In her new book, Cyberwar: How Russian Hackers and Trolls Helped Elect a President, Kathleen Hall Jamieson argues that the Russian-government directed interference campaign likely provided Donald Trump with his winning margin in the crucial states that allowed for his 2016 Electoral College victory. But in an election with so many moving parts, how can we really know, or even be confident, that the Russian operation made the difference?
I recently spoke by phone with Jamieson, who is the director of the Annenberg Public Policy Center at the University of Pennsylvania. During the course of our conversation, which has been edited and condensed for clarity, we discussed the specific aims of the Russian social media campaign, how WikiLeaks helped shape the second presidential debate, and whether her book is skeptical enough about what we still don’t know.
Isaac Chotiner: What leads you to be confident enough to make the judgment that Russian interference likely provided Trump’s winning margin?
Kathleen Hall Jamieson: There are so many different pathways by which the effect on the electoral outcome could have been achieved that the likelihood that some combination of them achieved it is relatively high, and two of the three individual pathways are of themselves significant enough that they could have accomplished that end.
What are those pathways?
The first, which is the one that is the weakest of the arguments, is the Russian social media interventions.
First, they reached 126 million Americans that we know of through Facebook, and there was reach as well, although not that substantial, on other platforms.
Second, the message aligned with Donald Trump, because my argument isn’t that they put new messaging in, but rather that they amplified messaging that was already there. For example, you find very strong anti-immigrant appeals. You also find attacks on Hillary Clinton, or I should say on her candidacy, but also on Hillary Clinton the person, that are consistent with the Trump attacks. She’s corrupt. She’s lying. She should be in prison.
Isn’t this then a way of saying, “Well, you know. Trump could have spent slightly more money, and amplified his message, and that would have made the difference?”
Yes, because the argument about the impact of social media is that it increased the balance of the messaging in a way that on the margins would have made the difference. It is not an argument about new messaging.
What I argue is that there’s strong evidence that they were trying to mobilize evangelical Christians who were white, conservative Catholics, and they were trying to mobilize veterans and military households. Those are traditional Republican constituencies, so if they’re going to vote, they’re much more likely to be voting for the Republican. But if you look at the data in August, what you see is they’re not terribly enthusiastic about Donald Trump. The question is, can the combination of the Trump messaging and also whatever the trolls added be enough to help mobilize those constituencies?
What if these Trump messages, and I’m just asking a hypothetical, were actually Trump’s least effective messages, and the way Trump got the 46 percent was by keeping his mouth shut and quieting down occasionally, so actually amplifying these messages did not do his campaign good? How would you control for a scenario like that?
The first thing you’ve got to ask is: Are these messages attempting to reach the right people? And what I am arguing is that they are aligned with his need to mobilize and that the messages are consistent with that need. Now, if those two things are true, to argue that doing that didn’t have an effect would be to suggest that there was some countervailing factor that undercut them so they were not strategically apt. They were not really persuasive. It creates a backlash. We don’t have any evidence of that. They also wanted to demobilize of African Americans, demobilize Sanders supporters, or shift them off to a third-party candidate.
Then the question is, “Well, did the messages follow what we would reasonably say is … the theory of persuasion?” They’re certainly visually evocative. They elicited sharing and liking. There’s the fact that their English is not good on some of the sites … but if you look at the way in which content works in social media, we usually click and like based on looking at the visual images and the print that’s overlaid on them, and there was sharing by people who are sophisticated conservatives, who would not have been sharing things if they thought they were from the Russians. So they were persuasive enough to get sharing from influential conservatives.
What are the other two arguments, or pathways?
The second argument is the hacked content changed the media agenda, and it changed it in ways that were substantial. For example, by virtue of putting the hacked content out as the convention approaches, the ability of the Republicans and also the capacity of reporters to report on evidence that the Democrats and the Democratic National Committee had their thumb on the scale in essence was higher than it otherwise would be, and that content is used at a time in which Hillary Clinton needs to be consolidating the Sanders supporters behind her.
The second place in which they’ve got a change in the media agenda occurs on Oct. 7 when what you have is the confirmation from the DNI and Homeland Security, Johnson and Clapper, that the Russians were behind the hacking. Within the hour, you then have the Washington Post posting the Access Hollywood tape with its lewd remarks by Donald Trump. Then within the next hour you have the drop of the Podesta emails from WikiLeaks’ Julian Assange.
What you see on [the] Sunday news shows, and this is the change in the media structure that wouldn’t be there had the hacked content not been there, is Rudy Giuliani playing those two against each other and saying, “Look. They’re both flawed candidates.” In essence, discount Access Hollywood because it’s counterbalanced.
You may be correct that the media narrative changed in some way, but it seems like there are still two questions. The first is how much the change in the media narrative mattered, and the second is would the media narrative have changed regardless?
Yes, but you set up all of the alternative hypotheticals. I’m not dealing with alternative hypotheticals. We assume that everything those journalists could have done they actually did. You’ve got to constrain this in time or you can’t do the analysis. The question is if you took the Podesta hack out of there, what’s your hypothesis about what would have come in its place in that two-day period?
Why is this two-day period so important?
Because it’s the night of the second debate.
You’ve got Republicans who are actively talking about taking him off the ticket, putting up an alternative ticket. If the hacked content hadn’t been there, he might not have survived the weekend. You’re a journalist. Wouldn’t the fact of the Russian hacking being confirmed by the intelligence folks have been in the news cycle in a way giving you two anti-Trump stories and no anti-Clinton story?
I know, but Trump has bad news stories it seems like every single day, and I’m not sure how much effect it has. But what I meant—
Trump didn’t have any stories that were equivalent in the disruptive effect on his campaign, and the demonstrative public statements by Republicans who had been supporting him, who basically withdrew from him. Now, they came back to him after the fact, but they came back to him after the fact, after the Podesta emails were out there.
Well, right. Or you could argue that they were going to come back inevitably, because that’s what Republicans do.
They always come back.
You could, but you can’t tell me that that weekend is not the most traumatic moment for the Trump campaign, given that we had journalists [establishing] that they were thinking about trying to replace him at the top of the ticket.
What’s the third pathway?
We’re not there yet. The night of the second debate you’ve got this framing in the news that has the hacked content in it, and the hacked content is used to frame a question in the debate that asks Hillary Clinton whether basically she’s two-faced. This is a press effect based on what they say is WikiLeaks content. They don’t attribute it to the Russians. They don’t attribute it to Julian Assange. And in that debate moment what you have is a large mass audience able to be exposed to the content and a question they couldn’t have gotten there had there been no hacked content, and because the question is taken out of the context of the speech. [Editor’s note: Clinton was asked whether it was OK for politicians to be “two-faced,” because a WikiLeaks release of her paid speeches included her saying, in a celebration of Abraham Lincoln, that politicians need “both a public and private position” on certain issues.]
Hillary Clinton’s answer looks disingenuous because nothing in the news coverage has said she was talking in the context of Lincoln, Spielberg’s film, and she was not saying, “Here’s what I really am going to do. I’m not gonna do the things I say in public. Wall Street, I’m going to get rid of Dodd-Frank.” Or “Here’s what I’m really going to do when I become president. I’m not going to do what I’m promising about trade. I’m going to do something different.” That was a statement in the context of an interpretation of a film about the nature of political engagement. Here’s the reason this is important: because that is playing into “She says one thing in public and another thing in private.”
I know, but it just seems so hard to kind of actually appraise how much one debate question and the follow-up on that mattered.
But what you can say is we have polling data that suggests that debate viewers differed from nonviewers in the presence of controls which try to account for the difference between viewers and nonviewers in their assessment of the dimension “She says one thing in public and another private,” after the second debate.
How do we know they attribute that to one question?
We know that we can attribute it to debate viewing. We don’t know we can attribute it to one question. But we know that it isn’t the things that are leading up to that, because the things that have hurt her on that dimension presumably before that happened in September, they’re already baked into the polling data, and that’s the “deplorable” thing. It’s saying she had allergies when she had pneumonia.
What’s the third pathway?
We’re still on the debates. The hacked content is then used in the debate.
No, I get it.
And then it’s used in the third debate as well, and there again you’ve got a press statement that takes something out of context from what is in the WikiLeaks content, and again it’s the same dimension. Here the assertion is that she’s come out for open trade and open borders.
No, I get that.
And you see another polling data point that suggests that that is reflected differently in the assessment of debate viewers and nonviewers. At the same time, you’ve got the drop in her perceived qualification across the period after the first debate and before Comey reopens the investigation, and if there’s something in the news that’s damaging her perceived qualification other than the hacked content, that would be an explanation, but I couldn’t find it.
Have you read the book?
The third one is the hardest one to digest into simple sentences.
But I thought the most convincing. Go on.
I think it actually is, but this one takes work because first people don’t remember all the context surrounding it. If James Comey, as he says, was influenced to make the public statement he made in July, because there was “classified information,” and also the Lynch-Clinton tarmac conversation, and if as Lindsey Graham suggested, the classified information was the primary reason for [Comey] making that statement, and if, as Washington Post and other reporters have suggested, [he] was concerned that the [effort] to discredit the FBI and Justice Department was Russian disinformation, alleging that Loretta Lynch said to someone that someone on the Democratic side that they would not be pursuing the Clinton investigation in a way that ultimately would be problematic, then the Russian disinformation had a first effect in making that public statement.
If it did, then on Oct. 28, 2016, it is plausible to say that for the same reason, if it were leaked, people could plausibly believe that Comey had not done what he needed to do in order to ensure that the public had all of the information it needed to have to judge her before the election. So if there is Russian disinformation at play in that decision, and she drops two and a half points in the polls as a result of press coverage over that nine days of the time there’s an investigation, then the Russians’ disinformation played a very important role in ultimately shaping the outcome. Because you’re very close to the election at that point. That drop in the polls is pretty clear.
I think one thing we’re seeing in America right now, especially with the Republican Party, is that there are a lot of Republican voters who will find some reason to excuse the president’s behavior or to support Republican policies no matter what. And the reasons are often, “Oh, Republicans are angry because the left did this,” or “They’re angry because the media’s unfair to the president.” There’s always kind of some reason given that they are gonna get to where they’re gonna get. When you’re looking at an election and you’re saying all these different things mattered, this had an effect, how do we know that? The Republicans were gonna come home. Maybe Trump was always going to get to near 46 percent and slightly better just because of demographics in the Midwest and Florida, and Hillary was always going to be at 48 or whatever. I’m not saying that’s true. I’m just saying how do we wrestle with that?
It’s a really important point because historically, party identification predicts vote outcome. Historically, by the time you get to the convention, you pretty much have locked in your votes. They’ve gotten their party views and you can predict what’s going to happen in the election. By that argument then, the Democrats would have come home too and African American turnout would have not dropped as substantially as it did. You can’t say on that argument, “Well, it’s gonna work only for Trump.” If those forces are at play and inevitable, then the African American votes at its historic average, not its Obama average, at its historic average in recent times, and that’s higher than it was now.
The premise underlying that argument has to account for then why didn’t the Democrats do what that argument presupposes Democrats and the most reliable Democratic constituency, African American voters, did?
Let me step back. This is a different kind of election from those that we can predict coming out of the convention, and it’s because first we have a higher proportion of independents. They are less anchored to party. Secondly, we have a higher proportion of people who just don’t like either of the two major party candidates. They are really conflicted about casting a vote, and there are Republicans in that group and Democrats in that group. They’re staying conflicted, and as a result, coming into the last period in the election, we have almost 1 out of 8 who are still undecided. That’s extraordinarily unusual. And we have unprecedented levels of early voting, which means that across that last month, any time you come in with a very strong stimulus, there are people who are actually casting ballots.