Nobel Prize in Economics 2021: Natural Experiments and Causal Analysis

This note covers the Nobel Prize in Economics 2021: who won it, what "natural experiments" are and how they let economists answer cause-and-effect questions, how David Card used them to study minimum wages, immigration and education, how Joshua Angrist and Guido Imbens built a method to interpret such experiments correctly, how the discovery unfolded, why it matters and quick facts for exams.

What was the Economics prize 2021 awarded for?

The official citation for the prize reads, in two parts. David Card received one half of the prize "for his empirical contributions to labour economics", while Joshua D. Angrist and Guido W. Imbens shared the other half jointly "for their methodological contributions to the analysis of causal relationships".

In plain words, the three laureates showed how economists can answer questions about cause and effect even when they cannot run a controlled laboratory experiment on people.

Card used real-world situations, called natural experiments, to measure how things like a higher minimum wage or a wave of immigration actually affected jobs and pay.

Angrist and Imbens worked out the precise mathematical conditions under which such natural experiments allow a researcher to draw a trustworthy conclusion about cause and effect, rather than a misleading correlation.

Together, what they received is officially called the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, widely known as the Nobel Prize in Economics.

Who are the laureates?

David Card

David Card was born in 1956 in Guelph, Canada. At the time of the award he was the Class of 1950 Professor of Economics at the University of California, Berkeley, CA, USA. He received one half of the prize.

Card's contribution was empirical: using natural experiments from the early 1990s onward, he studied the labour market effects of minimum wages, immigration and school resources, and his results often challenged what economists had previously believed.

Joshua D. Angrist

Joshua D. Angrist was born on 18 September 1960 in Columbus, OH, USA. At the time of the award he was the Ford Professor of Economics at the Massachusetts Institute of Technology (MIT), Cambridge, MA, USA.

He received one quarter of the prize, shared with Imbens. With Alan Krueger, Angrist pioneered the use of birth-date rules as a natural experiment to study the returns to education, and with Imbens he developed the statistical framework for interpreting such studies correctly.

Guido W. Imbens

Guido W. Imbens was born on 3 September 1963 in Geldrop, the Netherlands. At the time of the award he was the Applied Econometrics Professor and Professor of Economics at Stanford University, Stanford, CA, USA.

He received the remaining one quarter of the prize. Working with Angrist from the mid-1990s, Imbens helped show exactly what can be learned about cause and effect from natural experiments where people choose whether to take part, rather than being forced to.

What problem were they trying to solve?

Many of the biggest questions economists and policymakers face are about cause and effect: does raising the minimum wage destroy jobs? Does immigration push down wages for workers already living in a country? Does an extra year of schooling raise future income? The trouble is that we can never watch the same group of people live through both the policy and its absence at the same time.

In medicine, researchers solve this with randomised controlled trials, where a flip of a coin decides who gets the real treatment and who gets a placebo.

But you cannot randomly force some children to drop out of school and others to stay, or randomly assign a minimum wage to one town and not another, for ethical and practical reasons.

So for decades, economists relied on comparing groups that differed in many uncontrolled ways, and any link they found between, say, education and income, might really be caused by some other factor, such as natural talent or motivation, that affects both.

The laureates' shared insight was that the real world sometimes does the randomising for us.

A natural experiment is a situation, outside anyone's planning, in which chance events, government rules or policy changes place similar groups of people on different sides of a divide, as if by a lottery.

If researchers can find and correctly use such situations, they can get close to the reliability of a controlled trial without ever running one.

How does Card's work with natural experiments work?

David Card's approach was to find a real situation where a policy changed for one group but not for a very similar comparison group, and then compare how the two groups evolved afterwards.

His best-known study, with the late Alan Krueger, examined the minimum wage. In the early 1990s most economists believed raising the minimum wage would mean fewer jobs, because it raises the cost of hiring.

To test this, Card and Krueger used a natural experiment: in the early 1990s New Jersey raised its minimum hourly wage from 4.25 dollars to 5.05 dollars, while neighbouring Pennsylvania did not change its wage.

Because labour markets on either side of a state border tend to move together, any difference in job numbers that appeared only in New Jersey could reasonably be blamed on the wage rise.

Card and Krueger focused on fast-food restaurants, an industry where many workers earn close to the minimum wage. Contrary to the conventional view, they found that the wage increase had no effect on the number of workers employed.

The press release notes that increasing the minimum wage "does not necessarily lead to fewer jobs."

1. Identify a real policy change that affects one group but not a similar neighbouring group.
2. Check that the two groups would otherwise have evolved similarly, so differences can be attributed to the policy.
3. Collect data on the outcome of interest, such as employment, for both groups before and after the change.
4. Compare the change in the treated group with the change in the untreated comparison group.
5. Interpret the difference as the causal effect of the policy, while checking that no other factor explains it.

Card used the same logic to study immigration. In April 1980, Fidel Castro unexpectedly let any Cuban who wished to leave do so; between May and September that year, 125,000 Cubans emigrated to the United States, many settling in Miami, raising the Miami workforce by about seven per cent.

Card compared wages and employment in Miami after this sudden shock with four comparison cities and found no negative effects on wages or employment for Miami residents with low levels of education.

He also studied school resources, comparing adults who had grown up with different levels of school funding, and found that greater resources raised the later returns to education, an effect that was particularly strong for students from disadvantaged backgrounds.

Draw two parallel timelines, one labelled "New Jersey" and one "Pennsylvania", each with a line for employment before and after the date the minimum wage rose.

Mark the policy change only on the New Jersey timeline, and show that the two lines move together afterwards, illustrating that the wage rise did not pull New Jersey's employment line down relative to Pennsylvania's.

How did Angrist and Imbens fix the interpretation problem?

A natural experiment is messier than a clinical trial, because people are not forced into a treatment group; they usually choose for themselves whether to take the opportunity a policy offers. This creates an interpretation puzzle.

Suppose a law lets students leave school a year earlier if they were born early in the year. Not everyone born early actually leaves early: some would have stayed in school regardless.

So when researchers compare the "early birth quarter" group with the "late birth quarter" group, only some people in each group actually changed their behaviour because of the rule.

In 1994, Joshua Angrist and Guido Imbens worked out exactly what such a comparison can and cannot tell us.

They showed that the effect we measure from a natural experiment only applies to the people whose behaviour was actually changed by it, a group they called compliers.

They named this specific effect the local average treatment effect, usually written LATE.

Their method works in a sequence of linked steps, known as the instrumental variables approach:

1. Measure how the natural experiment (for example, being born early in the year) changes the probability that a person actually receives the treatment (leaving school early).
2. Measure how the natural experiment is associated with the final outcome of interest (future earnings).
3. Divide the second measurement by the first, to scale up the diluted effect to the size it would have if everyone affected by the natural experiment had actually complied with it.
4. Interpret the resulting number strictly as the effect for the group of people whose behaviour actually shifted because of the natural experiment, not for the whole population.

Angrist, working earlier with Alan Krueger, had applied exactly this kind of birth-date natural experiment to schooling.

Because children born early in a calendar year reach the legal school-leaving age sooner than children born later in the same year, the researchers could compare students born in the first quarter of a year with those born in the fourth quarter.

They found that each extra year of education raised income by about nine per cent, a somewhat larger effect than the roughly seven per cent gap seen in raw population data.

The Angrist and Imbens framework, described by the Academy as merging the instrumental variables approach from economics with the potential outcomes framework from statistics, clarified that the resulting estimate applies only to compliers and set out when such a result can be trusted.

Draw three boxes in a row labelled "Natural experiment", "Actual treatment received" and "Outcome (earnings)".

Draw an arrow from the first box to the second, and from the second to the third, but no direct arrow from the first box to the third, showing that the natural experiment is assumed to affect earnings only through its effect on whether a person actually gets the treatment.

How did the work develop?

Why does this matter?

The committee's chair, Peter Fredriksson, said that "Card's studies of core questions for society and Angrist and Imbens' methodological contributions have shown that natural experiments are a rich source of knowledge." Because of this work, economists now have much more confidence in answering questions that affect real policy: whether raising the minimum wage costs jobs, how immigration affects the wages of existing residents, and how much school funding matters for a child's later success.

The approach has also spread far beyond labour economics. The presentation speech describes how, during the COVID-19 pandemic, Swedish researchers used a natural experiment, created because upper secondary schools moved online while lower secondary schools stayed open, to measure that school closures reduced reported infection rates among teachers by 50 per cent, with a smaller fall among parents.

The same logic has been applied to questions in political science, such as whether being the incumbent helps a politician win re-election, and to healthcare, such as whether extra medical treatment saves the lives of underweight newborns.

Open questions remain. The scientific background notes that it would be wrong to conclude that a higher minimum wage never reduces employment; the honest conclusion is only that such negative effects are usually smaller than economists once assumed.

Likewise, the local average treatment effect only tells us about the people who actually changed behaviour because of a particular natural experiment, so results from one natural experiment do not automatically apply to everyone in a population.

How does this connect to what you study?

If you study statistics or economics at school, this prize connects directly to the difference between correlation and causation, a core idea in any data-handling syllabus. The entire note is really about how to tell the two apart when real life, not a laboratory, is the source of the data.

When two things move together, such as more schooling and higher income, it is tempting to assume one causes the other, but as the committee's own example shows, hidden factors such as natural ability can cause both without any direct causal link between the two measured variables. Spotting this trap is a skill tested in data-interpretation questions.

The laureates' tools, natural experiments and the local average treatment effect, are a rigorous way of testing whether a cause-and-effect claim actually holds, a skill useful whenever you read a news report claiming that one thing "causes" another, whether in economics, health or social policy.

Card's minimum-wage study is also a useful case for anyone studying basic microeconomics, because it shows that a simple textbook prediction, that higher wage costs always cut jobs, can fail to hold once it is tested against carefully chosen real-world data rather than theory alone.

More broadly, the idea of a control group, something you may meet in science practicals, applies just as well to economics: without a fair comparison group, it is hard to know whether a change in outcomes was really caused by the policy being studied or by something else entirely happening at the same time.

Quick facts for exams

The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2021, widely called the Nobel Prize in Economics 2021, was announced on 11 October 2021 by the Royal Swedish Academy of Sciences.

It was divided: David Card of the University of California, Berkeley received one half "for his empirical contributions to labour economics", while Joshua D. Angrist of MIT and Guido W. Imbens of Stanford University shared the other half jointly "for their methodological contributions to the analysis of causal relationships".

Card used natural experiments to study minimum wages, immigration and schooling, while Angrist and Imbens showed how to correctly interpret the results of such natural experiments using the local average treatment effect.

Note: Source. The prize facts in this note are from the Nobel Prize's official site, nobelprize.org.

Glossary

• Natural experiment — a real-world situation, not designed by a researcher, that divides people into groups as if by random chance
• Causal effect — the actual change in an outcome that is produced by a specific cause, as opposed to a mere association
• Correlation — a statistical pattern in which two things tend to change together, without necessarily one causing the other
• Treatment group — the group of people exposed to the policy, event or programme being studied
• Control group — a similar group not exposed to the treatment, used as a point of comparison
• Randomised controlled trial (RCT) — an experiment in which a researcher randomly assigns subjects to a treatment or a control group
• Instrumental variables — a statistical method that uses a factor affecting the treatment but not the outcome directly, to estimate a causal effect
• Compliers — people in a natural experiment whose actual behaviour changed because of it
• Local average treatment effect (LATE) — the causal effect estimated specifically for the compliers in a natural experiment
• Selection bias — a distortion that arises when the groups being compared differ in ways other than the treatment itself
• Minimum wage — the lowest hourly or weekly pay an employer is legally allowed to give a worker
• Monopsony — a situation where an employer has enough market power to influence the wage it pays, unlike in a fully competitive labour market

Common errors and misconceptions

• Misconception: A natural experiment is the same as a laboratory experiment. Correct: A natural experiment arises from real events, policies or rules, with no researcher assigning who is treated.
• Misconception: Card proved that minimum wages never reduce jobs. Correct: He and Krueger found no negative effect in their specific study; later research shows negative effects are usually small, not absent in every case.
• Misconception: The local average treatment effect applies to an entire population. Correct: It applies only to the compliers, the people whose behaviour was actually changed by the natural experiment.
• Misconception: Angrist and Imbens won for an empirical finding about the economy. Correct: Their citation was for a methodological contribution, the framework for analysing causal relationships.
• Misconception: This prize is officially called the "Nobel Prize in Economics". Correct: Its official name is the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel; "Nobel Prize in Economics" is the common name.
• Misconception: Correlation between two variables always implies causation. Correct: The committee's own example shows higher income and education are correlated partly because of unobserved factors like ability, not purely cause and effect.

Exam-style questions with model answers

Q1. What was the citation for David Card's share of the 2021 Economics prize? [1 mark]

1. The citation was "for his empirical contributions to labour economics".

Q2. Which three economists shared the prize, and how was it divided? [2 marks]

1. David Card received one half of the prize, while Joshua D. Angrist and Guido W. Imbens jointly shared the other half, each getting one quarter.

Q3. Explain what a natural experiment is and why economists use it. [4 marks]

1. A natural experiment is a real-world event, policy change or rule that places similar groups of people on different sides of a divide without any researcher planning it, so it behaves as if people had been randomly assigned. Economists use natural experiments because many important questions, such as the effect of a minimum wage or immigration on jobs, cannot be studied with a true randomised trial for ethical or practical reasons. By comparing a group affected by the natural event with a similar unaffected group, researchers can estimate a causal effect while avoiding the selection bias that occurs when comparing groups that differ in many uncontrolled ways.

Q4. Describe the minimum-wage natural experiment used by Card and Krueger and its main finding. [4 marks]

1. In the early 1990s, New Jersey raised its minimum hourly wage from 4.25 dollars to 5.05 dollars, while neighbouring Pennsylvania did not change its wage. Because the two states' labour markets were expected to move similarly otherwise, Card and Krueger compared employment in fast-food restaurants on each side of the border before and after the change. Contrary to the conventional view that a higher minimum wage reduces jobs, they found that the wage increase in New Jersey had no effect on the number of employees compared with Pennsylvania.

Q5. Explain the local average treatment effect and why Angrist and Imbens introduced it. [5 marks]

1. In a natural experiment, people usually choose whether to actually take up an opportunity rather than being forced to, so only some of the people exposed to a policy change their behaviour; these people are called compliers. Angrist and Imbens showed, in 1994, that the causal effect measured from such a natural experiment can only be interpreted reliably for this group of compliers, not for the whole population, because non-compliers were unaffected by the natural experiment in the first place. They named this effect the local average treatment effect, or LATE, and derived it using a two-step instrumental variables method that first measures how the natural experiment changes the chance of receiving treatment, then uses this to scale up the measured effect on the outcome. This framework merged the instrumental variables approach from economics with the potential outcomes approach from statistics, giving researchers a transparent way to state the assumptions behind their causal claims and to check how sensitive their results are if those assumptions fail.

Q6. Discuss how the laureates' combined contributions changed empirical economics. [6 marks]

1. Before the 1990s, economists trying to establish cause and effect from real-world data relied heavily on structural models built on strong theoretical assumptions, which were often difficult to justify and could give misleading results. David Card's studies from the early 1990s, on the minimum wage, immigration and school resources, demonstrated that natural experiments could provide convincing evidence on important policy questions, and his findings repeatedly challenged what had previously been assumed, such as the belief that a higher minimum wage always costs jobs. At the same time, Joshua Angrist and Guido Imbens solved the deeper methodological puzzle of how to interpret such natural experiments when people's responses vary and not everyone complies with the assignment, by developing the local average treatment effect framework. Card's applied work showed the richness of natural experiments as a source of evidence, while Angrist and Imbens clarified exactly what conclusions such evidence supports; together, according to the prize committee, their work has substantially improved researchers' ability to answer key causal questions, benefiting fields well beyond labour economics, including education policy, political science and public health research such as pandemic school-closure studies.

Key takeaways

• David Card received one half of the 2021 Economics prize for using natural experiments in labour economics.
• Joshua Angrist and Guido Imbens shared the other half for methods to analyse causal relationships correctly.
• A natural experiment is a real event or policy change that divides people into groups as if by chance.
• Card found that New Jersey's minimum wage rise had no measurable negative effect on fast-food employment.
• Card also found no negative wage or employment effect on existing Miami workers with low levels of education after the 1980 Cuban emigration.
• Angrist and Imbens introduced the local average treatment effect to interpret natural experiments with imperfect compliance.
• Their instrumental variables framework merged ideas from economics and statistics into one transparent method.
• The approach has spread to education, health, political science and even pandemic school-closure research.

Test yourself

Where was David Card born, and where did he work when he won the prize?

David Card was born in 1956 in Guelph, Canada, and worked at the University of California, Berkeley, USA, when he won the prize.

What share of the prize did Joshua Angrist and Guido Imbens each receive?

Joshua Angrist and Guido Imbens each received one quarter of the prize, sharing the other half between them.

What natural event did Card use to study the effect of immigration on wages?

Card used the sudden 1980 wave of Cuban emigration to Miami, which raised the city's workforce by about seven per cent.

What is a complier in the Angrist-Imbens framework?

A complier is a person in a natural experiment whose actual behaviour, such as leaving school early, genuinely changed because of that experiment.

What did Card and Krueger find about New Jersey's minimum wage rise?

They found that raising New Jersey's minimum wage had no measurable negative effect on employment at fast-food restaurants compared with Pennsylvania.

What is the official name of this prize?

Its official name is the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel, commonly called the Nobel Prize in Economics.

What does LATE stand for and why was it needed?

LATE stands for local average treatment effect, needed because natural experiments only reveal a reliable causal effect for the compliers, not everyone.