Mega data

Chicago Booth economist Matthew Gentzkow sifts insights about the media from massive amounts of digital information.


Economist Matthew Gentzkow’s award-winning work could not have been done 20 years ago. His analysis of media bias and the perceived ideological echo chamber of online news, to name two of his more prominent research topics, required what Chicago Booth colleague Austan Goolsbee called the “unfathomable data sets” that Gentzkow gathered.

His generation of economists is the first to have access to such a wealth of media information, both from increasingly digitized historical records and real-time contemporary data on what people read, watch, think, and buy. Few of his peers have put the raw material to more productive use.

The 2014 recipient of the John Bates Clark Medal, presented annually to the top American economist under age 40, Gentzkow demurs. To hear him tell it, his graduate school experience at Harvard, where he completed his PhD in 2004, merely steered him toward an opportune intersection of this vast information and his own interests. “Most things, if they’re super interesting, either someone’s answered them already,” Gentzkow says, “or if they haven’t, there’s a reason they haven’t.”

(Photography by Drew Reynolds)

His curiosity about media—from journalism to advertising, in print, on television, and online—provided potential research topics. Increasing data accessibility paved a broad new avenue to pursue them. “I’ve been able to look at some questions that people have thought about for a long time and maybe make more progress on them,” Gentzkow says, “because now the scale at which you can do things is much bigger.”

Still. Try searching for relics of his postundergraduate year with a Maine theater company and it’s clear that the mere existence of the internet doesn’t necessarily yield information conducive to fruitful research.

Digital archaeological digging turns up nothing from those days, only a fragment from his college directorial career. A 1997 review in the Theater Mirror—“New England’s LIVE Theater Guide”—of the Gentzkow-directed Goose and Tomtom at the Harvard-Radcliffe Summer Theatre commends him as an “excellent referee” for a strong cast and script. That’s about it—and probably the right amount.

Before and since his theatrical dabbling, Gentzkow has been an economist at his core. He arrived at Harvard as an undergrad with a vague interest in social science, particularly as it applied to the problems of poverty and development. Economics, he discovered, offered a means to study a wide range of questions relevant to those issues while applying his interest in mathematics.

In retrospect Gentzkow’s professional path seemed set with that choice of concentration. Only his extracurricular excursions into theater varied his course. And as he reflects on that diversion from nearly two decades removed, he finds that theater and research really involve overlapping synapses.

“They’re both entrepreneurial. You’re making them up as you go along.” Economics, like theater, is not a programmatic application of learned skills, he says, but a constant act of innovation, whether the framework is a script or a data set.

Announcing Gentzkow’s Clark Medal last year, the citation from the American Economic Association sounded not unlike a review, complete with blurbs worthy of being exclaimed from a theater marquee: “Great data hustle!” “Frontier methods!” “Creative without sacrificing quality!”

The Clark Medal also brings Nobel buzz. Past UChicago winners include eventual laureates Milton Friedman, AM’33; Gary Becker, AM’53, PhD’55; and James Heckman, and there’s more than just an anecdotal correlation. A 2009 Economics Bulletin study presented “the first statistical evidence that John Bates Clark medalists and individuals affiliated with the University of Chicago have a higher chance of winning the Prize.”

Gentzkow prefers not to think about that. In an interview with Harvard’s Neiman Foundation, he noted the uneasy idea of winning the Clark Medal, by definition at a young age, and then not producing work worthy of the Nobel. Leaning back on the couch in his Harper Center office, feet up on a coffee table, he seems most comfortable in the weeds of his research, which includes the theory of persuasion, consumer brand preferences, and health care spending.

Originally growing out of the brand studies, the health care spending work has led him “a little distant from what I’ve done in the past,” Gentzkow says, but in a way that hews to his sense of innovative inquiry. His celebrated research on the impact of emerging media—including how the introduction of television affected test scores and voting habits, and the internet’s effect on political divisions—likewise came from a thought process he considers similar to a playwright’s. “You wake up in the morning and you have to ask yourself, ‘What am I going to try and create?’”


For Gentzkow, the Richard O. Ryan Professor of Economics and a Neubauer Family Fellow, the creative aspect of economics refers more to the process than the outcome. It’s not about the answers produced, but the questions asked. Graduate school taught him that. Classwork offers a necessary foundation to understand the discipline, but “the really hard skill is, how do you identify good research questions?”

He has a formula of sorts. The subject has to be personally interesting—“you’re going to spend most of your life thinking about it”—important to the field, and with the potential for meaningful progress. American media offered all of that.

For 150 years, Gentzkow notes, newspaper, radio, and television companies have conducted private market studies because of the value to advertisers of detailed audience information. “There’s a tremendous amount of measurement that this industry had done,” he says, “not for research reasons but for commercial reasons.”

The data exists and technology has made it increasingly available to researchers. But what to do with it? One question among many that occurred to Gentzkow and Jesse Shapiro, a former Chicago Booth colleague and frequent collaborator, involved the impact of the internet on how people got their news. Was it true that new media had splintered the long-standing edifice of “mainstream” reporting and amplified more polarized views, drawing audiences away from traditional outlets and driving Americans apart politically?

(Photography by Drew Reynolds)

“We were really playing directly off of a hypothesis that other people had put out there,” Gentzkow says, specifically that the availability of more information sources led people to segregate their media consumption into ideological camps. Liberals watched MSNBC, conservatives Fox News, and they lived in distinct and strident online media universes, tuning out anything contradictory. This sorting, the argument goes, creates an “echo chamber” with a potentially pernicious effect on civic life. Cass Sunstein, a former Law School professor now at Harvard, wrote about the subject in (Princeton University Press, 2001). Online echo chambers, he argued, limited the exposure to conflicting points of view that are “central to democracy itself.”

Survey evidence that Gentzkow and Shapiro had seen indicated that the perceived effect did not exist. When they mentioned those surveys, though, skeptics pointed out that the results were based on self-reporting, suggesting that the subjects exaggerated or lied about their habits.

Online data made more precise research possible. In addition to sheer volume, digital information offers another advantage to the study of media consumption: it reduces the inaccuracies that self-reported surveys permit. Internet browsing leaves a data trail that cancels out the effect of misleading answers that could skew results, and that’s where Gentzkow and Shapiro went searching for empirical answers. 

They found that those self-reported surveys were actually pretty accurate. Increased media segregation was not happening. If anything, Gentzkow says, the trend is drifting in the other direction, although he emphasizes the difficulty of making comparisons across years.

They did find more partisan segregation in online news readership than among television viewers or readers of local papers, but not as much as in national newspapers. The New York Times, the study shows, has a significantly more liberal readership than USA Today or the Wall Street Journal. Overall the data did not support the idea that the internet created the echo chambers that have been a common cause for concern in both academic studies and popular commentary.

Partisan sites accounted for only a small proportion of online readership, which by and large is “concentrated in a small number of relatively centrist sites.” And readers of online outlets on the ideological poles were more promiscuous browsers. “Visitors of extreme conservative sites such as and are more likely than a typical online news reader to have visited,” the study reported. “Visitors of extreme liberal sites such as and are more likely than a typical online news reader to have visited”

Although evidence shows most people gravitate to relatively few common sources, the real-life political conflicts that have given rise to the internet echo chamber hypothesis do exist. Gentzkow refuted online fragmentation as a potential cause, but his paper raised new questions that he seems to relish asking. “Is it that people are interpreting what they see differently? Is it that even though they’re seeing similar stuff, they’re paying attention differently to it? Is it that everybody’s views are fixed and [their partisan response] depends on who they talk to and who their friends are?”

Gentzkow’s data hints that the answers to those questions could well be yes. More than any form of media, areas that showed the highest rates of political self-segregation included the workplace, family relationships, and the most by a significant amount, networks of trusted friends. The General Social Survey, conducted every two years by NORC at the University of Chicago, provided the information about personal interactions in Gentzkow’s study, drawn from respondents answers to questions about the political leanings of those in their social orbits.

One potential interpretive filter that the echo chamber paper did not address was social media. When Gentzkow and Shapiro published their research in 2010, Facebook and Twitter didn’t have the cultural traction they do now.

He hasn’t studied the subject himself, but several researchers reported their results at a Becker Friedman Institute panel last spring. Gentzkow was left with the impression that social media is highly polarized but a minor part of the bigger picture. “Most of the political stuff that’s consumed through social media is opinion, not news stories,” he notes, “and the opinion that gets passed around by conservatives is very different from opinion that’s passed around by liberals, but that’s still a very small share of a broad diet.”

Gentzkow has a simple explanation for the persistence of traditional outlets atop the media food chain: it’s expensive to produce news and few companies have the resources to reach an audience wide enough to support firsthand reporting. Technology has transformed the methods of delivery, but media economics dating back at least to the 19th century remains a barrier to entry.

“One of the big underlying insights for me from that echo chambers paper is the internet actually isn’t very different from traditional media,” Gentzkow says. Because sites with only niche appeal typically cannot afford to cover the news, major outlets remain most people’s predominant source of information. “Writing the Buddhist vegetarian perspective on the Iraq war, and the Buddhist vegetarian perspective on what’s happening in Ukraine, and sending reporters over there to write that, that would be a lot of cost for a very small audience.”


Television has always attracted mass audiences, although it too has now grown and fragmented into a dizzying array of options, its established business practices suffering from internet-driven pressure. Not so long ago, though, TV was new media, and the era of its piecemeal introduction in the United States offered Gentzkow a foothold on complicated research questions.

Disentangling viewing habits from other personal circumstances makes studying the effect of television on, say, student test scores tricky at best. “Because kids who watch six hours of TV are different in all kinds of ways,” Gentzkow says. “You can bet that their parents have different levels of education, you can bet that they have different levels of income, you can bet that the environment they are living in is different.”

One point in time offered insight into the effect of increased television viewing that could be distinguished, to some degree, from those other factors: the introduction of TV into American homes. The process did not follow an obvious geographic pattern, creating randomness conducive to research. Although major markets like New York and Los Angeles were predictably at the forefront, because of the regulatory system “there was a lot of variation mixed in that was pretty random,” Gentzkow says. “That seemed like a good kind of natural experiment to look at the effect of TV.”

There’s an assumed negative correlation between the amount of television children watch and their results on standardized tests: more TV, it appears, equals lower scores. But like the perceived ideological segregation of news sources, that connection turned out to be false, according to Gentzkow’s analysis in a 2008 paper coauthored with Shapiro—with whom Gentzkow has said his Clark Medal should be shared. “What we found was, one, there’s no effect that television reduced the kids’ test scores, and two, there’s some evidence that it actually had a positive effect, especially for more disadvantaged kids.” Those whose parents did not speak English, for example, showed benefits from the exposure to the language.

At issue, Gentzkow argues, is not whether watching television itself is a positive or negative, but what children might be doing otherwise. The impression he gets is that people perceive the alternatives to be mentally enriching, like reading or reviewing homework with a parent. In fact, he says, for many kids “TV could be relatively more rich educationally” than their other options. He notes that many people view television’s impact through the lens of what their own family might do instead: visiting museums, for example. Children who are not exposed to those options might otherwise play with friends or toys, activities that provide less intellectual stimulation than they could find on TV.

The importance of gauging television’s impact in the context of what it replaces really registered for Gentzkow in a 2006 paper on voting patterns. There was a sense in the early days of the medium that TV might increase political participation. Its efficiency in delivering information, many thought, would reach more people and translate to higher turnout at the polls. The opposite happened. Especially in local elections, he found, lower turnout could be attributed to television’s introduction.

What did it displace in people’s media diet? Newspapers or radio with more local coverage. “You’re switching from reading your local Santa Fe newspaper … to national NBC programming,” Gentzkow says. “So a lot of the effects on voting seem to relate to what it’s crowding out.”

Beyond the nature of the information, television also altered the American entertainment universe and people started “spending a lot more time watching I Love Lucy.” Time devoted to such expanding entertainment options, he notes, might have previously been spent in a more thorough reading of the local newspaper.

His findings present a mixed media message. Television might not have been the educational bane for children it has been thought to be, but it also wasn’t a path to increased civic engagement. Likewise with the internet, the gateway to instantaneous information from anywhere in the world—and also cat videos.

“I think it’s pretty hard to argue that it’s not an improvement,” Gentzkow says of the access the internet offers to a diverse and global array of sources. “So you might think therefore everyone must be much more informed now than they were before. Well, it depends.”

As evolving media has increased the flow of information and entertainment, and improved the means of education and distraction, studies about how informed the public is have held relatively steady over time. Technology makes access to everything easier, Gentzkow says, but it’s as if two competing forces are “balancing out,” changing the way in which people consume news and find amusement, but not necessarily for better or worse.

The Economist, in fact, found his body of work heartening. The theme running through Gentzkow’s research, the magazine said, “reinforces the simple but reassuring point that what readers want most is to be informed.”

Join The Discussion

Log in with Disqus to automatically enter your contact information.