Millennium Post

Champions of 'natural experiment'

Empirical and methodological applications introduced by David Edward Card, Joshua David Angrist and Guido Wilhelmus Imbens enabled effective delineation of cause and effect of interventions — leading to improved policy analysis

Champions of natural experiment

The Nobel Prize in Economic Sciences for 2021 was awarded to David Card, of the University of California, Berkeley (one half of the prize) "for his empirical contributions to labour economics". The other half of the prize was shared between Joshua D Angrist, of Massachusetts Institute of Technology and Guido W Imbens, of Stanford University "for their methodological contributions to the analysis of causal relationships".

Card completed his undergraduate in economics from Queen's University in 1978 and his PhD in economics from Princeton University in 1983. His doctoral dissertation was on long-term labour contracts. After his PhD, Card joined the faculty at the Chicago School of Business, briefly before moving to Princeton University, where he stayed from 1983 to 1997. In 1997, he moved to the University of California at Berkeley, where he continues. He also did a brief stint at Columbia University.

Angrist completed his undergraduate in economics from Oberlin College, Ohio, in 1982. After a three-year stint with Israeli armed forces, he returned to complete his Masters in economics in 1987 and PhD in 1989 — both from Princeton University. After his PhD, Angrist joined Harvard University as faculty in 1991 for a brief period. He then joined Hebrew University in Israel, where he stayed till 1996. In 1996, he joined MIT as an economics faculty and continues there.

Imbens completed his undergraduate studies in economics from Erasmus University, Rotterdam, in 1983. He did Masters in econometrics and economics from Hull University in UK in 1986 and his PhD in economics from Brown University in 1991. Imbens was a faculty member at Harvard between 1990-1997, UCLA between 1997-2001 and the University of California at Berkeley from 2002-2006. He returned to Harvard University in 2006 and stayed there till 2012. After that he moved to the Graduate School of Business at Stanford University.

In this article, we will discuss the main works of the three laureates and see how these have influenced public policy the world over and continue to do so.

Main works of David Card

Card is best known for his work on the labour market and related issues such as immigration, minimum wages and education. He used natural experiments, i.e., real or chance events and their impact on variables such as minimum wages and employment. This was a clear departure from the way empirical research was done in economics — natural experiments provided a strong fundamental basis for getting into cause and effect in social and economic policy. As we know, causal relationships are difficult to establish because it is difficult to create control groups (control groups are those that are similar to the experimental group, but are not subjected to a specific change 'treatment' or 'intervention' and the two groups are then compared after the intervention). Natural experiments offer a simple and elegant way to create control groups, such as through policy intervention.

Card published his main research in the 1990s, but the work that comes to mind is: his work with Alan Krueger where the authors used a natural experiment to examine the causal effect of increase in minimum wages on rates of employment. They looked at labour in the fast-food market in New Jersey, USA, with a similar market in neighbouring Pennsylvania as the control group. They found that raising minimum wages did not reduce fast-food employment in New Jersey (employment remained at similar levels in Pennsylvania where no raise in minimum wage had occurred).

Card also looked at the effect of the Mariel Boatlift of 1980 on the Miami labour market in a paper published in 1990. Card noted that the Mariel immigrants from Cuba (after Castro had temporarily allowed Cubans to leave the country in 1980) increased the Miami labour force by seven per cent, and labour supply to less-skilled occupations and industries was even greater because most of the immigrants were relatively unskilled. Card found that this influx had no effect on the wages or unemployment rates in Miami. More significantly, it had no effect on wages and employment rates of the Cubans who had immigrated earlier. Card used the wage and employment figures in four other cities as a control group. Card suggested that Miami had absorbed large waves of immigrants previously, which increased its capacity to accept immigration into its labour force.

Card and Krueger also looked at the cause-and-effect relationship between state investments in public schools and the academic performance of students, and their later success in labour markets. They found that resource levels in public schools have little impact on individuals' academic performance or economic opportunities in later life.

Card's book, 'Myth and Measurement: The New Economics of Minimum Wage' (1995) and other books that he co-edited, namely 'The Handbook of Labour Economics' (1999), 'Seeking a Premier Economy: The Economic Effects of British Economic Reforms' (2004), and 'Small Differences that Matter: Labour Markets and Income Maintenance in Canada and the United States' (1992), have used natural experiments to unravel the cause and effect of various social and economic policies.

Main works of Imbens and Angrist

Imbens and Angrist also used natural experiments to study causal relationships. In a paper published in the mid-1990s, 'Identification and Estimation of Local Average Treatment Effects', they looked at the problems of identifying a causal relationship between interventions and effects in situations where effects vary between subjects. Further, the researchers don't have full control over which subjects get the intervention and which don't (for example, if one were to study the effect of education on earnings, the researcher has no control over the educational level of the subject). Imbens and Angrist suggested a solution named LATE (long average treatment effect), which enabled them to calculate an average causal effect for a given intervention.

Angrist, while working on his PhD, wanted to see whether future earning potential of veterans was lesser than that of non-veterans. While a simple way would be to compare the wages of the two groups, this would ignore the fact that some men were more likely to enlist in the armed forces than others. Further, it is the less educated and those with fewer opportunities who were more likely to get enlisted. This problem was resolved by the Vietnam war and a lottery announced by the US government to enlist men into the armed forces (or the draft). This was a natural experiment that divided the men into two groups — a control group that wasn't drafted and a treatment group that was drafted. However, there was a further problem; there were many who would have enlisted anyway, notwithstanding the lottery. Angrist found a way to isolate the effects of the draft on future earnings of the veterans as compared to non-veterans. He found that veterans were likely to earn 15 per cent less than non-veterans.

Angrist also co-authored a paper with Krueger, published in 1991, where they looked at the impact of more education on higher income in peoples' working lives. It would be simple to compare those who study longer with those who don't and see to what extent their earnings differ. However, this is simplistic because those with longer years of education may have additional attributes (a higher IQ or better quantitative skills), which could boost their earnings. Therefore, to answer this question, they looked at a natural experiment, which came in the form of compulsory schooling. They saw that people born at different times in a given year were in the same class at school. That provided a natural cohort to measure the relationship between how long someone spends in school, and their earnings. They found that years of schooling do not impact earnings much.

Imbens and Angrist also worked on the effect of a guaranteed income on the labour force. For this, they looked at a natural experiment, which was basically the Massachusetts lottery. In the lottery, winners were paid over a number of years — resembling a guaranteed income. When they compared lottery winners with those who did not win, there were minor effects on the labour supply. However, it didn't significantly change how much people worked.

Imbens and Angrist designed methods to take care of the problems referred to above, and encountered by Angrist in his PhD work, so that the cause and effect were cleanly delineated. These problems were also acknowledged by the Nobel Committee. To quote:.

Natural experiments differ from clinical trials in one important way — in a clinical trial, the researcher has complete control over who is offered a treatment and eventually receives it (the treatment group) and who is not offered the treatment and therefore does not receive it (the control group). In a natural experiment, the researcher also has access to data from treatment and control groups but, unlike a clinical trial, the individuals may themselves have chosen whether they want to participate in the intervention being offered.


It is interesting to note that natural experiments that were championed by the three laureates are quite different from randomised control trials (RCTs). RCTs are more suited to fields such as natural sciences and medicine. For example, while testing for the efficacy of vaccines, one group is administered the vaccine, while the other is injected with water and the difference is noted between the two groups. While RCTs have been used in economics by the 2019 Nobel laureates Abhijit Banerjee, Esther Duflo and Michael Kremer, it is not always possible to do so and is even unethical in many cases. For example, you can't really impose different educational attainments or different wages on two groups. Moreover, as the 2015 Nobel laureate Angus Deaton has noted, RCTs only capture a particular instance. It is not necessary that a policy found effective in a RCT would be effective in another instance also. It is for this reason that natural experiments are a more attractive tool for analysis of a policy or a regulatory intervention.

To be sure, natural experiments also have shortcomings, and it is sometimes impossible to get all the data needed to come to a precise conclusion about cause and effect. For example, natural experiments divide participants into two groups based on a policy change as noted above. However, this assignment to groups is not randomised, and therefore, the research design and statistical methods used in random variables don't always apply. Hence, inferences about causation are not fool-proof. The work of Card, Angrist and Imbens got around this problem by using the 'difference-in-differences' approach and the instrumental variable approach, which injected some randomness in the assignment of participants to groups in a natural experiment. These contributions made it easier to understand the impact of policy interventions and clearly delineate cause and effect.

The natural experiments approach was used primarily in labour economics by the three laureates and Angrist and Imbens also used it in other policy areas as we saw above. However, its applications have increased manifold in social sciences. It is no accident that since 1990, at least half of the Clark Medals (a prize given by the American Economic Association to the best economist under 40) have gone to those who have used natural experiment methods.

The writer is an IAS officer, working as Principal Resident Commissioner, Government of West Bengal. Views expressed are personal

Next Story
Share it