clock menu more-arrow no yes mobile

Filed under:

Study: social media bubbles might not be making us more polarized after all

Don’t log off yet.

Photo by Chris Jackson/Getty Images

After Donald Trump’s stunning electoral victory, lots of people wondered how, exactly, this had all happened. One possible solution, it seemed, was that everyone was trapped in an information bubble of their own making — numerous articles alleged that self-segregated social media feeds had become an echo chamber for people’s own thoughts and beliefs. The theory was popular, even if empirical evidence was scant.

Now a new paper from the National Bureau of Economic Research argues the opposite. Looking at nine measures of polarization across different age groups, the researchers found that increases in polarization over 16 years were the greatest among people least likely to use the internet and social media. Polarization has increase the most among voters 75 and older, the group least likely to use the internet.

“These facts,” write authors Levi Boxell, Matthew Gentzkow, and Jesse M. Shapiro, “argue against the hypothesis that the internet is a primary driver of rising political polarization.”

The researchers chose nine metrics of polarization that other studies have used in the past. Some looked at polarization in voting patterns and ideology, like the percentage of straight-ticket voters for presidential and House races in each election, or how people’s party affiliation predicted their views on a series of issues. They also looked at measures of perceived polarization, like how warmly or coldly supporters of each party felt about people in the other party.

Rise in polarization by age group.

They then compared these measures of polarization within three age groups — voters 18 to 39, 65 or older, and 75 or older. In eight of the nine metrics they examined, polarization increased the most among those who were least likely to use the internet and social media, those 65 or older and 75 or older.

Similarly, they examined measures of polarization by predicted internet access, which they calculated by looking at age in conjunction with other demographic factors such as race, education, and geography. Again, the researchers found that those with the least likelihood of being online experienced the highest rate of polarization.

Rise in polarization by predicted internet access.

“We don’t take this evidence in any way to mean that what’s happening with social media is not important or that we shouldn’t worry about if people are increasingly finding themselves in filter bubbles,” says Gentzkow, a professor of economics at Stanford and one of the researchers behind the study. “But just that this evidence argues against [social media] being quantitatively the main force behind these trends that we see.”

Of course, the paper’s authors are open to the idea that indirect effects not picked up by the study could be having an effect on polarization. Perhaps social media has been having an effect on younger individuals, and then they in turn polarize older voters by talking with them or electing more polarized politicians. But were that the case, Gentzkow says, we’d likely see that these indirect effects would have less of an impact on polarizing older voters than social media would on younger ones, a finding not supported by the data.

“My best guess as to what’s going on is simply that although digital technologies may be playing a role, they’re not the driving force,” he added.

To Gentzkow, the rise in polarization is less a problem of “easy culprits” like fake news on social media, and more a result of deeper divisions within American society. Addressing filter bubbles on social media is important, but it doesn’t address the root cause of polarization.

“The realities of inequality, of the way the workforce is changing, of the way gains to technological changes are being distributed across different groups … there’s deeper things going on,” Gentzkow says. “I think public policy needs to address those deeper problems if we’re going to make progress.”