:

    Thanks for the link to the other maps. Also, in answer to your question, Yair and I are still working on our paper; I hope we will finish it soon.

  13. At least for the New York map, I didn't find the color scheme confusing at all. As a New Yorker, I know very well which racial group lives where in the city. So when I saw blue dots in neighborhoods I knew to be African-American, my brain automatically coded "Black", same for the other racial groups. It think its more important that the colors be visually distinct from each other.

    And anyway, in the US you will usually see the color associated with Whites more than the other colors. You can figure it out from where. Genuinelly multi-racial areas like New York are pretty rare in the US, they are basically limited to the big four metro areas of New York, Los Angeles, Chicago, and the Bay Area. Biracial metro areas are more common.

    I was a bit confused at first by the big green dot in lower Manhattan, admidst all that red, then realized I had completely forgotten about Chinatown.

  14. According to the graphic, the Marina (and Castro) district of San Francisco do not appear to be any more white than the Richmond or Sunset are Asian, or the Mission and Bernal Heights are Hispanic. Yes, Kuang changed his wording, but the message remains the same.

    As for the use of color, the intent of Rankin when creating these graphics is to depict varying degrees of difference and transition in urban areas; so it is more important to choose colors that are distinct from each other. If Rankin wanted to focus on race more specifically, then an informed likeness of skin color might make more sense.

  15. Does the NYC graph really show how segregated NYC is? Especially in the more densely populated areas of NYC, which is the most densely populated city in the USA, it's clear what the dominant group is. But there seems to be a problem with overplotting.

    Take the upper west side of Manhattan. On the map, it looks like it's all White. This is not reality. Certainly a large majority of residents in this neighborhood are White, but it's not close to the domination shown here. In particular, a stroll along Amsterdam Avenue will reveal a LOT of non-White people.

    I think they need to adjust the dot size, and probably the number of people per dot, to reflect the density of the population.

  16. Peter:

    Good point. Usually I don't have much use for so-called "cartograms"–those maps that distort areas in proportion to population or some other variable–but in this case maybe they'd be a good idea.

    The other problem is that everyone "knows" that Manhattan is segregated, and people see what they're expecting to see. Recall the original, ludicrously-wrong, claim that San Francisco is "very, very white."

Comments are closed.