A “CREDIBILITY REVOLUTION” has transformed economics since the 1990s. Before that, theory ruled the roost and empirical work was a poor second cousin. “Hardly anyone takes data analysis seriously,” declared Edward Leamer of the University of California, Los Angeles, in a paper published in 1983. Yet within a decade, new and innovative work had altered the course of the profession, to such an extent that the lion’s share of notable new research today is empirical. For helping enable this transition David Card of the University of California at Berkeley shares this year’s economics Nobel prize, awarded on October 11th, with Joshua Angrist of the Massachusetts Institute of Technology and Guido Imbens of Stanford University.
The messy real world can often defy economists’ attempts to establish causality. Working out how a rise in the minimum wage affects employment, for example, is complicated by the fact that some other influence (a chronically weak labour market, say) may have contributed to changes in both policy and employment. In other fields researchers establish causation by designing experiments where subjects are randomly assigned to different groups, only one of which receives a particular treatment, so that the effect of the treatment can be clearly seen. More economists are also using randomised controlled trials—indeed, the Nobel prize in 2019 rewarded such efforts. But many questions cannot be studied this way for reasons of politics, logistics or ethics.
This year’s prizewinners surmounted such hurdles by using “natural experiments”, in which some quirk of history has an effect similar to an intentional trial. In a landmark paper published in 1994, Mr Card and Alan Krueger studied the impact of a minimum-wage increase in New Jersey by comparing the change in employment there with that in neighbouring Pennsylvania, where the wage floor was unchanged. Although theory predicted that a minimum-wage rise would be followed by a sharp drop in employment, such an effect, strikingly, did not seem to hold in practice. The paper inspired further empirical work and injected new energy into thinking about labour markets. Krueger, who died in 2019, would probably have shared the prize had he lived.
The use of natural experiments quickly spread. Mr Card analysed another unique circumstance—Fidel Castro’s decision in 1980 to allow emigration out of Cuba—to examine the effects of immigration on local labour markets. About half the 125,000 Cubans who fled to America settled in Miami. By comparing the city’s experience with that in four other places that were similar in many respects, but which did not receive an influx of migrants, Mr Card found that neither the wages nor the employment of native workers suffered as a result of the migration.
Mr Angrist, together with Krueger, used a similar technique to examine the impact of education on labour-market outcomes. Because students of a more scholastic disposition are likely both to spend more time in school and to earn more in work, what looks like a return to education could in fact reflect natural aptitude. In order to determine causality, the researchers made use of odd characteristics of America’s educational system. Although laws typically allowed students to drop out of school when they turned 16, all students born in the same year began school on the same date, regardless of their birthday. Those born in January, therefore, received more schooling, on average, than those born in December—and, the researchers found, also tended to earn more. Since the month of a student’s birth may be assumed to be random, they concluded that the added education caused the higher earnings.
The study of schooling found that an extra year of education raised subsequent earnings by 9%. Such an effect seemed implausibly large to many economists. But that reflected a difference in definition, concluded Mr Angrist in work with Mr Imbens. The two scholars noted that the effect of a “treatment” would not be the same for everyone in a natural experiment. If the age at which students could drop out were raised from 16 to 17, for example, some would be forced to receive an additional year of schooling; others, who had always intended to stay in school for longer, would be unaffected.
Together, the researchers developed statistical methods to make the conclusions from natural experiments more useful. Economists refer to the quirky factor used in natural experiments (like the birth month of a student) as an “instrument”. Messrs Angrist and Imbens explained the assumptions that need to hold in order for an instrument’s use to be valid: it must, for instance, only influence the outcome being studied (earnings, in this case) through its effect on the treatment (years of schooling), and not through other channels. By laying out these assumptions, the researchers allowed for more sophisticated analysis: the boost to earnings in the case above, for instance, only applies to students born early in the year and who are forced to stay in school for longer than they would have otherwise. Moreover, the methodology the researchers described improved the transparency of economists’ findings. The reader of a paper can judge for themselves how well an instrument satisfies the necessary assumptions, and discount the result accordingly.
It’s only natural
The credibility revolution, like any big upheaval, has had its excesses. Critics point to careless work and the mining of data in search of results that seem meaningful. Scholars are occasionally too eager to extrapolate findings from a particular natural experiment in ways that may not be justified, given the uniqueness of the circumstances. Yet the innovations developed by this year’s prizewinners unquestionably changed the field for good, illuminating questions once shrouded in darkness and forcing economists to push theory in directions that better describe real-world experience—a cause, indeed, for celebration. ■