The Victory Lab

1. The Birth of Political Science Experiments

"Politics is an observational, not an experimental science."

Early skepticism. For decades, political science largely dismissed the idea of controlled experiments, with prominent figures like Woodrow Wilson and Abbott Lawrence Lowell asserting that human relationships and politics were not subjects for scientific experimentation. Scholars typically relied on historical documents and observational studies, analyzing correlations rather than establishing direct cause and effect. This intellectual inertia meant that fundamental questions about voter behavior remained speculative.

Gosnell's pioneering work. Harold Foote Gosnell, a University of Chicago graduate student in the 1920s, challenged this orthodoxy. Inspired by urban politics and social psychology, he conducted the first known field experiments in political science. His 1924 study, "Non-Voting," used random sampling and face-to-face interviews to understand why people didn't vote, and his 1927 follow-up, "Getting Out the Vote," tested the impact of postcards on voter registration and turnout.

• Gosnell's 1924 study cost $5,000.
• He used punch cards and a city hall comptroller's machine for data tabulation.
• His postcards increased registration by 9 percentage points.
• His follow-up cartoon mailer increased turnout by 9 percentage points.

A forgotten legacy. Despite initial acclaim, Gosnell's experimental approach was largely abandoned by political science for half a century. The discipline shifted towards large-scale surveys and statistical regressions, focusing on broad theories of voter behavior rather than the granular, causal insights that experiments could provide. This left a significant gap in understanding what truly motivated individual voters.

2. 2000 Election Ignites Republican Data Revolution

"One good thing about the closeness of the last presidential election is that it erases the need to convince people how important two percent can be."

A wake-up call. The razor-thin 2000 presidential election, decided by just 537 votes in Florida, served as a stark realization for Republicans: their ground game was lagging behind Democrats. Despite George W. Bush's victory, internal analyses revealed that Democratic field operations had effectively tied the race, highlighting a critical tactical gap in mobilizing known supporters. This forced a re-evaluation of traditional campaign strategies.

The 72-Hour Task Force. Karl Rove and Matthew Dowd, key strategists for Bush, initiated the "72-Hour Task Force" to address this deficit. Their goal was to systematically research and develop new best practices for voter turnout in the crucial final days of a campaign. They recognized that relying on anecdotes and traditional methods was no longer sufficient in an era of increasingly polarized and closely contested elections.

• Bush's 2000 campaign lost the popular vote by half a million.
• Democrats showed surprising strength in 1998 midterms due to turnout.
• Republicans had been doing "the same thing over and over again for twenty years."

Embracing experimentation. Dowd, influenced by the work of academics like Alan Gerber and Don Green, advocated for adopting randomized-control experiments to rigorously test voter contact methods. This scientific approach aimed to replace "gut politics" with empirical evidence, convincing skeptical consultants to reallocate resources from mass media to more effective, personalized field operations. The task force's findings would become the foundation for a new era of Republican campaign science.

3. Microtargeting: Pinpointing Voters with Consumer Data

"You mean, you don’t do this in politics?"

A commercial inspiration. Alex Gage, a Republican pollster, observed that commercial marketers used sophisticated data-mining techniques to profile individual consumers and tailor messages. He wondered why politics, which also sought to influence individual behavior, wasn't doing the same. This led him to propose "super-segmentation," later dubbed "MicroTargeting," as a way to move beyond broad demographic or precinct-level targeting.

Merging data for precision. Gage's innovation involved combining publicly available voter registration and history data with vast commercial consumer databases (like Acxiom's), which contained hundreds of variables from marital status to pet ownership. By commissioning large-scale surveys and applying algorithms, he could create "segments" of voters based on their political beliefs and lifestyle traits, regardless of their geography.

• 80% of Michigan voters matched an Acxiom record.
• Acxiom data included hundreds of variables.
• Gage's models could predict issue priorities (e.g., tax plan vs. education pitch).

Early successes and skepticism. Gage's methods showed promise in Michigan and Massachusetts, helping candidates like Mitt Romney identify "subtribes" of persuadable voters. However, the high cost and the radical departure from traditional precinct-based targeting met resistance from established consultants and party officials. Karl Rove, initially skeptical, was eventually convinced by the potential for more efficient resource allocation, leading to a $3 million investment in microtargeting for Bush's 2004 reelection.

4. Challenging Campaign Gurus with Empirical Evidence

"The same people wouldn’t be right from month to month. And that just reinforced my belief that this empirical approach was far better than a guru approach where someone came in and said ‘This is the piece you had to send’ because of some theory they had about the election or the candidates or about how human beings think."

The "guru" problem. The political consulting industry was dominated by "gurus" like Mark Penn and Bob Shrum, who commanded high fees based on reputation and anecdote, often without rigorous proof of their methods' effectiveness. Campaigns, especially those with wealthy candidates, often spent millions on strategies that were never truly tested, leading to a culture of unexamined practices and blame-shifting after losses.

Podhorzer's "insiders poll." Mike Podhorzer, deputy political director of the AFL-CIO, grew frustrated with this lack of accountability. He conducted an "insiders poll" among party operatives, asking them to predict the effectiveness of different mail pieces. The results consistently showed a random distribution of opinions, suggesting that even experienced professionals lacked a reliable understanding of what truly worked.

• Podhorzer's poll showed about one-third of votes for each of three mail pieces.
• This highlighted the lack of consistent, evidence-based decision-making.

The need for rigor. Podhorzer, with his background in statistical analysis and a "Moneyball for progressive politics" mindset, sought to introduce empirical rigor to campaign spending. He recognized that traditional methods like focus groups and standard polls provided limited insight into actual behavioral change. His efforts, including early, albeit unscientific, field experiments, laid the groundwork for a more data-driven approach on the left, challenging the established "guru" culture.

5. Democrats Build a Unified Data Infrastructure

"Losing is a great tonic for internecine bullshit."

Post-2004 urgency. George W. Bush's 2004 reelection, widely attributed to his campaign's microtargeting prowess, spurred Democrats to action. Laura Quinn, a former DNC official, became obsessed with reverse-engineering Rove's methods, recognizing that the left needed to catch up in data and targeting. The defeat fostered a willingness among disparate Democratic groups to collaborate and overcome long-standing rivalries.

Demzilla and Catalist. DNC Chairman Terry McAuliffe initiated "Demzilla," a project to build a national voter file by consolidating state party lists and enriching them with commercial data. Simultaneously, Quinn and Mark Steitz founded Catalist, a data "utility" designed to serve as a shared, continuously updated voter information hub for Democratic campaigns and liberal organizations. Catalist aimed to democratize access to sophisticated data, making it affordable and usable for smaller groups.

• DNC's initial email list had only 70,000 people.
• Demzilla aimed for 100 million new names.
• Catalist built a file of 200 million voting-age adults.
• Catalist now maintains one-half of a petabyte of data.

Overcoming resistance. Both Demzilla and Catalist faced significant political hurdles, including state parties' reluctance to share their prized voter lists and DNC officials' skepticism about external data initiatives. However, the shared goal of winning elections and the promise of enhanced data capabilities eventually prevailed. Catalist's "double bottom line" model, prioritizing electoral success over profit, allowed it to operate across legal boundaries between campaigns and outside groups.

6. Behavioral Science: Nudging Voters Through Psychology

"Changing how people see themselves can change behavior."

Rogers's psychological insights. Todd Rogers, a social psychologist from Harvard Business School, brought a fresh perspective to voter mobilization. He argued that voting was not merely a rational choice but a "self-expressive social behavior," influenced by subtle psychological cues and social norms. Rogers sought to apply behavioral science principles, often tested in commercial or laboratory settings, directly to political campaigns.

Testing "social norms" and "plan-making." Rogers's research, inspired by Robert Cialdini's work on conformity, explored whether framing voting as a popular activity could increase turnout. He also investigated the "plan-making effect," where people are more likely to perform an action if they visualize the steps involved. His early attempts to run experiments in campaigns, like Wesley Clark's 2004 primary, were often thwarted by campaigns' short-term focus.

• Rogers's bus station experiment aimed to increase registration by emphasizing popularity.
• His Kentucky experiment tested the "plan-making effect" through automated calls.
• He found that asking voters when and how they would vote increased turnout.

The Analyst Institute. Rogers's unique blend of academic rigor and practical application led him to become the executive director of the Analyst Institute, a "secret society" of liberal data geeks. The Institute became a hub for designing and sharing randomized field experiments, bridging the gap between academic research and campaign practice. Rogers's work shifted the focus from merely identifying voters to understanding the deeper psychological levers that could motivate them.

7. Social Pressure: The Potent, Controversial Turnout Tool

"Although it still turns out that making them ballistic—boy, is that powerful!"

Grebner's "report cards." Mark Grebner, a Michigan political consultant and data pioneer, independently stumbled upon the power of social pressure. Frustrated by voters lying about participation, he conceived of sending "voting report cards" to entire blocks, publicly listing who voted and who didn't. His 2006 experiment, conducted with Gerber and Green, revealed that this "neighbors" letter was astonishingly effective, tripling turnout compared to traditional mailers.

• Grebner's "neighbors" letter increased turnout by 8.1 percentage points.
• It was roughly three times more effective than any other mailer tested.
• The cost per marginal vote was an enviable two dollars.

Ethical dilemmas and refinement. While powerful, Grebner's method provoked outrage, leading to threats and complaints. This highlighted the ethical tightrope of using shame as a motivator. Hal Malchow, a direct-mail veteran, sought to refine the technique, developing "Promise" mailings (asking voters to pledge to vote) and "Thank You" letters (congratulating past voters). These softer approaches, while less potent than direct shaming, still delivered significant turnout boosts without the backlash.

• Malchow's "Promise" technique produced new votes at $18 each.
• Panagopoulos's "Thank You" letter increased turnout by 2.4 percentage points.
• The "Honor Roll" (listing consistent voters) increased turnout by 2 points.

Widespread adoption. By 2010, social pressure tactics, in various forms, became widespread across both parties. The Analyst Institute actively promoted these methods, and campaigns, recognizing their effectiveness, quickly adopted them. This rapid diffusion, however, also meant that the novelty and psychological impact of such messages would likely diminish over time, pushing innovators like Malchow to seek new frontiers.

8. Obama's 2008 Campaign: A Data-Driven Masterpiece

"It was the overwhelming culture of the campaign."

A new political corporation. The 2008 Obama campaign, under the leadership of David Plouffe, became an unprecedented model of data-driven electioneering. It integrated sophisticated analytics into every aspect of its operation, from voter contact to delegate allocation, fostering a "laboratory attitude" among its staff. This approach allowed the campaign to question traditional assumptions and optimize resource allocation with empirical rigor.

Iowa as a blueprint. The campaign's success in the Iowa caucuses, where it mobilized a record 239,000 participants, served as a blueprint. Ken Strasma's "virtual IDs" and "Caucus Math Tool" used algorithms to predict individual support and turnout, guiding field organizers to target specific voters and even influence delegate realignments. This allowed Obama to overcome initial skepticism and build a coalition of new, young, and minority caucus-goers.

• Obama's Iowa turnout goal was 180,000, but achieved 239,000.
• Half of Iowa caucus-goers were first-time participants.
• The campaign used ACT scores and high school directories to find 17-year-olds.

In-house expertise. Unlike previous campaigns that outsourced data analysis, Obama's team brought its targeting operations in-house. Analysts like Dan Wagner and Ethan Roeder, often campaign neophytes with strong quantitative backgrounds, became integral to daily strategic decisions. This internal capacity allowed for continuous refinement of models and rapid adaptation to changing campaign dynamics.

9. The "Matrix" and Continuous Voter Intelligence

"The presidential campaign was just nothing like anything else that ever exists."

A centralized data hub. The Obama campaign developed the "Matrix," a centralized repository that tracked every interaction with a voter, from mail and phone calls to door knocks and even exposure to broadcast ads. This unprecedented data collection allowed for a continuous feedback loop, enabling analysts to measure the impact of various campaign activities and refine their strategies in real-time.

Dynamic microtargeting. Unlike previous microtargeting efforts that produced static voter scores, Obama's system continuously updated individual support and turnout probabilities. This dynamic approach allowed the campaign to identify "shifters"—voters whose preferences were changing—and tailor messages accordingly. It also provided early insights into shifts in public opinion, often faster than traditional polls.

• The Matrix tracked every "touch" a voter received.
• Strasma's algorithms updated scores weekly.
• The campaign identified "shifters" to refine targeting.

Auditing phone vendors. The campaign's data-driven culture extended to auditing its phone vendors. By embedding "fictive alter egos" with verifiable information (like age) into call lists and tracking cost-per-opinion, the campaign ensured accountability and quality control. This rigorous approach weeded out underperforming firms and improved the reliability of voter ID data, which was crucial for feeding the predictive models.

10. Beyond Broadcast: Optimizing "Lo-Fi" Voter Contact

"Now if I know that there are twenty-seven people I want to reach and they all cluster around this bus bench, I’ll buy that bus bench."

Rethinking media buys. Obama's campaign, despite its technological sophistication, found new value in "lo-fi" and unconventional advertising. Data analysts discovered that mass-transit ridership correlated with Obama support in Wisconsin, leading to targeted bus ads in cities like Milwaukee, Philadelphia, and Akron. This demonstrated a willingness to invest in overlooked media channels if data indicated their effectiveness.

The "Starbucks model" of field. The campaign's ubiquitous storefront offices, dubbed the "Starbucks model," and its emphasis on "motivational organizing" (inspired by Marshall Ganz) combined humanistic outreach with data-driven efficiency. Field organizers used microtargeting scores to draw neighborhood boundaries, ensuring volunteers focused on the most promising voters and tailored their conversations.

• Obama opened nearly 100 field offices in Ohio.
• Field organizers classified precincts as urban, suburban, exurban, or rural.
• Canvassability and callability scores predicted voter responsiveness.

A/B testing and "Jigsaw" projects. The campaign rigorously A/B tested its online content, from website button text ("Learn More" outperformed "Sign Up") to email subject lines, to optimize engagement and fundraising. "Special projects" like Moffo's "Jigsaw" unit explored unconventional youth marketing, including ads in video games and chartering planes with banners, all measured by text-message response rates. This empirical approach led to reallocating a $30 million youth advertising budget away from traditional TV.

11. Unmasking Hidden Biases with "Openness Scores"

"We knew who these people were going to vote for before they decided."

The "Bradley Effect" challenge. A persistent concern for Obama's campaign was the "Bradley effect," where white voters might misrepresent their true preferences to pollsters when a black candidate was on the ballot. While early primary results in white states suggested this wasn't a major issue, later primaries and AFL-CIO robopolls indicated that racial attitudes could still subtly influence voter behavior.

Quantifying prejudice. Ken Strasma's team sought to quantify this hidden bias. They identified voters with high Obama-support scores who nonetheless backed McCain in ID calls. By adding a "neighbors" question (asking if respondents thought their neighbors would vote for a black candidate) to surveys, Strasma developed an "openness score." This score, integrated into the algorithms, helped predict a voter's susceptibility to racial prejudice.

• Obama's support was higher in polls conducted by black callers.
• "Openness scores" helped identify voters likely to be biased against a black candidate.
• Low openness scores could lead to removal from Obama's contact universe.

Strategic adaptation. The openness score allowed the campaign to refine its targeting, avoiding voters who might react negatively to Obama contact and instead focusing on those who could be swayed by economic messages. This statistical fix for a complex social phenomenon demonstrated the campaign's commitment to understanding and adapting to the nuanced realities of the electorate, even when those realities were uncomfortable.

12. The Enduring Quest for Campaign Accountability

"What if elections were simply less about shaping people’s opinions than changing their behaviors?"

Obama's legacy and new challenges. Obama's 2008 victory cemented the triumph of data-driven empiricism in politics, inspiring a global demand for "Obama-style" campaigning. However, the underlying question of how to truly turn a person into a voter remained. The 2010 midterms, which saw a decline in participation from key Democratic constituencies, highlighted the ongoing need for innovation, especially in voter registration.

Innovating voter registration. Hal Malchow, now focused on Women's Voices Women Vote, continued to experiment with registration methods. He tested variations of mail packages, including those using subtle social pressure or emphasizing the historic significance of voting. His work, leveraging PLANET codes for real-time tracking, dramatically reduced the cost of registering new voters, from $15 to $7 per registration.

• Malchow tested 14 variations of registration packages.
• The "birthday" program, emphasizing the historic moment, was highly effective.
• PLANET codes allowed real-time tracking of registration form returns.

Beyond persuasion: behavior change. The shift in focus from persuasion to behavior change, and from opinion shaping to turnout, marked a profound evolution in campaign science. Researchers like Alan Gerber explored the psychological barriers to voting, discovering that many nonvoters simply didn't trust the secrecy of the ballot. Experiments showed that reassuring voters about ballot confidentiality could significantly boost turnout among first-time voters. This ongoing quest for deeper psychological insights continues to push the envelope of what campaigns can achieve.

Last updated: February 21, 2026