Presumptive Republican presidential nominee Donald Trump recently argued that data profiling of prospective voters was “overrated.” He said he plans only “limited” use of data in his quest for the presidency, believing voters care more about the candidates. As Trump put it, “Obama got the votes much more so than his data processing machine, and I think the same is true with me.”
Meanwhile, Hillary Clinton’s campaign has hired a number of former Obama 2012 data and analytics staffers. She is relying on a network of Democratic organizations to coordinate her technology, data and analytics work.
These contrasting approaches to political data and the technology and expertise that support it offer an experiment. They make 2016 a proving ground, of sorts, for political uses of data and analytics.
Understanding political data
First, it’s important to understand the significance of political data to contemporary campaigning.
To motivate supporters, convince undecided voters and counteract efforts of opposing campaigns, politicians must seek to understand the voters themselves. Data help draw portraits of individuals and groups, helping to better aim campaign attention, communication and spending.
As I document in my new book, campaigns use many different types of data. Electoral strategists are most interested in public records. These contain voter registration, party affiliation, turnout history and demographic data. The parties then supplement this with commercial and consumer records, and information gathered through surveys, to get a fuller view of the electorate. Parties also keep track of what they learn when talking to voters during door-to-door or street canvassing.
While Trump is indeed correct that many variables shape electoral outcomes, data can make campaigns more efficient at identifying the right voters to contact and how to contact them, as well as what to say.
The 2012 Obama campaign used data and analytics to guide electoral strategy, improve the efficiency of voter contacts and assess the comparative results of various communications strategies. For example, the campaign conducted polls asking individuals about their political preferences, matched respondents to identifiable individuals in the voter file, and thenlooked for patterns in the data.
These patterns described detailed probabilities for individual voters, going well beyond showing whether someone with certain characteristics was likely to support Obama. They also predicted how likely a voter was to turn out to vote, how persuadable specific groups of voters were and how responsive they would be to particular campaign messages. Guided by the data, the Obama campaign was able to target groups of likely supporters with messages aimed at their concerns and interests. Importantly, the campaign used repeated polls and on-the-ground voter contact to continually update the models throughout the election season.
Data-driven experiments with public opinion
Beyond voter-modeling data, media analytics involves analyzing data on television viewership to purchase targeted advertising more efficiently, appealing to small groups of voters when they are tuned in.
In online communications, campaigns continually run experiments, too. They test the effectiveness of email and website content, including details like formatting and color. The goal is to increase the probability that users will take the actions that campaigns want them to – such as donating or volunteering.
To take but one example, on the digital side, testing resulted in the odd, and much lampooned, emails with subject lines such as “Hey.” Through experimental tests, Obama campaign staffers knew that this content and format outperformed other, flashier, or more polished material. They estimated that testing and optimization of email netted the campaign an additional US$100 million in donations.
The people behind the technology
It takes time, money and expertise to gather and store all this information. Campaigns also must keep it up-to-date – as voters move or die – and ensure that it’s available to many staffers, from street-level people handing out flyers to the very top strategists. All those people also need to be trained on how to use all the information that’s available.
The two parties have addressed these problems very differently. Since the 2004 election, the Democratic Party has invested considerably more than the Republicans. Recent Democratic presidential campaigns have also hired more data-related people than their Republican counterparts – 507 Democratic staffers in the areas of technology, digital, data and analytics compared to 123 Republican staffers.
And, after these campaigns ended, Democratic staffers founded many more organizations to carry the campaign-developed innovations across election cycles and to races for lower offices: 67 Democratic firms and organizations compared to 15 for the Republicans. That has built a robust extended network dedicated to producing technologies, expertise and a culture that values data-driven decision making.
One example is BlueLabs, cofounded by several Obama data and analytics veterans. The firm brings data analysis methods pioneered during the campaign to lower-level races and nonprofit clients. Company co-founder Elan Kriegel, a senior modeling analyst at the Democratic Party during the 2010 cycle and the battleground states analytics director for Obama 2012, also serves as the director of analytics for Hillary Clinton’s 2016 run.
The contrast in 2016
We may often think of election cycles happening separately, years apart. However, the campaigning capacities that parties and their candidates have are the result of work done many years before any one cycle began.
All that Democratic investment – including from the party’s allied organizations – trained key Clinton staff, laid the foundation for their data workand improved the party’s prospects at the state level.
Trump’s organization, meanwhile, only recently hired its first pollster. He has invested little in data or analytics. And he lags Clinton in tracking data on donors – both of large dollar amounts and small, grassroots contributions.
Heading into the general election, we should expect these trends to continue. Some attacks on Hillary Clinton might come from super PACs, but in general their spending is focused on broadcast advertising. Staffers working in data and analytics told me during my research that super PACs don’t have the data-intensive practices of campaigns and parties, nor the mobilizing power of a candidate. As a result, super PACs are less efficient at targeting voters and delivering messages, which likely results in less effective persuasion and turnout efforts.
Beyond this election cycle
The long-term consequences of Trump’s run may not be limited to the speculation about the Republican Party’s breakup or realignment. As I have recently argued, Trump’s lack of investment in data, analytics and technology will mean comparatively fewer GOP staffers with high-level experience who can work on future campaigns. It will also mean less data produced from field organizing to be available for future candidates.
There will be fewer GOP experiments in persuasion and turnout to inform voter modeling, and fewer Republican-leaning firms launched after the election to gather and analyze more data. In addition, those organizations that do handle data for the Republican Party will miss out on large contracts with the nominee. As a result, they will be less able to expand their experience in ways that would be useful in the future.
Trump’s disavowal of data may have much longer-lasting consequences for the Republican Party’s ability to contest elections during future cycles than his actual candidacy. At the same time, Clinton’s embrace of a technology and data-forward strategy will continue to build on the Democratic Party’s infrastructure, which increasingly offers competitive advantages in contemporary electioneering.