U.S. Elections And Politics: Four Questions For The Data Miners

Stephanie Overby

Data analysis has been part of political campaigns for more than a century. As early as 1891, the chairman of the Republican National Committee spent two years painstakingly creating a file featuring the age, occupation, birthplace, residence and other relevant data for eligible voters.

But in-depth analysis of data is a much more recent advance, made possible by more affordable and effective analytics. Microtargeting emerged in 2004. In the 2008 U.S. presidential election, what we now call Big Data took a more central role as processing power and analytics capabilities increased so that by 2012, data-driven campaigning was the norm.

In this election cycle, Big Data is a big deal, and it stands to become even more integral to American politics. We talked to four experts about the biggest advances this cycle, current challenges, voter opinions about data analysis and what lies ahead.

The Experts

Patrick Ruffini
Echelon Insights

Patrick Ruffini is a strategist, focused on data and technology’s impact on politics and business. He was among the first digital strategists in American politics, working for George W. Bush’s 2004 campaign and directing the Republican National Committee’s digital strategy in 2006. Ruffini has led technology efforts in three successive presidential cycles and advised clients around the world. He co-founded the political research and intelligence firm Echelon Insights in 2014 to merge opinion research and data analytics.

Matt Lackey
Vice President of Research and Development
Civis Analytics

Before taking over R&D at Civis Analytics, the consulting and services firm founded by Obama for America 2012’s Chief Analytics Officer Dan Wagner, Matt Lackey developed an analytics platform to help state and local campaigns make data-driven decisions as senior political strategist at the AFL-CIO. 

Elea McDonnell Feit
Drexel University’s LeBow College of Business

Elea McDonnell Feit’s research focuses on leveraging customer data to make better product design and advertising decisions, particularly when data is incomplete, unmatched or aggregated. Much of her career has focused on developing new quantitative methods and bringing them into practice, first working in product design at General Motors and most recently as the executive director of the Wharton Customer Analytics Initiative.

Richard Skinner
Policy Analyst
The Sunlight Foundation

Before joining the Sunlight Foundation, Richard Skinner served as a research analyst for the Campaign Finance Institute and authored the book “More Than Money: Interest Group Action in Congressional Elections.” He is an expert on “dark money” and other campaign finance issues.

The Questions

1. What has been the biggest advance in the application of data analytics this election cycle?

Lackey: The biggest advance has been the emergence of technologies that automate the rote responsibilities of analysts [e.g., cleaning up data or creating standardized reports]. This has opened up the door for campaigns to spend more time on data-driven strategy—giving analysts more time to work on hard problems while also supplying campaign leadership with more transparency into the data.

Feit: A lot of the focus has been on polls and predicting election outcomes. But that’s not where the biggest impacts can happen. Some of the campaigns are working on how to measure how different types of marketing communication affect voters. That’s the type of analytics that will be super actionable. They’re making baby steps toward understanding, but in some ways they’re probably ahead of the private sector. The political folks have more pressure on them to move the needle forward with these messages. People who are selling diapers are going to be selling diapers this month and next month, and if the problem takes a little while to solve, it’s OK. The way election cycles work, campaigns have to figure this out now.

Ruffini: In the last election cycle we saw the impact of analytics on television ad buying. The advances now are related to applying analytics to the digital world—scraping the Web to find out more about individual voters. The rise of unstructured data allows us to learn more about voters—what they like on Facebook, who they follow on Twitter, what they talk about in social networks.

Skinner: [Campaigns] are in a much better position to target voters in a more precise way and use a lot more different types of data—targeting though digital media and figuring out what types of digital content voters are likely to consume.

2. What are the biggest challenges campaigns face in implementing analytics?

Ruffini: There’s so much data out there and so many ways to analyze it, being a good editor of data is very important. Otherwise there’s the risk of analysis paralysis. More people have skills with statistical software and can draw out correlations. But first you have to be focused on what the key questions are that you need to answer, what hypothesis needs to be proven true for the candidate to win. You need to set aside all the other stuff that is not going to impact a win or a loss.

Feit: In the private sector, companies have developed long-term customer databases so that they can use those assets going forward. The most effective data is collected over years, not weeks. Presidential campaigns happen once every four years, and there are different people running the campaigns every time. The parties need to do a better job of passing that voter data on from one campaign to the next.

Lackey: The incentives for campaigns are all upside down. Since campaigns end after an election, the campaign organization doesn’t gain anything from research that can’t be applied this cycle. When this is coupled with the infrequency of elections, you don’t have the opportunity for continuous measurement.

Skinner:  I think there are going to be a lot of campaigns this fall—particularly Republican incumbents—that are going to be trying to figure out what voters might be open to splitting their tickets and what subgroups of Clinton voters they should target.

3. How do voters feel about the use and analysis of their personal data by political parties and candidates? Do they think it benefits them?

Feit: The parties have pretty detailed information about households that they’re beginning to use to target them, but most voters aren’t aware that it’s happening. The campaigns need to get ahead of the issue. In the private sector, when customers all of a sudden become aware that their data is being used in ways they didn’t expect, they get upset and it causes a media storm. Companies that do this best are very transparent with their customers about how they are gathering and using the data.

Lackey: It’s unclear. We know that there was some backlash when the Ted Cruz campaign sent out voter histories, but we also know that volunteers and activists want to know that the time they commit to campaigns is not being wasted on ineffective tactics. The benefits to the electorate are pretty huge. By identifying which citizens will need to register for the first time or re-register, they’re helping to make the electorate more representative of the country. Additionally, as campaigns get more effective at exciting voters about elections, we’ll see higher participation rates.

Skinner: Data mining and voter targeting does seem to have some effect on turning people out to vote. On the other hand, tailoring messages targeted to their specific concerns could further exacerbate the polarized nature of the electorate and the long-term trend of there being fewer and fewer swing voters up for grabs.

Ruffini: If you ask people in a vacuum whether they’re concerned about the fact that there’s all this personal data out there about them, they’ll say yes. But at the end of the day they haven’t changed their behaviors. Nonetheless, we need to be good arbiters of the public trust.

4. What Big Data trend do you expect to emerge in political elections in four years?

Skinner: I think we’ll see much better communication targeting and a general shift of advertising from TV to digital.

Ruffini: So much of any campaign happens in real time, but polls are a reflection of what’s happened days ago. There’s going to be a shift toward more real-time analysis and a much clearer, shorter feedback loop between action and reaction. Traditional research isn’t good at that. If there’s a Trump gaffe or big Hillary news, campaigns need a way to understand and respond to that in real time with more rigor.

Feit: The 360-degree view of the voter. We’re going to see more data points about voters tied together just the way retailers, for example, have tied together records of customer interactions online, via smartphones [and] visits to physical stores. The political parties can start gathering data on voters across many touchpoints: when they voted, when they’re canvassed, when they visited a campaign website, when they donated.

Lackey: Undergraduate programs in most disciplines are beginning to integrate data science and analytics into the curriculum, so we’re going to see an emerging class of workers who are subject-matter experts first and incidentally know how to answer questions in a statistically rigorous way. As the barrier to entry on data and analytics lowers, we’ll see accelerated adoption and learning in ways that we can’t even imagine now.

If you enjoyed this post, you might enjoy One Platform Has Already Won The 2016 Elections: Mobile.

This blog was written through a partnership with Thompson Reuters. To learn how SAP is helping them Run Live, click here.


Stephanie Overby

About Stephanie Overby

A Boston-based journalist, Stephanie Overby has covered everything from Wall Street to weddings during her career. She is currently focused on the implications of digital transformation.