Hillary Clinton's presidential campaign and the Republican National Committee had one thing in common toward the end of this election season: They both predicted that the former Secretary of State would win between 300 and 350 Electoral College votes on Election Day.
But according to data journalist and Wall Street Journal columnist, Dante Chinni, low African American turnout, outdated polling models, pro-Trump union households in the Rust Belt, and - in some states - a new Trump coalition, led to the Tuesday night election shocker.
Chinni spoke with Chuck Todd in the latest episode of "1947: The Meet the Press Podcast."
All told, Donald Trump's raw vote tally was lower than Mitt Romney's 2012 results. So how did Trump run away with an Electoral College victory? By winning in places where nobody was looking.
To see how far off the voting model really was, Chinni points to two battleground states. "The story in Pennsylvania and Florida is that he really did something remarkable," Chinni says of Trump's voter turnout effort. Clinton "got her numbers," but Trump "remade the electorate," rendering the Clinton campaign's strategy toothless. Both Florida and Pennsylvania went Trump's way Tuesday night.
Initially, Chinni predicted a win for Clinton in Florida on Election Night after looking at data favorable to her in Miami-Date County. And in Hernando County, home to an increasingly large number of Hispanic immigrants, the majority of voters supported the Democratic candidate on the ballot. But Trump was able to sneak out a win in the highly competitive Sunshine State due to "domestic migration," or large numbers of older white voters moving to the region after retirement.
"It was death by a thousand cuts for her in Florida," Chinni says.
And in Pennsylvania, everyone was looking to Chester County, a Philadelphia suburb that leans slightly Republican. In 2008, President Barack Obama won the county and the state, but Romney took Chester by less than 2,000 votes in 2012.
So how wrong were the models for a victorious campaign? Clinton won Chester County by 10 points - but she lost the entire state in what would become the first of Clinton's safe states to fall.
That doesn't mean that Republicans have figured out a new path to victory. In the same way that the Obama coalition required Obama's presence on the ballot, Trump's new voting bloc might not be replicable in 2020 or further down the road. Even if those voters were drawn to the message and not the man, the number of white voters without college degrees is shrinking.
Given the difference between predictions and the outcome on Election Day, any rational observer might ask: should voters ignore national polling data in future elections?
Chinni points to the toplines on national polling, which turned out to be fairly accurate, putting Hillary Clinton's percentage resting at around 48 percent - the number she ended up yielding after votes were cast.
But that's not to say that the polling business can't improve in the future. "I think we spend way too much time thinking about what's going to happen, and not why they happen," Chinni says.