What Google Tells Us about the Election (and Ourselves)

Jason Micheli —  October 26, 2012 — Leave a comment

“On Election Day in 2008, roughly 1 in 100 searches that included “Obama” also included “KKK” or “nigger.”

Our thoughts are often superficial. “Paul Ryan shirtless” is currently Googled 9 times more often than “Paul Ryan budget.” Don’t ask me why, but “Paul Ryan shirtless” is Googled more frequently in blue states than in red. When we search for “Michelle Obama,” we include the word “ugly” three times as often as the word “beautiful.”

Politicians can map the geography of their popularity by looking at what they’re called on Google. “Obama” is Googled more frequently in blue states, but “Barack Hussein Obama” is Googled more often in red states — just as “Willard Mitt Romney” is in the blue states.”

It’s widely known that people tend not to be truthful or forthcoming to pollsters when it comes to sensitive topics like race, religion and sexuality. Especially when your opinions on those topics tend towards the lesser angels of our nature. Turns out we’re less guarded when we’re on the computer, googling and thinking no one is watching us or recording every search term and key stroke. And what we see about ourselves when we don’t think anyone can see isn’t very flattering.

Check out this article in the NY Times about what Google tells us about our elections. Some of the data is amusing. Much of it is ugly and disheartening. What’s interesting is just how reliable Google analytics can be in predicting voting patterns.

————————————————————————————————————————–

Every election season, pollsters try to figure out the demographic makeup of the electorate in an election that hasn’t happened yet. And every election season, pollsters are greeted with charges that their estimates are wrong. Republicans criticize 2012 polls that assume that African-American turnout will remain at its 2008 level. Democrats criticize 2012 polls that assume African-American turnout will be lower than it was. And that’s just one demographic group.

It’s hard to predict voter turnout because people are reluctant to admit that they will not vote. How reluctant? One recent estimate suggests that as many as two-thirds of people who will end up not voting tell pollsters that they will.

In my work in economics, I use anonymous, aggregate data from millions of Google searches in hundreds of media markets in the United States to measure variables on sensitive topics — racism, drug dealing and child abuse, for example — where people tend to be less forthcoming in surveys (to put it mildly).

My research suggests that by comparing Google search rates for voting information so far this year with search rates on comparable dates from previous elections, we might already be able to get a pretty good idea of the composition of the 2012 electorate.

Despite the ubiquity of Google searching, and searchers’ demonstrated willingness to share their true feelings and unbridled thoughts on Google, what Americans are typing when they search remains surprisingly underutilized in political analysis. But Google can often offer insights unavailable elsewhere.

Some of what we learn is pretty silly. Every month, about 5,000 people ask Google about Mitt Romney’s underwear choice (devout Mormons wear temple garments). But some of what we learn is disturbing. On Election Day in 2008, roughly 1 in 100 searches that included “Obama” also included “KKK” or “nigger.”

Our thoughts are often superficial. “Paul Ryan shirtless” is currently Googled 9 times more often than “Paul Ryan budget.” Don’t ask me why, but “Paul Ryan shirtless” is Googled more frequently in blue states than in red. When we search for “Michelle Obama,” we include the word “ugly” three times as often as the word “beautiful.”

Politicians can map the geography of their popularity by looking at what they’re called on Google. “Obama” is Googled more frequently in blue states, but “Barack Hussein Obama” is Googled more often in red states — just as “Willard Mitt Romney” is in the blue states.

How frequently people in a state searched for “Obama jokes” almost perfectly predicted the vote share of Mr. Obama’s 2008 opponent, John McCain. “Romney jokes,” which typically focus on his wealth, are popular in Iowa and Ohio, two swing states in which Mr. Romney has struggled to connect with working-class voters. Never mind favorability; maybe what we need is a jokeability index.

Google search data also give some evidence suggesting that last-minute rumors had negative effects in the 2008 election. There were a number of states, like Oklahoma, Tennessee and Kentucky, in which Mr. Obama slightly underperformed in the final polls. Google search data offer one rather interesting correlation: these states had some of the largest search volumes for “Obama Muslim.” And those searches, while not uncommon throughout the summer and early fall, rose substantially in the final days of the campaign, after many of the final polls were conducted.

Comparing the timing of our Google searches to outside events is often intriguing. Searches for “McCain life expectancy” rose to unprecedented levels the day of his controversial choice of the Alaska governor Sarah Palin as his running mate. They rose again after Ms. Palin’s poorly received interview with Katie Couric.

Google data may also help us predict the composition of the 2012 electorate. Individuals may systematically deceive pollsters regarding their intentions, but actual voters are far more likely to Google phrases like “how to vote” or “where to vote” before an election.

By the middle of October, taking the frequency with which Google searches include “vote” or “voting” and comparing the number to those from the same days four years earlier strongly predicts where turnout will rise, stay the same or fall. If search rates for voting information were higher in the first half of October 2008 than in the first half of October 2004, voting rates tended to be higher in 2008 than in 2004. It’s true for midterm elections, too. If search rates for voting information were higher in the first half of October 2010 than in the first half of October 2006, voting rates tended to be higher in 2010 than in 2006.

Click to see the rest of the article.

Jason Micheli

Posts

No Comments

Be the first to start the conversation.

Leave a Reply

Text formatting is available via select HTML. <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.