The ‘natural experiments’ approach to economics that won three researchers the 2021 Sveriges Riksbank Prize in Economic Sciences has helped to make the field more robust, say economists.
Joshua Angrist at the Massachusetts Institute of Technology (MIT) in Cambridge, Guido Imbens at Stanford University in California and David Card at the University of California, Berkeley, received the award for work that shows how causation can be inferred from observational data in real-world natural experiments. Their work has been used to examine, for example, how different minimum wages affect jobs and businesses; and the economic impacts of immigration.
“Causality is the holy grail in economics,” says Diane Coyle, an economist at the University of Cambridge and author of a new book, Cogs and monsters: what economics is, and what it should be. The decision will be widely welcomed among economists, she adds. “These are people who have driven forward the ‘applied turn’ in economics through their work on methods for detecting causal relationships.”
Understanding cause and effect in social science is hampered because controlled experiments — such as randomized controlled trials (RCTs) — might not always be practically or ethically possible. But economics has undergone “a credibility revolution”, says macroeconomist Lisa Cook at Michigan State University in East Lansing, “and these folks were central to it”.
The award came as a “complete surprise”, Card told Nature. “I thought that there are many very deserving alternatives.” Card thinks the main impact of the trio’s research is “in getting people to spend a lot more time and effort developing credible, high-quality empirical evidence, and in ‘calling out’ weak evidence”.
“Before these three and [the late Princeton University economist] Alan Krueger”, economists “were content to say correlation is not causation, and then just toss that maxim out and make causal claims without the evidence for them”, says Cook.
By making the conclusions of economics research more robust, the laureates have helped to secure observations as a foundation for developing reliable policy decisions. “You don’t want to use a bunch of correlations to inform policy,” says Cook. “You want to be as sure as you can be that there’s a causal connection between whatever policy you want to put in place and the outcome you’re interested in.”
She stresses, however, that the approach of using natural experiments is no panacea. “Sometimes,” she says, “we don’t have enough facts to put together the puzzle” to make a reliable causal story.
An ethical solution
In the natural sciences and medicine, researchers might seek to establish causal relationships by conducting experiments, such as RCTs, in which one variable is altered while others are kept the same.
Drugs, for example, are tested using RCTs, with one group of people being randomly assigned to receive the drug and another assigned to receive a placebo. RCTs can sometimes be used in economics, too — they were central to work that saw the 2019 prize awarded to Esther Duflo and Abhijit Banerjee, both at MIT, and Michael Kremer, now at the University of Chicago in Illinois. But often, this approach is impossible or even unethical — for example, it wouldn’t be appropriate to impose unequal wages on a group of people to assess the relationship between income and productivity. All three laureates have been crucial to the development of an alternative strategy.
In 1991, Angrist, working with Krueger, considered the question of whether people with more education are more likely to earn more throughout their working lives1. They pointed out it can't be assumed more education causally drives higher incomes. This is because people who choose to study for longer periods might alternatively have associated motivations or attributes that boost their earnings, too. To establish whether educational attainment does drive incomes, the two researchers looked at the natural experiment provided by compulsory schooling. In the United States, people born at different times in a given year will be in the same class at school. That provides a natural cohort to measure the relationship between how long someone spends in school, and their earnings.
Card and Krueger, meanwhile, considered what happens to employment levels when a minimum wage is introduced, or altered. In the early 1990s, they used the natural experiment provided by comparing employment levels in two neighbouring US states, New Jersey and Pennsylvania, which had different minimum wages, but which could otherwise be considered comparable2. Card also looked at the effects of immigration on wages — an issue relevant to debates today in the United States and Europe.
In 1994, Angrist and Imbens developed a mathematical formalization for extracting reliable information about causation from natural experiments, even if their ‘design’ is limited and compromised by unknown circumstances such as incomplete compliance by participants3. Their approach showed which causal conclusions could and could not be supported in a given situation.
Thus there are close connections between the three recipients’ work, says Card. “Much of the work by me, Angrist and Krueger in the late 1980s and early 1990s was motivated by the same broad set of concerns over the credibility of standard approaches for studying things like minimum wages, the return to schooling, the costs of military service, and the effects of immigration,” he says.
Card, like many others, thinks that if Krueger — who chaired the Council of Economic Advisers in the administration of former US president Barack Obama — had not died in 2019, he would have shared the prize, too.
Angrist, J. D. & Krueger, A. B. Quart. J. Econ. 106, 976–1014 (1991).
Card, D. & Krueger, A. B. Am. Econ. Rev. 84, 772–793 (1994).
Imbens, G. W. & Angrist, J. D. Econometrica 62, 467–475 (1994).