A while back, hackers linked to the Russian government stole a cache of emails from Hillary Clinton campaign chair John Podesta and furnished them to WikiLeaks. One of those emails has generated a lot of interest in conservative circles because — conservatives say — it shows that polls are being systematically biased in Clinton’s favor.
In the email, Thomas Matzzie, a Democratic operative, asked advisers at a progressive group called the Atlas Project to “recommend oversamples for our polling.” Bloggers like Zero Hedge and Gateway Pundit pounced on this as evidence that the Clinton campaign was working with mainstream media organizations to rig public polls in Clinton’s favor.
That would be a big story if it were true. But as I’ll explain below, it’s not — people were misinterpreting a banal discussion about Democrats’ internal polling.
But even though the story is obviously, comically, wrong, it won’t die. It’s been circulating for days among conservative blogs and social media accounts since it became public. And its longevity points to a troubling development in our media environment.
Social media sites like Facebook have democratized the media landscape, allowing anyone to create and distribute content to their friends and family. There are a lot of good things about this, but it’s also proving to have a serious downside: Without the quality filters traditionally supplied by mainstream media outlets, there’s a lot more room for total nonsense to circulate widely.
The increasing polarization of news through social media allows liberals and conservatives to live in different versions of reality. And that’s making it harder and harder for our democratic system to function.
Why nonsense stories spread so quickly on social media
The Washington Post’s Philip Bump has a thorough debunking of the conspiracy theory about “oversampling.” But to summarize, there are at least three reasons experienced reporters knew the story made no sense:
- Oversampling isn’t a nefarious tactic for manipulating poll results. It’s a standard technique for collecting extra data about a particular group (Hispanics or millennials, for example) to enable more detailed analysis of that group. Pollsters automatically re-weight the members of the oversampled group to ensure the overall result isn’t skewed.
- The email was about internal polls being conducted by Democrats to guide campaign strategy. They weren’t intended for release to the public. Manipulating these polls would have only hurt Democrats.
- The email was written in January 2008. So if there were a conspiracy to rig polls in favor of Democrats, it would have shown up in poll results in the 2008 and 2012 general elections. Yet that didn’t happen. Polling averages got the final margin almost exactly in 2008 (missing by only 0.3 percent in the RealClearPolitics average), and in 2012 they actually underestimated Obama’s final margin by about 3 points.
On the other hand, if you’re an ordinary Trump supporter scrolling through your Facebook feed, none of this would necessarily be obvious. In that case, you’ve probably never heard of oversampling, you don’t know that campaigns run internal polls, and you might not notice that the email was from 2008 rather than 2016. And so if someone shares a Facebook post arguing that the email shows the Clinton campaign is rigging public polls — and that this explains why Trump is behind in the polls even though most of your friends support Trump — you’re going to be inclined to believe it. And you might even share it with your friends.
So this conspiracy theory has rocketed around the conservative internet, passing from one conservative Facebook user to another. Its spread has been amplified by aggregators like Matt Drudge, who is happy to direct viewers to juicy stories without worrying too much about whether they’re true.
Social media distorts how partisans understand the world
This isn’t just the fault of online media. Even before the advent of Facebook and Twitter, partisan media platforms like talk radio and Fox News sometimes promoted conspiratorial stories with little evidence to back them up. But it’s a problem that social media is now making much worse.
The Washington Post and other mainstream media organizations have written stories explaining why the “oversampling” conspiracy theory is wrong. But because these stories challenge conservatives’ biases instead of confirming them, they’re much less likely to get traction among conservatives on social media.
The key thing to note about this process is that it’s not apparent to the average Facebook user. When people opened a traditional newspaper, they got a representative sample of the previous day’s news. They also got stories that had been written by professional reporters who had at least a passing familiarity with the stories they were writing about. So obvious nonsense like this rigged polling story wouldn’t have shown up in the news.
The Facebook newsfeed isn’t like that. It’s a sampling of stories heavily skewed toward the kinds of stories your friends and family like to share. And many stories are produced by amateurs with no real expertise in the topics they write about. So stories that are inaccurate but confirm people’s biases (like “the Clinton campaign and the mainstream media are conspiring to rig the polls”) are more likely to show up in people’s Facebook feeds than stories that reach an accurate but banal conclusion on the same subject.
The result: Partisans can feel like they’re very well informed on a particular subject because they’ve read dozens of stories about it. What they often don’t realize is that thanks to social media’s filter bubble, they’re only hearing stories from one side — and that these one-sided stories may all be based on the same mistaken reasoning or may all be ignoring the same inconvenient evidence.
Social media misinforms people across the political spectrum
So far I’ve focused on a particular conspiracy theory that’s in vogue among Trump supporters. But a similar dynamic can be found on the political left. When I last wrote about this topic back in March, I described how supporters of Bernie Sanders had found themselves trapped in a similar kind of filter bubble.
Reddit is a popular news aggregator in which users vote about which stories to promote to the front page. And by March of this year, the politics section of the site had become dominated by Bernie Sanders supporters. As a result, if you relied on Reddit for your political news this spring, you basically only saw stories that praised Sanders and criticized Clinton and the Republican candidates:
Sanders supporters didn’t just read a lot of pro-Sanders opinion pieces — they wound up living in a different factual universe. As Clinton consolidated her lead in the primary campaign, Sanders fans concocted increasingly far-fetched theories to explain why he could still win. And as it became clear he couldn’t win, Reddit and other social media helped to convince many Bernie fans that the Clinton campaign had stolen the election.
That wasn’t true. The Democratic primary process is complex and probably has room for improvement, but the basic fact was that Clinton won the race because she had more supporters.
Conspiratorial thinking has been less common among Clinton supporters, but they shouldn’t be too smug about it. If the campaign had turned out differently — if Sanders had won the primary or Clinton were currently trailing Trump in the polls — we likely would have seen the same kind of anger and paranoia among Clinton’s most hardcore supporters.
Filter bubbles are a long-term danger to our political system
In the short run, this isn’t a big problem. Most likely, Hillary Clinton will win the election in two weeks, and Trump supporters will grumble about it but won’t try to defy the will of the voters. The toxic environment of the 2016 campaign will make it harder for Clinton to compromise with congressional Republicans, but the country is likely to muddle through the next four years.
The problem is that this probably isn’t a one-time problem. If anything, it’s likely to get even worse in future elections. The same technological forces that embittered both Sanders and Trump supporters against Clinton are likely to do the same thing to future presidential candidates. Many Democrats will convince themselves that the Republican nominee poses a mortal threat to the republic, and vice versa.
In a decisive election like we’re likely to see in 2016, this won’t matter, since the numbers don’t lie. But in a close election, it can matter a lot. The 2000 election was bitter enough already. Imagine how much worse it would be if we had an equally close election in today’s much more polarized political environment — and with an evenly split Supreme Court.
All democracies require a certain amount of compromise to function well. Our system, with its elaborate system of checks and balances, requires more compromise than most. And compromise requires that people trust that those on the other side of the political aisle are as committed to following the rules as they are.
But the internet — and social media in particular — is destroying that trust by filling people’s heads with prejudicial nonsense. If we don’t find ways to help people find more reliable information, it’s going to cause big problems in the long run.
Correction: I originally said the “oversampling” controversy was about polls conducted for Hillary Clinton’s 2008 primary campaign. In fact, according to Tom Matzzie, the polls were conducted by an independent Democratic group. “Our group included both Clinton supporters and Obama supporters,” Matzzie says. “Our goal was to stop the Republican nominee regardless of who the Democrat was. “