About the Politics

The Votes

“Just the facts, Ma'am.”

- Sergeant Joe Friday, Dragnet

The voting data, as compiled from the 2000 and 2004 presidential votes and 2000 census data, suggests that the popular media has it pretty much wrong. To get it right will require a bit of careful thinking outside the box. The Red and Blue narrative is both true and false and we need to recognize the truth in order to dispense with the myths and fallacies.

The most familiar piece of evidence that captures the 2000 and 2004 elections is the infamous Red-Blue map of the states colored according to the Electoral College results.

If these aren't burned into your brain, here they are again:

Since EC results are winner-take-all according to whichever candidate wins a plurality (the most votes), the states were colored either all red or all blue. It shows a stark regional divide between the Red heartland and the South versus the Blue coasts and Great Lakes region.

One interesting point to make here is that regional divides in US politics are not new. They have recurred frequently in our history. [A Political Puzzle? Video link - 1896 clip]

A great example is illustrated by this map of Electoral College results from the landmark election in 1896 between the Democrat William Jennings Bryan and the Republican William McKinley (McKinley won).

By now this should look familiar: the coasts and Great Lakes regions versus the South and Midwestern heartland. Does this mean our current divisions go back to 1896? Not exactly.

But the divisions that existed then are instructive to understanding what divides us now. (Note: the colors of the parties are reversed for comparison purposes: McKinley was actually blue and Bryan red. Isn't that interesting. Partisan affiliations have flipped geographically so the former Democratic strongholds are now Republican and the former Republican strongholds are now Democrat. Have the voters changed or the parties?)

Looking at these Electoral College maps we should understand that this is not how the popular vote was cast. (Pros and cons of the Electoral College later.) The problem is that these state maps are very deceiving in how they depict the results and thus may point us to false explanations like race, wealth, education, religion, just because these factors may differ across regions. We need better information.

Here, for example, is the 2004 election with states colored according to vote shares:

Looks kind of purple, no? Some people have used these kinds of illustrations to prove that we're really a “Purple America,” but this is also misleading because it homogenizes the voting results (do we really believe all Californians are the same shade of purple?). This obscures the truly odd and persistent patterns across the country. (In academic jargon we would say we're looking at the wrong level of analysis.)

A better representation would give us more detailed voting information by congressional district or county. Here are the two election maps colored red or blue according to 3142 separate counties:

This shows us something more interesting. First, the county maps, like the state maps, are highly similar, suggesting something significant. It also shows that densely populated areas (cities!) are voting one way while everyone else appears to be voting another. Now we're getting closer to an accurate depiction, but this is still a misrepresentation of the results because each county is actually divided according to vote shares, not winner-take-all. In other words, counties should be depicted in shades of red and blue shades of purple, not pure red and blue.

So, finally we have these color-blended maps which portray party vote shares by county:

Purplish, but with deep shades of red and blue in specific areas. This gives us a much better picture of how people are actually voting. This reinforces our impression that cities, suburbs and rural areas are voting differently, reflecting different political preferences that have carried over from election to election.

If we show the results in three dimensions to reflect voter or population density we get this map for 2004:

Blues are highly concentrated in cities while reds are dispersed across more sparsely populated areas

Introducing County Census Data

Tabulating voting data by county rather than congressional district has the advantage of comparing voting results to the census profile of the county. Since there are 3142 counties in the US, this gives us a large population of cases from which to draw statistical inferences.

Without going into more detail than necessary, when we run statistical analyses of the 2000 and 2004 elections and compare these to data from the 2000 census, we learn some interesting things.

Summary of Data Analysis:

• The true voting divide is mostly determined by two county characteristics: population density and proportions of married households.


• Densely populated counties lean Blue, while less populated rural and suburban counties tilt Red.


• Counties with high proportions of married households trend Red, while counties with high proportions of non-married households mostly choose Blue.


• Other county census variables such as race, ethnicity, age, sex, education, and income levels tend to correlate with voting patterns, but are not as determinant.


• Values issues as measured by opinion polls identify several principle issues: homosexuality, abortion, gun control, religion and family values. Opinions on these issues are highly correlated with the locational factors of population density and household composition.

Here is a table that gives a breakdown of these variables across counties:

Table 1 - Comparisons of Average Vote Share by County

Votes	Pop/sq.mi.	% of Married Households	% of Female Head of Household	% of White Inhabitants	% of Black Inhabitants
Gore	739	50.7%	13.6%	73.8%	17.6%
Bush2000	108	57.4%	9.6%	87.7%	6.4%
Kerry	836	49.9%	13.5%	73.0%	17.5%
Bush2004	110	57.4%	9.8%	87.4%	6.8%

We can zoom in on this population density and vote preference relationship by focusing only on those counties that comprise the 50 largest metropolitan areas.

Table 2 - 417 Metro Counties (All figures expressed as Average Vote Share)

MetroCode	Pop/sq.mi.	%Married Households	Bush2000%	Gore%	Bush2004%	Kerry%
Core	14429	34.7%	24.4%	72.4%	25.0%	74.6%
Inner suburb	2514	46.0%	41.7%	55.8%	43.1%	56.5%
Mature suburb	1432	52.3%	46.9%	50.3%	49.9%	50.8%
Emerging suburb	423	60.3%	57.4%	40.4%	60.2%	39.5%
Exurb	89	60.0%	57.6%	40.8%	62.1%	37.7%

As we move from the urban core outward to the suburbs, into the outer exurbs and finally into the rural areas, voters shift from strong Democratic support to strong Republican support. These data are remarkably consistent and, in statistical terms, highly significant, giving us greater confidence that we're seeing a more accurate representation of the reality of voting patterns over the past two elections.

So, why is this?

JOIN THE RSBS Mailing List TODAY!

RSBS Productions ©2008 | All Rights Reserved |site by>> MACLEAN DESIGN