When Georgia’s secretary of state, Brian Kemp, announced the weekend before this year’s midterm elections that Democrats in the state had tried to illegally access the state’s voter database, the announcement was treated with a heavy dose of skepticism. In the days that followed, that skepticism seemed increasingly warranted, as reports suggested that the claim by Kemp’s office was at best misinformed and, at worse, willfully disingenuous.
Over the weekend, the Atlanta Journal-Constitution reported that, as suspected, there was no attempted hack by the Democratic Party — or, for that matter, by anyone. There was a security flaw that had been reported to the state, and the office of the secretary of state appears to have responded by blaming Democrats.
During the midterm elections, the Republican candidate for governor narrowly defeated Democrat Stacey Abrams by 1.4 percentage points, or about 55,000 votes. That Republican candidate, of course, was Brian Kemp.
We’ve seen this pattern before: Last-minute allegations in close races that, only after the fact or only quietly, are shown to be unfounded. The same thing happened this year in Florida, where Democratic Tallahassee Mayor Andrew Gillum was tied to an FBI probe that, it turns out, may not involve him at all. Gillum lost by 0.4 percentage points, or 32,000 votes.
The most obvious example is the election of Donald Trump in 2016. He is president thanks to thin margins in three states — 11,000 votes in Michigan, 44,000 votes in Wisconsin and 23,000 in Pennsylvania — and thanks, to some extent, to a last-minute announcement by FBI Director James B. Comey about an investigation into Trump’s opponent. Or, perhaps, he’s president because of a coordinated effort on the part of the Russian government to dissuade Democrats and energize Republicans before Election Day, as a new Senate report suggests.
But did these things actually shift the results in these close races? It's almost impossible to tell — or, more accurately, it's hard to say conclusively that, had these things not happened, the results would have been different.
Take Kemp. Polls at the end of the contest showed that it was close; one, from Emerson College, had Kemp up two points. That poll ended a couple of days before Kemp’s office blamed hacking on the Democrats. The Journal-Constitution had the race as a tie, making a 1.4-point result well within the margin of error.
What’s more, nearly every Democrat voted for Abrams and nearly every Republican voted for Kemp, according to exit polls. Abrams won independents by 10 points. Would Abrams have won independents by 11 had Kemp’s office not dropped that untrue bombshell? Would 36 percent of the electorate have been Republicans instead of the estimated 38 percent if it hadn’t been announced? It’s hard to say — but seems unlikely in the context of that Emerson poll.
In Florida, Gillum’s narrow loss was different. Polls had him up by nearly four points on average, with Emerson putting him up five. That margin was one reason that observers were surprised when he conceded on election night. But those polls also came well after questions about that FBI investigation were first raised. Was it the investigation itself that powered his opponent, Ron DeSantis, to victory? Take those questions away, and does Gillum win?
A very concrete example of the did-this-make-the-difference question played out in the race for North Carolina’s 9th Congressional District. There, Republican Mark Harris seems to have been the beneficiary of an effort to fraudulently collect absentee ballots and cast votes on his behalf. But did it make the difference in the 900-vote race? We looked at this last week, considering the actual votes involved and . . . it remains hard to say.
FiveThirtyEight’s Nate Silver did an analysis of the effects of Russia’s social media efforts on the 2016 election, reaching several conclusions. One is particularly important: On the scale of what the candidates' campaigns were doing, those social media efforts were small. A few million dollars vs. hundreds of millions. There’s another complicating factor, too. When a race (or several races, in Trump’s example) are so close, there are a lot of things that might have been determinative. Might.
Silver suggests that the more likely candidate for a decisive last-minute factor was that announcement by Comey. Why? In large part because it leveraged the same thing that Kemp’s announcement did: media amplification.
While Kemp's claims were treated with skepticism — he'd falsely blamed the administration of Barack Obama for trying to hack the state the prior year — that wasn't universally true. The Journal-Constitution pointed to a CNBC headline: “Georgia Sec. of State Calls for FBI Vote Hacking Investigation.” It can be hard to convey nuance in a headline, but that headline is not skeptical.
CNBC wasn’t alone. USA Today had a similar headline. So did CNN. Even when the lack of evidence was noted, as at the Journal-Constitution’s own blog, there’s still necessarily an element of he said, she said. Kemp’s office knew that the accusation would get coverage and, even if that coverage was tempered, it still pushed the story out.
We keep having this same conversation. Did Russia swing the election to Trump? Did fraud swing North Carolina’s 9th District to Republicans? Or more-complicated questions about things that didn’t happen: Did hush-money payments to adult-film actress Stormy Daniels and Playboy model Karen McDougal boost Trump to victory?
Think about the number of factors that would need to be evaluated to answer that last question. Who would be influenced by that news? Evangelical voters? Would they have been affected by Daniels’s story of a one-night stand where they weren’t deterred by allegations of sexual assault leveled against Trump? Are there enough of those voters in the three key states who would have been swayed to make a difference? Would enough Republicans have been demotivated to cast a ballot or enough Democrats energized to show up to actually change the outcome? How do you assess this? Even if you poll people about what they would have done — is that self-reporting even reliable?
It’s important to remember, too, that Kemp’s victory also depended on his office in another way. He effectively leveraged existing tools that disenfranchised heavily Democratic voting groups to tamp down turnout. In Florida, outgoing Gov. Rick Scott was shown to have restored voting rights to a higher density of Republican former felons and a lower density of black former felons than any governor in 50 years. What role did that play?
None of these tactics are new. Voter suppression, burying negative information, seeking earned media (that is, news coverage) for questionable stories — all of these have, to varying extents, been part of campaigns forever. What's changed, in part, is the existence of social media, increasing awareness of questionable practices and also allowing stories like that of the alleged hacking of Georgia's voter rolls to quickly propagate.
In a statement after the accusation was made, Kemp made an apparently unintentional allusion to the defense Comey offered of his own last-minute announcement.
“I can assure you if I hadn’t done anything and the story came out that something was going on, you’d be going, ‘Why didn’t you act?’ ” Kemp said. Comey, who later admitted that his decision to announce the discovery of new emails in the Clinton case stemmed from concern about being seen as not acting, similarly argued that he erred on the side of full disclosure.
The result was a flood of attention in the closing days of the campaign focused on a negative story for Hillary Clinton. She was cleared on the final weekend, but the negative attention, Silver believes, helped drive her poll numbers downward.
Comey, at least, seemed chagrined at the idea that he might have affected the outcome. Kemp, as he prepares to be inaugurated as governor, has expressed no such qualms.