Google’s Marissa Mayer, VP of search products & user experience, proposed Google Trends as a way of polling the populace, suggesting in a webcast yesterday that Trends could help predict who will win an election.
Mayer showed how Google Trends accurately predicted George W. Bush’s dominance over John Kerry in 2004 and Nicolas Sarkozy’s win in May of this year over Segolene Royal in the French presidential election. Current Google trend lines show Clinton beating Obama and Edwards, though I wonder how anti-Hillary sentiment plays into this, given it seems stronger than any anti-Barack or anti-fancy haircuts feeling.
On the webcast, Mayer also said that Google will eventually provide an API for Trends and allow download of the data, but didn’t commit to a time frame for either of those.
Seriously now, how can Google Trends predict the winner. I understand it measures the search query; but the interest could be a negative one and not positive. As I find out more, one might change the opinion and go the other way around. After all, wasn’t Osama one of the top 10 searches in 2001?
Aswath-
I agree completely. Google’s approach misses the boat. It’s simply too much of a stretch to assume that anyone searching say for “Hillary Clinton” is ready to pull the level on her behalf come election day.
Alternatively, Compete.com has created a very interesting method for tracking the popularity of candidates online based on the amount of time people spend across the candidates’ websites and related sites on myspace, youtube, flickr, etc. I found this on their blog…it’s very interesting:
http://blog.compete.com/2007/11/15/facetime-president-contenders-myspace-facebook-youtube-meetup/
It’s just kind of a fun way to use Google Trends, not entirely serious. It’d be interesting to study, though, if it does have any predictive value. Of course people doing the searches might be doing so because they don’t like the candidate — that’s why I pointed out the issue with anti-Clinton sentiment.
Hmm, so who’s going to win the presidential race? I threw in the top 4 candidates, and then the dark horse (Ron Paul). It’s funny to see that the amount of searches (aka interest) are marked by the lowest news reference volume. Sad, actually.
http://www.google.com/trends?q=rudy+giuliani%2C+mitt+romney%2C+hillary+clinton%2C+barack+obama%2C+ron+paul&ctab=0&geo=US&geor=all&date=ytd&sort=0
What about the other side (the Republicans)? Let’s see what Google Trends suggests:
http://www.google.com/trends?q=huckabee%2C+%22ron+paul%22%2C+%22giuliani%22%2C+%22romney%22
Finding “predictions” after the fact is always so easy. Give me a break.
Maybe Marissa Mayer wants to wager a bet on this? Come on Om, take her on!
:)
Had another thought: maybe Google can actually predict their own share price: http://www.google.com/trends?q=google%2C+microsoft%2C+apple&ctab=0&geo=all&date=all&sort=0
Well, maybe not
Great post i agree Google trends is a great tool to predict and view reality of things.
Unfortunately great tools need thinking minds, to feed them with none bias information.
Lets take a closer look; you made search for “Hillary Clinton” (thats good since most people would mistake and input “Clinton” a name shes shares with the most Popular man in the world, her husband ex president Bill Clinton.
For The other candidate you entered “Barack Obama” most people don’t search for Barack since well they have no idea about his first name.
If you run this query you will clearly see who the winner really is ,
http://www.google.com/trends?q=%22hillary+clinton%22%2C+%22obama%22&ctab=0&geo=US&geor=all&date=all&sort=0
Folks – does anyone know if Google Trends does “sentiment” analysis?
Thanks,
@Josh Treadwell: I think the interest in Ron Paul compared to the others is fascinating. Too bad it won’t equate to votes. Still, it’s encouraging to see people thinking outside the 2-party box.
for a french reader looks almost incredible that the sarko sego fight is in the radar of gigaom…
@GeorgeZ: that’s just speculation… we don’t know how many people would use “hillary clinton” or “hilary clinton” or “clinton” or “hillary” or “hilary” for hillary just like we don’t know how many people would use “barack obama” vs “obama”
jenslapinski has the better point: it’s always easy after the fact to find things that predict what happened. I’m sure you could find trend graphs that were completely wrong in their “predictions.”
Still, I think it’s fun :) And I really like Google Trends.
jm
for a french reader looks almost incredible that the sarko sego fight is in the radar of gigaom…
for the French i dedicate this Google Trends search
http://www.google.com/trends?q=Nicolas+Sarkozy%2C+france+strike&ctab=0&geo=all&date=mtd&sort=0
[...]Daylife adds:“Once again, the Daylife Presidential Press Tracker compares favorably with other metrics. Earlier we blogged about the Press Tracker’s correlation with a Harvard Study. First Harvard, now Google…. Anne Zelenka, writing at GigaOM, fleshed out some bits of a Google press conference. Zelenka wrote: … So let’s compare that with what we have on our Presidential Press Tracker…”[...]
Are you sure that it’s not just a fluke that can be attributed to Christmas shoppers making a mad rush for their Hilary Clinton nutcrackers?
http://hillarynutcracker.com/completelynuts.html
Supposing Google has a control group that they monitor before and after, get their disposition and then project to the whole search group, then they can predict what the “posterior” (I hope this doesn’t vector into something else) distribution. They can become the next Gallop, with so much data they are sitting and so many PhDs. Does the webcast hint at anything like this? Is there a link to that webcast?
@Anne Zelenka
Yes it is speculation, but then again speculation drives are economy
lets try this one “hillary” and “Obama”
http://www.google.com/trends?q=%22hillary%22%2C+%22obama%22&ctab=0&geo=US&geor=all&date=all&sort=0
I totally agree with the majority of the comments. Google, is not in position to predict the result of the elections. Even if the track record indicated is in Google’s favor. Bush vs Kerry I would have predicted it also.
@GeorgeZ: How many people know that Hillary is spelled that way and not “hilary”?
How about this: “hilary” or “hillary” (use a vertical bar for or) vs. “Obama”
http://www.google.com/trends?q=hilary+%7C+hillary%2C+obama
Ha! She wins!
Is that what we’re arguing about? ;)
@Anne Zelenka
For arguments sake, you need to take a closer to your search of HILARY
http://www.google.com/trends?q=hilary+%7C+hillary%2C+obama
the majority of results come form countries below
Venezuela
New Zealand
Mexico
Canada
Australia
You need to exclude those countries and run the search string properly within the USA, obviously only americans can vote and obviously we need to add all mispells of names of both candidates
hilary | hillary | hillary clinton | senator hillary clinton, obama | barack obama | barac obama | barak obama | barrack obama
http://www.google.com/trends?q=hilary%2C+hillary%2C+obama&ctab=0&geo=US&geor=all&date=all&sort=0
Hey, given the success Trends has had in predicting the future, I just figured out what I am always going to play in “rock, paper, scissors”!
http://servicelevelautomation.blogspot.com/2007/12/off-topic-its-official-rock-beats-paper.html
Anne -
I suspect your culling of only “USA” will not include military personnel abroad — and am curious what would then happen to the stats; have these folks traded what is their traditional pro-GOP / anti-Clinton (Bill AND Hillary) play for a pro-Hillary get us out of Iraq play ?
Hey thanks for this. I discovered about google trends through your blog. Its interesting. Btw one question, do they take into account misspelled searches too? Like in analytics they do not.
Hirani
http://www.goforads.com
By this measure Ron Paul will be president. He is far and away the most searched/popular candidate according to Google trends, Tech President, Technorati, Alexa and so on.
I’m not sure interest alone can predict an election, but unfortunately, name recognition does play a huge role in American politics so it’s possible Google Trends has a chance.
Another option that has been incredibly accurate over the years is The Iowa Electronic Markets from The University of Iowa Tippie College of Business. The markets have shown much greater accurately than the polls you’ll see in newspapers and on television. The main reason for this is because they are real-money markets. People tend to be a little more honest with themselves when their hard earned money is involved.
I have been tracking the candidates based on various metrics over at this post: http://thetomoreport.blogspot.com/2007/11/2008-presidential-primary-data-tracker.html
The democrat side the equation seems to be fairly uniform (with Clinton basically winning out in most metrics). The republican side is completely up for grabs…
Looks like it might be time to revisit the Google Trends data, Obama is now the leading search results winner.
http://www.firetop.co.uk/2008/02/04/barack-obama-to-win-super-tuesday/