Support The Bulwark and subscribe today.
  Join Now

Today’s Reminder That the Plural of ‘Anecdote’ is Not ‘Data’

A few interviews with reluctant passengers of the Trump train do not say anything about 2020.
May 31, 2019
Today’s Reminder That the Plural of ‘Anecdote’ is Not ‘Data’
(Photo by Scott Olson/Getty Images)

Where are all the regretful Trump voters? All the uneasy conservatives, undecided moderates, and crossover Bernie Democrats who held their noses to vote for Trump and are embarrassed to say so are nowhere to be found! At least, they’re nowhere to be found among a handful of people in one county in Michigan, writes Daniel Allott in the Washington Examiner.

He ventured up to Macomb County, a swing county of 840,000 people in the southeast of the Lower Peninsula, to find these elusive creatures. He came up short: “Among the two-dozen people I interviewed in and around Macomb County, not a single one could think of a Trump voter who no longer supported him.”

Well that settles that.

Except it doesn’t, because if each of the two dozen people Allott interviewed had instantaneously accounted for the political views of their hundred closest friends, he’s only gathered data from less than 0.2 percent of the population. And this far into the Big Sort, Allott’s interviewees, like all Americans, are likely to surround themselves with people who think like them.

Needless to say, his sources are not a representative sample. “I attended a local business awards dinner in Port Huron,” writes Allott. “A middle-aged woman rushed over to me when she heard about the story I was working on.” She was a non-reluctant Trump supporter, and a classic example of selection bias.

None of this would matter if all Allott aimed to do was to gather the opinions of some voters in a swing county, but he overreaches, making predictions that don’t follow from his data.

More accurately, he tacitly endorses tenuous predictions made by his interviewees. Predicting another Trump victory in Michigan in 2020, one man says, “If [Trump] was going to the optical store to get some new glasses, he’d have 20/20 vision.” Mark the state Safe R.

After pointing out that Republicans lost the governor’s race, the attorney general’s race, the Senate race, and two (net) House seats to Democrats in 2018, Allott provides bromidic reassurance. “But Republicans say 2018 shouldn’t be seen as a harbinger for 2020, because 50,000 Trump voters didn’t show up in 2018.”

So why didn’t those people go to the polls? Could they possibly maybe be disillusioned with the president and those who supported him? Hey, we’re just asking questions!

Someone in Michigan must be unhappy with the president, because his net favorability in April was -10, per Morning Consult. It’s been as bad as -15 several times since his inauguration, when he enjoyed +8 net favorability in the state. His disapproval numbers have increased from about 40 percent to somewhere between 53 percent to 55 percent since January, 2017, while his favorable numbers have fallen from 48 percent—about the same percent who voted for him—to between 40 percent and 43 percent. OK, so maybe it’s possible that none of those voters are in Macomb County. But Trump still has to win the whole state..

The fundamental problem with Allott’s analysis is that he misunderstands why Trump won Michigan in the first place. Even if there were no regretful Trump voters, he could still lose Michigan in 2020. It’s true that Trump was able to turn out about 8 percent more votes in Michigan than Romney did four years earlier—an impressive feat given that the state’s population increased by only half a percent in Obama’s second term. Trump simply outperformed Romney in Michigan—but it wasn’t the decisive factor in the election.

While Trump was driving out more voters, Clinton did the opposite. She lagged behind Obama’s 2012 performance there by nearly 12 percent, or 300,000 votes – enough that the total number of votes cast in Michigan declined from 2012 to 2016 by about 100,000, despite Trump’s success. Trump could even add more votes to his outlier 2016 total (if he’s able) and still be on track to lose Michigan to a halfway competent Democrat.

If, as Allott concedes, 50,000 Trump voters didn’t turn up in 2018, the midterms might indicate regression toward the mean for Trump. More worrisome for the president, Michigan Governor Gretchen Whitmer received more votes in a midterm election than Trump did in the general. Democrats may also be reverting to their mean.

None of this is to call Michigan preemptively for the Democrat in 2020. The point is we just don’t have enough information to make a prediction.

Benjamin Parker

Benjamin Parker is a senior editor at The Bulwark.