To what extent did Russia’s social media effort in 2016 suppress the black vote?

Published in Politics
1 week ago   by Philip Bump

The Associated Press’s Nick Riccardi, like me and other reporters, has heard a theory about the 2016 race that has proved hard to shake: Russian disinformation tamped down black turnout. The implication is that lower turnout among African American voters may have been decisive, particularly given that some 1.6 million black voters who cast ballots for Barack Obama in 2012 didn’t vote at all four years later.

When Riccardi shared that reporting in response to a new report from the Senate Intelligence Committee looking at Russia’s efforts on social media, FiveThirtyEight’s Nate Silver offered the caveat that black turnout was relatively high, from a historic perspective. In other words, the idea that turnout was suppressed by the Russians is undermined in part by the point that black turnout was still high.

The exchange and that Senate report resurfaced a number of useful questions about what happened in the last presidential election. Specifically:

We can answer the first question fairly easily. The Census Bureau conducts analysis of voter turnout after each federal election. Michael McDonald, who runs the United States Election Project, adjusts those numbers to correct for data errors.

The result? Black turnout in 2016 was about 60 percent, just under turnout in 2004 and well under turnout rates in 2008 and 2012. Black (non-Hispanic) voters also made up a smaller share of the electorate than they had in 2012.

So it’s clear that there was a decline in turnout, though as Silver points out, it was still higher than in elections before 2004.

The question is why. There are a few obvious factors, such as the absence of Obama on the Democratic ticket. Black turnout surged in 2008 and was still over 67 percent four years later. In 2016, turnout receded. Hillary Clinton was not the draw that Obama was.

Part of that might be that Clinton was less popular in 2016 than Obama was in 2012. (In October 2016, 75 percent of black respondents in a Post-ABC News poll viewed Clinton favorably. Immediately before the 2012 election, Obama’s favorable rating with black respondents was 97 percent.) Part may also have been that Clinton was widely expected to win in 2016, which can tamp down turnout among a candidate’s supporters.

All of that aside, did Russia’s effort, conducted through a group called the Internet Research Agency, also play a role?

The Senate report highlights the Russian focus on race.

“[N]o single group of Americans was targeted by IRA information operatives more than African-Americans,” the report states. “By far, race and related issues were the preferred target of the information warfare campaign designed to divide the country in 2016.”

This isn’t a new revelation. When Facebook released its collection of ads purchased as part of Russia’s effort, it was obvious that race was a central component. A number of the ads targeted black Facebook users specifically, and many contained content clearly geared toward exacerbating issues of race.

In January, I looked at the distribution of those ads by location over time. As the chart below shows, Russia’s efforts on highlighting race on Facebook had a heavy initial focus on Maryland and Missouri — states that were central to the emergence of the Black Lives Matter movement in 2014 and 2015 after police-involved deaths in Ferguson, Mo., and Baltimore.

This mirrors what the Senate report found:

“Evidence of the IRA’s overwhelming operational emphasis on race is evident in the IRA’s Facebook advertisement content (over 66 percent contained a term related to race),” it reads, “and targeting (locational targeting was principally aimed at ‘African-Americans in key metropolitan areas with well-established black communities and flash points in the Black Lives Matter movement’).”

Notice something else important on that chart. In the first week of November 2016, the Russian effort included no targeted ads in the three states that swung the election to Trump (Michigan, Pennsylvania and Wisconsin) and none in the five states that ended up being the next closest. In fact, the only ads were in four of the country’s most populous states: California, Illinois, New York and Texas. Nearly all of them targeted New York — which Clinton won easily and was always expected to win easily.

This is a critically important point. Voter suppression is sort of the mirror opposite of get-out-the-vote, a campaign’s effort to turn out voters it hopes will support its candidate. Suppression isn’t as time-dependent as GOTV, which can happen only while voting is active. But it’s safe to say that a robust attempt to suppress votes would be particularly active while voting was happening. That’s not what happened here.

Sure, we’re looking only at targeted ads (which is why we can say whom the Russians were trying to reach). What this suggests, though, is that Russia wasn’t actively focused on suppressing black voters in particular places from voting in the days before the 2016 race. Remember that “in particular places” caveat; it will be important again in a second.

First, though, we should note that these campaigns weren’t even terribly effective. Of the ads included in Facebook’s release last year, the most effective ads were purchased in summer 2015 — well before Trump was even the Republican nominee. Again, those ads mostly targeted Missouri and Maryland.

If we zoom in a bit, we see how modest the effort was right before the election. In October 2016 and the first eight days of November, the Russian ads focused on black issues and from black-focused groups earned only about 92,000 impressions, most of them from ads targeting places other than the states that ended up making the most difference in the 2016 election. Over the last five or six weeks of the campaign, there were about 4,700 impressions from paid ads in the eight states that had the closest presidential results.

Research cited in the Senate report reinforces that targeted advertising by the Russians was modest.

“[L]ocation targeting of ads was not used extensively by the IRA,” the report states, “with only 1,673 different instances of location targeting, by 760 ads.” That undercuts the theory that Russia used Facebook to target specific places where the 2016 vote was most up for grabs. There’s no evidence that its effort was that sophisticated.

There were also ads that targeted black Americans or those interested in black issues nationally. Over the course of 2016, most of the Russian ad campaigns targeted the United States broadly and not specific states or cities.

In October 2016 and the first week of November, there were about 2 million impressions from ads targeted nationally.

Facebook figures that as many as 126 million Americans came into contact with information spread by the IRA either through ads or on issue pages. But that includes a substantial amount of contact that occurred only after the election. According to the Senate report, Facebook estimates that 11.4 million people saw paid Russian ads — more than half of them after the election was over. (Our analysis suggests that more nationally targeted race-related ads ran after the election than before.)

Maybe this was enough, though. Maybe those loosely targeted ads or the organically spread content the IRA was producing was enough to suppress the black vote — perhaps not in key states but generally speaking.

Maybe. But there are a few reasons to be skeptical of that idea.

The first is that not many of the ads specifically mentioned a political candidate. Only 5 percent of the Facebook ads before the election mentioned either Clinton or Trump, according to the Senate report. There’s no reason to think that the non-targeted content was much more openly partisan. What’s more, only a small amount of the content that was produced was specifically supportive of a suppression effort. There were posts that pretended to represent black voters who were arguing against voting for Clinton, but they were only a small percentage of the total.

The second is that scale is important. Millions of Americans see lots of things on Facebook (and on Twitter and so on). A Facebook ad that has thousands of impressions over a three-day period is an ad that makes up only a minute portion of everything that’s seen on Facebook. Exact figures are hard to come by, but consider how many of those 126 million people who saw IRA-sponsored content 1) saw it before the election, 2) saw content centered on a suppression effort and 3) were persuaded by it. Being convinced of something is admittedly a sliding scale, but it’s obvious that by distilling that top-line number through those three criteria, the effect may not be discernible.

The third is that Russia’s informal suppression effort existed within a legitimate and broad-based effort to keep people from voting for Clinton. There were supporters of Sen. Bernie Sanders (I-Vt.) who argued against casting a ballot for Clinton, as did Republicans and the Trump campaign. Shortly before the election, in fact, Bloomberg News reported that Trump’s team was explicitly working to suppress the black vote for Clinton — an effort that, if realized, would almost certainly have been more robust and more targeted than the Russians’. Anti-Clinton propaganda shared with black Americans in 2016 wasn’t only or primarily Russian, though the focus on Russian interference may make that propaganda seem as though it was necessarily a function of Russia’s effort.

There have been statistical efforts to link the IRA’s effort to significant results, including a faulty study earlier this year alleging that Trump saw substantial poll gains when Russia’s team tweeted more things. The study has a number of gaps, but it did fill a demand: Many Americans are eager to attribute to Russian interference what might be explained in any number of other ways.

The answer here is unsatisfying. It’s probably not the case that many black voters stayed home in 2016 primarily because of what Russia was doing. It’s almost certainly the case, though, that the Russian effort on social media didn’t substantially affect the results of the election.


Other Items