The last time chief analytics officer Becca Siegel had been inside Joe Biden’s sprawling campaign headquarters in Center City Philadelphia, her department had about 15 staffers on its payroll. That was in March. In late October, her team had ballooned to about ten times that size, but now only nine of her colleagues could join her in person — from a safe distance, wearing N95 masks, subject to a regular virus-testing protocol — for the final sprint to Election Day. Siegel, who had spent most of the pandemic working from a basement in tiny Manitou Springs, Colorado, had begun preparing for the first Tuesday in November months earlier. “Normally, I’m pretty low-key about it, because when the polls are closed, the votes have been won,” she said, speaking a few days after the AP and the networks had finally declared Biden the winner. “But this one was obviously a different situation. Because of the occupant of the White House, and the way the votes were going to be counted, it was clearly much more complicated.”
Their task was to figure out, in real time, if Biden was on track to win. Others in the Biden orbit, scattered around the country but mainly concentrated in Wilmington, Delaware, and Washington, D.C., were in charge of countering the expected confrontational messaging and legal fights from Trump’s campaign. In Philadelphia, Siegel and her colleagues patched in over 100 co-workers whose assignments covered 17 targeted states and laid out a to-do list of 300 items to get through before polls opened the next week. They’d need to figure out exactly how each county would be reporting its results and also how they should set up their internal vote-total projections, down to the precinct level, based on what kinds of ballots (in-person? Vote by mail?) the local governments would report and when. And then, they’d have to consider how to adjust their expectations and projections throughout Election Night and potentially the next few days too. “We spent a long time thinking about, What does it mean to overperform expectations in vote by mail? Does it mean we will overperform on Election Day too?” Or would it suggest an underperformance? They’d learn the answer when votes were counted. But until then, they had to get ready to navigate an information black hole.
Siegel, who is 28 and joined the campaign at its inception in 2019 after stops in the Democratic data world including Hillary Clinton’s campaign in Ohio, had been briefing Biden’s top aides on his progress every morning since early voting began earlier in the fall. These meetings turned hourly, and then continuous, on Tuesday night. Beaming into the Wilmington war room soon after polling places began closing, she ran the former VP’s inner circle through the vote counts state by state, discussing what kinds of votes were still outstanding, what that likely meant for his chances, and what to expect next. (You could do worse than to think of her, at this point, as the campaign’s private Steve Kornacki, armed with much more data but similarly little sleep.) There were few surprises, and not much to celebrate, when she first dialed in around 7:30 p.m. Biden appeared to be hitting expectations in the highly educated Indianapolis suburbs — which Siegel said was likely a good sign for similar areas in battleground states like Michigan and Pennsylvania — but he was obviously lagging badly in rural, heavily white areas in Florida, where other warning signs were also popping off the map, most obviously in Miami-Dade County, where Trump looked like he was crushing expectations.
The previous evening, Siegel had walked the campaign’s top brass through her team’s final expectations. She had two overriding messages: first, that the count could take a while, and second, that the public would likely not have a full understanding of what was happening based solely on the numbers on TV or the New York Times’ interactive map, so they might have to do a lot of explaining. She said that her team had multiple scenarios mapped out — a good one, an okay one, and a bad one — for each hour, so they could easily track Biden’s pathways to 270 electoral votes depending on how the actual numbers matched up with their projections. And she laid down some markers. For weeks, Siegel had been souring on Florida and had in recent days tried convincing colleagues to lower public expectations for it. “That’s probably not going to happen. If it does, great,” she remembered thinking about the state — if Biden won it, he would almost certainly win the presidency early on Tuesday night. That was the simplest path. But that wasn’t the expectation: Siegel predicted more success not only in the swing states Clinton had won, like New Hampshire, Minnesota, and Nevada, but also in the “blue wall” states that Biden’s team considered the most likely avenue for getting him to 270 electoral votes: Pennsylvania, Michigan, and Wisconsin. If those didn’t work out, he might be able to look to some other states, all of which Trump had also won in 2016, and whose recent history was far more Republican. Arizona, for one, was well within Biden’s grasp, too, she predicted. North Carolina and Georgia were a bit more of a stretch, she said, but wins were also within the realm of possibility there.
But she didn’t need to remind them that the race might still end up being pretty close. Early in the summer, as many Americans got used to their locked-down lives and others effectively ignored the pandemic, Siegel began raising alarms internally about public polling that looked unrealistically friendly to Biden. The analytics team, which worked closely throughout the campaign with Biden’s pollsters, John Anzalone and Celinda Lake, thought it was seeing a pair of dynamics that now — months later — have the polling industry once again questioning its methods. Whenever Biden’s team insisted publicly that the race would be close, it was dismissed as a fundraising tactic or an expectations-management tool. But “our stuff was always much, much more pessimistic than the public stuff,” Siegel said, explaining that her department believed public polls were underrepresenting non-college-educated voters and underappreciating a partisan nonresponse effect during the pandemic. Democrats were more likely to be home and to answer the phone.
In few places was the gulf between internal expectations and the public narrative greater than Florida. “On Election Night, there was some time, for a few hours, where [the race] was obviously far closer than anyone expected,” remembered Jen O’Malley Dillon, Biden’s campaign manager. “I knew the tougher states would come first, and we felt like we were projecting in a way where we were going to be okay.” But as Florida turned more and more in the direction of Trump, O’Malley Dillon called on Siegel to make sense of what was happening.
Instead of trying to find reasons for optimism within Florida itself, Siegel’s team dug into the two areas of obvious underperformance there to gauge whether such results were likely to carry over into other states. “Miami-Dade is not a county that exists elsewhere in America,” Siegel said, referring to the heavily Cuban- and Venezuelan-American population center where Trump ran far ahead of his expectations. “We were like, ‘Okay, well, that’s less of a problem.’” The bigger worry was Biden’s anemic showing in Florida’s more rural zones. “The rural piece was obviously concerning, though that was along the lines of what we’d been seeing the whole time: that [the] white non-college [vote] would be worse than people were saying.” For one thing, it meant that Biden was in serious trouble in North Carolina, too.
In order to make up for that, it was clear Biden would need to maintain his strong performances in suburban areas full of highly educated voters, Siegel figured, turning her and her team’s focus to Wisconsin. They worked out just how much of a buffer he’d need in areas around Madison and Milwaukee to make up for likely brutal results in rural stretches. Before long, they were buoyed by results from just outside Columbus, then in Colorado and Omaha.
Still, it was obvious that without early positive results in Florida, North Carolina, and Georgia — which looked painfully close — they likely wouldn’t have final clarity on the race anytime soon. In Wilmington, Biden’s brain trust decided to delay his Election Night speech, figuring he could no longer reasonably hope to declare victory around 11 p.m. Around the country, his data staff redoubled its efforts to figure out when to expect more results from its targeted areas, not only monitoring local news outlets in Georgia for updates but even refreshing the personal Twitter page of a government clerk in Yavapai County, Arizona. When, at 11:20 p.m., Fox called Arizona for Biden, many of his top aides rejoiced and exhaled, seeing no path left for Trump. In Philadelphia, however, Siegel was surprised. She didn’t disagree with the call — she felt comfortable about the state’s trajectory — but it was still looking close. Still, she figured, she’d take the good news. (In Washington, meanwhile, Trump’s top aides exploded and tried pressuring Fox to reverse course; over the next few days, Siegel herself would get on the line with networks’ decision-desk teams to make sure they weren’t wavering on their state calls.)
Siegel had long thought of her role on Election Night as trying to bring some clarity, and calm, to Biden’s brain trust. But as Tuesday night gave way to Wednesday morning, after Trump emerged at the White House to absurdly declare victory, she decided to try and get some sleep.
Things were looking pretty good in Arizona and Michigan, and Pennsylvania appeared to be on the right track, but it was too soon for the number-crunchers to relax. For that, they needed to be confident that Biden was on track in his simplest path toward 270 electoral votes, and for that they needed all three Rust Belt states trending his way. Around 3 a.m., she rang a deputy posted on the West Coast and said she was going to close her eyes. But, she said, she’d only be gone briefly — and, if it happens, call me the second it’s clear we’re going to win Wisconsin. I’ll wake up, she promised.
When Siegel joined Bidenworld, it was a delicate time for the political-data world, between ongoing recriminations over 2016’s disaster, the Democratic National Committee’s messy reinvestment in its own program, and Republicans’ continued insistence that they had somehow mastered some dark voter-targeting arts (and that they would never relinquish their advantage). For a while, it was also a difficult time for Biden. Though he consistently led the Democratic field and Trump in national polls, a much grimmer landscape met him in the earliest-voting states, where Siegel — who had been in the campaign’s Des Moines boiler room and was one of the first to recognize the Iowa-caucus reporting system was melting down — had to help make the case that he wasn’t yet doomed. After his fourth- and fifth-place finishes in Iowa and New Hampshire, she pointed out to his communications staff, which was desperate for talking points, that 99.9 percent of Black and 99.8 percent of Latino voters had yet to weigh in on the primary. After New Hampshire, she dug through the vote results to find a slight overperformance from Biden compared to Elizabeth Warren in the towns on the border with Massachusetts, which helped Biden’s aides make the case that he was actually in a fine position for Super Tuesday, when the senator’s home state would vote. (He wound up — shockingly — winning Massachusetts. Warren dropped out two days later.)
Siegel was hardly playing the stereotypical role of a campaign’s data chief, especially compared to the popular bros-in-the-data-cave mythology of the 2008 and 2012 Obama operations. One of the highest-ranking women on the campaign from the start, Siegel found herself working closely with O’Malley Dillon, who took over the campaign in March, in a prominent strategy role. “This is not an easy business for women, but especially doing data and analytics. The stories that were told in ’08 and ’12 were all about men,” said O’Malley Dillon, who worked high up on both of those campaigns. “She has paved a path.” For months, that meant not just relaying voter-data updates to the senior-most operatives but informing the campaign’s messaging and fundraising strategies and, at one point, convincing higher-ups to include a positive persuasion message about Biden in its pitch to liberal voters who they’d internally classified as only needing a prod to go vote. In the final weeks, she also worked with O’Malley Dillon and the campaign’s surrogates operation to determine where to send its top messengers: Did they need to have Biden do an interview with Black-focused press in Detroit or Latino media in Miami? With early-vote numbers looking lighter than expected in Guilford County, North Carolina, she pushed for Biden to sit for a late interview there. As Georgia became a clearer target toward the end of the campaign, she advised the communications staff to prioritize radio stations with large Black audiences in Fulton County and helped push for both Barack Obama and Kamala Harris to make final-stretch visits to Atlanta.
On Election Night, however, Siegel put Florida and North Carolina out of her mind early — they were firmly in the ‘it would be nice to have them’ category, not ‘must have.’ “In a scenario where the election was close, Florida was not going to be the state” to determine the election, she explained, so she urged her colleagues not to worry too much about it. As the night stretched on, though, she did keep an eye on Georgia and Arizona, which were central to Biden’s backup path to 270. In the months leading up to that night, she had run through various “doomsday scenarios” — like a polling skew as bad as 2016’s, or low turnout in midwestern cities that would hand Pennsylvania, Wisconsin, and Michigan back to Trump — and found that neither Arizona nor Georgia would be affected. Internally, Arizona had always been considered a top-tier swing state, but it was at the insistence of Siegel and others looking at campaign data in the final months that the campaign went all in on Georgia. “I’m a lifelong Georgia skeptic because we’ve tried so many times and never got over the hump,” said O’Malley Dillon. But, she explained, she was convinced to give it a shot because of its unique demographics, “where, if we saw some movement down in the blue wall, we would see it in other states in the electorate but not see that in Georgia.”
Still, as long as the blue wall remained standing, it was Option A. And by the time Siegel went to sleep, it was the only thing on her mind. Results in Milwaukee, Madison, Detroit, Philadelphia, and Pittsburgh were likely going to flood in at some point and push Biden ahead of Trump in those states, but no one knew when that would happen, and it seemed likely that Pennsylvania would still take days longer to fully report its results. But Michigan had always looked Biden-friendly, so it was Wisconsin that looked likeliest to serve as a tipping point.
About two hours after she fell asleep early on Wednesday, Siegel got the call: after some late-night vote dumps, it was now clear to her team that Biden was going to win Wisconsin. She rushed to look over the latest numbers there, in Michigan, and in Pennsylvania, and dialed back into Wilmington around 6:30 a.m. Her message this time: It was truly over. They may have been all but sure Biden would win the night before, after Arizona was called. But now the numbers looked undeniable: Biden would beat Trump by flipping back the midwestern states he’d stunningly won in 2016, though the Democrat may not officially cross 270 for a while, considering how long it was still taking to count every vote. The only hard part now was trying to explain this.