Understanding the Geography of the 2008 Election

morrillshiftmapinset.png

Scholars as well as pundits and politicians will study this remarkable election exhaustively. Many, including me, will use county data, because they are convenient and available. From a statistical point of view, counties are lousy units, because of huge variation in size and excess internal variability. But we can’t resist, so here are some at least suggestive findings.

First, what correlates with the percent voting for Obama? By far the strongest variables are negative – characteristics associated with voting Republican: a county’s share of husband-wife families (-.64), the rate of home ownership (-.55), percent working in craft occupations (-.52), and religious membership (-.51) all work against Obamamania. Other high negative correlations were with percent rural (-.48), with percent white (-.47), other positive were median rent (.45) and percent foreign born (.45). These are not at all surprising, and are what the exit polls told us.

The highest positive correlations for Obama lay in percentages of non-family households with 2 or more persons (partners, roommates, .50), percent in urbanized areas (.49), or using public transit (.48), and percent with a BA or higher degree (.46). What these figures highlight is the continuing basic polarization between large metropolitan (+ variables) and non-metropolitan (- variables) areas, and simultaneously between the more modern and diverse character of the big city and the more traditional and conservative values of much of non-big city America.

But, you may protest, we thought race, ethnicity and age played a big role in this election? Indeed, they did, but the correct dependent variable should be the degree of change in the share voting Democratic. In other words, what helps distinguish the 2008 from the 2004 results? The largest effect, of course, is simply the quite large 5-6 percent shift in national sentiment because of economic uncertainty and disillusionment with the Republican regime.

But beyond that, the pro-Obama variables tend to be the percent of women in the labor force, percent with a BA degree, median household income (yep, time to toss out the traditional wisdom of Republicans being the party of the ‘rich’), non-family households, professional-managerial occupations, and, yes, percent Hispanic, percent Black and percent aged 25-34. In contrast variables leading to a lesser shift, no shift, or even more Republican, were again church membership, percent rural, percent in crafts jobs, and percent 45-64 or over 65, and percent with less than a 9th grade education.

Overall, education, occupation, age, race and ethnicity help us understand Democratic strength in large metropolitan America and also in rural and small-town American Indian, Black and Hispanic areas, especially in parts of the South and West. But areas and regions with a less educated and professional populace, with higher rates of religious persuasion, with fewer women in the labor force, and with older populations remained loyally Republican. This helps us understand the resistance to Obama and the Democrats in Appalachia and across the border South, from WV, through KY and TN, AR, LA and OK.

An interesting geographic phenomenon should be noted: the emergence of Chicago and the upper Midwest as part of the new Democratic coalition. Metropolitan Chicago provided Obama with a margin of almost 1.5 million votes, more than New York or Los Angeles. This presaged a gigantic increase in Democratic margins throughout the upper Midwest, including IN, IL, MI, WI, IA, and MN. In this one part of the country more than 150 counties moved from the Republican to the Democratic column. In addition to the big shifts on the coasts, this is where Obama gained the most ground.

If this pattern continues, the Democrats may well have achieved a critical mass in their core support, adding a powerful upper Midwest base to their almost total control of both coasts. These would leave the GOP with little more than the heart of the Old Confederacy – even that is threatened in places like North Carolina and Virginia by modernization – as well as more socially conservative regions such as Appalachia and parts of the Great Plains. It’s not a pretty picture if you are a Republican.

Richard Morrill is Professor Emeritus of Geography and Environmental Studies, University of Washington. His research interests include: political geography (voting behavior, redistricting, local governance), population/demography/settlement/migration, urban geography and planning, urban transportation (i.e., old fashioned generalist)

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

"Hyperlocal" election maps

Thanks for this analysis. At the local level, we're also looking at precinct-level data to get a sense of the 2008 results, as well as how that might presage any shift in the electorate for upcoming local races. In New York City, we've prepared several maps and tables along those lines, available at http://www.urbanresearch.org/resources/2008-voting-results

Steve