Home > United Nations Online Network in Public Administration and Finance (UNPAN)
1. Global
Global
2. Africa
Africa
3. Arab States
Arab States
4. Asia & Pacific
Asia & Pacific
5. Europe
Europe
6. Latin America & Caribbean
Latin America & Caribbean
7. North America
North America
UNPAN North America
Public Administration News  
Share
How Big Data Saved the Democrats: What U.S. Elections Are Teaching Canadian Politicians
Source: ottawacitizen.com
Source Date: Saturday, October 18, 2014
Focus: Knowledge Management in Government
Country: Canada
Created: Oct 20, 2014

For U.S. Democrats, the power of data to target individual voters became painfully clear when Republican strategists put a blue chip senator from Georgia in their crosshairs.

Max Cleland was a Vietnam veteran and a triple amputee who had served two terms in the Senate. Even in Republican-friendly Georgia, Cleland’s military service seemed to make him a secure Democrat candidate heading in the 2002 midterms. He opened his re-election campaign with a 22-percentage-point lead in the polls over Republican challenger, Saxby Chambliss.

The election came just a year after the 9/11 attacks on the U.S., so the Chambliss campaign used controversial TV ads featuring imagery of Osama bin Laden and Iraqi strongman Saddam Hussein to suggest that Cleland was soft on terrorism. The ads drew condemnation for their tone but narrowed Cleland’s lead substantially.

Then Chambliss got a ballot-box bump from the concurrent gubernatorial race in Georgia. Republican strategists had identified a group of voters — mostly older white men — who had not voted much in the past but who had strong feelings about preserving the design of the Georgia state flag. At the time, there was debate about removing the X-shaped Confederate battle flag design from the flag.

Using direct mail and phone calls highlighting the flag issue, the Republicans successfully rallied these voters to vote for the Republican candidate for governor, Sonny Perdue, who favoured a referendum on the flag. And once in the polling booth, enough of these voters backed the Republican ticket that they also chose Chambliss for the U.S. Senate. He won by 139,000 votes over Cleland.

It was the first compelling example of using what would later be called “micro-targeting” of voters to change the outcome of an election.

For Democrats, the disaster in Georgia was a wake-up call and it triggered a ground-up rethink of their campaign strategy that, 12 years later, is now resonating in Canada as the federal Liberals and NDP play catch up with the Conservatives.

By the time of the 2002 midterms, the Republicans were making dexterous use of voter data to outperform the Democrats in fundraising by extracting small contributions of money from a larger number of donors.

The Democrats, who had tended to rely on larger donations from a smaller pool of donors, were seriously outgunned. And their prospects were getting worse: that same year, campaign finance reforms were implemented, outlawing the contribution of so-called “soft money” that had brought in large sums to the party. Donations were capped at $2,000 per donor.

Now the Democrats faced an existential crisis, cash-strapped and beaten in even traditional strongholds such as Georgia. The new chairman of the Democratic National Committee, Terry McAuliffe, asked a former Al Gore aide named Laura Quinn to figure out what needed to be done.

Her “gap analysis” for the Democrats identified a glaring problem: the party’s efforts to track voters electronically was in a shambles.

Valuable data from fundraising, volunteers, polling and GOTV (get-out-the-vote) campaigns were scattered in various silos within the party. The party’s database of email addresses could be measured only in the single thousands.

Worse, Democratic campaigns across the country made millions of voter-contact calls or door-knocks every election cycle, but the crucial data harvested from these efforts were invariably discarded at the end of the campaign.

Hard drives full of useful information about voters languished in the backs of filing cabinets. Most campaigns had to spin up their voters’ list each cycle from scratch.

Quinn helped organize the data into a central repository, allowing progressive campaigns across the country to draw on and update it as they made their outreach to voters. The more campaigns that used the data, the more robust it became. More important, the data were retained and updated with each election, beginning in 2006.

Quinn turned this database into a business, Catalist, which today maintains a vast database of voter profiles. The company is a run as a trust and provides its services only to progressive campaigns.

In Canada, political conservatives were paying close attention to U.S. campaign tactics. As early as 2004, the Canadian right was attempting to emulate the Republicans’ success with using databases to raise money and deliver voters.

The federal Conservative party began pairing a new voter-tracking database, the Constituent Information Management System (CIMS), with an aggressive campaign of voter-contact calls made by a phone bank company, Responsive Marketing Group. The sophisticated fundraising and voter-contact effort helped the Tories win victory in 2006.

Almost immediately upon forming the government, the Tories moved to leverage their data advantage further by lowering the limit on political contributions to $1,000 — echoing the campaign finance reforms that had helped U.S. Republicans outmuscle Democrats four years earlier. Using CIMS to solicit more smaller-dollar donations from a large pool of supporters, the Tories realized a clear edge over the Liberals, who relied on bigger cheques from fewer contributors.

The fundraising arms of the Liberals and New Democrats had grown lazy, content to suckle on the steady $1.75 per-vote parliamentary subsidy implemented by the Jean Chrétien government. Neither made any serious effort to match the Tories’ fundraising success with CIMS.

In the U.S., however, the Democrats were scrambling to close the gap on data with the Republicans. With their voter contact databases beginning to unite, the Democrats had twigged to the power of data-tracking.

Former Vermont governor Howard Dean showed in his 2004 bid for the presidential nomination that Democrats could effectively raise large amounts of money through co-ordinated online fundraising campaigns that also helped mobilize voters.

But it was the eventual winner of that race, John Kerry, who gave the Democrats a sniff of the power of predictive analytics to target individuals not just for cash, but for votes.

Led by statistician Ken Strasma and his company Strategic Telemetry, the Kerry campaign identified people in Iowa likely to support the candidate but who didn’t usually participate in the crucial caucuses. By focusing campaign resources on these people, Kerry vaulted from fourth place to win the caucuses and, eventually, the Democratic nomination.

The lesson wasn’t lost on advisers to Illinois Senator Barack Obama, who recruited Strasma to do the same in the 2008 presidential primaries. Facing the well-financed campaigns of the two others front-runners, Hillary Clinton and John Edwards, the Obama campaign built a fundraising machine, leveraged heavily on data, to focus on small-dollar donations solicited online.

Like Kerry’s team, the Obama campaign also effectively targeted voters who were more likely to support Obama but who didn’t have a strong history of participating in the caucuses, then repeatedly contacted them to motivate them. One undecided voter received 100 phone calls from Obama volunteers and when the campaign noticed, Obama himself made the next call.

Based on analysis of their own numbers, the Obama campaign was confident in their models predicting victory. But there was one untested variable that still made them nervous: Iowa voters were mostly rural, mostly white and it was uncertain if they would actually get behind an African-American candidate, despite what they were telling campaign workers on the phone.

The concerns, the outcome showed, were largely misplaced: Obama won Iowa, then the Democratic nomination.

Strasma and his growing team of data scientists began performing the same functions for the national presidential campaign. With its heavy emphasis on data, the campaign became a virtual lab for testing theories of behavioural economics.

Messages to voters were rigorously tested with A-B comparisons: The campaign would send two slightly different versions of emails to supporters asking for donations, then compare the results to see which message performed better.

The same approach of testing and retesting messages was applied to scripts used to make voter-contact calls, to broadcast media advertising, and to online ad campaigns. These experiments also showed that people who had a poor record of voting in the past were more likely to respond to a message that said all their neighbours had cast ballots in the past rather than one that emphasized their civic obligation to vote.

But the big shift between the Kerry and Obama campaigns was moving the campaign’s outreach efforts from geography to individual voters. Where once a campaign would focus efforts on individual precincts — the equivalent of polls in Canada — that were believed to be potentially favourable, now specific voters were the target.

Obama canvassers wouldn’t knock on every door on a street. Instead, guided by data, they would skip addresses less likely to be supportive and focus on those where voters might be persuadable. At homes identified as supportive, the message would be only about remembering to come out to vote.

So effective was the technique in 2008 that Obama brought the data analytics operation in-house for the 2012 campaign, dedicating substantial staff and resources to data on the way to re-election.

The Democrats had leaped from a distant second in the Big Data race to a commanding lead.

Heading into the 2015 election campaign in Canada, the Liberals and New Democrats face a similar challenge. To maximize their chances of winning, they need to match the Tories and their rich data set in CIMS and jump out ahead.

To Laura Quinn, who help rebuild the Democrats’ data muscle, the key is compiling data over multiple campaigns.

”You’ve got to be able to have large universes to prospect to and you’ve got to be able to track what you’re doing in a very rigorous way,” she said.

“And you have to do it longitudinally over time. That’s the only way to be successful.”

The great irony of the new big data techniques pioneered by U.S. campaigns is their capacity to make politics personal again.

For more than 50 years, television advertising has been the dominant medium for persuading voters. But compared to micro-targeting, bulk-advertising buys on network television are blunt instruments. They reach large numbers of voters, at great expense, and everyone gets the same, single message.

But by targeting individual voters with customized messages, a campaign can respond more directly to their issues — an online ad highlighting the candidate’s policy on tuition fees aimed at student-age voters, for example, or a direct-mail blitz aimed at middle age workers over changes to pension plans.

“That was the original way politics worked,” Quinn says.

“You went to people’s doors, your remembered their kids, what they were interested in and what they were mad about.

“This is a little bit full-circle but with a lot of machinery underneath it to keep it organized.”

(BY GLEN MCGREGOR)
News Home

 Tag This
 Tell A Friend
del.icio.us digg this Slashdot
Rate:
0 ratings
Views: 370

Comments: 0 Bookmarked: 0 Tagged: 0



0 Comments | Login to add comment

Site map | FAQs | Terms and Privacy | Contact Us
Copyright 2019 by UNPAN - United Nations Public Administration Network