September 13, 2017


One of the most popular explanations of Donald Trump’s electoral victory is that the white working class, victimized by Rust Belt de-industrialization, rallied to his calls for repudiating free-trade deals such as NAFTA and the proposed Trans-Pacific Partnership; for restricting immigration; and for creating jobs to rebuild American infrastructure. This economic theory of Trump’s victory is attractive because it tracks many Trump voters’ words (as relayed by media interviews), and thus, presumably, their subjective perceptions. So it’s more interpretively charitable than other popular theories about Trump’s supporters—e.g., that they are authoritarians or xenophobes—which portray them as victims of irrational fears and aversions.

Early evidence for the economic theory of Trump’s support came from a county-level analysis of early primary voting conducted by the New York Times in March 2016. This analysis showed that the most significant demographic correlate of Trump’s support was the proportion of white voters in a given county who had not graduated from high school; the correlation was .61. People who lacked high-school diplomas used to be able to find plentiful industrial work that has, over the course of decades, become scarce. By the same token, the correlation of counties’ labor participation rate with Trump votes was -.43, meaning that the higher the non-participation rate, the higher the Trump vote. This is consistent with the deindustrialization theory, along with the hypothesis that there was general dissatisfaction with the pace of the economic recovery after the Great Recession.  

Two Types of Economic Explanation

Economic theories of Trump’s election were resisted on the grounds that Trump voters in the primaries were not predominantly members of the working class: their median income was considerably higher than that of most Americans, and was comparable to the income of Republicans in general, who tend to make more money than Democrats. This argument for skepticism was bolstered by a November paper from two Gallup pollsters, who ran 52 demographic variables against Trump’s primary support. They found that Trump supporters were “significantly further to the right” than other Republicans: 44 percent described themselves as conservative, 16 percent as very conservative (p. 11). They also tended to be over 40, white, male, and Christian. But while their relatively low educational levels were consistent with the “white working class” stereotype, being unemployed produced no statistically significant correlation with support for Trump, and working in industries exposed to global trade competition produced only a small correlation.

However, economic explanations don’t necessarily require that Trump’s supporters were themselves in dire economic straits. They may simply have heard about economic problems that seemed to require “change.” In other words, the lack of correlation between the Trump vote and personal economic indicators does not mean that Trump voters did not care about the economic conditions of other Americans. And exit polling suggests that they did.

Voters who wanted a president who would “bring about needed change” favored Trump over Clinton by 83 percent to 14 percent. At the same time, among voters who wanted a president who would “care about people like me,” Clinton prevailed, 58 to 35. These two findings suggest that while Trump voters wanted change, it did not necessarily have to be change designed to help them personally, or even people like them.¹

What type of change did Trump voters want? Economic change seems to have been important, although other types of change were even more important.

Among voters who rated the condition of the economy as “poor,” Trump prevailed, 79 percent to 15 percent. But he lost, 42 to 52, among those who rated the economy as the most important issue. In contrast, among voters who viewed immigration as the most important issue, Trump won, 64 to 32. And among those most concerned about terrorism, he won, 67 to 39. Economic dissatisfaction seems to have played a major role in Trump’s victory, but not the largest role.

Still, it’s important to bring economic issues into the picture. A razor-thin electoral victory has many parents, and we need to consider all the significant ones.

Moreover, if we’re trying to draw larger lessons, we should recognize that all three of the main factors identified by the exit poll—the economy, immigration, and terrorism—fit under the rubric of what political scientists call “sociotropic voting.” 

Sociotropic Voting

In the political science literature, sociotropic voting originally meant economic voting that’s guided by perceptions of the state of the economy as a whole, not by voters’ own financial situation. As opposed to “pocketbook voters”—who vote their economic self-interest—sociotropic citizens vote for what they think will serve the economic interests of everyone, or the majority, or those who most need help, in their society.

The sociotropic understanding of voting flies in the face of academic orthodoxy in economics, but this orthodoxy is a mere dogma. There’s no reason to think that people are everywhere and always self-interested. The assumption of self-interest does make sense as a starting point in analyzing economic behavior, because in modern societies, people are taught that self-interest is acceptable in their employment, business, consumer, and financial affairs. But they’re taught the opposite when it comes to government affairs. The standard, culturally accepted view is that public policy should advance the common good. So it’s not surprising that when non-economists talk about politics, the common good is what they talk about.

When you read interviews with actual voters—such as those excerpted in the political science classic, The American Voter (1960), or in more recent works, such as Katherine J. Cramer’s The Politics of Resentment (2016) and Arlie Russell Hochschild’s Strangers in Their Own Land (2016)—you’ll find the interviewees thinking hard about whether given policies, parties, and politicians serve the common good. Rarely will you find them contemplating how policies, parties, or politicians might serve the voters’ own interests. To be sure, the common good, in many voters’ estimation, usually includes the interests of “people like me.” But rarely do they seem to think of politics in terms of “me” alone.

Sociotropic voting has three profound implications.

  1. Sociotropic voting gives a more empiricist bent to ordinary politics than many political observers realize. In trying to determine what will serve the common good, values, “identities,” and a priori principles don’t go very far. Sociotropic voting requires thinking about facts—such as the unemployment rate, the employment effects of immigrants, and so on. And it requires theorizing, however inchoately, about how these facts interact with each other. Is an economic stimulus program likely to fix a recession? Will tariffs protect American jobs? Do immigrants reduce Americans’ wages? The answers to such questions depend on answers to broader questions about how the economy works, how the society works, and how the polity works. I’m not suggesting that more than a tiny fraction of voters pays explicit attention to such questions, but their political opinions and actions rely on tacit answers to them.
  1. Sociotropic voting is likely to lead to disagreement, because the facts and theories we use to answer such questions are obscure and contestable. 
  1. Sociotropic voting is—in the world as it is now—nationalistic. The national society is the context in which we most naturally think about the common good, given our nationalistic acculturation. The people governed by the nation-state inhabited by a sociotropic voter are the people about whose welfare the voter “naturally” thinks.


Media Effects

If the sociotropic effects of public policies aren’t self-evident, then where do people’s perceptions of them originate?

An obvious answer is that they originate with an array of specialists, ranging from academic experts to policy wonks to pundits, who spend their waking lives thinking about social and economic problems, politics, and government. These thoughts are relayed—primarily through the mass media—to members of the public, who tend not to think about politics and society very much. This division of labor adds, to the potential for disagreement among any two citizens about the obscure issues involved, the potential for disagreement among groups of citizens exposed to information from polarized groups of specialists.

That’s just an elaborate hypothesis, and unfortunately there is little research to back it up (or refute it). The possibility that ideas guide politics eludes most political scientists because they haven’t recognized the implications of sociotropic voting. And the possibility that the media convey these ideas from specialists to the public, while occasionally considered by political scientists, is difficult to study, because many of effects of the media would tend to be cumulative. So much is happening at once in a complex society that cumulative causes of long-term effects are difficult to isolate.

For a long time, then, research on “media effects” was scant, a situation that political scientist Larry Bartels, then of Princeton, once called “one of the most notable embarrassments of modern social science.” In the last decade, however, a small band of political scientists and economists have been making up for lost time.²

A 2007 study argued that when Fox News Channel entered a new cable market, its programming “convinced 3 to 28 percent of its viewers to vote Republican.” Another study in the same year found that northern Virginia residents randomly assigned a free ten-week newspaper subscription to the Washington Post voted for the Democratic gubernatorial candidate by 8-11 percentage points more than those who were randomly given a free ten-week newspaper subscription to the (conservative) Washington TimesA 2009 study of the effect of British newspapers showed that usually Conservative newspapers that switched to favoring the Labour party in the 1997 parliamentary election seem to have persuaded 10-25 percent of their readers to switch, too.

The most recent contribution to the new literature on media effects is even more stunning. One of the methodological problems in media-effects research is trying to figure out whether people choose a media source because they already agree with its political bias, or whether, having chosen it, it conveys its bias to them through the messages it transmits. But in a study appearing in the American Economic ReviewGregory J. Martin of Emory and Ali Yurukoglu of Stanford took advantage of the fact that when Fox News Channel was being added to various local cable lineups, it was assigned a channel number randomly. The authors show that lower channel numbers automatically tend to get higher viewership, enabling them to isolate the effect of watching FNC from the effect of choosing to watch it because of one’s conservative predispositions. (The Martin and Yurukoglu study is 108 pages long, but Dylan Matthews has a good writeup at Vox.)

During the period Martin and Yurukoglu studied (2000-2008), they found that “being induced to watch an additional hour per week of FNC by the channel position . . . would lead to an approximately 7.2 [percentage] point increase in the probability of voting Republican in presidential elections” (p. 17). The authors don’t claim that a lower channel position induced the average FNC viewer to actually watch an additional hour per week; that’s just a way of stating the magnitude of the effects that they discovered. But they do claim:

  • that “the effect of Fox news on centrist voters is consistently [conservative], ranging from 0.41 [percentage] points to 1.0 points in 2008” (p. 29);
  • that as more FNC channels were added during the 2000s, and as FNC became more conservative, it may have increased the Republican share of the presidential vote by up to 0.46 percentage points in 2000, 3.59 percentage points in 2004, and 6.34 percentage points in 2008 (p. 32); and
  • that between FNC, CNN, and MSNBC, “cable news can account for “all of the increase in [political] polarization” during the period, as measured by the distance between conservative and liberal citizens’ positions on an index of policy views (p. 32). (I don’t like to shout at readers, but that finding is so remarkable that it deserves emphasis.) The policy views in the index concerned “tax and spending levels, abortion, drug and crime policy, the environment, and attitudes towards business, labor, homosexuality, and religion” (p. 33).

These conclusions are built on very complicated models, and each complication adds room for error. But that’s a danger inherent to quantitative social science. For political scientists who insist that only quantification is scientific, the study lays down a challenge: turn your attention to the media, pronto. (I’d also suggest turning attention to the possible sources of media bias, such as the ideas taught to journalists by their own prior exposure to news media, cultural media, and formal instruction.)

However, for those of us who believe that social science should start from the realities revealed by introspection, even as it may end up challenging these realities, the Martin and Yurukoglu study merely confirms what Bartels was getting at years before the new wave of experimental media research got underway. We’re swimming in mediated messages about politics, so it stands to reason that these messages might influence our opinions; introspection suggests that they do. Again, I’d add that this applies not only to news media but entertainment media and formal instruction in elementary schools, high schools, colleges, and graduate programs. An epistemologically sensitive political science would investigate the possible effects of all of these mediators of messages about politics, society, and the economy, even if such research is sometimes incompatible with quantification.

The Logic of Media-Driven Polarization

Given the conservatism of Trump voters and their age—Fox News viewers skew older—the Martin and Yurukoglu study suggests that exposure to Fox News Channel may be a crucial reason for Trump’s victory. The same reasoning would also apply to consumers of other conservative media, such as talk radio. The talk-radio effect (assuming there is one) might also explain the rurality of so many Trump voters. People in low-density areas tend to spend lots of time driving over long distances just to do their daily rounds. As a former resident of south-central Texas, I can testify that talk radio is often the only thing to listen to in the wilds of ruritania.

To spell out how sociotropic voters’ reliance on the media can lead to political polarization, let’s return to the theory that Trump voters were particularly pessimistic about the trajectory of the economy. This has often been taken to show that they are irrational, inasmuch as the economy was performing well by the time the presidential campaign was underway; unemployment declined to 4.9 percent in 2016. However, while some media outlets might have emphasized this optimistic fact, others may have pointed out, more pessimistically, that the so-called real unemployment rate—the rate that includes those who have stopped looking for work—remained high. If MSNBC, CNN, and the non-cable mainstream media emphasized the positive aspect of the unemployment picture, while Fox emphasized the negative aspect, we would expect the polarized coverage to drive polarized economic perceptions in their respective audiences. Likewise with foreign trade. According to the exit poll, 65 percent of Trump voters agreed that “trade with other countries” “takes away jobs,” while 59 percent of Clinton voters thought that trade “creates more jobs.” Could this clash of perceptions, too, be a product of media polarization?

Turning to the interwoven issues of immigration and terrorism, a recent literature review concludes that Americans’ “immigration attitudes show little evidence of being strongly correlated with personal economic circumstances.” Instead, attitudes toward immigration “are shaped by sociotropic concerns about national-level impacts, whether those impacts are cultural or economic” (p. 1). Being concerned with “the nation as a whole” (p. 3), voters tend to favor the immigration of well-educated skilled workers or professionals who speak English; they oppose immigrants who lack these traits. They also oppose immigrants from terrorist-plagued countries such as Iraq (and presumably, now, Syria). Thus, they seem to want immigrants who will help the American economy rather than bringing down wages; immigrants who won’t be isolated from American society by a language barrier; and immigrants who don’t harbor violent designs on Americans. These preferences bear out the view that the opponents of immigration tend to be sociotropic nationalists.

But how can a nationalist voter decide, at a given time, whether immigrants will tend to have sociotropically desirable traits or not, and therefore whether he or she should support or oppose immigration? Similarly, what might have prompted Trump voters to think that the most important issues were immigration and terrorism?

Thanks to Martin and Yurukoglu, we don’t have to look far for answers. If Fox broadcasts the “news” that immigrants bring down Americans’ wages while other media outlets don’t; if Fox highlights terrorist attacks around the world more than other media outlets do; if Fox reports that French banlieus are “no-go zones” for the police, where shariah is the law of the land—and if, moreover, Fox promotes the view that immigrants drain the public treasury; and if it makes a household name among its viewers of Kate Steinle, murdered by an illegal immigrant in San Francisco; then it is perfectly logical that Fox viewers would oppose immigration for sociotropic reasons, that they’d think that terrorism and immigration are crucial interwoven issues, and that they’d respond positively to a politician who expressed the same ideas—and who criticized the mainstream media for being too politically correct to report on these realities (the realities portrayed by Fox).

To suggest that Trump voters were, to a significant extent, acting in accordance with what they learned about American society from Fox, and perhaps from talk radio, is to view them as rational human beings confronted with a complex world and doing what they can do deal with it well. We are all in the same situation.

But why would anyone deal with this situation by choosing to watch Fox? That, too, has a rational explanation. We approach any decision with an existing web of beliefs. This web determines for us which messages are plausible and which messengers are credible. Conservatives are people whose webs of belief have been significantly shaped by conservative messages, so they will naturally find conservative messengers credible. Liberals, moderates, and others use a similar process to decide which mediators to trust. 

The picture of politics that emerges from recognizing the implications of sociotropic voting aren’t restricted to the ones I’ve discussed. Nor am I done with the analysis of Trump voters, whose populism has yet to be discussed. Stay tuned.


¹However, it’s possible that Trump voters did want a president who cared about people like them in deciding what change to make; but that, when confronted by the survey researcher with the choice between designating this desire as a wish for change and a wish for a president who cared about people like them, they chose the former option.  

²Larry M. Bartels, “Messages Received: The Political Impact of Media Exposure,” American Political Science Review 87(2): 267–285; 267.


Jeffrey Friedman, the Director of the Niskanen Center’s Institute for the Study of Politics and a Visiting Scholar in the Charles and Louise Travers Department of Political Science, University of California, Berkeley, is the editor of Critical Review: A Journal of Politics and Society and of The Nature and Origins of Mass Opinion Reconsidered (Routledge, 2014).