Tuesday, June 7, 2011

What People Read (or Buy, or something)

Thanks to colleague Karen Russell for pointing this out to me on Facebook, and colleague Karen King for her comments.  In a desperate need of something to blog about, I shamelessly lift this from my FB wall and plop it here.  In it,this New Yorker piece briefly discusses Amazon.com's analysis of cities and book buying.  Here's the list below.

1. Cambridge, Mass.
2. Alexandria, Va.
3. Berkeley, Calif.
4. Ann Arbor, Mich.
5. Boulder, Colo.
6. Miami, Fla.
7. Salt Lake City, Utah
8. Gainesville, Fla.
9. Seattle, Wash.
10. Arlington, Va.
11. Knoxville, Tenn.
12. Orlando, Fla.
13. Pittsburgh, Pa.
14. Washington, D.C.
15. Bellevue, Wash.
16. Columbia, S.C.
17. St. Louis, Mo.
18. Cincinnati, Oh.
19. Portland, Ore.
20. Atlanta, Ga.

The article's author, Macy Malford, offers some interesting explanations for the data -- which can in some instances make perfectly good sense (Cambridge, Mass., Berkeley, Calif.) and at times puzzle to the point of distraction (Miami, Fla., Orlando Fla.).  Keep in mind these are per capital results, thus they control statistically for population size. There are any number of good explanations for why Atlanta makes the Amazon book-buying list, but New York City does not.  Great available bookstores is a good one.  I

Here's where the confusion comes in:
  • When we think of books we automatically think of fiction, perhaps even serious fiction, but this Amazon list is all books.  Boulder, Colo., for example, leads in books in the Cooking, Food, and Wine category.  Alexandria, Virginia, is tops in children's books, therefore pushing it high on the overall purchase list.  For all we know, Atlanta may make the list thanks to Christian fiction or non-fiction books.
  • Measuring per capita makes good sense.  It does control for population size.  We do this in crime statistics too, such as murders per 10,000 people, so apples are compared to apples, oranges to oranges, and murders to murders.  But you can screw the pooch here.  Where do you draw the lines for "Atlanta" makes all the difference in computing crime (and book sales) statistics.  Even the pros who do this for a living make mistakes.
  • This is a direct measure only of buying some type of book (good or crap) from Amazon.com.  Not reading books.  So the good bookstore argument above may confound the results.  The only way to get at "reading" versus "buying on Amazon" is to use survey data that asks specifically how often you read a book.
All in all, I love lists.  They give us something to argue about, or to make fun of (Madison, Wisc., where are you?  And where's Athens, Ga.?).  But as I often preach on this blog when talking about media and political knowledge, it's the methodology that gets you most in trouble. 

Amazon calls this a measure of the "best read cities."  It's not.  It's a measure of "most Amazon-buying cities."

No comments: