Expert Commentary

How management by the numbers is changing American election campaigns: Research brief

2014 research brief on the impact and implications of “computational management” of election campaigns, by Daniel Kreiss of UNC Chapel Hill and the Scholars Strategy Network.

(Pixabay/public domain)

From the Scholars Strategy Network, written by Daniel Kreiss, University of North Carolina, Chapel Hill

 

Recently, a new approach I call “computational management” has become dominant in U.S. electoral politics. Campaign directors now use detailed, ongoing data analysis to make decisions about everything from contacting voters and scheduling candidate events to the content of advertising and the focus of social media outreach. Effective campaign offices are stuffed with computer people crunching numbers, as all kinds of decisions about the allocation of resources come to be guided by regularly updated data collection and analysis.

My research probes these practices, and finds that the rise of computational management has cross-cutting implications for U.S. electoral democracy. On the one hand, campaigns use data to boost participation by volunteers and voters. On the other, data-driven decisions are highly focused on specific, immediate results, and they raise many concerns about privacy and the segmenting of the electorate by levels of previous political engagement. Campaign managers face many pressures to shape the outcome of contests by mobilizing ever-smaller and more closely targeted subsets of voters.

The rise of computational management in election campaigns

Over the last two decades, election campaigns and political parties have become increasingly reliant on digital media.

  • During the presidential cycle of 1999 to 2000, the primary campaigns of insurgent candidates Bill Bradley and John McCain began leveraging the Internet for fundraising and volunteer mobilization. New Internet techniques were deployed alongside a revival of old-fashioned shoe leather approaches in which volunteers went door-to-door to contact voters on behalf of their favored candidates.
  • Over the next three presidential election cycles, data analysis quickly became central for both digital and field efforts. Campaigns experimented with messages and collected and analyzed data about citizen responses to different appeals — including messages sent to email inboxes and delivered by mail to household doorsteps. Such data helped staffers determine precisely which messages and mediums furthered fundraising and turnout goals.
  • In 2004, the re-election bid of President George W. Bush used micro-targeting of particular slices of potentially supportive voters to tie together digital and field operations in innovative ways.
  • After 2004, Democrats leapt ahead in the digital race. Staffers on the 2008 and 2012 Obama campaigns made key decisions and structured voter contacts around rigorous data and analytic practices. They relied on sophisticated national voter databases developed by the Democratic Party and independent firms such as Catalist, and deployed technology platforms to harvest regular online behavioral data from supporters. Internally, these campaigns used computational management to make staffing and budgetary decisions; and externally they used it to stimulate potential supporters to sign up on email lists, make donations, and become volunteers.
  • In President Obama’s 2012 re-election campaign, further refinements occurred. Campaign analysts gleaned hundreds of data points from public, commercial, and behavioral sources on every member of the electorate, and used that information to model whether individuals were likely supporters of Obama and determine whether they were likely to turn out to vote or were open to persuasion by various sorts of appeals.

What number-crunching campaigns can — and cannot — do

Data-based computational approaches work because U.S. election campaigns need money, volunteers, and, at the end of the day, more votes than their rivals. Better information more thoroughly and quickly analyzed can serve these ends. Campaigns profile supporters to learn the appeals that will encourage them to open their wallets and donate their time; and they use data to allocate staff and volunteer time to contact supporters and persuadable voters efficiently.

Because data-rich practices have led to more contacts with voters through every medium of communication, they have probably helped to boost participation. That said, it is worth noting the relatively modest changes brought by the new approaches. The 2012 Obama campaign captured the imaginations of those hoping, or fearing, that data would transform elections, but the reality was not revolutionary. Many Americans have a tendency to sit out elections. Even in the highly contested 2008 presidential election, only 63% of the electorate turned out to vote. The steady advance of computational management may have held back forces pushing toward voter disengagement, but during the 2012 campaign it did not prevent turnout from falling back below 60%. Going forward, it is highly unlikely that campaigns will have the staff or volunteer resources to craft and deliver appeals sufficient to overcome widespread voter disengagement. What is more, the best data in the world cannot enable campaign managers to turn firm Republicans into Democrats or vice versa. At most, this approach helps campaigns improve in-person contacts and craft meaningful targeted appeals. With these techniques, campaigns can mobilize low propensity voters and perhaps persuade some of those not yet decided.

Along with these real accomplishments come heightened concerns about voter privacy and the security of the new datasets now traded on the open and unregulated political market. Campaigns are not transparent about how they gather, store, and act upon voter data. As recent outcries over experiments on Facebook and other social media platforms suggests, the possibility of manipulating votes is not outside the realm of imagination. What is more, campaigns are likely to keep using data analysis to fight over the 60% of the electorate already engaging in regular voting, ignoring the rest. Data mining will likely matter most in low-salience primaries and general contests with small turnouts, where campaigns and activist groups can swing the outcome by targeting small slivers of voters. Clearly, data-rich election management holds both modest promise and possible downsides for U.S. electoral democracy.

Related research: Kreiss is also the author of a related work, Taking Our Country Back: The Crafting of Networked Politics from Howard Dean to Barack Obama, published by Oxford University Press in 2012.

 

The author is a member of the Scholars Strategy Network, where this post originally appeared.

Keywords: election, presidential politics, campaign issues, research brief, presidency