Can Google Search Trends Predict Election Results?
American voters are perplexing animals. How easily persuaded are they, what issues matter most to them, what causes them to scream for blood, on what issue do they keep their head in the sand – these are questions that pollsters thrive on when studying the American voter. These electoral spectators literally make their daily bread on being able to accurately weigh the answers to these questions because the result creates a (mostly) reliable dataset that can be used to predict the outcome of elections. Even then, a reliable prediction of which candidate is projected to win can produce this brain-knotting meta effect that may sway a voter’s opinion on a candidate.
In short: political polling is the math of meaningful chaos.
Some of the most sought after data are responses voters give in exit polls, polls taken after voters exit the voting booth that ask them who they just voted for. Pollsters can usually predict the results of the election based on these election polls, so it’s pretty coveted information. When asked about which candidate they voted for, pollsters will sometimes also ask voters other questions, like what issues guided them towards voting for their chosen candidate. It could be the economy, immigration, national security, hair cuts, ad negativity, or any of a thousand other variables that hold some ephemeral credence in the minds of American voters.
I’ve always suspected there may be a response bias in exit polls; that is, a voter will reply with answers that they think pollsters want to hear, not with answers that are actually sincere to the voter. Such a phenomenon can muddy the statistics and thus actual election results might not reflect the exit poll. The exit poll metric is less debatable when an election is a blowout, but in narrow races – which seems to be all that happens in the United States these past few presidential elections – it can be a highly contentious issue. That’s what prompts recalls, recounts, legal challenges and so on (*ahemFloridahum*). At any rate, can we ever be sure that voters are really being truthful about what issues concern them when responding to exit polls?
In the exit polls of Florida’s Republican primary earlier this week, people said that the economy was the issue that concerned them most (62%), followed by the federal deficit (23%), abortion (7%), and illegal immigration (3%). But are voters actually concerned with any of these issues or are these terms mere buzzwords that voters are parroting when asked by nosy exit pollsters? Google, who always looks to wave their titanic quantity of search data at a question in order to pan out an answer, may have the answer.
Google took a look at the search trends in Florida leading up to the election to see what people in The Sunshine State were searching for prior to the election. Through the use of Google Insights for Search, Google found that the issues weighted most heavily in the exit polls actually reflect the popularity of search trends in Florida. Turns out people in Florida really do consider the economy the most important issue.
What’s more, the exit polls show that 81% of voters consider home foreclosures in their community to be a problem. To check this out, I searched for “foreclosure” in Insights for Search and found that searches involving that word spiked up on January 30, the eve of the election primary. Even before the 30th, though, searches involving the word “florida foreclosure” have remained high in the state over the past year.
Google’s analysis of the search trends in Florida seems to corroborate the reliability of the exit polls. What’s more, this reveals a pretty potent fact: people probably aren’t just making things up or embellishing answers when responding to exit polls. They’re actually searching the issues and their answers in the exit polls seem to reflect their interests.
Given the positive correlation in the relationship between exit polls and search trends, this raises a pretty fascinating question: could search trends actually be used to predict the outcome of elections? Google’s been sharing the search trends of presidential candidates this year on their Google Politics & Elections News page, so it’ll be very peculiar in mid-November if the then-known election results (at least, let’s hope we know the results at that time) reflect the patterns observed in search trends leading up to November 6.
So is this the way we’re going now, crowd-sourced electioneering via Google? How do you think knowing the search trends among voters might affect the campaign of a candidate? Could it affect how you weight certain issues or candidates? Add your thoughts below in the comments.