Google correlations review

NPR’s Morning Edition did a story today about my Google Correlate explorations on NPR  — focusing on politics and food searches — by Shankar Vedantam, I thought it might be helpful to pull together the correlations posts I’ve done. Someone — probably not me — should get serious about using this kind of data to connect search behavior with demographic trends, politics, culture, and other aggregate patterns of social behavior.

Ideally, we need Google to give up the real data, warts and all, for real research. For example, this disturbing bit of text has appeared on their FAQ page:

In December 2011, we added support for time series correlations on a number of new countries. As part of this change, we reduced our sample size for US states and US time series to match that of the other countries. While this does not have much of an effect on popular queries, it may cause a noticeable increase in variance for queries with lower volumes.

In fact, some of the searches I’ve reported on in the past no longer produce the same results, apparently because of this change. Since they have apparently reduced the sample size, I favor the results I reported, but this is not the kind of thing we’d do if we ran the zoo. If anyone knows Google, you should try to work this out with them, get some grant money, and really wow us.


Here’s where the blog has been:

Wow, that’s a lot of posts. (It reminds of when Maine’s recycling law led Stephen King to start stockpiling beer cans instead of throwing them out, and the size of the pile illustrated the depth of his alcoholism.) Maybe I should cool it for a while.

The story is here, with an audio link.

NPR addendum:

Here are the foods that appear in the top-100 correlations with searches for “npr” across states:

  • couscous
  • no knead
  • cooks illustrated
  • knead bread
  • no knead bread
  • flourless
  • flourless chocolate
  • creme fraiche
  • curried
  • chard recipes
  • polenta
  • wheat allergy
  • caster sugar
  • bittersweet chocolate

4 thoughts on “Google correlations review

Comments welcome (may be moderated)

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s