Canadians were rightly alarmed earlier this year when details of a secretive figure named Pierre Poutine first came to light. Using an auto-dialing service in Quebec, an anonymous partisan operative allegedly sent voters identified as non-Conservatives to the wrong polling stations during the 2011 federal election. But while the so-called “robocall affair” exposed the underbelly of today’s political campaigns, it also opened a door into a world where political parties exploit our ever-growing webs of personal data.
Mobilizing your supporters and discouraging your opponents, the bread and butter of any election campaign, was once a matter of recruiting enough volunteers to canvass neighbourhoods and drive people to the polls. These days, it’s increasingly the work of data analysts and behavioural scientists who collect reams of publicly available personal information and use computer algorithms to exploit it. Their goal: nothing short of pinpointing the fears and hopes that motivate individual voters, and using that information to target them for donations and votes on election day.
How you vote may seem like the last bastion of individual agency, but political campaigns say they can predict what messages will move you with unnerving accuracy by studying everything from your home address, to your magazine subscriptions, to what you like to watch on TV on a Saturday night—or even whether or not you own a TV in the first place. Dubbed “microtargeting,” these new techniques promise to have profound implications for the political process. “The idea of Pierre Poutine, it was funny,” says Carleton University professor and former Reform party pollster André Turcotte. “But the real story is in what parties are doing and not doing with their data and about how that technique is hijacking the political process.”
Nowhere is this emerging field of data mining more advanced than south of the border, where political campaigns are investing millions to hire consultants and buy individualized consumer databases that track fishing licences, people’s web-browsing history and credit card transactions. Barack Obama’s presidential re-election campaign reportedly has more than a dozen data analysts, including a “chief scientist” whose previous job was mining consumer data for private sector clients. The Obama campaign’s latest technological innovation is an algorithm that searches for patterns in voters’ written submissions to the campaign’s website. They call it “microlistening,” and while the Democrats have been quiet about exactly how their algorithm works, they say it has the ability to make predictions about someone’s voting behaviour based on the words they use to talk about such things as unemployment and the economy. In theory, if one person were to write on the campaign website about the “socialist” 2009 auto bailout, while someone else used terms like “saving jobs” or “sacrifice,” the algorithm could predict which person is likely to vote Republican, and which one is worth targeting for donations or a follow-up to make sure they’re on the registered voters list.
In fact, American political parties now collect thousands of identifying details to develop individual profiles of the nearly 175 million registered voters in the U.S., along with profiles on millions of unregistered voters to decide whether it’s worth spending money to encourage them to vote. U.S. voter registration lists already provide the basic information—names, addresses, voter turnout histories and, in some states, birthdates, race and ethnicity. On top of that, parties add in the voting history of your registered poll to decide whether your neighbourhood tends to always vote for one party. More recently, parties have been purchasing consumer data collected in private, for-profit data warehouses. If you’ve ever filled out a warranty form, entered a sweepstakes, filled out a customer-service questionnaire, subscribed to a magazine or taken a cruise, that information can be purchased by American political campaigns to develop your voter profile.
Campaigns then use computer algorithms programmed to weigh the relative importance of each piece of individual information and calculate a series of “probabilities” about you: a 72 per cent chance you’ll cast a ballot and a 49 per cent chance that you’ll support Obama.
The key to data mining, say experts, is the crucial demographic detail buried in the sheer volume of available consumer data. On its own, knowing that you took a Hawaiian cruise last September doesn’t necessarily say much about what you might think of Obama’s health care reform. But knowing that you take three cruises a year might suggest that you have a fair bit of disposable income. Couple that with data that says you earn between $50,000-60,000 a year, have children, voted in every election for the last 15 years and live in a neighbourhood that swings between Democrat and Republican, and suddenly you’re seen as a prime target for a political message about tax breaks for the middle class. “There’s an assumption that when you have a thousand data points about an individual, many of them related to their past political behaviour, you can start to make informed predictions about [how they’ll vote], even if they’re not upfront about that with you or with themselves,” says Sasha Issenberg, a journalist who spent the last few years researching the data-mining techniques of U.S. campaigns for a new book, The Victory Lab.
By the time a politician knocks on your door, he may know more about you than most of your friends do, says Issenberg. It’s allowed campaigns to make informed predictions about how you might vote before you think you have even made up your mind. The parties can then target you by phone, mail or email with a message best suited to you—one that might be different than the message sent to your neighbour.
In using such tactics, political parties are borrowing from the corporate world, which has been successfully collecting and mining data on its customers for years. Retailing giant Target employs 50 data analysts who, by examining subtle shifts in the buying habits of female customers, can predict with startling accuracy which women are pregnant, down to their due date. Meanwhile, companies like Google and Hewlett-Packard analyze data on their employees to figure out which ones are most likely to quit in the future—potentially even before the employees themselves decide it’s time to go.
Former Columbia University computer scientist Eric Siegel, an expert in the emerging field of predictive analytics, says data-mining has helped uncover all sorts of behavioural trends, from the fact that fans of Rihanna are likely to vote Democrat, to the revelation that vegetarians are less likely than meat-eaters to miss a flight. “Data is a tremendously rich resource because it’s encoding what’s happened in the past and so you’re literally learning from experience,” he says. “It enables you to make a prediction rather than a guess.”
In Canada, campaign spending limits and privacy laws mean parties have less access to the expensive private consultants and warehouses of consumer data that have become central to American politics. But that doesn’t mean parties here aren’t diligently building vast databases on individuals that can be used to solicit for donations and mobilize voters on election day.
Parties regularly scan obituaries to weed out deceased voters and scrutinize letters to the editor to flag voters who are passionate enough about certain issues to write to their local newspaper. They use networks of volunteers to canvass neighbourhoods and call voters, and hire pollsters to conduct detailed daily polls measuring the impact of an attack ad or a campaign promise. But where those political volunteers were once armed with pens and clipboards, they now have mobile apps to record the answers you give at the door and feed them into centralized databases.
Need a ride to the polls? Care about health care more than the economy? Want a lawn sign? All of that is recorded next to the barcode assigned to each individual voter. “If you write a letter to a political party or show up at an event or register your opinion in some form or another, they can track you,” says journalist Susan Delacourt, who has researched the marketing techniques parties use during elections for a book, Shopping for Votes, due out in March. “If I’ve somehow communicated with the Liberal party that my cable bill is bugging me, they can log that. It’s just this rolling database of contact and information and anything else they can accumulate.”
For instance, parties could use the information they’ve gathered to contact voters for telephone town halls on issues believed to be important to them. Once you’re on the phone, the party will track how long you stay connected as a measure of how engaged you are with the campaign and the core issues. Increasingly they’re also tracking the words you may use during question-and-answer sessions and monitoring them for keywords that will indicate your views.
Already, the Liberal database tracks about 150 different voter issues, says Delacourt. It has room to log 260 different answers to questions beyond just “yes” and “no” and can tell a volunteer before they even pick up the phone whether a voter has already hung up on past canvassers, or whether they spent 20 minutes talking about their passion for the environment.
Capitalizing on the growing demand for data on voters, private sector firms are rushing to help parties sort through the mountain of personal information Canadians post about themselves on the Internet. Pitney Bowes sells software that can analyze a database of voter names to automatically detect gender and ethnicity. New Brunswick’s Radian6, now owned by U.S. cloud-computing service Salesforce, can track everyone on Twitter who has retweeted a message from a party or politician. Say someone isn’t a member of the NDP, but always retweets a candidate’s messages around the party’s platform on poverty. That information can be logged to identify the voter and then target him for a traditional membership drive by promoting the party’s platform on poverty reduction.
The company also uses software to determine whether someone is writing positively or negatively about a certain issue, say, the Keystone pipeline or a tax cut for small business. It offers its “sentiment analysis” in seven languages.
At the same time, Environics Analytics’ Prizm database has crunched reams of demographic data from Statistics Canada, consumer surveys, and existing Environics research to sort Canadians into 66 “lifestyle groups” based on their postal code. The company also breaks down federal ridings into 87 different “social value” attributes.
The pictures of voters that emerge from this can be unnervingly specific. The Prizm database can tell parties, for example, that voters in the Davenport riding in Toronto tend to be less interested in nature than the average Canadian, are much more skeptical toward both big and small business, and place a higher importance on physical beauty. Coupled with details on individual voters, the company says its data-crunching resources allow parties to identify the lifestyle preferences and value systems of their most dedicated supporters, and then identify postal codes with similar demographics to target for votes or donations. “It’s about trying to understand the voter,” says Rupen Seoni, vice-president of Environics Analytics. “You’re peeling back the onion.”
Another favourite of Canadian political data miners is the Bureau of Broadcast Measurement, which tracks people’s TV, radio and media consumption by postal code and can help parties target their limited advertising budgets. The NDP gathered data on where Canadians paid the highest cellphone rates and then targeted those voters for a campaign promise to cap cell charges.
“If someone reads Field and Stream they’ll probably be a hunter and a Conservative,” says Delacourt. “So if they know where all the Field and Stream voters are, they can send them cards on gun control.”
Among the ways the bureau tracks media habits is through a “portable people meter,” a small pager-like device that detects inaudible codes broadcasters embed into television shows and radio, along with advertising in movie theatres and in-store entertainment. While the information gathered is anonymous, it can signal that a male in his 20s who works in manufacturing in Toronto regularly listens to a particular radio station at 3:15 p.m. on Tuesdays. That can be hugely helpful for parties looking for the right radio station to advertise a campaign promise aimed at young working-class men.
The next frontier in microtargeting is services that match the offline data parties collect on voters with their online behaviour. Private companies now offer services that can match cookies—small text files that the websites you visit store on your web browser—to names and email addresses stored in political databases. “If you want to send an ad to women in Missouri with an Obama support score between 40 to 60 per cent, you can take that category of names, go to a network of [Internet] sites and tell them for the next week if any of these people come to your site, show them this ad and don’t show it to them more than five times,” says Issenberg.
Despite its popularity among political campaigns, microtargeting has its share of skeptics. Conservative strategist and University of Calgary professor Tom Flanagan says the focus on microtargeting is overblown. “Yes, there are sometimes special letters that go out when some issue is in the news. You may know from your identification records that there’s a certain number of people who respond to an appeal related to gun control,” he says. “But that’s a kind of sideline. Most of the money is raised just by asking people to give because it’s the party they support.”
Even so, the growing use of technology to identify and target voters has alarmed federal privacy advocates. Both Elections Canada’s chief electoral officer Marc Mayrand and the federal privacy commissioner, Jennifer Stoddart, have raised serious concerns about high-tech political campaigns, which are neither government departments nor private corporations and therefore aren’t covered by existing privacy rules governing what information can be collected and whether it must be disclosed.
“I can go to the Bay and demand to see all the information they have on me and I can go to Revenue Canada and demand to see all the information they have on me,” says Delacourt. “I can’t do that with the Liberals.”
Parties argue that their data analysis makes governments more responsive to the electorate, their policies more grassroots. But it also means they focus on the interests of only a small slice of the population since they can accurately identify whom those swing voters are and where to find them. Out of 23 million eligible voters, says Carleton University’s Turcotte, the Conservative voter-tracking software enabled the party to identify just 500,000 that could hand them a majority.
It’s at least partly responsible for the rise of election campaigns full of niche issues that may be important to groups of targeted voters, but short on national vision. “It’s hard to deliver on vision, but it’s easier to deliver on practical, specific policies,” says Turcotte.
That trend was thrown into stark relief last month, when Citizenship and Immigration Minister Jason Kenney’s office sent a mass email promoting the government’s support for gay refugees from Iran. The email addresses of the recipients had been gathered from a petition they had signed on the site Change.org supporting gay refugees, which was sent to Kenney’s office.
Some were outraged, suggesting the goverment could be tracking their sexual orientation. Others were simply outraged that the government was trying to pick them off as single-issue voters. Either way, all got the message that signing an online petition had landed them in the database of at least one federal political party, if not all of them. Such stories conjure up images of political operatives hidden away in a dark room somewhere in Ottawa trolling the Internet for lists of Canadians. But in a future where technology can help peel back the curtain at the ballot box, that image may not be far from the truth.
Watch Patrick Tucker, deputy editor of The Futurist, on how our personal data will be used to explain not just who we are but who we’ll become:
And here’s Lee Rainie, director of the Pew Research Center’s Internet and American Life Project, on how society will grapple with the coming data revolution: