Why Google Knows How You’ll Vote in the 2016 Election
If you’re like a lot of Americans following the 2016 presidential election, you’ve probably downloaded an app or two to stay up to date on the news, or you turn to Google to search for the latest on your candidate of choice. But did you know that Google, the search engine giant that’s basically impossible to avoid on the web, probably knows who you plan to vote for?
David McCabe reported for The Hill that Google is leveraging the excitement surrounding the 2016 election to try to establish itself in the political polling industry. The company has been promoting its survey products to staffers and operatives with presidential and congressional campaigns, plus the journalists who cover them. Google Consumer Surveys — a product that espouses the philosophy, “To find out what people really think, just ask the internet” — has proven appealing to presidential campaigns, reporters, and editors.
At least one presidential campaign has put a poll in the field using the tool, and after Google launched a website in January to explain the service to journalists and let them access data from polls created by Google’s team, publications including The Wall Street Journal and The Washington post have used Google’s tools for political polling. Additionally, Google has “struck a longer-term partnership” with the Independent Journal Review, a right-leaning site that’s created viral videos starring presidential candidates.
Google makes money from the surveys, since the cost of running customized polls can run in the thousands of dollars, and campaigns that want to take advantage of advanced targeting options sign longer-term contracts. But Justin Cohen, a product marketing manager for Google Consumer Surveys, tells The Hill that the effort “obviously has great branding impacts as well.”
The system delivers results much more quickly than traditional polling methods, but many have been skeptical about online polling because most polls rely on self-selecting groups of respondents, which makes obtaining a representative sample difficult. Some groups of people, such as elderly or low-income Americans, are less likely to have reliable internet access. Google says that it manages to survey groups that are representative of the population on the internet, in part because of the information it can infer about you based on your browsing history.
As to whether Google has determined which candidate you intend to support, it probably has if you’ve answered some survey questions before reading a news article online, or through an Android app that offers Google Play store credits for answering a few questions. As the Google Consumer Surveys website explains, “people answer questions in exchange for access to premium online content,” which enables campaigns or journalists to “reach thousands of validated, representative respondents” who are “everyday people, not just those who chose to participate in research panels.”
But even if you haven’t answered questions in exchange for Google Play credits or access to an article, the search engine giant is still likely to have identified your favorite candidate. Google tells The Hill that it can infer a person’s age and gender from their browsing history, using technology similar to what it uses to target ads, and can determine a user’s location based on their IP address. Using insights culled from your browsing activity, and the demographic questions that users answer in its mobile app, Google can craft a representative sample of the internet population.
“Google Consumer Surveys selects potential responses for each survey using inferred demographic characteristics to get as close as possible to the census for the internet population,” Cohen told the publication. “This ensures a representative and statistically significant sample.” According to Google’s help page for Google Consumer Surveys, the company has found that it doesn’t need to replicate traditional techniques that pollsters use “to adjust their results to reduce bias introduced in the sampling methodology or the design of their survey.”
That’s because “GCS’s reach of internet users is so large that the sampling methodology we use doesn’t need many of these tricks to give us accurate results, as proven in the 2012 US Presidential election.” Translation? Google has access to so many people (and to enough information about their browsing activity) that it can “accurately find likely voters and measure their voting intentions,” without the help of techniques to reduce bias introduced by the sampling methodology or survey design.
The Hill’s McCabe points out that Google isn’t the only tech company trying to gain a foothold in political polling, or other services for journalists and campaigns. SurveyMonkey, for instance, has a partnership with NBC to produce polls for the 2016 election. And The Next Web’s Kirsty Styles reports that Facebook has launched a “Politicswire” tool for journalists to access user-generated videos, images, and stories about politics in America. (Such content will also appear inside Signal, Facebook’s news-gathering platform.) But as Styles points out, it’s starting to look “a little creepy” that tech giants have a such a view into democratic processes which are supposed to be secret ballots.