The 2021 Economics Nobels: The Importance of Empiricism, and its Limits

The economics Nobel is given to individuals, but it often really intends to recognize schools of thought or methodological approaches. That is the case with this year’s prize, given to three economists who emphasize natural experiments in their research. David Card of the University of California-Berkeley won half of the award, while MIT’s Joshua Angrist and Stanford’s Guido Imbens split the other half. In the eternal debate between theory and experiment, the Nobel prize committee decided to award a point to the experimenters’ side. Somewhere, Francis Bacon and René Descartes are smiling.

This does not settle the matter, however. Good analysts use both theory and experiment, not just one or the other. More subtly, good analysts are also aware of the limits of both, as well as their virtues, and seek to find a healthy balance. Theories are useless without data to test them against. And data are useless without theories through which to interpret them. At the same time, both are subject to all kinds of human foibles, from sloppy thinking to cognitive biases to spreadsheet typos.

Card has had a long and varied career, but he is best known for his controversial study that found that a minimum wage increase in New Jersey actually increased employment in local restaurants—the opposite of what theory predicts. It was later revised as claiming the increase had no effect on employment. That study was coauthored with Alan Krueger, who might have shared Card’s half of the award had he not passed away in 2019 (The Washington Post’s Catherine Rampell, who was Krueger’s research assistant as an undergraduate, wrote a moving tribute to Krueger after his death). More about their minimum wage study below.

Angrist and Imbens have found their own natural experiments. Angrist, in collaboration with Krueger, found a new argument that education does, in fact, affect a person’s income. People born in the first quarter of a year are no different from people with birthdays elsewhere on the calendar. But because they are slightly older than their classmates, they are eligible to drop out of school earlier in the school year—at least in places with compulsory education laws. That naturally isolates an important variable, and even provides a control to check the results. And they found that people with first-quarter birthdays do, in fact, have slightly lower lifetime income than people with fourth-quarter birthdays, who must wait longer to drop out of school if they choose. In this case, the difference of about 1/10^th of a year of formal schooling is associated with an income difference of about 1 percent. Over the course of a lifetime, that can add up to thousands of dollars for many people.

Imbens’s best-known research is about how effective it is for doctors to encourage people to take flu shots. This is of obvious interest in the COVID-19 age. Imbens has also done extensive work on the theory of natural experiments, helping to bridge the gap between theory and experiment.

Many economists confine themselves to writing out thought experiments on blackboards. Card, Angrist, and Imbens ask: How well does this blackboard economics hold up in real life? Does the law of demand hold up as well as it does in textbooks? This isn’t a new idea. But economists, especially in the post-Samuelson era of abstract mathematics, need the occasional reminder.

Card and Krueger’s minimum wage study is one of the most famous attempts of the last 30 years to leave the blackboard behind. That is largely because it found that, contrary to two centuries of theory, a minimum wage increase in New Jersey did not reduce employment in restaurants, compared to next-door Pennsylvania, which did not increase the minimum wage. It remains the single most cited study by proponents of increasing minimum wages.

While its methodology is groundbreaking, this instance of it has serious problems. First, it relies on self-reported survey data. Restaurant owners do not want to appear stingy to other people, even on anonymous survey questions, so their answers are likely colored by social desirability bias. When the data going in are not reliable, the results are also often unreliable.

Second, although Pennsylvania and New Jersey are neighbors, they still have enough differences—and a border that thousands of workers cross every day—where the minimum wage difference is not exactly an isolated variable. The study does not account for changes in other industries such as retail, or for relevant changes in local tax rates, regulations, political leadership, or other factors.

Third, minimum wages affect more than just wages. Every worker also earns non-wage income. At a restaurant, this can mean complimentary meals or parking, employee discounts, tuition assistance, flexible hours, investments in better working conditions, or many other things. Many of these defy measurement, which is why many economists, including Card and Krueger, defy incorporating them into their research.

That matters, because owners will often go to great lengths to avoid layoffs and firings. If their wage costs go up, they will often offset them by cutting non-wage pay. In some cases, it means workers will get no pay increase, despite getting a larger formal paycheck. Firings are a last resort, which is on reason why so many studies find that minimum wages have smaller-than-expected employment effects. Card and Krueger are hardly the only analysts to forget about non-wage tradeoffs.

To be blunt about the Card and Krueger minimum wage paper, neither its data nor its conclusion hold water. But its methodological innovations, such as its differences-in-differences method, still make it worth studying. Their creativity in looking for natural experiments, if applied carefully, can enrich our understanding of any number of policy issues. The trick is to be careful and humble, rather than go for a headline-grabbing finding that makes a paper easier to publish and politically popular to cite.

None of this means that natural experiments do not deserve a Nobel prize, or that Card is not a worthy representative of that approach. Card’s creativity, and Angrist and Imbens’s, in finding natural experiments is impressive, and should influence other economists’ approaches.

The same is true of other empirically oriented prizes, such as former CEI Julian Simon Award winner Vernon Smith, one of the founders of experimental economics, or Elinor Ostrom, who did extensive field research on different ways people have found to solve the tragedy of the commons. Ronald Coase upended years of naysaying about private provision of public goods by going out into the world and asking a few questions. He found that many lighthouses—a classic public good—were, in fact, privately owned and managed. Future prizes will be given to other empirical innovators. This is good, and important.

It is also important to learn the limits of these approaches. People don’t always give honest answers in surveys. Not all results can be replicated independently, which is an important check built into the scientific method. If a researcher isn’t asking the right questions, what statisticians call the “dreaded third thing” might drive their results, rather than the variable in which they are interested—and they would never even know it.

With that in mind, congratulations to this year’s deserving winners. May they continue to find new ways to explore the real world, to be mindful of their human limitations, and to use sound price theory to interpret what they find.

For more on this year’s winners, see Alex Tabarrok over at Marginal Revolution, and David Henderson in today’s Wall Street Journal. For more on the non-wage effects of minimum wages, see my paper “Minimum Wages Have Tradeoffs.”