Tracking Obama's Momentum: Google Trends
Wed Feb 27, 2008 at 02:23:54 PM PDT
For those who have not yet noticed Google Trends, let me introduce a new addiction. It's like a new poll from every state, every day! Just think of all the BREAKING diaries that could be written!
In this diary I will take you on a tour through the primary season and show you why, so far, Google Trends does a pretty good job of showing which way the momentum is going.
As a tease, here's the Google Trends results for Clinton (blue) and Obama (red) over the last two months or so, showing the volume of searches in the United States over that time period.
Click to enlarge.
The polls are pretty clear what has happened in the Democratic primary, nationally: Clinton drifted up through most of 2007, and Obama rocketed skyward starting about December 2007, just recently overtaking her.
Click to go to Pollster.
What do we see in the search volume from Google Trends over the same time period? For almost all of 2007, the volume for Clinton is about twice that of Obama. Then, the number of news stories about the primary picks up in December, and the number of searches shoots through the roof (even overcoming the nation's disturbing fascination with Britney Spears. Now, however, the number of searches for Obama is twice the number of searches for Clinton. (Note Clinton is always blue and Obama is always red in these graphs.)
Click to go to Google Trends.
Does Google Trends tell us who's got the momentum?
These two graphs suggest that in order for a candidate to be gaining in the polls, they must also be getting more Google searches. Also, the more searches, the faster the potential gains. However, there's no guarantee - a spate of very bad news for a candidate would probably also result in high Google search volume, but their poll numbers would surely drop.
Let's zoom in on the primary season:
Click to enlarge.
It's pretty clear that the day after almost every primary, there is an increase in the news and a proportional increase in Google searches for both Obama and Clinton. There are some exceptions: NV, MI, and FL. There wasn't too much news after the Nevada caucuses, but there's still not much search activity, perhaps because it was more or less a tie. MI and FL, of course, were not contested and the press didn't pay much attention; consequently, neither did the public.
Do the polls agree with Google Trends over the short term?
Let's switch to the Rasmussen and Gallup tracking polls for a minute:
Click to enlarge.
I've added a smooth line through the data; the graph on the right shows just this line, along with some key dates. First, aside from a bump when Edwards withdrew, there's little change in Clinton's numbers here. Obama shows clear gains between IA and NH, and after SC. The last week or so he has leveled off a bit. Part of this may be because half the country has voted already, so there's less room for change. For the most part, Obama's gains correspond to the time periods when he was getting about twice as many Google searches as Clinton.
Here's Google Trends and the polling data together:
Click to enlarge.
Does this work for states too?
At this point, then, let's say that outperforming your opponent on Google Trends is a necessary but not sufficient step in gaining in the polls. So let's test this hypothesis by looking at Google Trends for individual states.
We know from the polls that Obama had momentum in almost all the states that have voted so far, even if he ended up losing them. But there are two where he did not compete: FL and MI. Let's see what they look like:
Click to go to Google Trends. FL left, MI right.
From the search volume, we would expect perhaps some momentum for Obama in the weeks before the election in FL, because of his wins in other primaries. Pollster shows in fact a slight decrease in Obama's poll numbers during January, although contrary to this exit polls show he had a slight advantage among those who made up their mind during this time period. Likewise with Michigan, Google Trends would lead us to believe no candidate had momentum, and what polls we have, as well as exit polls, seem to agree.
Now let's turn back to states where the polls showed clear momentum for Obama:
Click to go to Google Trends. Left to right: CA, WA, VA.
In all these states, and indeed in most states, Google Trends says Obama had momentum. Generally, the peak search volume is the day of the election, not the day after as we saw for the national search trends. This is because, I assume, searches are driven by campaign activities and the resulting press coverage, not by press coverage of results.
Conclusion: Google Trends probably does show momentum.
If you find yourself hovering over the refresh button waiting for the next poll to come out, try visiting Google Trends while you wait. It looks like, for the Democratic primaries so far at least, search volumes are pretty good predictors of who is gaining and who is not.
Here's some links: Texas, Ohio. Sadly, Vermont and Rhode Island don't have enough searches to plot. From the recent data, however, it looks like Ohio is where the action is the last few days....
P.S. Tip o' the hat to poblano, who wrote an earlier diary about Google Trends, it turns out.
Cross posted at Open Left.
Permalink | 15 comments