In this April 19, 2018, file photo, Paul Manafort departs U.S. District Court after a hearing in Washington. (Alex Brandon, File)

We know, thanks to a failed redaction by lawyers working for Paul Manafort, that Manafort, then the chairman of Donald Trump’s presidential campaign, shared polling data in 2016 with a former colleague named Konstantin Kilimnik. Kilimnik, born in the former Soviet Union, is believed by government investigators to have links to Russian intelligence.

The transfer of that information likely occurred at a meeting between the two and Rick Gates, then deputy chairman of the campaign, on Aug. 2, 2016, a few blocks south of Trump Tower. It’s a meeting that a lawyer for special counsel Robert S. Mueller III described as getting “very much to the heart of what the special counsel’s office is investigating”: A link between the Trump campaign and Russia’s effort to interfere in the election.

How detailed that polling data might have been is not yet clear. But on Monday, journalist Marcy Wheeler divined one detail from a document filed by Manafort’s attorneys: The document that was shared likely ran for more than 70 pages.

Some background is in order here. It appears, based on various court filings, that Gates told Mueller’s team about the poll data being shared. An email obtained by Mueller indicates that Manafort asked Gates on Aug. 2 to print out a document. Manafort’s team alleges it was for a campaign scheduling meeting, not Kilimnik.

But Gates also told investigators that Manafort walked Kilimnik through . . . something — it’s redacted — on that date. When Manafort’s attorneys challenged that claim during a Feb. 4, 2019, hearing set to determine if Manafort had lied to Mueller’s team, Judge Amy Berman Jackson said she thought that “the timing of [the email] and the substance of it is consistent with what Mr. Gates said was going on.” Manafort’s lawyers countered that Gates was untrustworthy, which Jackson waved off.

What Wheeler points out is how Manafort’s team referred to that email in various filings. During the Feb. 4 hearing, she notes, Manafort’s attorney suggests there’s a copy of the document that was printed in the government’s exhibits. A document filed by his attorneys in response to that hearing appears to identify which exhibit.

It mentions the email from Manafort to Gates (blue text, below) and identifies it as Government Exhibit 233. It appears to suggest that the email asked Gates to print out an attachment — presumably the polling data. Manafort’s lawyers then refer to that attachment more specifically (yellow text), apparently identifying a time frame during which it was created. (The Republican convention began on July 18, 2016, about two weeks before the Aug. 2 meeting.) A footnote attached to that line is mostly redacted but includes page numbers: 4 through 79.


(Court filing submitted by Paul Manafort) (Philip Bump/(Court filing submitted by Paul Manafort))

The implication? The attachment to the email, part of the same exhibit, ran for more than 70 pages.

More than 70 pages, then, of polling data.

On the surface, that probably seems like an awful lot of polling data. But it’s worth pointing out that when printed out, data from polling is often voluminous for a very simple reason: The complexity of it grows exponentially as you start looking at demographic data.

Consider the data below, taken from a Suffolk University poll released in December in partnership with USA Today. This is a page of what are called “cross-tabs,” a comparison of how different demographic groups answered a particular question. The page below shows how members of six different demographic groups answered one question in the Suffolk poll; namely, if the country is headed in the right direction.


(Suffolk University/USA Today) (Philip Bump/(Suffolk University/USA Today))

Let’s zoom in so you can see how the data works.


(Suffolk University/USA Today) (Philip Bump/(Suffolk University/USA Today))

In Suffolk’s cross-tabs, the results are shown in two formats: The number of responses and the percentage of the group that the number represents. So in the yellow text, we see that 1,000 people were asked this question, 488 of them men and 512 of them women.

In the blue text we see their responses. Overall, 35 percent of respondents think the United States is headed in the right direction; 54 percent think it isn’t. Among men, the split is 41-48; among women, it’s 30-61. Above the percentages are the actual counts. For example, 198 men who were polled thought the country was headed in the right direction.

We can look at the results by party, too. Unsurprisingly, most Democrats think the country is headed on the wrong track (86 percent) and most Republicans think it’s headed in the right direction (81 percent).

Suffolk’s cross-tabs for a poll that consists of fewer than 40 questions runs for 336 pages.

Much of that information isn’t particularly useful, showing that various demographic groups vary only slightly in their views of a particular thing. Sometimes that’s useful to a campaign, showing that a message is or isn’t working well. Where a broad discrepancy can be identified — such as between men and women on the question above — polling has its own uses, informing campaigns about problem spots or particular demographic groups that are receptive to specific messages.

Manafort’s lawyers, Wheeler writes, at times framed the polling data as hopelessly complex, with the apparent implication that it would not be particularly useful to Kilimnik. That also suggests we’re not talking about 70 pages listing specific demographic groups that should be targeted with specific messages in specific places. It suggests we’re talking about big tables of numbers like what you see above.

For Kilimnik, though, it probably wasn’t that complicated. While every pollster has a different way of presenting cross-tabs, once you know how to read them, you can generally figure any of them out. That is likely even less of a problem here: Kilimnik worked with Manafort on polling repeatedly, including as recently as 2018.

We don’t know how densely packed the data contained in that email attachment might have been. We don’t know what questions were asked, how specific they were or what demographic groups were broken out. We don’t know what geographies might have been involved. We don’t know what Manafort presented to Kilimnik or why.

We do know, though, that there’s no evidence that Russia actually embarked on a sophisticated campaign targeting closely tailored groups in small demographic areas — the sort of thing that polling would allow. There was no explosion of new targeting in the weeks after Aug. 2, 2016. July, August and September, in fact, were a lull in the Russian targeting efforts.

None of that changes the claim made by Mueller’s team in court, though, that questions about Manafort potentially passing polling data to Kilimnik are at “the heart” of Mueller’s probe. They certainly are.