Matt Levendusky is associate professor at the University of Pennsylvania; Neil Malhotra is associate professor at the Stanford Graduate School of Business

Since the disputed 2000 U.S. presidential election, there has been a surge of scholarly and popular interest in partisan political polarization. While there is a lively debate among political scientists about the extent of polarization among ordinary Americans, the mass media typically portray the country as deeply divided along partisan lines, as that makes for more exciting journalism (take the recent Duck Dynasty fiasco, for example). What kinds of effects does the media’s reporting of political polarization have on Americans? Are we in a self-reinforcing cycle, in which news coverage emphasizing polarization itself has a polarizing effect?

Thanks to the support of Time-sharing Experiments in the Social Sciences, we have conducted numerous surveys of the mass public to address these questions, discussed in this paper. We find that the media’s coverage of polarization—and the ensuing increase in perceived polarization among Americans—influences political attitudes, but in some surprising ways. Specifically, media coverage of polarization increases antipathy of the other party, even as it encourages moderation among those who were more moderate to begin with.

To begin, we note that there is reason to suspect that ordinary citizens think the electorate is more divided than it actually is. Work in social psychology finds that humans tend to over-estimate the distinctiveness of rival groups—men and women, blacks and whites, Israelis and Palestinians, and, in our case, Democrats and Republicans. This stems from the underlying psychology of categorization: merely labeling groups makes people see them as more distinctive than they actually are. So when people think about where “Democrats” and “Republicans” stand, they will tend to place Democrats too far to the left, and Republicans too far to the right, which psychologists term “false polarization.” Previous psychological studies found this pattern, and we find that in our data as well, as illustrated in the figure just below.

The figure plots the actual distance between Republican and Democratic respondents in our surveys as measured by seven-point issue scales (the solid lines) against how our respondents perceived the divide between Republicans and Democrats (bracketed line). Across several different surveys, we find a large degree of false polarization. That is, when we ask subjects about where they think the “average Democratic voter” and “average Republican voter” stand, they think they are further apart than the average Democratic and Republican voters actually are. For example, on the issue of capital gains tax cuts, respondents think ordinary Americans are 84 percent more polarized than they actually are (see the second row of the graph above). Because we have data from a high-quality probability sample from GfK/Knowledge Networks, these effects are not simply driven by politically motivated online survey takers, but rather reflect deeper underlying psychological tendencies.

We argue that media coverage of the electorate—which tends to emphasize polarization and discord when discussing voters, a finding first noted by Fiorina and co-authors–will exacerbate false polarization. Using a population-based survey experiment, we find exactly this pattern. We randomly assigned some subjects to read media accounts of a polarized electorate and others to read accounts of a more moderate electorate. When subjects are exposed to media coverage suggesting electoral polarization and division, they perceive greater electoral polarization–as measured by where they place typical Republican and Democratic voters on issue scales (readers interested in the details of the analysis can consult our paper). This suggests that media coverage can make people think the U.S. is a politically polarized country even if it is not.

But what are the political implications of this false polarization? Counterintuitively, we find that it actually moderates citizens issue positions. Most Americans support bipartisanship, compromise, and consensus, and all but the strongest partisans respond negatively to violations of these norms. So when subjects read about polarized citizens who disagree with the opposition and eschew compromise and consensus, they react negatively to them. These individuals then become a sort of foil, telling subjects what not to believe, and as a result, subjects moderate their issue positions. Using these experimental data, we find that when subjects see polarized media coverage (relative to media coverage suggesting the electorate is more moderate), they become approximately 5 percent more moderate on traditional seven-point issue scales. False polarization is not a “self-fulfilling prophecy”: the media reporting on a polarized electorate will likely not create one.

But such false polarization has a more pernicious effect as well: it increases what Shanto Iyengar and his colleagues have termed “affective polarization,” or dislike for the opposing party. By the same logic as above, subjects see the polarization and division depicted in the media, and they respond negatively to it, and in particular, they especially dislike members of the other party. We find, for example, that after seeing the polarized article, subjects rate the opposing party 4 degrees lower on a feeling thermometer, and they are almost 10 percent more likely to assign them the lowest rating on the feeling thermometer scale. They also report more things they dislike about the other party, and evaluate its members less positively on a number of different dimensions. They view their own party less favorably, too, but to a more limited extent.

We also find that there is important heterogeneity in these effects. We find that much of the attitudinal moderation occurs among those who are already moderate ex ante, while the affective polarization occurs among all subjects. So the end result of this process is not a nation of moderates that dislike each other, but rather, an increasingly moderate core with extreme wings, all of whom have increased antipathy for the opposition. Our findings therefore suggest that the media likely are not shrinking mass polarization, as their moderating effects are centered on those who are already middle-of-the-road. Rather, the media help to further segment and stratify the electorate into a more moderate core turned off by polarization (as suggested by Fiorina and his co-authors) and a more extreme segment (as suggested by Abramowitz and others). Media coverage of polarization, then, has important, and non-obvious, consequences for American politics. Though it does not create a self-fulfilling prophecy, it shapes public opinion in key ways.

Further Reading

This is the latest post in our ongoing series on political polarization. The previous posts are listed below.  -Dan Hopkins