Seth Stephens-Davidowitz writes in a NY Times blog, How Racist Are We? Ask Google:
Quantifying the effects of racial prejudice on voting is notoriously problematic. Few people admit bias in surveys. So I used a new tool, Google Insights, which tells researchers how often words are searched in different parts of the United States.
Can we really quantify racial prejudice in different parts of the country based solely on how often certain words are used on Google? Not perfectly, but remarkably well. Google, aggregating information from billions of searches, has an uncanny ability to reveal meaningful social patterns. “God” is Googled more often in the Bible Belt, “Lakers” in Los Angeles.
[...]
Once I figured out which parts of the country had the highest racially charged search rates, I could test whether Mr. Obama underperformed in these areas. I predicted how many votes Mr. Obama should have received based on how many votes John Kerry received in 2004 plus the average gain achieved by other 2008 Democratic Congressional candidates. The results were striking: The higher the racially charged search rate in an area, the worse Mr. Obama did.
[...]
Add up the totals throughout the country, and racial animus cost Mr. Obama three to five percentage points of the popular vote. In other words, racial prejudice gave John McCain the equivalent of a home-state advantage nationally.
I'm not statistician enough to check this Harvard doctoral candidate's methods. What do Kossers think?