He saw his friendships strained, his methods challenged, but in the end he was right again. Allan Lichtman has now predicted nine consecutive presidential election results based on a model he created — topped off by Donald Trump.
Lichtman is deeply critical of polls and the political journalists who rely on them, and he insists his 13-part questionnaire does a far better forecasting job than the squadrons of data nerds who pore over digitized maps of battleground states.
Not that he's happy about it.
"I take no satisfaction in being right (about a Trump win)," the American University history professor said Wednesday.
Still, he's happy to point it out.
Lichtman believes election analysts err by studying campaign as a series of battlefields — northern versus southern Florida, western versus eastern Pennsylvania, and so on. He prefers likening it to dominoes, toppling one way or the other.
He says U.S. campaigns are contests of momentum.
In the early 1980s, he and a colleague examined election results dating back to the Civil War and found a pattern. They drew up 13 true-false statements, and if the answer to six or more was "False," the incumbent party always lost the White House.
Examples include: "The economy is not in recession"; "There is no serious contest for the incumbent-party nomination"; "The party has won seats over the two previous midterms"; and "The candidate is charismatic or a national hero."
It's worked every time since 1984.
This time, he says, the sixth and fateful domino toppled onto Hillary Clinton during the primaries — when she had an unusually strong challenge from Sen. Bernie Sanders. Lichtman called upon President Barack Obama to intervene in the primaries, and throw his support to Clinton.
He predicted a Trump win in media interviews — and drew outrage.
"Not (with) hate mail. But I got a ton of criticism," said Lichtman, 69, who once ran for a Democratic Senate seat in Maryland.
"I think I lost every Democratic friend I had, at least for a while. I got a lot of pressure to change my prediction... Yeah, this was a very difficult thing for me to stick with... I was under tremendous pressure to change my prediction."
He made his prediction before the scandals involving the FBI investigation and the vulgar videotaped conversation featuring Trump. He said his prediction held steady, even as pundits and polls bounced around in their predictions.
Don't get him started on the polls.
"We don't need the polls. Polls promote the laziest kind of journalism. So much of journalism is based on slavishly following the polls. You don't even have to get out of bed to follow the polls. It's lazy."
Then again, he confesses, he qualified his earliest predictions earlier in the year. The reason was the unconventionality of the man his model had anointed as the 45th president.
"Because we had such an out-of-the-box, dangerous candidate in Donald Trump," he said. "A history-smasher."
Over time, his prediction grew bolder. And when the results rolled in from Florida, then Wisconsin, he knew the old model had done the trick again.
The clincher, to him, was the primaries. He insists the pressure on Clinton to adopt trade protectionism was contrary to the history of progressive politics — and ultimately doomed her.
National polls didn't actually fare that badly: the average of head-to-head surveys on Real Clear Politics showed Clinton winning the popular vote by 46.8 per cent, and she wound up winning about 48 per cent. They undersold Trump's result by just over three percentage points.
It's several state polls that were wildly off-base.
Pollsters more accurately predicted the result in the last Canadian election.