Analytics: key to Obama’s victory
Lessons learned from 2012 U.S. presidential campaign sure to play more important role in future elections.
By Douglas A. Samuelson
While pundits and operatives are still chewing over the lessons of the 2012 U.S. presidential campaign, one theme is clearly emerging: analytics, especially microtargeting, played a critical role in both campaigns, and Obama won in significant part because his analytics performance was better. Leading analytics practitioners in both camps advance this claim.
As reported in OR/MS Today four years ago, political campaigns for both parties and at all levels use microtargeting – highly precise polling, marketing data and computerized data analysis – to identify small groups of people who will respond to carefully crafted messages aimed at their particular interests. OR/MS Today also noted the increased use of social media, especially by the Obama campaign and administration, to gather information about voters’ preferences and to send carefully targeted messages to them. The proliferation of data sources, the expanding capabilities of computers and social media now make it possible to take analysis and messaging literally to the individual level.
The essential activities are “persuadables targeting” and “get out the vote (GOTV) targeting.” Persuadables targeting consists of identifying uncommitted or weakly committed voters and determining what appeals, policy positions, personal appearances, walk-through canvassing or other actions might win their votes. GOTV focuses on likely supporters who might not vote, reminding them to vote, sometimes offering rides to the polls, and other appeals to increase turnout among these people.
Another new feature of campaigns, especially the Democrats’, is the use of social media to deliver precisely targeted messages. The Obama campaign made extensive use of social networks, identifying and recruiting people to send messages to people like themselves, while the Romney campaign relied more heavily on TV. As also reported in OR/MS Today, this was part of a coordinated effort that began with the 2008 campaign and continued as an aspect of policy development. These efforts, in turn, built up the data and networks the 2012 campaign then used.
Refining the Methods
Strategic Telemetry did Obama’s microtargeting in 2008, along with a number of state and local races in the past decade. The Obama campaign brought the effort in-house in 2012, continuing and expanding the earlier work. Andrew Drechsler, Strategic Telemetry’s vice president for strategic services, explains, “We just did more of the same [from 2008], being smart about handling data sets and being thorough. The big advancements are computing power and speed. What we have now blows away what we had in ‘08, and we had the top of the line we could afford then. Now we can run models longer and can tweak them more to make them better and more efficient.
“Also, we can turn models around in as little as 24 hours,” he adds. “We couldn’t do that four years ago. The normal turnaround time in 2008 was around two weeks. In 2012 we could turn them around in one week, some faster. And we can run 18,000 or 19,000 combinations of factors, compared to more like 1,000 in the past.”
So what were the keys to victory? “We did better at making the campaign on a personal level: offices, recruiting and utilizing volunteers,” Drechsler says. “Credit Obama for America and the Democratic Party for building a strong grass-roots effort.”
The counterpart on the Republican side was Target Point Consulting, which has supported many successful campaigns in the past decade, notably the 2004 presidential campaign and a number of Senate and congressional races. Alex Lundry, vice president and director of research, acknowledges, “We weren’t very satisfied this year, but it was a great year for analysis, analytics and big data. We actually had many of the same data resources and tools. They [the Democrats] had an enormous advantage, though, in being able to put their organization together two years in advance, preparing for one election. We had primaries and a contested nomination, so we had six months to do what they had two years to do.”
Lundry’s high opinion of microtargeting on the Republican side is supported by a number of successes by Target Point in past years, including some messages about healthcare and education to African-American voters in Ohio and some messages about education to middle-aged Hispanic women in New Mexico that helped President Bush narrowly carry those two states in his 2004 re-election campaign.
Another key to the result was a failure that also, ironically, demonstrated the power of analytics. The Republicans had a computer program called ORCA to develop schedules and instructions for large numbers of volunteers in their Election-Day GOTV efforts. They apparently did not test this software and the computer network at production scale. Sources close to the campaign report that the actual Election Day volume of messages and requests was so overwhelming that it resembled a denial-of-service attack, causing the servers to crash. In fact, the hacker group “Anonymous” claimed to have carried out a denial-of-service attack, but a Republican source denied this, assessing that it was simply a failure induced by the huge difference in volume from development to production mode. Whatever the explanation – and this reporter found no inside source who was willing to comment on the record or to elaborate even on background – it is clear that this analytics software was important and might well have made a difference if it and the resulting human activities had worked as intended.
Changing the Balance?
“This election really makes me wonder, under what circumstances will we see an incumbent lose?” Lundry adds. “Barring an economic crisis, war or scandal, the power of incumbency is significant and extraordinary. It takes a huge technological investment – time, money and talent.”
Ah, yes, money. What about the huge increase in campaign spending and the changes in how funds can be raised and spent? Lundry is unimpressed. “Even with this billion-dollar campaign, we spent more money marketing Halloween candy in 2012 than we did marketing the presidential candidates,” he points out. “We always have both sides hitting swing voters. Both sides use focus groups and panels. Now we have to track and monitor what the Super-PACs are doing, but it’s always been a dense information environment. Even for an issue like gun control in 2013, it’s not all that different – you always have well-funded institutions making their case. It’s always a data and analytics arms race.
“What is really important is that we’re in the world of big data. There does need to be a more meaningful effort put into data literacy. What questions can be asked? What answers can you get? How do you interpret those answers? What makes these actionable? It comes down to training, both for people at the senior level and for people coming up. How do you be a smart consumer?”
Lundry mentions several organizations that do such training: the New Organizing Institute on the left, the Leadership Institute on the right, the Republican National Committee’s “campaign school.”
“The right can improve in data-related training programs,” he says. “They [the left] have a class of operatives and volunteers who are ahead of us. But in 2014 there’s no president at the top of the ticket, turnouts are down, the opportunity is there for us.”
Drechsler makes the same point. “In hindsight,” he says, “pursuing the Wisconsin recall when we did (June 2012) was bad strategy. The college students were out. The other side could concentrate their resources on that election. We could likely have won if we’d waited and done the recall in conjunction with the presidential campaign. And [Governor Scott] Walker wouldn’t have had the same amount of money in a presidential year.”
Drechsler notes that, “Nate Silver said there was a 92 percent probability the Republicans would take that North Dakota U.S. Senate seat in 2012, but Heidi Heitkamp won. She had a good message, well targeted. In the state of Washington, [Democrat Jay] Inslee won the race for governor by focusing on persuadables. These races highlight the power of microtargeting. But the Democratic National Committee had made a huge investment, around $25 million, in 2008, and after what we did then, they brought the effort in-house. Ken Strasma [president of Strategic Telemetry] was in the lead, with eight desks in Chicago, and well over 50 internal staff total.”
It remains to be seen whether the Democrats will make a comparable effort and investment in the off year. So, taking these lessons into account, Drechsler concludes, “the Republicans have an off-year advantage right now.”
Another key factor to watch is the extent to which the Obama organization will share data and analytics with state and local Democratic campaigns. In 2009, in Virginia, Creigh Deeds, the Democratic candidate for governor, did not have this help, and he ended up losing badly, with a turnout around half of what the presidential election had generated the year before. This year, without a presidential re-election campaign in prospect, the Obama organization may be more willing to share – and the likely Democratic nominee, Terry McAuliffe, may be in a better position to make the request, as he is much better acquainted with the key people in the national Democratic Party. Sources close to McAuliffe’s campaign have expressed strong optimism about obtaining this cooperation.
Privacy and Propriety Concerns
The collection and use of so much data, on such a detailed level, obviously raises issues about privacy and confidentiality protections. “Taking social media out of the picture, we’re not looking at anything much different from four years ago,” Drechsler responds. “We follow the law: the information is volunteered, and very little information about identifiable individuals goes to the campaign. We show recommendations, not personally identifiable information. Social media present a whole new picture, and we’re just beginning to explore that. But it’s not as easy as you might think to tie on-line profile information back to the voter file. Often, agreements with data sources prohibit giving certain kinds of information to the campaign. So we come up with a handful of scores.” The campaign, that is, must then figure out messages and targeting in more vague terms, without necessarily knowing a specific individual’s concerns and tendencies as well as a highly analytics-based retailer would.
These legal protections still leave untouched the more general concern about the ways in which modern methods may be contributing to the nasty tone of politics at present. The precision and tone – and honesty and decency – of microtargeting-based appeals are pretty much up to those who design them. Political scientist (Cal-Berkeley) and best-selling novelist (“The Ugly American” and “Fail-Safe”) Eugene Burdick described targeting in his first novel, “The Ninth Wave,” in 1956, and described it in more detail in “The 480,” in 1964. The latter title refers to the number of clusters into which the method of that time divided the U. S. electorate. While the methods he depicted, based on California campaigns in the 1950s and 1960s, are primitive compared to the current approaches, the essential idea is the same: identify small, relatively homogeneous sub-groups of potential voters and create computer models to indicate how to move voters your way.
As Burdick vividly illustrated, finding particular groups’ hates and fears, and playing on them, can be part of the package. In “The Ninth Wave,” winning by playing on hate and fear leads to more of the same. Over time, this can corrupt the politicians and analysts who use these tactics and diminish people’s trust in government, undermining the winners’ ability to govern.
Political scientist and quantitative historian Allan Lichtman argues that such tactics are not only destructive, but also ineffective. Discussing the 2012 election in advance, citing his “13 Keys” model and its implications, he stated – again – that this model, the most accurate yet developed, indicates that campaign factors actually do not seem to influence outcomes much, at least at the presidential popular vote level. Still, he concedes that these findings do not apply to state and local races, nor perhaps even to the close contests for the electoral votes of a few key states. Nevertheless, he did accurately predict the voter burnout, induced by over-saturation with political messages, widely reported from the 2012 campaign, and the resulting decline in the two parties’ willingness and ability to work together to make decisions.
Target Point, however, sees it differently: a way to reverse the trend toward TV-centered campaigns and back toward person-to-person contact, “grass-roots” politics. Certainly the 2012 results can be read both ways, which is one of the many things that make this subject so interesting.
Expanding computational capabilities and related developments in software and analysis have greatly increased the power and utility of analytics methods in political campaigning. Highly placed operatives from both parties agree that these methods made a critical difference in recent elections, in particular Obama’s victory in 2012, and that analytics will play an even more important role in elections to come.
Douglas A. Samuelson (firstname.lastname@example.org), a contributing editor of OR/MS Today, is president and chief scientist of InfoLogix, Inc, a small R&D and consulting company in Annandale, Va., and senior statistical subject matter expert for Great-Circle Technologies, Chantilly, Va., another small analytics company supporting national security applications. Earlier in his career, he worked professionally for several political campaigns, including some voter targeting work in the 1970s.
- http://en.wikipedia.org/wiki/Microtargeting http://www.newpolitics.net/node/395
- “Microtargeting 101” link on Target Point Consulting website
- Eugene Burdick, “The Ninth Wave,” Houghton Mifflin, 1956.
- Eugene Burdick, “The 480,” McGraw-Hill, 1964.
- Allan Lichtman, “The Keys to the White House: A Surefire Guide to Predicting the Next President,” 2012 Edition, Rowman & Littlefield Publishers, Lanham, Md., 2012.
- Douglas A. Samuelson, “Big Brother’s Campaign is Watching You: How Political Campaigns Target Us and What That May Be Doing to Governance,” OR/MS Today, October 2008
- Douglas A. Samuelson, “Change We Can Blog In: Obama’s Use of On-line Social Networking to Help Him Govern,” OR/MS Today, February 2009.
- Douglas A. Samuelson, “Election 2012 Update: The ‘13 Keys’ to the White House,” Analytics, September/October 2012.
- Ken Strasma, “Targeting the Most Unusual Electorate in America: How Innovative Targeting Helped Win a Record Third Term for New York City’s Mayor,” Campaigns and Elections, Feb. 1, 2010.
- Elizabeth Svoboda, “Digital Exposure,” Discover Magazine, November 2009.