It’s heady company. When he won the John Bates Clark Medal earlier this month, University of Chicago economics professor Matthew Gentzkow suddenly found himself among legends such as Paul Samuelson and Milton Friedman. Both are past recipients of the award, which the American Economic Association bestows on the American economist under the age of 40 who “who is judged to have made the most significant contribution to economic thought and knowledge.”
Plenty of past winners have worked in familiar areas, such as wage dynamics or health economics. Gentzkow’s work is less orthodox: an interesting mix of the history and micro-economics of the media world. For instance, he’s studied the drivers of political “slant” in American newspapers. (Short version: Political slant tends to play to the views of readers, not owners.) Along with his frequent collaborator and University of Chicago colleague Jesse Shapiro, he’s investigated tendencies among consumers to read only online news sites that square with their own ideological biases. (Short version: They found no evidence that segregation among consumers of online news was becoming more pronounced.) His research has also found that television—and the television news which supplanted politics-heavy newspapers—has helped drive down US voter turnout.
Gentzkow sat down to talk about the future of economics, the state of the media ecosphere and virtues of “data hustle.” Here are some edited excerpts of our discussion.
An ever-expanding data toolbox
Quartz: First off, congratulations on the award. You’re in some pretty great company: Samuelson and Friedman and Feldstein and Krugman. So that’s got to feel pretty good.
Matthew Gentzkow: Thank you. Yeah, it’s a little intimidating but I’ll take it.
Your research deals with the media and, often, the push and pull between public opinion and the press. It’s not usually an area of study for economists, is it?
The truth is economists have been thinking about these questions and looking at the media for a long time. What changed really is the availability of data and technology that allowed you to get traction on those questions in a different way. I was lucky enough to kind of wander into this field at the time that that change was taking place. And so my work builds on previous work by lots of other economists who have asked these same kinds of questions.
What have you been able to do with data that you think has given you more tools to look at these questions?
Well, let me give you two examples. One is studying the content of news. That’s something people have been doing for a long time. In the past, if a researcher wanted to study the content of newspapers, what that typically meant was hiring some poor graduate students to sit down in front of a stack of hundreds and hundreds of copies newspapers and read through them, coding and checking off every time some word is used or coding various features of the content. That’s obviously a really labor-intensive, difficult kind of endeavor. A lot of people and a lot of graduate students did a lot of heroic work in that direction.
But the change to having digital text available for hundreds or thousands of media outlets covering many many years and thousands of different issues of each newspaper or each outlet, means you can do the same thing in an automated way on a much much larger scale. … So if you want to think about measuring content, studying what drives content, studying the effects of content, digital text changes that game dramatically. And the second thing that’s just changed is that questions that we’ve thought about in theory, we can now attack with data. And the whole enterprise can move in a more empirical kind of direction.
The death of theory
It’s interesting, this question of whether we’re entering some kind of post-theoretical age. Some people have talked about the “death of theory.” There was a big Wired magazine piece on that a couple years ago. Do we need theory anymore?
Absolutely. I disagree completely with the view that theory is dead or that even theory is less important than it used to be.
I think what is not always so productive is theory in the absence of data.
We can debate for a long time my model versus your model… Having data to discipline theory helps make that a much more productive process. But also having theory to guide the way you look at and understand data makes empirical analysis a much more productive process.
And I think if you look around economics today, it has certain become a much more empirical, data-driven field than it was 30 or 40 years ago. But a lot of the best work combines theory with empirical analysis; going back and forth between the two is where the really big gains are to be had.
On the virtues of “data hustle”
One of your papers was described in your citation as having “great data hustle,” which I thought was a pretty terrific compliment. I picture you, sort of, on the floorboards going for a loose ball or something. How important is doing the sort-of artisanal work of assembling data?
If you were giving advice to graduate students in economics today I think “data hustle,” entrepreneurial, creative gathering of new data, or getting access to data that people haven’t looked at before, is a huge part of success in the profession right now.
If you looked at top economic journals 20 years ago, or 30 years ago, a large share of the empirical papers were all basically using the same data. They were analyzing the census or some other large publicly available data set. If you look at a top journal in economics now, just about every paper has a different source of data and many of those, a very large share of them are proprietary data sources that the authors have either collected themselves or gotten access to in ways that took a lot of work.
One of the big exciting things on the frontier right now is using big administrative data sets. There’s been for example a lot of really exciting work done by Raj Chetty, who’s at Harvard, and Emmanuel Saez, who’s at Berkeley and a group of co-authors using data that they got access to from the IRS. That’s the first time, looking at the US, that people actually have administrative records of people’s earnings, all of the sort of individual-level economic outcomes, and that’s been incredibly fruitful in terms of generating new insights. But getting access to that meant them working really hard to talk to the right people to convince them that what they wanted to do would be valuable and wasn’t going to jeopardize privacy or cause any other concerns.
The return of the partisan press…
You’ve looked a lot at the history of American newspapers, going back to their roots as ideological party organs in the 19th century, as well as the advent of television, and more recently online news. Is there some sort of grand unified theory or thread running through all that work that you were surprised at?
In some ways, the US media today looks increasingly like the US media of the 19th century. Back in the day we had fiercely competitive, partisan newspapers going after each other, wearing their ideological views on their sleeve … not pulling any punches talking about scandals and using all kinds of inflammatory language. That is very much like what we see if you turn on cable TV or you look at political blogs.
And really the exception, historically, is the period that I grew up in and the period that many people grew up in. We had three broadcast networks and everybody got their news from the same places. People would argue about the political slant of the broadcast networks, [but] they certainly presented themselves as very objective and sort of partisanship-free. That was really the unusual period. When you go back and look at partisan newspapers in the past, things look awfully similar to what we see today.
…and the “wonk bubble”
A self-serving question—because Quartz is part of this—I wonder if you’ve been keeping up with the new media start-up flurry that we’re seeing right now, and if you have any thoughts on it.
I’d say it’s sort of what we would have expected. Because the internet clearly disrupted the business model for journalism that meant a lot of incumbent firms that had built businesses around older technologies were hurt. And it also meant there were big opportunities for new people to walk in and try new things.
Whenever there’s a big change like that it creates a lot of disruption and a lot of hand-wringing… But usually what happens with those disruptions is, after a little while, there’s then a burst of innovation and you look around and you say, “Wow this is actually a lot better than we had before.” So, that doesn’t have to be the way things play out. But it often is the way things play out. I have often suspected that that’s the way things were going to play out in terms of online news.