Federal election 2015: How data mining is changing political campaigns
Parties collect data on voting intentions to target their message
Canvassing door to door, a time-honoured ritual of any political campaign, is more than it appears.
It's a candidate's chance to press the flesh, of course, but increasingly, it is the entry point for data mining, which is how parties collect as much data as they can about you and your voting intentions.
It's kind of scary in a way, because what are they going to do with all this in the future?- Jill Mills - voter
Whether canvassers enter it into an iPad or scribble it in a notebook, the information you inadvertently give a prospective politician will end up in large databases jealously guarded by each of the parties – and closed to the prying eyes of privacy commissioners.
"I think we have Big Brother watching us," says Jill Mills, a voter in Scarborough Centre, a hot riding in Ontario, who was recently canvassed.
"You know it's kind of scary, in a way, because what are they going to do with all this [information] in the future?"
Analytics is fast becoming an essential tool in Canadian elections. It means the use of big data to micro-target voters in order to get out the vote, recruit volunteers and dig for donations.
The Liberal Party of Canada's software is called Liberalist and the accompanying app is MiniVAN. The Conservatives, long the frontrunners in data collection, have CIMS, with a mobile version called C2G. The New Democrats have rolled out Populus, a smaller version of campaign analytics.
The parties are reluctant to share much detail about their databases. The NDP and Conservative party headquarters turned down our interview requests. The Liberals wanted to assure Canadians its database is not intrusive.
"I think sometimes people get worried about analytics, that it's something scary," says Jeremy Broadhurst, national director for the Liberals.
"This isn't about creating secret files. This is about making sure we understand what voters are talking about and what their concerns are."
But analytics is more than that. It uses algorithms to segment a constituency into bytes: who firmly supports a given party, who is wavering and who is solidly with the opposition. It also helps track campaign volunteers.
'I really believe in a ground war'
In the upcoming federal election, Liberal candidate Salma Zahid is trying to wrestle the riding of Scarborough Centre away from the Conservative incumbent, Roxanne James. Zahid has been canvassing door to door at a breakneck pace since January.
"I am a daughter of an army officer and I really believe in a ground war. We have to get our message across to each and every door," Zahid says.
Trailing her is volunteer assistant Jeff Jedras, armed with an iPad and the MiniVAN app. At every door, he keys in who's in support, undecided or against the Liberals; their preferred language; the issues most important to each voter; and whether they'll take a lawn sign or agree to volunteer.
Each night, that data is uploaded to a server.
"I feel confident it goes through safe hands," Zahid says, "Jeff has the Liberalist [software] – he tracks it."
Plotting a party's support isn't new – that's been going on for generations. It's the amount of data nowadays that's remarkable, and how it's manipulated to profile voters.
It's a surveillance system. It's a monitoring system. The question is whether it's a good monitoring system.- Colin Bennett, University of Victoria
"It's a surveillance system. It's a monitoring system. The question is whether it's a good monitoring system," says Colin Bennett, a University of Victoria professor who tracks privacy issues.
"What is the appropriate balance between the capture of data for good, public purposes and the capture of data for purposes, which many people may find creepy?"
In addition to door-to-door conversations, parties look at electoral lists, the census, past donations, polling and increasingly your social media trail.
They ingest all that into databases to make predictions on what you care about – and how persuadable you are.
"We don't change the policies, but we might change what we emphasize at the doorstep," says Broadhurst.
'It's big data'
While canvassing recently in Mississauga Lakeshore, one of Canada's 30 new ridings, Conservative incumbent Stella Ambler met Alan, a homeowner at his front door. Alan is friendly but not currently supporting the Conservatives, saying he doesn't like their record on the environment.
Ambler politely listens, hoping he is a voter she can persuade in the future. The information he gives her is documented by Ambler's patient campaign assistant.
"It goes into a database, and we'll be able to generate lists," Ambler explains, leaving Alan's doorstep.
"Alan here will be on the list of people who care about the environment. So let's say the prime minister makes an announcement about the environment. We'll make sure we tell Alan."
When asked about the scale of data collection, she pauses.
"I guess if you look at it that way - it's big data, it's sophisticated, but it's also very simple," she adds.
This kind of micro-targeting of voters began with the U.S. Republican campaign in 2004. But the use of big data kicked into high gear during Barack Obama's first U.S. presidential election campaign in 2008, which added data accumulated from social media.
Fast-forward seven years and analytics is chewing up more of campaign budgets in the U.S. and in Canada.
The federal Liberals, for example, are tripling their spending on it over the last election. Investing in data also allows parties to target political advertising spending more precisely.
Tailoring the message
New models of data mining in the U.S. show just how useful voter information can be to a political campaign.
U.S. privacy laws allow Americans to gather more data on individuals than in Canada. They can get income, number of homes, kinds of vacations, number of children and education history. They can also buy other consumer demographic data in order to predict voters' leanings.
Last year, Campaign Grid, one of Washington, D.C.'s top digital advertising companies, used data mining to run a campaign for U.S. clients in favour of changing state laws to decriminalize marijuana.
"What we looked for was people who were supportive of decriminalization and also people who were unlikely to vote," says Campaign Grid's president, Jordan Lieberman.
Using cookies on computers, Campaign Grid's client sent messages to that untapped group encouraging them to get out and vote.
"This isn't about creating secret files. This is about making sure we understand what voters are talking about." - Jeremy Broadhurst, national director, Liberal Party of Canada
The campaign was successful in two out of three states.
Given developments in technology and data mining, Lieberman says in the near future, political campaigners will be able to send messages directly to voters' television sets and even customize them.
"The more segmented the audience, the more tailored your message can be," says Lieberman. "We can run three different sets of ads, all at the same time, into different houses across the [same] street."
Campaign Grid has worked with Canadian political clients, although it won't reveal which ones.
Suspicions of spying
Big data can evoke suspicions that parties are spying on voters, and no one likes that.
"We don't have any rights to know what political parties are collecting on us," says the University of Victoria's Bennett.
"They're not subject to any of our privacy laws and the privacy commissioners can't get inside the databases that political parties hold," he warns.
"I'd be very surprised at some point in the future if there wasn't a data breach."
Scarborough, Ont. resident Jill Mills echoes that concern.
"I mean, what are they going to do with [the data they collect]? Will they sell it to someone else? They tell you it's secure, but a lot of websites that say they are secure aren't."
The Liberals' Broadhurst maintains it's the parties' responsibility to regulate the security of information.
"We take very high industry standards in terms of how we limit access to any kind of information, and ultimately if a party doesn't do that and isn't responsive, then they'll pay a price for that."