Roy Moore speaks at a campaign rally, in Dora, Ala., on Nov. 30. (Brynn Anderson/AP)

Let’s say you have a book club. Is that too outdated? A wine-drinking club. You have a wine-drinking club that meets on the 12th of every month, and you need to figure out how much wine to buy. Normally, you can expect about 30 people. When the 12th is on a weekday, though, it’s more like 20: school night and all that.

This month, though, the 12th is Easter. And there’s a blizzard. And many of those attending balked at your suggestion that, instead of wine, you taste different vodkas. How many people do you expect to attend?

The tortured analogy above is to the special election in Alabama, which is the electoral equivalent of a blizzard-blanketed Easter Sunday meeting of a non-wine-drinking wine-drinking club. It’s a race between a Democrat and a Republican, sure, nothing weird there. But it’s taking place two weeks before Christmas in a deep-red state during a year in which Democrats are mobilized by opposition to an unpopular president and the Republican in the race has been accused of sexual misconduct with girls as young as 14.

Any one of those mitigating factors by itself would mean a hard-to-predict race simply because it’s hard to estimate how many people will turn out. Combined? It’s all but impossible.

Consider recent polling in the race.

The Post and our partners at the Schar School conducted a survey at the end of November, finding Democrat Doug Jones with a three-point lead over Republican Roy Moore. All other considerations aside, that’s basically a tie, given margins of error. On Sunday, CBS and YouGov released a poll of their own, showing Moore up by six points. Over the course of the past few weeks, the RealClearPolitics polling average of polls has shown the lead flipping back and forth; as of writing, it shows Moore up by about three points.


How do you reconcile two polls that were completed within a day of each other that show a difference of more than nine points in the margin? By recognizing that pollsters have the same problem as you and your wine-drinking club: It is tough to figure out who’s actually going to show up to vote.

Last November, Donald Trump won Alabama by a 27.7-point margin. That was a presidential general election, though, in which a lot of infrequent voters went to the polls. In total, about 2.1 million people voted. In March, during the presidential primary, about 1.26 million people voted, about two-thirds of them in the Republican primary. That’s a big drop-off. In the special election primary this August, only about 588,000 people voted. There are about 411,000 people who voted in each of the primary and general elections in 2014 and 2016, according to L2 Political Data. Will they all turn out next week? Probably not; only 250,000 of them voted in August. Turnout in special elections is always tough to predict, much less in this very special special election.

So who will come out? For our poll, we cut to the chase: We asked people. This is not a foolproof way of assessing turnout, certainly, but it allowed us to consider the responses by the likelihood that people would actually go out and vote. From our story about the poll:

Republicans have a clear advantage in party identification in Alabama, and Trump defeated Hillary Clinton by 28 points in the state last November. But Democrats say they are more enthusiastic about turning out for the special election. By 47 percent to 38 percent, more Democratic-leaning voters than Republican-leaning voters say it is “extremely important” to vote in the election. Democratic-leaners are also 12 points more likely to say they are following the race “very closely” and 10 points more likely to say they are “absolutely certain to vote.”

Emphasis added. Those more-enthusiastic Democrats — maybe thanks to Trump’s unpopularity, maybe thanks to the allegations against Moore — are a wild card. If they turn out more heavily, that would sway the result of the race.

CBS’s poll, despite showing a wider gulf in favor of Moore, makes a similar point about turnout.

Moore has a lead over Democrat Doug Jones, 49 percent to 43 percent, among the likely voters who are most apt to vote on Dec. 12. Among all registered voters, the contest is even. And nearly a quarter of voters still describe themselves as “maybe” or “probably” going to vote.

If infrequent voters turn out to vote for Jones, that six-point margin winnows away. This is why Jones’s campaign is focused on turning out Democrats, with a special focus on black voters. Last year, a small dip in support from black voters in key states could have helped cost Hillary Clinton the presidency. The higher the black voter turnout in Alabama, the more likely it is that Jones wins. An effort from a pro-Jones super PAC to increase turnout by using the proven tactic of publicizing voting records met with backlash from Moore.

As FiveThirtyEight notes, even that CBS poll showing Moore up by only six points demonstrates a huge shift to the Democrats. Last year, Sen. Richard C. Shelby (R) won reelection by about 28 points. An incumbent, yes, but the RealClearPolitics average shows Moore doing about 25 points worse.

In other words, the race is indeed unusually close for Alabama, and we don’t know exactly who will turn out next week. Polls — from different pollsters making different assessments of who they think will vote — are on average giving an advantage to Moore. It’s an advantage, though, that’s in the margin of error. There’s no perfect poll, and imperfect turnout estimates are often one reason that polls are off the mark. In this case, though, the turnout estimates are much harder to make.

In the case of the wine-drinking club, there’s another option: canceling the thing, or at least postponing it. That fits with the analogy here, too. Republicans pushed an effort to delay the election to get Moore off the ballot. But that never amounted to much, and so now we’re left trying to figure out how much vodka we’re going to need late on Tuesday night.