• by Rob Garner , March 12, 2008

Back on Dec. 4 of last year, I listened in on a press call hosted by Google's Marissa Mayer, who discussed the 2007 Google Zeitgeist, and how marketers can better use Google Trends to measure search buzz. Mayer spoke of Google Trends as a "political oracle," in which user search trends can potentially predict the outcome of major political elections (covered here: "Google Zeitgeist 2007: The 'Fastest Rising' Social Network is... Webkinz"). Considering this interesting assertion, I was tempted to compare the Democratic candidates' search popularity in Google Trends in order to predict the outcome of last week's heated primaries in my previous column.

Instead, the post-primary analysis appears to be much more interesting and revealing for future political search campaign strategies. If Mayer is correct, then the candidates who are more popular in searches within their respective geographic areas would also be the winning candidates in those regions. If not, maybe we just learn something linguistically about how people in Texas and Ohio search for the Democratic candidates -- a valuable finding in and of itself. So let's have some fun and see how the candidates compared in last week's primaries.

Three search variations in two states

For this test analysis, I ran a Google Trends comparison for three name sets: First names ("barack vs. hillary"), last names ("obama vs. clinton"), and firstname/lastnames ("barack obama vs. hillary clinton"). These terms and phrases were compared in Texas and Ohio only -- two of the most important states for winning Democratic delegates in the March 4 primaries. Keep in mind that only Google really knows all the variations used to search for candidate information, and this one-to-one comparison of searches is conducted by name only. Google Trends also does not provide quantitative search data. The real story about how people search and perceive candidates may also lie in the sum of the broader queries that are three, four, five, and six or more words long (you know, that long-tail thing).

I also used March 2008 datato measure the search buzz momentum around the candidates as we were going into the polls. Note that the third notch on the horizontal line of all Google Trends diagrams is March 4, the day of the primary, and is the primary point I refer to in the observations below.

How Obama and Clinton searches compared in Texas

This link shows Google search patterns for the phrases "barack obama" (blue line) compared to "hillary clinton" (red line) in March 2008. Clearly Obama maintained more popularity than Clinton on voting day, though Clinton had a slight spike in the post-primary news cycle, before both candidates trail off significantly in the days following the primary. Interestingly enough, even after search interest takes a nosedive, Obama still led in Google search interest in this chart.

If you click on this Google Trends link, you will also see that Clinton and Obama were in a dead search heat in Austin and San Antonio, while Obama trounced Clinton in Ft. Worth, and also led in Houston and Dallas.

Those Texas searchers who preferred to seek info on candidates by last name in the early days of March ("obama vs. clinton") also preferred Obama as well. Obama searches were significantly higher in every city in Texas, indicating either Obama's dominant popularity, or a preference in searching for him by last name. Obama stayed on top with an even greater divide in comparing "obama" vs. "hillary clinton".

"Barack vs. Hillary" was a different story. Clinton (let's take a leap and say that 99% of "hillary" searches were for her) was more popular than Obama in this search, and a further comparison of last names vs. first names shows that searches by last name are the most popular name variation searchers use to find information about the candidates -- and in this example, Obama had a huge lead at polling time.

The final Texas tally: Clinton had fewer searches for the primary name variations in Google, but beat Obama for the popular vote in Texas, 50.9% to 47.4%. Both candidates tied with the number of delegates at 92 each.

How Obama and Clinton searches compared in Ohio

In Ohio, last names were also more popular than first name searches, and while Obama led by last name, there seems to be an equal trend for cumulative searches when viewing from top to bottom -- again, another reflection of the dead heat that political pundits and pollsters have talked so much about.

Looking at last name searches, Obama also held a significant lead overall, and also in every major city in the state. Obama also spiked at a higher trajectory on the day preceding the Ohio primary. Overall in Ohio, Google Trends indicates that there were higher search volumes for Obama than Clinton for the three common variations.

The final Ohio tally: Obama won the Google Trends frequency comparison, but lost the popular vote to Clinton, 44.8% to 55.2%, with a split of 229,873 votes. Clinton also took the majority of delegates - 74 votes to Obama's 65.


Overall, the relative Google Trends search patterns in Texas and Ohio do not mirror how the actual primary and caucus results turned out. Maybe this is a first aberration for the Oracle, but what is possibly more revealing is that Obama followers and interested searchers may be more frequent users of Google as a search engine. To this point, campaign strategists of both parties may find many insightful stories in the search data - they will just have to know what to look for, and know how to use it. Search engines are a place where minds and ideas are challenged, and quite possibly the candidate who is top of mind in search -- or has the best search campaign -- could have an extra edge going into the November election.