The Visual Display of Quantitative Information

by Henry on November 7, 2004

As a complement to Kieran’s post, Michael Gaster, Cosma Shalizi’s and Mark Newman’s electoral map where area is proportional to population is fascinating, as well as weirdly beautiful – like butterflies exploding. Gaster-Shalizi-Newman also have a really interesting analysis of the distribution of votes for for the Republican candidate – go read it now.

Update: according to Cosma, the histogram on his site showing 300 odd counties with 99% Democratic support was the result of a coding error – however the map is accurate.



junius ponds 11.07.04 at 1:19 am

“The number of counties in which more than 99% of voters voted Democrat was 307.”

99% for Kerry? That’s astonishing. What counties would these be?

And check out Appalachia:


eudoxis 11.07.04 at 1:35 am

Nearly 400 counties voted almost 100% Democrat? Which counties are these?


prometheus 11.07.04 at 2:29 am

cf. Geomblog


Tom 11.07.04 at 3:14 am

What about Appalachia? I see a wide range of colors there.



junius ponds 11.07.04 at 4:24 am

>What about Appalachia?< Perhaps I'm geographically confused, but there's a red swath running from northern Alabama to central Pennsylvania. WV is mottled, but it only recently switched camps.


aaa 11.07.04 at 4:41 am


Barry Freed 11.07.04 at 5:47 am

Yeah, I can dig it. If I just hang out in my room. And turn off all the lights but the black light. And drop a couple of hits of purple microdot. Things don’t look so bad after all.


lago 11.07.04 at 6:44 am

I think Goffman called these sorts of maps “secondary adjustments.”


Keith M Ellis 11.07.04 at 9:25 am

I, too, am having a little trouble with there being many ~100% blue counties and few ~100% red counties. Maybe it’s because I’m in the southwest, but I would have guessed the reverse. (I’d have thought that there’d be quite a few low-population, rural heavily-red counties, especially in the west; while the heavily-blue urban counties would always have a minority of conservatives.)

As I ponder this, one thought that occurs to me is that because urban areas are highly populated, are their counties correspondingly geographically small and thus more likely to represent clustering of very like-minded individuals? I dunno.


Danny 11.07.04 at 11:11 am

The researchers must have some messed-up data. There cannot possibly be 400 counties voting 99% Kerry. DC voted 90% Kerry. Manhattan 82%. SF 83%. I defy anyone to find me one county Kerry won 99% of the vote.


mitch p. 11.07.04 at 11:31 am

The commentary I’ve seen on the election results so far has all been about how the ‘religious right’ won the election for Bush. But look at the blogosphere. Prominent pro-Bush blogs there would include InstaPundit, LGF, Belmont Club. The first two are quite secular, and Belmont Club appears to have a stronger affinity for Tolkien than for King James. So what’s the story here? Is there a secular pro-Bush tendency that’s being underreported?


mona 11.07.04 at 12:18 pm

Yes, Mitch, you’re right but I don’t think it’s that secular – the war on terror has spawned a religion all of its own.


Tom 11.07.04 at 2:11 pm


It is hard to tell without the state lines, but I think that red swath includes a lot of the Piedmont region, and that there’s a lot of purple and blue in the uplands, as well as red.


Rob 11.07.04 at 2:14 pm


County size in the US vaies greatly. Not with population as such but with when the area was first settled. So in general as you go further west, the size of counties increases.


Mrs Tilton 11.07.04 at 2:20 pm

Hmmm. It looks like China on acid.


Brett Bellmore 11.07.04 at 2:21 pm

I’ve asked to see his data set, so I can do some analysis of my own.

I suspect it may be some kind of tipping point phenomena; That when the percentage of “tolerant” Democrats gets over some critical value, they make life for the minority of Republicans so hellish that they all flee. Manhattan and D.C. are never going to go 99% Democratic, because no matter how bad it gets for Republicans, they’ve got strong reasons to remain there.

Alternatively, I suppose there could be 400 or so counties out there where Democratic ballot fraud is just erasing all the Republican votes. ;)

Anyway, I do believe that graph, barring some evidence that there’s a mistake. After all, how else to explain the Democratic Underground?


nnyhav 11.07.04 at 2:28 pm

“Clever algorithm” indeed. Try a Russia overlay.


Tom 11.07.04 at 2:29 pm

Junius — I looked at some results from eastern Kentucky. Harlan County was 60% Bush, and Pike was %52% Kerry. Both counties are more than 95% white, and white people nationally went 58% Bush, so Harlan is pretty typical of white people, while Pike is much more democratic than white people natinally.


Giles 11.07.04 at 3:35 pm

Looking at the map, you’d have to think that the Democrats have some serious diversity issues to deal with. their heartlands are so overwellmingly dmocrat that its probably likely that many of their members have never even met a republican. They’re therefore going to have more than a little difficulty prususading them to vote any other way.


andrew cooke 11.07.04 at 4:23 pm

does anyone have a histogram that is coloured with the same scheme as the map? it would help tie up whether the two are consistent (for example, it’s difficult to understand what the solid red areas on the map are – presumably the colour scheme is red of 80% or higher republicans? a shaded histogram would show this).


Zizka 11.07.04 at 4:25 pm

I live in a Blue county (Multnomah, Oregon). Hard as I try, I meet LOTS of Republicans, but I regret it.

Microanalysis of country results can be a can of worms. My home town was 50/50, the two counties it’s in were 60/40 Bush. Town population is 1400, so it asn’t really an urban/rural split. No idea why. A group of very similiar counties 50 miles South all voted for Kerry. No idea why there either.


Katherine 11.07.04 at 4:38 pm

“The commentary I’ve seen on the election results so far has all been about how the ‘religious right’ won the election for Bush. But look at the blogosphere. Prominent pro-Bush blogs there would include InstaPundit, LGF, Belmont Club. The first two are quite secular, and Belmont Club appears to have a stronger affinity for Tolkien than for King James. So what’s the story here? Is there a secular pro-Bush tendency that’s being underreported? “

The blogosphere is grossly unrepresentative, is what’s going on.

I mean, if you took a poll on gay marriage in the blogopshere, it would be approved overwhelmingly. That wouldn’t even be representative of Massachusetts’ attitudes on it, let alone the country.


andrew cooke 11.07.04 at 4:41 pm

and how about making the y axis of that histogram number of votes rather than number of counties?


Katherine 11.07.04 at 4:47 pm

You know what might be skewing the “county” results? All six New England states–New Hampshire, Vermont, Rhode Island, Maine, Massachusetts, Connecticut–have what we call “wall to wall” incorporation. Everyone is in a village, town/township, or city, so the county provides fewer services and is a less important unit of government than elsewhere. And they report their election results by TOWN, not by COUNTY.

99% is still weird though. I’d be 90% easily enough, but 99? Even Cambridge & Somerville Mass. have 15% or so Republicans.

Maybe intense racial segregation in certain parts of the country–either central cities or counties in the south–is a better explanation. And give me a break with the “liberals make life a living hell so everyone else must flee” explanation.


Jack 11.07.04 at 4:52 pm

Do the 99% Democrat counties have anything to do with gerrymandering/boudary reform?


Angry Moderate 11.07.04 at 5:00 pm

Danny is right. Their county-level data must be screwed up. A quick look at Dave Leip’s site shows ZERO counties with Kerry over 90% much less 99%. Kerry won only ten counties with over 80% (San Fran, Bronx, NY, Phil, Prince Georges, DC, Macon Al, Claiborne MS, Jefferson MS, Shannon SD). Bush won 15 counties with over 80% of the vote in Utah alone. This, of course, doesn’t say anything about voter concentration, just that Bush dominates in rural areas with small counties where by random variation you get more lopsided results. If you divided up the Bronx or San Francisco into 500 vote segments, you’d get plenty of Kerry over 90%.


Katherine 11.07.04 at 5:15 pm

I was about to make the same point. I think the data is just wrong. I would certainly buy that there are more 90-10 or 80-20 Democratic counties than Republican counties, and that would explain the preponderance of red & also contribute to the explanation of why we do worse in the House than the Senate when it seems like the opposite should be true. (Redistricting is also part of the explanation, obviously.)But the data just looks wrong.


Keith M Ellis 11.07.04 at 6:08 pm

It does seem like something is wrong. This is being discussed over at MetaFilter (by way of someone seeing it here), and there’s skepticism there, too. Here’s what I wrote there (since it’s long and it would stupid to quote myself).


Keith M Ellis 11.07.04 at 6:12 pm

Also: I question their bar chart and the conclusions they draw related to it. But the cartogram seems fine, doesn’t it?


dsquared 11.07.04 at 6:46 pm

This would look so much better if it was done in Excel :-)


eudoxis 11.07.04 at 7:02 pm

The shape of the cartogram is fine, but the colors, if they are using the same source data is are wrong. Besides, it’s a small, small world. Everybody knows a republican. The bar chart and the explanation (decided and divided) must be a joke.


harry 11.07.04 at 7:31 pm

Surely Katherine is right and we’re talking about overwhelmingly black counties.


Cosma 11.07.04 at 7:43 pm

Since the 307 county bit met with such skepticism, we rechecked it. Unfortunately, it was due to a bug in one of our programs. (It got confused by county names with spaces in them.) I’ve updated my post to reflect this.

The maps were produced by a separate program, not subject to that bug, and those results are right (so far as we can tell). Anyone who wants the data, please get in touch.


fleacircus 11.07.04 at 8:08 pm

Links are down, but people are trying to draw the conclusion that it’s the Democrats who are insular and xenophobic?

People from the hinterlands lecturing the city dwellers about cultural and political diversity?


No matter how many votes you get or how much data you skew, up is not down.


fleacircus 11.07.04 at 8:10 pm

Links are down, but people are trying to draw the conclusion that it’s the Democrats who are insular and xenophobic?

People from the hinterlands lecturing the city dwellers about cultural and political diversity?


No matter how many votes you get or how much data you skew, up is not down.


Felix 11.07.04 at 11:18 pm

my guess at the moment is his program turned ‘no votes’ (0/0) into ‘100% for Kerry’, and he fetched the data when lots of counties hadn’t reported in yet.

I just fetched and parsed the data myself from USA Today. there are 3 counties/districts more than 90% for Kerry, 1 with no report yet, 5 reporting no votes (small townships in Maine), and 10 more than 90% for Bush.


Ethesis 11.07.04 at 11:35 pm

Pretty work, though a Blue-Yellow or Red-Green color pattern would give a better visual result in terms of accuracy (red-blue creates artificial boundry impressions).

Still, Red-Blue is very pretty.


katherine 11.08.04 at 2:48 am

is there really so much more red than blue? I think there may be an optical illusion because the red (almost) all connects up and the blue doesn’t.


John Faughnan 11.08.04 at 3:09 am

How about a representation that adjusts for wealth? I’d be curious to know the relative percent of GNP produced by red vs. blue states. Handy in case states’ rights catches on.


JD 11.08.04 at 4:31 pm

Since we are actually a Representative Republic using the Electoral College as our method for determining the Executive, shouldn’t we use congressional and senatorial districts as our allocation method, not states or counties? Actually I suppose the top map is the best since most states use the “winner-take-all” method for counting electors. But the districting method would be interesting to see.


Brett Bellmore 11.09.04 at 3:04 am

I would assume, using the composition of the House and Senate as a proxy, that distributing electoral college votes in that manner, (Not a bad idea, IMO, so long as it were adopted by amendment to avoid coordination problems.) Bush would still have won, perhaps by a somewhat larger margin. Winning by huge margins in urban areas may be impressive as heck, but it seriously hurts your representation in the House.


anon 11.09.04 at 6:15 pm

Here’s a list of states listed with the percentage of graduate degree holders, color coded to show which candidate they voted for in the 2004 elections.

The list is from:
Election results were from:

1. Washington, D.C.: 23.6 percent (Kerry)
2. Massachusetts: 14.5 percent (Kerry)
3. Maryland: 14.1 percent (Kerry)
4. Connecticut: 13.7 percent (Kerry)
5. Virginia: 12.9 percent (Bush)
6. New York: 12.6 percent (Kerry)
7. Vermont: 12.3 percent
8. Colorado: 11.5 percent (Bush)
8. New Jersey: 11.5 percent (Kerry)
10. New Mexico: 11.0 percent (Bush)

11. Illinois: 10.6 percent (Kerry)
12. New Hampshire: 10.5 percent (Kerry)
13. Washington: 10.1 percent (Kerry)
13. California: 10.1 percent (Kerry)
15. Rhode Island: 9.9 percent (Kerry)
16. Alaska: 9.8 percent (Bush)
17. Oregon: 9.5 percent (Kerry)
18. Minnesota: 9.3 percent (Kerry)
19. Delaware: 9.2 percent (Kerry)
20. Kansas: 8.9 percent (Bush)
20: Missouri: 8.9 percent (Bush)
20. Pennsylvania: 8.9 percent (Kerry)
23. Hawaii: 8.8 percent (Kerry)
24. Florida: 8.5 percent (Bush)
25. Michigan: 8.3 percent (Kerry)
26. Arizona: 8.2 percent (Bush)
27. Texas: 8.0 percent (Bush)
28. Georgia: 7.9 percent (Bush)
28. Utah: 7.9 percent (Bush)
30. Maine: 7.8 percent (Kerry)
31. North Carolina: 7.7 percent (Bush)
31. South Carolina: 7.7 percent (Bush)
31. Indiana: 7.7 percent (Bush)
34. Ohio: 7.6 percent (Bush)
35. Montana: 7.5 percent (Bush)
35. Kentucky: 7.5 percent (Bush)
35. Alabama: 7.5 percent (Bush)
38. Nebraska: 7.4 percent (Bush)
39. Wisconsin: 7.2 percent (Kerry)
39. Wyoming: 7.2 percent (Bush)
41: Idaho: 7.1 percent (Bush)
42. Tennessee: 7.1 percent (Bush)
43. Louisiana: 6.7 percent (Bush)
43. North Dakota: 6.7 percent (Bush)
45. West Virginia: 6.6 percent (Bush)
46. Oklahoma: 6.5 percent (Bush)
46. Arkansas: 6.5 percent (Bush)
48. Iowa: 6.3 percent (Bush)
49. South Dakota: 6.1 percent (Bush)
50. Nevada: 5.9 percent (Bush)
51: Mississippi: 5.8 percent (Bush)


lth 11.10.04 at 10:47 am

Here’s a bright idea… why doesn’t every person in the US have one vote with which they can vote for a single presidential candidate? Thus neatly eliminating the electoral college system, and giving everyone equal say. This has the handy side effect of the person winning most of the votes being the person who actually gets elected…


Anthony Robinson 11.11.04 at 1:24 pm

If you’d like to look at election data against a wide range of other socioeconomic and even environmental data, check out the tools and data I’ve made available here:

No cartograms at the moment, but we do provide a fully interactive visualization environment that you won’t see elsewhere. Also, I have a link to the database I created, so you can check it out and use it yourself.

Comments on this entry are closed.