How population density affected the 2012 presidential election

There are lots of reasons why the 2012 presidential election broke the way it did, but one that’s not often reported—but particularly germane to Per Square Mile—is the divide between cities and the country. I’ve been thinking for a while now about this split as a driving force behind the polarization of U.S. politics, and I know I’m not alone. (On election night, Adam Rogers tweeted as much.)

But I was curious. Can we actually see the divide between cities and the country in the electoral map? In short, yes, but I’ll let the maps to the rest of the talking.

Swipe back and forth to see how population density relates to each candidates’ electoral result.

Thanks to Andy Woodruff of the always interesting Bostonography for the shapefile of the election results.

Related posts:

Income inequality, as seen from space

Redrawing the United States of America

Ghosts of geography

Thanks to the support of readers like you, Per Square Mile remains independent and ad-free.

If you enjoy what you just read, please consider supporting the site with a monthly recurring donation as a Sustaining Member.

You can also support Per Square Mile with a one-time donation.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

  1. How can you do a population density heatmap based on colour without providing a key? This just looks neat, but doesn’t convey any real information or data. Cool idea, but the execution is just incomplete…..

    1. actually the key is not needed.

      the problem is the density map isn’t the same on each.

      it needs to be the same for the maps to be useful. as it is they might as well be for different elections

      1. The density map is the same (underlying data is identical), it’s just that the colors are different (blue vs. red). Same breaks, same number of shades, just different colors.

  2. Very interesting. It would also be interesting to see a scatter plot of population density vs percent votes, so one can see how close the correlation is, and how the curve is shaped.

    1. I second this. Without a nice scatter plot and regression line, its really hard to see whether there is actually a relationship here. You have all the county level data, so this shouldn’t be very hard.

  3. Also, the “county” level of granularity, used in the maps above, still isn’t quite enough to do justice to the correlation between density and political preference, as many large counties include liberal cities as well as conservative suburbs and unincorporated areas. However, “precinct” would be too granular. I don’t know what would be an ideal granularity: maybe the US Census concept of “urban area”?

    1. I think “city” would be the right level to bring out the correlation, though the large size of the data set might get unwieldy.

  4. Clearly this isn’t absolute population density, right? Otherwise the density maps for Romney and Obama should be identical. The fact that “density” of Obama voters in northern Minnesota and NYC seem the same makes me wonder what you mean by “density” here.

    1. It appears to me that the density maps for each are identical. The only difference being red for Romney, Blue for Obama.

    2. I think you are confused, as I was initially and as other commenters above were, because the captions on the maps are the reverse of what I would have expected. The captions look as if the density map is on the left and the vote map is on the right. Actually you need to SWIPE left to see the density map and right to see the vote map.

      1. (It appears that the maps have been switched since I wrote that last reply. Anyway, it’s easy to figure out which is which.)

  5. Can we please overlay the county-by-county 2012 electoral map overlayed on a 3d graph of county population density in the US, with the population density on the Z-axis… so it looks like a topographical map opf population density with the 2D-ness of the electoral map overlayed? would much be much appreciated.. please contact (journalist working on story)

  6. I saw an interesting formulation the other day that hadn’t previously occurred to me: places with more sidewalk tend to vote more liberal.

  7. I don’t find this display illuminating at all. I live in Tennessee, and know full well the salience of the urban-nonurban divide. But here it’s Nashville and Memphis that stand out like sore thumbs. The suburbs around Nashville show up on this map as equally dense [The Nashville MSA is a poster child for urban sprawl] but they’re intensely Republican. The same can be said for lots of other counties that show up as “dense” on this rendering. Moreover, in the South the striking divergence is between Greater Appalachia and the Black Belt–but that bears little relationship to density; in fact, there are probably a lot more “dense” counties in the Republican Upland South than in heavily black old plantation areas.

  8. I’ve always believed in a similar but not identical indicator.

    I’d bet that voting behavior is strongly correlated to whether one’s house is hooked up to city sewer or not. If you flush the toilet, and it goes to municipal sewage treatment, you vote Democratic. If you literally have to take care of your own shit, you vote Republican.

  9. Great information. It might be interesting to see another map of broadband internet connections overlaid with election results. I would assume broadband connectivity would correlate with population density. If broadband connectivity does correlate with population density then we might assume information is giving the Democrats a boost.

  10. There is an interesting cluster of rural counties that went for Obama in NE Iowa, SE Minn, SW Wisconsin and NW Illinois. It is notable for being one of the few places outside of New England or New Mexico where a sizeable, contiguous group of rural counties went blue. I would be interested in your thoughts as to what might explain this. The area is geographically distinctive (“driftless,” non-glaciated) and has some agricultural similarities to places like Vermont.

    1. The other rural places that went for Obama were in North Dakota and various places west, from Wyoming stretching into New Mexico. I suspect these areas are mainly Native American reservations, but I do not know for sure. Is that so?

      1. I’d venture that much of the Obama vote in the anomalous Eastern portions of the Dakotas is actually Scandinavians who flowed Westward from their initial settlements in Minnesota and Northern Wisconsin.

      2. In South Dakota, the Native American reservations and Clay county tend to vote Democrat. Clay county is home to the University of South Dakota, and contains most of SD’s democrats. In Minnesota, the Iron Range is a long-standing union hub and votes Democrat, as does the Twin Cities. Minnesota was considered a swing state by the end, but in that, republicans were ignoring the fact that while most of the state geographically leans Republican, a majority of the population leans Democrat, and carried the state for Obama.

    2. Yankees, ’48er Germans, and Nordics.

      Some of it is, literally, transplanted Vermonters (the cotton gin put Vermont’s wool-based economy out of business, and sent these original republicans flowing west.

      In a larger sense, I think what we’re seeing in these voting patterns is a further iteration of Kevin Phillips ‘the Cousins’ wars,’ reflecting the fact that much of the Western rural settlement post-Civil war was white southern.

  11. What would be most interesting is not the regression itself, but a map of the residuals: in other words, which counties deviated from the predicted value based on density? Certainly Reservations would show up, as well as the heavily Latino counties near the border; on the flip side, military bases and Salt Lake (just guessing).

  12. I’ve studied the actual results by county on this against 2010 census data. The general rule seems to be this:

    In places with population density greater than 88 ppl/sq. mi. (the national average), there is a 66% chance that they went for Obama.

    In places with population density less than 88 ppl/sq. mi., there is a 66% chance that they went for Romney.

    All of the Top 25 most densely populated counties went for Obama. Of the top 50, 96% went for Obama; the only two in the top 50 that didn’t were Orange County (naturally) and Colonial Heights, Virginia.

    There is more of a correlation than these maps let on and I think there are likely better ways to visualize it, which I’ll be working on some this weekend.

  13. Some interesting outliers: rural new England, Northern Minnesota and the Eastern Dakotas, as well as New Mexioc.

    Funny thing about the slider: slide Obama’s vote counts from right to left, and it looks like a time lapse of the (New England/Northeast based stream of Western settlement).

    Which is to suggest that there are significant cultural issue at play.

    1. Interestingly, parts of rural Arizona are overwhelmingly Obama as well, reflecting a high Obama vote among Native Americans and Hispanics, I expect.

  14. Neat. Romneys is roughly the inverse of pop density, Obamas roughly corresponds to pop density.
    As a crude theory, poorer people live more crammed together and are more likely to vote Dem than GOP, maybe?

  15. Cool maps and I see some overlap, but there are huge inter-regional differences where the trends are exactly reversed. There is no doubt that there is something to this, but pundits always oversell this as a major factor. There are plenty of high-density urban/suburban areas that went heavily for Romney (especially in the south) and plenty of rural areas that voted for Obama (upper midwest). An accompanying scatterplot with a good trend would make it more convincing, but I think you will be surprised at how noisy the scatterplot will look.

  16. Any chance you could share the raw data? Or point me to public sources? I tried a similar analysis a few weeks ago, but I’m not sure I had everything sitting quite right.

    Basically, I want to see what it looks like in a scatter plot, and if there are other county-level variables that might strengthen the relationship.

  17. I agree that the higher the density the more for the Democratic Party. Lower income Americans tend to live in the cities, e.g. the higher density environment. Lower income Americans want the government to continue to provide them with government support such as food stamps, etc. I don’t blame them – I would too.

    Another way of looking at it is by looking at private vs public employment. If I was working in a federal job, would I vote for the party that wants to reduce the size of the government or vote for the party that believes in big government. Naturally, the latter.

    As shown recently in the NY Times, all the big population centers of America (except possibly Salt Lake City) went Democratic. For the reasons stated above – this should not be a surprise to anyone in America.

    1. It’s not about “food stamps” or gov’t benefits, for Cthulu’s sake: it’s about opportunity. No one moves to a town for the free stuff, it’s for the chance to make it. The fact that there is a social safety net is due to the density and dynamic nature of cities: some number of people will need help and those people may be transplants or otherwise unconnected to a personal support network. So the community provides, rather than have them begging or reverting to crime.

  18. So this is just further evidence of the Urban Archipelago, first defined 8 years ago? Communities vote Dem, sprawl votes GOP, even though sprawl is more expensive to maintain per capita.

  19. So,… if we want to see a more Democratic governance, we need to support density and all its bits: walkability, transit, bike facilities, parks and plazas. If one wants to see a more Republican governance, one would support sprawl-based development of car supported subdivisions, with their malls, highways, pollution, and privatized public functions.

  20. Nice work!

    I found some related maps, showing the 2012 results plotted on “cartograms”, maps in which the sizes of states (or counties) are rescaled according to their population, here:

    http://www-personal.umich.edu/~mejn/election/2012/

    Might be cool to combine the two: a swipe-able dual cartogram of population density and election results. (Although that would in a way be presenting the density data twice… Ed Tufte might disapprove! :-)

  21. Obama also has great support on the beaches, and on the Canadian and Mexican border. People who have a less narrow sight perhaps?

  22. Regardless of statiscal modeling perfection opinions, these are pretty cool images — does anyone know if a poster is available? Thanks, Mike

  23. all the chatter below about regression analysis makes me chuckle. you dont need a regression analysis. we already know which candidate was more regressive.

  24. Population density is a surrogate piece of information for exposure to people who are different from you (race, sexual orientation). To me this kind of map shows that if you only know white people, straight people, etc., you fear those who are different and you vote Republican. If you live and work around many people who are different than you, you don’t fear the “other” and you vote Democratic.