海角大神

A few weeks ago on Meet the Press, Chuck Todd proposed that the fate of the Senate could come down to the distribution of Starbucks coffee shops (an elitist, urban chain) vs. Chick-fil-A restaurants (propagated by rural, social conservatives). He predicted 鈥渋t could be advantage to the chicken鈥� this November. Click here for a video of Todd鈥檚 segment.

David Wasserman of FiveThirtyEight聽went one step further, taking the time to actually map Chick-fil-A and Starbucks locations. Because, you know, data! He then compared the distribution of both restaurants and compared them with Obama鈥檚 vote share in 2008. Here鈥檚 what Wasserman found:

Wasserman proceeds to remind voters of an earlier post of his (see here) showing that, unlike Chick-fil-A and Starbucks locations, the number of聽of Whole Foods stores and Cracker Barrel restaurants does correlate highly with Obama鈥檚 vote share. 聽He notes:

Now, yes,聽food聽can be political. Look no further than pizza, where we can easily find not one, but two examples of this. BrandIndex (a company that tracks the reputations of food chains) reports that perceptions聽of Godfater鈥檚 Pizza soared with Republicans and dropped with Democrats after their former CEO 鈥� the one and only Herman Cain 鈥� decided to run for president. We saw the same聽pattern聽with pizza chain Papa John鈥檚, after it鈥檚 CEO, John Schnatter, promised to reduce workers hours and increase pizza prices after Obama鈥檚 2012 victory.

But while food can be political, I鈥檓 not satisfied with either Chuck Todd or David Wasserman鈥檚 dichotomy. Let鈥檚 have a food fight!

What follows is a summary of a 2011 post of mine on this very issue. In that post, I proposed a better way to correlate peoples鈥� culinary habits with their voting behavior. Rather than examine the volume of restaurants in a geographic聽area, let鈥檚 look at Google search traffic for these eateries.

I see three advantages with this approach. First, it鈥檚 easy! Just play with Google Trends for a few minutes and you鈥檒l see why. Second, the number of restaurants in a geographic area is pretty far removed from the unit of analysis (people). In contrast, voting behavior and聽search traffic both derive from individual level behaviors. And third, the geographic location of a restaurant (Cracker Barrel) vs. a large grocery store (Whole Foods) has more to do with geographic dispersion and population density than anything else. It鈥檚 not surprising that Cracker Barrels are more numerous in less populated, rural settings: They cost less to build and operate! 聽In contrast, you wouldn鈥檛 put a 35,000-sq.-ft. Whole Foods in a small town (irrespective of whether people in that town are staunch social conservatives or big government lefties).

In this old post, I proposed two more reliably partisan food chains:聽Chick-fil-A聽(like Chuck Todd) and Ben and Jerry鈥檚? 聽Chick-fil-A, after all, gave Tea Party protestors free iced tea on April 15, has donated nearly $2 million to anti-gay rights groups, and is closed on Sundays so their employees can 鈥渞est, spend time with family and friends, and worship if they choose to do so.鈥� Ben and Jerry, by contrast, are two self-described 鈥渓iberal lefties鈥� who fed ice cream to聽600 Occupy Wall Street protestors and adopt (gasp) fair trade practices. In short, these two chains exhibit clear ideological differences (rather than simply聽correlate聽with it).

In order to examine this issue and pit the two typologies against each other on a comparable scale, I coded Google Trends data for the states with the top 10 search results for each of the four chains. I then merged that with Obama鈥檚 share of the two-party vote.聽 Here are the results.

It appears that my typology with Chick-fil-A聽vs.聽Ben and Jerry鈥檚 stacks up favorably 鈥� if not better 鈥� than the Cracker Barrel vs. Whole Foods divide.聽 By my count, Obama received 64% of the two-party vote in Top 10 states for Ben and Jerry鈥檚 searches and only 62% in the Top 10 states for Whole Foods searches. Clearly this difference isn鈥檛 statistically significant, but it鈥檚 a marginal improvement. In聽contrast, Obama only received 46% of the two-party vote in top 10 states for Chick-fil-A searches and 50% in top 10 states for Cracker Barrel searches.

Of course, there are perfectly good explanations for these effects.聽 And to his credit, in his old post Wasserman notes one study, Bill Bishop鈥檚 鈥淭he Big Sort,"聽which argues that Americans are geographically sorting into homogeneous clusters (and the factors which drive this clustering correlate with political ideology).

Jordan Ragusa publishes his Rule 22 blog at http://rule22.wordpress.com/.