‘Intuition or instincts’ do not assist in winning elections

How should politicians act when gut feelings conflict with data analysis?

“Rely on the analytics,” stated North Carolina Sen. Thom Tillis.

The two-term Republican visited campus last week to help launch Political Analytics 2025. Arranged by the Center for American Political Studies, the one-day event brought together leading minds on voter patterns, election projections, and behavioral studies. Tillis, previously an executive at PricewaterhouseCoopers and IBM, credited a public opinion survey for elevating his legislative career.

He recounted confronting a seemingly unbeatable incumbent in the 2006 Republican primary for a position in the North Carolina House of Representatives. “I hired someone to conduct a poll to assess whether the individual could be defeated — where his weaknesses lay,” Tillis mentioned. “We shifted from targeting someone who was thought to be invincible to defeating him by a two-to-one margin.”

In attendance were pollsters, advisors, and political scholars proficient in big data. “Politics deserves all the sophistication we can apply to it,” stated host Ryan D. Enos, a government professor and director of the Center for American Political Studies. “We can’t rely on instinct or intuition any more than we can leave healthcare to those things.”

Speakers addressed the challenges posed by today’s swiftly evolving political landscape, marked by declining public trust in institutions and a fragmented information ecosystem. “We realize even more the necessity for an analytical framework in politics,” Enos remarked. “This is why we revived the Harvard Political Analytics Conference after a seven-year break.”

In one session, experienced strategists presented findings from research and fieldwork. David Shor, a data specialist and consultant working with left-leaning candidates, countered claims that Democrats should nominate a white male in the upcoming presidential election, asserting that his data indicates women candidates frequently surpass men. Data and behavioral scientist Matt Oczkowski, previously of Cambridge Analytica, forecasted a sort of “civil strife” within the right, as establishment Republicans strive to reclaim the party’s ethos.

Enos and his co-organizers also orchestrated a panel to explore the nation’s youngest electorate. Rachel Janfaza ’20, founder of the qualitative research firm The Up and Up, explained her idea of “two Gen Zs,” separated by the pandemic’s disruption of K-12 education. Concerns regarding AI’s influence on the job market and an emerging gender divide in marriage and family planning were also emphasized.

Several panelists confronted the current state of political polling, noting that support for President Trump has been underestimated in three consecutive election cycles. They discussed whether artificial intelligence could assist pollsters in reaching more independent and Republican voters or if interviewers should revert to more in-person strategies.

Polling prior to last year’s general election was still more precise than historical averages, pointed out CNN chief data analyst Harry Enten. “The issue is,” he noted, “we have experienced more consecutive elections, at least in terms of the popular vote, decided by single digits than at any time since we began tracking the popular vote in 1824.”

Introduced as a potential counterbalance were prediction markets, or online platforms where users can bet on future outcomes. Jaron Zhou ’22 from the prediction market Kalshi noted traders on his site decisively favored President Trump last fall. The Q&A session witnessed attendees raising concerns, including the risk of further diminishing trust in noncommercial, nonpartisan political polls.

The Hispanic vote emerged as another provocative topic, with multiple panelists referencing the 2024 election as evidence of historical realignment. “I wouldn’t be shocked if in 2026, due to the typical nature of midterm elections, many of the gains Republicans achieved with Hispanic voters reverted,” speculated NBC News chief data analyst Steve Kornacki.

However, he likened it to the 1986 midterms when Democrats regained Senate seats in Alabama, Georgia, and North Carolina. It hardly signified a resurgence for the party in the South, Kornacki recalled. “It was a brief moment.”

He suggested that factors such as race, gender, and age were already inducing more enduring changes. “Trends we’ve observed with white voters are beginning to manifest among non-white voters,” Kornacki said, “but specifically, and rather dramatically and immediately, with Hispanic voters.”