One of the things I've been noticing lately is pundits have started using additional data tools to predict voter behavior. Google provides a product called Google Trends that allows you to compare search traffic for various search terms, such as the names of candidates in a campaign. Many Twitter users would commonly refer to these charts during the Republican Presidential Primary. I decided to evaluate the primary election contests I was following in Oregon with Google Trends to see what we can learn about the results of the primary election.
One of the most useful data sets available in Google Trends was that of the Republican Gubernatorial primary. As you can see, Alley and Pierce run neck and neck until the two or three days just prior to the primary when Pierce takes a commanding lead in Google searches. This suggests that early-voting Republicans broke almost evenly, but late-deciding Republican voters broke significantly for Pierce. This resulted in the commanding lead we saw on in the primary for Pierce.
HD 26 (R)
I will analyze more of the House primaries in a future post but for now, I wanted to show the HD 26 numbers because they were quite interesting. If you interpret the numbers below in typical fashion you might think Wingard was the candidate to beat. However, I can provide some context that will help us analyze these numbers.
Wingard complained like a weakling on his Facebook page about being attacked in the mail and with robocalls. There's no way to get a specific total but my guess is tens of thousands of dollars was spent by his opponents and outside groups to ensure Wingard's defeat. I think if you matched up the schedule of each negative Wingard mailer and robocall, that it would almost exactly correlate with each of the spikes in Wingard searches on Google.
Not all Google Trends are good Google Trends.
U.S. Senate (R)
The results in the Republican U.S. Senate primary caused a lot of head-scratching. Everyone was surprised to see Mark Callahan win a surprising six-point victory against favorite Sam Carpenter. Unfortunately, the Google Trends data on this race isn't particularly instructive.
My personal theory is that Faye Stweart and Sam Carpenter split the 50% of the vote that cast their ballot for a more traditional candidate which allowed Callahan to win with 38% of the vote. I don't have a lot of data to back this up, though.
Secretary of State (D)
The results of the Democrat Secretary of State primary were predicted by polling, but not by Google Trends. I don't have enough information to understand what happened in that race.
CD 5 (R)
The Google Trends chart for CD 5 is quite deceptive. You might think Ben West dominated this race. The election results showed Colm Willis was victorious with a 37 point victory. I don't know for sure why the numbers fail to match up, but my guess is that Ben West is a common search term and that many of the searches for "ben west" may not have been related to the Republican primary.
Interpreting data of any kind requires context. Polling, Google Trends, Facebook likes, and other kinds of data all have value, but only when understood in context. Here we notice that Google Trends are insightful when we understand the greater context of the race. We also learned that sample size matters here. Google Trends are useful for statewide races, but anything smaller and the value of the data declines considerably.