Josh Angrist is an acclaimed experimentalist who does not work in a lab. The economist keeps a modest office in MIT’s Building E52, where the most prominent object is often a bicycle leaning against one wall. Compact and athletic, tan and graying, Angrist, 52, rides to work most mornings. In recent years he has spent weekends tearing around mountain-bike trails with riders half his age.
From these informal quarters, Angrist has built a kind of virtual laboratory of economics, where he generates precise answers to difficult social questions. As much as any scholar, he has helped popularize the idea that microeconomic research can, and should, imitate the conditions of lab experiments. Many other microeconomists base their work on models that make large assumptions about human behavior. But Angrist uses only empirical data that illuminate causal relationships in society.
Consider an issue Angrist has been pondering a lot lately: the effectiveness of high schools. To evaluate schools, you might compare test scores, graduation rates, or college acceptance data. Yet it could just be that the top-rated school districts attract a greater proportion of families with well-prepared students.
Scholars can’t answer questions like this by randomly assigning students to schools themselves and studying the results. So to gain traction on such slippery problems, Angrist relies on natural experiments—cases in which two otherwise similar groups of people have been distinguished by one particular circumstance. If, say, a school district line is redrawn, instantly transferring one group of students to a new school, it might create what economists call a “clean identification” of cause and effect that isolates the schools’ own impact.
Over two decades, Angrist’s natural experiments have made him a prominent figure within economics. As of August, he was one of the 100 most-cited economists in the world, according to the Federal Reserve Bank of St. Louis, which keeps data on more than 33,000 authors. Among his best-known papers are studies on the relationship between length of schooling and income; the effect military service has on earnings; and the link between class size and student achievement.
Angrist did not invent his quasi-experimental methods; they were largely popularized from the 1980s onward by a group of prominent economists including Alan Krueger (currently chair of the White House’s Council of Economic Advisers), with whom Angrist has co-authored multiple papers; Lawrence Katz of Harvard University; David Card, now of the University of California, Berkeley, who was one of Angrist’s graduate-school advisors; and Angrist’s principal mentor, Orley Ashenfelter of Princeton University. But no one has been a more staunch advocate of lab-like economics.
“He’s had tremendous influence,” says Whitney Newey, PhD ’83, chair of MIT’s Department of Economics, who was one of Angrist’s graduate-school advisors.
Angrist’s citation ranking within economics probably understates that influence. Biostatisticians, who study cause and effect in medicine and biological research, regularly cite his methods, and political scientists and sociologists have adopted natural experiments as a basic research tool as well. Esther Duflo, PhD ’99, a prominent antipoverty researcher at MIT whom Angrist advised when she was a graduate student, says his relentless concentration on the problem of selection bias—for instance, the possibility that the better-rated high school is populated with better students—spurred her to go beyond his techniques and conduct actual field experiments. “Once you ask the question right, you can ask what the ideal experiment to answer it is,” she explains. “Josh [is] a champion of natural experiments, but that is also the work that [has] led many of us to think that true randomized experiments could be a very promising way to go.”
Yet if Angrist’s CV bears the markings of an academic star—PhD from Princeton, first job at Harvard, and a named chair at MIT, where he is the Ford Professor of Economics—his life could have been very different. Angrist left high school after the 11th grade, having completed the bare minimum of coursework needed to graduate. He took time off before deciding to go to college, dropped out of grad school, and then served in the Israeli army before pursuing a PhD in economics.
“I got a lot of lucky breaks in my life,” he says.
Angrist doesn’t think his research agenda, with its emphasis on the difference that education makes, has been driven by his own past. In economics, he says, it’s “a mistake to learn from your own experience” rather than being guided by data and a desire to study important topics. Still, it helps to know a little about Angrist in order to better understand his work, in part because he likes to probe the idea that contingent social circumstances can put otherwise similar people on different paths. After all, he himself could have gone in a few other directions—something that “colors my view of the world,” he acknowledges.
The twisting road to economics
Angrist grew up in Pittsburgh, with parents who taught at Carnegie Mellon. “I came from a very educated family,” he says. His father was an engineer who became a writer for Forbes and the Wall Street Journal; his mother was a sociologist who moved into private-sector jobs. For fun, Angrist would hop on the freight trains that rumbled through the Carnegie Mellon campus (“probably not the safest activity,” he reflects), and as a teenager he became more interested in cars than the classroom. He acquired his high-school diploma after meeting state requirements in English, health, and gym. While his former classmates were still in school, he got a job at a state mental hospital to pay for a car. His parents “thought it could have been worse,” he recalls.
After a year, Angrist did apply to college and persuaded Oberlin to admit him off the waiting list: “I went there and argued my case, and they saw that I was really into it,” he says. There, he finally began to flourish as a student; as a senior he wrote an honors thesis modeling how unemployment affects wage distribution. To help assess theses that year, Oberlin brought in Ashenfelter, a distinguished labor economist from Princeton. He was so impressed with Angrist that he invited him to become one of his PhD students.
“He was obviously a very fine student, and he had gotten into doing research at a very early age,” Ashenfelter recalls.
Opportunity was knocking, but instead Angrist went to Israel. He entered a master’s program in economics at Hebrew University in Jerusalem, but “I didn’t do well, and I dropped out,” he says. It was there that he met his wife, Mira.
Rather than try to get into Princeton, Angrist became an Israeli citizen and was drafted into the army. A paratrooper, he saw active duty toward the end of the war with Lebanon that had started in 1982; he was among the last troops to leave Lebanon in 1985. Military training was sometimes “fun,” he says, but combat was more daunting: a good friend was shot in Beirut. Angrist himself was fortunate, he says, that “the worst thing that ever happened to me was that I was scared.” Finally, before being discharged, he wrote to Ashenfelter to see if his offer still stood. A place in the economics program quickly materialized.
One day at Princeton, Angrist recalls, Ashenfelter mentioned to a class full of grad students that researchers at the University of California, San Francisco, had developed a natural experiment evaluating the effects of service in Vietnam on the mortality of former soldiers who had been drafted. Because of the random draft lottery, the researchers could assume that those men had not previously been more prone to suicide or fatal accidents than the general population, which served as the study’s control group.
“Orley said, ‘That’s such a great idea—somebody should do that for [former soldiers’] earnings,’” Angrist recalls. “I got to work that afternoon.”
The study became Angrist’s PhD thesis, though only after a lengthy process of extracting information from old government computers. “Very few people alive, including myself, would have managed to get that done,” says Ashenfelter. “It was like tunneling under the border or something—just one of these endless, tedious tasks.” Ultimately, Angrist found that white men who were drafted and served in the early 1970s earned about 15 percent less in the early 1980s than counterparts who were not drafted and never served.
Just the facts, man
Angrist’s thesis landed him a job at Harvard. Soon after, in 1991, he and Krueger published a paper on the relationship between education and earnings that remains a textbook example of a natural experiment—in fact, it appears in multiple textbooks. Many states, they observed, compelled children to start school during the calendar year in which they turned six and permitted them to leave upon turning 16. That meant not all dropouts spent the same amount of time in school: children born in December would start the preceding September, roughly at age five and three-quarters. Those born in January would start the following September, roughly at age six and two-thirds.
Thus dropouts born later in the calendar year, who started school younger, had to spend more time in the classroom before reaching age 16—and did in fact attend school longer, on average. Examining aggregate data for several decades of births, starting with the 1920s, Angrist and Krueger found that an additional year of education was worth about 7.5 percent more in annual earnings. As a natural experiment, the study works because people who dropped out at 16 were randomly “assigned” by birth month to receive different amounts of schooling.
Angrist has been adept at finding many different “instruments,” as economists call the variables they use to construct natural experiments. Consider another question: how does class size affect student performance?
Angrist and economist Victor Lavy used a policy quirk to shed light on this issue, publishing the results in a 1999 paper. In Israel, classes are capped at 40 students. If one school has 38 fourth-graders, those children will all have one teacher, while in an otherwise similar school that has 42 fourth-graders, the children will be split into two classes with an average size of 21 students. Examining 1991 test scores of students from more than 4,000 fourth- and fifth-grade classes in Israel, Angrist and Lavy found that smaller class sizes generated “significant and substantial” gains in math and reading achievement among the fifth-graders and smaller gains in reading among the fourth-graders.
Angrist has also produced important methodological work about the conditions under which natural experiments yield meaningful results, including a 1994 paper with Guido Imbens of Harvard on “local average treatment effects.” The paper formalized the idea that the average effect of, say, a new government policy is best measured by its influence on people who otherwise never would have found themselves in the circumstances it encouraged. Thus the average effect of Vietnam service on earnings, for instance, can only be determined for draftees; volunteers were excluded from Angrist’s natural experiment on veterans’ earnings.
In turn, this kind of estimate lets policy makers better forecast the potential effects of, among other things, expanding educational or health-care programs.
“That’s probably one of the very best papers over a 10-year span in econometrics,” says Newey, himself a leading econometrician. In 2009, Angrist and Jörn-Steffen Pischke of the London School of Economics also published a well-received book, Mostly Harmless Econometrics, that summarized much of their research about empirical methods in economics.
All this work has helped bring the idea of natural experiments from obscurity to the mainstream. “Josh has been ahead of his time in terms of wanting the world to be transparent when other people didn’t value that,” says MIT economist David Autor, referring to Angrist’s insistence on looking for empirical evidence rather than relying principally on models laden with assumptions. Through that work, Autor adds, Angrist has “figured out how to answer questions other people didn’t think could be answered” in empirical terms.
Angrist, for his part, says he is not opposed to model-based work for the purposes of, say, forecasting the effects of a policy change. But he maintains an “experimentalist mind-set” and believes that any such models should be based on significant amounts of empirical data.
To be sure, many economists, as Card puts it, still view their discipline as “a kind of mathematical philosophy” based on ideas about rationality and predictable responses to incentives. These scholars find pure empiricism “very alienating.” And some younger economists working in the mode of Angrist, Card, and Krueger have drawn criticism; they are sometimes depicted as opportunists looking for any topic that can yield a clear conclusion, even about something as seemingly inconsequential as the use of gym memberships. A 2007 article in the New Republic decried the “academic parlor game” played by new scholars using natural experiments.
“There has been some pushback in the last 10 years, that guys like me, or my students, or my school of thought—that we’re all about the tools and not about the questions,” Angrist says. “But I don’t think that’s fair.” He adds: “The reason the draft lottery was a good topic is not just because lotteries are cool. It’s because there’s something of substance there: guys being drafted, their careers being interrupted.” In general, he says, “it’s the combination of a cool tool applied to a central question that leads to good research.”
Case in point: Another economist whom Angrist helped advise, MIT professor Jonathan Gruber ’87, used natural experiments throughout the 1990s to study how people were affected by different health-insurance policies and programs, then used those results to project what would happen if affordable insurance were available to everyone. Gruber’s work played a key part in shaping the Obama administration’s Affordable Care Act of 2010. “Josh was a huge influence on me in grad school,” Gruber says. “He’s one of the key figures in this entire field [of natural experiments].”
In any event, Angrist rarely lets a little resistance flummox him. “One of the secrets of his success is his tenacity,” says Card. “He’s willing to work really hard and deflect criticism … You can go a lot farther in a tough field like economics if you have a little bit of that.”
Charting a new course
Recently Angrist has intensified his focus on education. In 2011, along with Autor and fellow MIT economist Parag Pathak, he launched the School Effectiveness and Inequality Initiative (SEII) to analyze issues such as the effectiveness of charter schools and the influence of financial aid on college performance. The idea is to inject lab-grade research into this civic debate. “School quality and human capital are major issues on the American policy agenda,” he says.
How far Angrist can take this research may depend on how many school districts will share useful data with him. Boston and Massachusetts, he says, are unusual in that “we have a great data infrastructure and wonderful coöperation with the city, the schools, and the state.”
That openness allowed him, Pathak, and other collaborators to produce some recent papers that he considers among his best. Boston has used a lottery system to determine which interested students will attend charter schools; when the number of applicants for a school exceeds the space available, the researchers are able to compare the performance of students accepted to a charter school with that of equally motivated students who were not chosen. In a 2009 report, they found that certain Boston charter schools had produced an average gain of roughly 15 percentile points for middle-school students on the state math exams.
Two years later, however, Angrist and colleagues found that in Massachusetts districts outside Boston, charter-school students did no better on average than students at other public schools. Angrist thinks charter schools may be too different from one another to justify sweeping conclusions about whether they provide better education, even though the best ones seem to adhere to the formula of extended instruction time, a focus on core math and reading skills, and an emphasis on good behavior.
“The charter schools’ idea is ‘Let a thousand flowers bloom,’” Angrist says. “Well, many of those flowers are dandelions … Charter schools are very heterogeneous.”
The SEII researchers emphasize that they are neutral on the subject of charter schools, which is politically charged because the schools draw upon public funding but largely use nonunion teachers. “We’re not charter advocates,” Pathak said after the first paper was released. “Our attitude is, let the data speak.”
The data may do more talking in the New Orleans area, where a lottery system assigns students to a mixture of charter schools run on varying principles. That may help the SEII scholars draw more conclusions about which kinds of charter schools seem to work best. Angrist and Pathak have also launched a related study of charter schools in the Rio Grande Valley of Texas. In another vein, an SEII team led by Angrist and Autor is analyzing the effects of a large college scholarship program for Nebraska students, funded by the Susan Thompson Buffett Foundation.
For now, the 2009 study on urban charter schools is widely assumed to have had a significant local impact. Both Massachusetts governor Deval Patrick and Boston mayor Thomas Menino adopted more charter-friendly positions soon after the research was released; Massachusetts limits how many students can attend charter schools, but a 2010 state law increased that number in districts with low test scores. Right now there are more than 70 charter schools statewide.
And as it happens, graduate students are researching the effects of that expansion and broadening the range of SEII studies. One of Angrist’s PhD candidates, Christopher Walters, has completed a paper estimating what would happen if Boston rapidly increased its number of charter schools. He concludes it would narrow the city’s racial achievement gap by 5 to 10 percent. But while previously low-achieving students gain the most from charter schools, Walters notes, they are the least likely to apply to them.
To reach his conclusions, Walters took empirical data from previous studies and, yes, built a model projecting future results. “Josh has been very open to the idea [of blending the data with a model],” he says. “I think that would be a surprising thing for his colleagues who know him as the king of the natural experiment.”
Angrist is also willing to suggest a larger conclusion emerging from his recent work: that students can learn a lot in their later school years. “We’re showing dramatic gains in middle school and later for kids who come in at a very low baseline,” he notes of the research on urban charter schools. The idea that children cannot compensate for disadvantages they face early in life is “a compelling narrative,” he says, “but it’s not true.”
On this point, Angrist is more than just a leading advocate of economics as a largely empirical discipline. His experience has also shown him firsthand how high the stakes are when it comes to, say, turning indifferent students into late bloomers like himself. Angrist’s view that secondary education can transform students’ lives happens to be shared by his children. His daughter, Adie, teaches at a Boston charter school, and his son, Noam, a senior economics major at MIT, cofounded an athletic and academic mentorship program for low-income Boston high school students.
Reflecting a bit before he gets on his bike for the ride home, Angrist sums up his unexpected career in economics: “I’m lucky I came to the conclusion that I would be better off if I went to college.”
The dark secret behind those cute AI-generated animal images
Google Brain has revealed its own image-making AI, called Imagen. But don't expect to see anything that isn't wholesome.
Inside Charm Industrial’s big bet on corn stalks for carbon removal
The startup used plant matter and bio-oil to sequester thousands of tons of carbon. The question now is how reliable, scalable, and economical this approach will prove.
The hype around DeepMind’s new AI model misses what’s actually cool about it
Some worry that the chatter about these tools is doing the whole field a disservice.
How Charm Industrial hopes to use crops to cut steel emissions
The startup believes its bio-oil, once converted into syngas, could help clean up the dirtiest industrial sector.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.