Yesterday, I took a look at Irregular Times’ sales records for the past four months to see which of the Democratic presidential contenders were garnering the biggest share of items sold. Every bumper sticker, campaign pin, fridge magnet, t-shirt, poster, or other item supporting a presidential candidate becomes a public expression of committed support for that candidate — and so measuring the distribution of committed support for different candidates across the country is one way of figuring out how the Democratic primary race just might turn out next year. In a whopping 45 out of 50 states, pro-Obama gear accounted for the biggest share of sales.
Today, I’d like to answer a different question, one that doesn’t pit contenders against one another but instead looks at the pattern of sales for each candidate separately. From where in the country does a candidate gain disproportionate support, and where in the country is a candidate getting less than proportionate support?
I’m thinking about proportions literally here, in the form of fractions with numerators and denominators. Imagine a fraction that takes this form:
(Percent of All Gear Supporting Candidate X Sold in State Y) / (Percent of U.S. Population Living in State Y)
The denominator here represents our expectation for the value of the numerator, assuming that Candidate X is supported equally well across the United States. For example, U.S. Census figures released last month tell us that 2.06% of Americans live in the state of Arizona. If support for Al Gore is neither stronger nor weaker in Arizona than it is anywhere else in the United States, then the percentage of items supporting Gore that are bought from someone living in Arizona should come out to around 2.06%. By chance alone, we might expect the actual number to be a bit different — say, 1.9% or 2.1% — but it should be close to 2.06%, especially as more items are sold. If the actual percentage share for Gore items came out to 2.06%, then our fraction would come out to (2.06% / 2.06%), or 1. The value of 1 in this fraction indicates that our expectation matches reality.
But if the actual percentage of pro-Gore items sold to Arizonans is especially low or especially high compared to our expectation of 2.06%, then we have some reason to suspect that the level of support for Gore in Arizona is disproportionately low or disproportionately high. If the actual percentage of pro-Gore items sold to Arizonans is 1.03%, for instance, then that’s half of what we would have expected by chance alone. In this case, our fraction comes out as (1.03% / 2.06%), or 0.5. If, on the other hand, the actual percentage of pro-Gore items sold to Arizonans is 4.12%, then that’s twice what we would have expected by chance alone. The value of 2 for our fraction (4.12% / 2.06%) reflects this.
Fortunately for my curiosity, we’ve sold a number of thousands of Election 2008 items in the past four months, making a large enough number of data points to address the question of the regional distribution of committed support with some degree of confidence. Of course, we have more data for those contenders who are most popular — Hillary Clinton, John Edwards, Al Gore, Barack Obama, and Bill Richardson. Apologies to Joseph Biden, Wesley Clark, Christopher Dodd, Mike Gravel, Dennis Kucinich, Bill Moyers, Al Sharpton, and Tom Vilsack — but not enough people bought items supporting you for us to spread the data out over 50 states and still have results that make sense.
Without further ado, let’s look at the results for the period of 9/19/06 to 1/23/07 for contenders Clinton, Edwards, Gore, Obama and Richardson. Maps showing the values of our fraction for all fifty states are below:
There are some similarities across all maps — Mississippians, Montanans, Texans, and West Virginians are universally low in the level of their support for any Democratic candidate. This isn’t surprising, because those states are pretty darned authoritarian and conservative (the Democrats’ Senate representation in West Virginia is based more on the delivery of pork than on support for progressive policy). But otherwise, there is a surprising amount of variation in the concentration of support for different Democratic contenders.
Sure, Hillary Clinton gains disproportionate support from the Northeastern states, but she also is the only candidate (with the lesser exception of John Edwards) to grab disproportionate support from a wide swath of America’s midsection. Clinton is saying something to the people of the Great Plains. I don’t know what that is, but it’s not an anomaly of the past few months — we’ve seen this geographic pattern for Clinton before.
There are other patterns, too: John Edwards is relateively strong in the South (as well as the old industrial Midwest), and reports of Edwards’ relative strength in Iowa are corroborated here, too. Al Gore and Barack Obama have a proportionately low portion of their sales coming from the South, and show the classic Democratic pattern of especially high strength on the Pacific coast and in the Northeast. Bill Richardson’s supporters come disproportionately from the Southwest, and other than strong showings in the DC area and Iowa (where he has been building buzz and courting support), he shows a proportionately low level of support in the rest of the nation.
What do you see here, and what might it mean for the 2008 primaries?