Candidates Sen. Cory Booker, former vice president Joe Biden and Sen. Kamala D. Harris before the Democratic presidential debate on July 31. (Scott Olson/Getty Images)

After political candidates finish a debate, there’s a vacuum. Before the debate, there’s all sorts of speculation about how people will do and what they will do and how those things will affect their candidacies. Right afterward, though, the accuracy of that speculation is hard to evaluate in any objective sense. So pundits and analysts like your humble writer rush to fill that vacuum with their own assessments of how things went.

These assessments are often way off the mark and usually fail to correlate significantly to how the field actually changes once new polls come out. Sometimes that’s because the assessments are dashed off. Sometimes it’s because they are poorly thought out.

Sometimes they’re just bad.

Kraushaar wasn’t alone. The New York Times’s Ross Douthat pronounced that if Williamson “doesn’t get to 5 percent in the polls, the Democratic Party should disband.”

Thinking Marianne Williamson did all right in the second Democratic debate is one thing. Thinking that she did well enough to become one of the handful of candidates in the Democratic field to earn more than 2 percent support from voters is another thing entirely.

There are a lot of reasons that Williamson wasn’t likely to see a surge to 5 percent. One is that Democratic voters are primarily focused on who can beat President Trump next year, and Williamson doesn’t seem like a likely candidate in that regard. Another is that assuming Williamson could surge is to some extent predicated on Trump’s having done so with Republicans in 2016 — a surge that involved a lot of factors that are obviously not present in Williamson’s case. (Like her parroting a taboo line of argument that had been percolating in partisan media, as Trump did with his immigration rhetoric.)

But one other significant reason that it wasn’t likely Williamson would surge is that poll numbers aren’t entirely driven by the number of people who watch the debates.

Less than 9 million people watched the first night of the most recent debate, hosted by CNN. That’s the night in which Williamson appeared. The second night had about 2 million more viewers.

Of those who watched either night, new polling from Quinnipiac University indicates, former vice president Joe Biden was the winner by a healthy margin, earning a third of the total vote. Williamson came in ... well, it’s hard to say. Among those who watched the debate, Williamson didn’t have enough support to register in Quinnipiac’s results, suggesting that the pollsters didn’t call the Kraushaar household.

There are interesting bits of data buried in those results. Here are the top four candidates at the moment, with the support for each indicated with three groups: overall, among those who watched the debates and among those who said they paid a lot of attention to coverage of the debates. We’ve included both the most recent Quinnipiac poll and the poll conducted after the first debate.

(Philip Bump/The Washington Post)

After that first debate, Sen. Kamala D. Harris (D-Calif.) did what Williamson didn’t: She jumped up in the polling. That was a function of her already being considered a viable candidate, but also, it seems, a function of her success at going after Biden on that very question of electability.

Notice that Harris did particularly well after that first debate among those who actually tuned in. Biden did worse with that group but did far better with those who said they paid attention to the debate coverage. His 29 percent support with that group was probably the only reason that he maintained any advantage over Harris in Quinnipiac’s July 2 poll. Given that much of the coverage of his performance in that debate was not complimentary, we are reminded of the complexity of the debate-influence ecosystem.

What I’d really like to talk about, though, is Sen. Bernie Sanders (I-Vt.).

After both debates, Sanders did worse among those who actually watched the debates than he did in the polling overall. The gap in support for Sanders between those who watched the second debate and those who tracked it in the press is as wide as the gap for Harris was between those groups after the first debate.

We’re reminded here of a bit of data that comes up a lot in primary polling: A lot of voters haven’t decided on a candidate. Which, sure: It’s August of the year prior. They have time. Another group of voters, though, has decided on a candidate but isn’t paying attention to the primaries.

Quinnipiac also asks respondents if they are paying a lot or a little attention to the primary contest. Among those not paying much attention, Sanders led in polling before the first debate. Since, his support with that group has faded, though he still gets more support from those paying only a little attention than from those watching the race obsessively. Harris and Sen. Elizabeth Warren (D-Mass.) are doing much better among those paying a lot of attention.

(Philip Bump/The Washington Post)

There’s some name recognition at play here. Sanders, as you know, ran three years ago. Biden, as you also know, was vice president. Warren is generally better known than Harris, but both are less well-known than Sanders and Biden. The better-known a candidate, the more likely someone not paying attention to the campaign will identify that candidate as their pick for the nomination.

But again: No one has heard of Williamson, relatively speaking. Another reason Harris surged after that first debate is that most Democrats had at least some familiarity with her. An Economist-YouGov poll released last month determined that more than half of Democrats weren’t familiar with Williamson.

All of this is talking around the central point, though. Williamson’s debate performance wasn’t actually very good — at least according to poll respondents. A HuffPost poll found that more viewers said their opinions of Williamson worsened than improved. Quinnipiac asked who had the worst performance; Biden and Williamson were each named by 9 percent of respondents. Both Biden and Williamson were identified as the worst performers by 10 percent of those who actually watched the debate.

That percentage was probably lower among the pundit class.