Search trends for U.S. presidential candidates Hillary Clinton and Barack Obama demonstrate how Google Trends could be a useful tool even though the information Google provides is only comparative and doesn’t show the number of searches made.
Google’s “Google Trends” feature is much in the news today: Google has improved its feature that tracks the most popular search terms by making daily results available. Most people are curious about what others are thinking, so it’s not surprising that information outlets from TV news to bloggers are picking up this development. Most of the news coverage focuses on titillating search terms, and indeed much of what people seem to be looking for on Google is related either to sex or to irrelevant celebrities.
Google’s feature is interesting, but how useful is it, really? I’m sure there’s money to be made one way or another by following or anticipating the direction of search trends — it sounds like a good way to fish for topics if you’re trying to bump up your blog traffic, for example. But the utility of the tool is flawed because of the lack of quantitative information. I can see, for instance, that searches for “Gavin Newsom”, the mayor of San Francisco, spiked in February of this year. And I can see that over the past 12 months, more people have googled that term in Pleasanton than in San Francisco. But there are no hard numbers. Was it 100 people who googled Newsom? 100,000? 10 million? Google probably has this quantitative information, or the ability to get it, but it’s certainly proprietary and perhaps not in Google’s business interests to release numbers. However, quantifying these trends certainly would make this feature a more useful tool for serious study of what people really search for, and I hope Google will make the data available.
Despite the lack of hard data, however, there is some value in using this tool, particularly when instead of looking at just one search term, you use it for comparative analysis of several terms.
It seems likely that PR flacks, investors and people in advertising will be keen on Google Trends. To me, one field for application that comes to mind is politics. Below are some graphs that result from entering both “Hillary Clinton” and “Barack Obama” into Google Trends, where Clinton’s trend line appears in blue and Obama’s in red.
The graph above shows Clinton in blue and Obama in red, with search volume constrained to the United States and unconstrained by time (the time period measured appears to start in 2004 and continue through to the date of the trend search). Obama’s got a nice spike in 2004, no doubt a result of his speech at the Democratic National Convention, that far exceeds Clinton’s search volume for any comparable period. But notice Clinton’s steady rise in 2007, particularly the recent rise while Obama has stayed relatively flat.
The graph above is also unconstrained by time, but the results in this case are constrained to searches made in California (I don’t know the method Google uses to determine the geographic locations from which searches are made or how they account for searches made by people using an IP anonymizer and such).
This graph shows the trends constrained to the United States and the time period of the past 12 months. There again are the relatively flat Obama line and the relatively slowly rising Clinton line.
Now here’s California for the past 12 months. My eyes tell me Obama is a more popular search subject in the state than nationally in this time period.
The graph above is U.S. national for the past 30 days. We’d have to run a different comparison to get any information about how these guys fare against Republican competitors in search trends, but in comparing Clinton and Obama, this looks tough for Obama. Every time he gets a bump, Clinton gets one too, and sometimes Clinton’s bumps are bigger.
Finally, just for the sake of completeness, here is California for the past 30 days. However, Obama doesn’t show up at all, so I’m inclined to guess that this graph probably is not statistically significant but simply shows a glitch in Google’s systems instead of a sudden disappearance of interest in a major presidential candidate.
How does search term popularity convert to either actual voter behavior or campaign contributions? That’s a good question, and unfortunately I don’t know yet. This is bound to become something that campaigns and analysts watch closely, though, and it wouldn’t be surprising to see campaigns push favorable Google trend information to the news media as a way of combating unfavorable results from traditional polls.
For some search terms, Google is able to provide not just trends for the volume of searches, but trends for mentions in the news as well. That information wasn’t available for my Clinton and Obama comparison, unfortunately, but one really interesting example would be to compare search volume and “news reference volume” for Google and Yahoo over the past 12 months.
At the moment, it’s likely that “news reference volume” plays an important role in influencing search volume, but at some point if Google search trends get more notice, searches and news references may very well become mutually influential feedback flows.
Graphs reproduced from Google for “fair use” purposes of reportage and commentary.
For more thoughts on using images to represent data, start with Edward Tufte.