The three winners of the Nobel Prize in Economics 2021, David Card (University of California, Berkeley), Joshua D. Angrist (Massachusetts Institute of Technology) and Guido W. Imbens (Stanford University) share the empirical approach used in their work: the use of natural experiments to study causal relationships .
What is the importance of those jobs? To answer this question, it is necessary to understand the importance of causal inference in science. The usual practice in science to determine the existence of a cause-effect relationship is to carry out an experiment.
For example, suppose we want to determine if a vaccine is effective, that is, if the person receiving the vaccine is less likely to get it. To determine if there is a cause-and-effect relationship between vaccination and the probability of infection, the experimental procedure establishes that the vaccine is administered to a group of subjects (the treatment group) and at the same time a placebo (an innocuous compound) is inoculated to another. subject group (the control group).
The experimental procedure succeeds in establishing causality because the assignment of a subject to the treatment or control group is carried out randomly, in such a way that all the experimental subjects have ex ante the same probability of receiving the treatment. Randomization of treatment to a sufficiently large sample of subjects ensures that subjects in the treatment group and those in the control group have similar characteristics.
For example, if among the experimental subjects there was a subgroup of individuals with a weak immune system, it is highly probable that the random assignment of the treatment would result in approximately half of the immunosuppressed being assigned to the treatment group and the other half to the control group. .
In this way, if the treatment group turns out to have a lower level of contagion than the control group, this difference may not be due to the weak immune system of some subjects, since they are equally distributed in the two groups.
Conducting an experiment can be very expensive, and at times it may not be feasible, it may be illegal, or it may simply be unethical. In economics and other social sciences, the possibility of conducting randomized experiments is usually highly restricted by the conditions of legality and ethics.
Despite these difficulties, in recent decades, the behavioral economics has made important achievements using laboratory experiments to study the behavior of subjects in situations recreated in the laboratory. Thus, the 2002 Nobel Prize in Economics was awarded to Vernon Smith and Daniel Kahneman for his contribution to behavioral economics, for which they used experimental methods.
Laboratory experiments make it possible to establish the cause and effect relationship, but extrapolation of the results thus obtained to the general population can be risky. The cause and effect relationship may only be maintained for the subjects analyzed and is not generalizable.
In order to generalize the results of the experiments, it is necessary to carry out field experiments, where the experiment takes place in the population of interest. Thus, other economists have used field experiments for their research, such as the recipients of the Nobel Prize in Economics 2019, Abhijit Banerjee, Esther Duflo and Michael Kremer for their contribution to development economics.
Despite these notable exceptions, the vast majority of relevant questions in economics would require conducting experiments that are simply impossible to perform. What can be done in such cases?
The winners this year with the Nobel Prize in Economics are so for their contributions to the study of causal relationships using observational data, that is, data that is collected in the field through surveys or public or private registries, and that were not collected in the course of of an experiment.
Immigration and the labor market
A few examples, authored by this year’s Nobel laureates, serve to illustrate how to infer causal relationships in a non-experimental context. In his pioneering study on the impact of immigration on the labor market, David Card observed that the massive influx of Cubans to Miami in 1980 could serve as a natural experiment.
Sure enough, Card collected data on wages and employment in Miami and other southern cities of the United States to determine the impact of the massive influx of Cuban emigrants. His conclusion was that the massive influx of Cuban immigrants did not have a significant impact on the wages and unemployment rate of the less skilled workers.
In another of his seminal papers, David Card, in conjunction with Alan Krueger, observed that the state of New Jersey planned to raise the minimum wage in 1992. As the political debate on the measure raged, Card and Krueger collected data of wages and employment in fast food restaurants (whose employees should be affected by the increase in the minimum wage) in New Jersey and other neighboring counties in the state of Pennsylvania where there was no increase in the minimum wage, thus serving as a control group .
Later, when the minimum wage was raised in New Jersey, they proceeded to collect the same data, thus obtaining data before and after the minimum wage increase (the treatment). They concluded that raising the minimum wage did not have the negative effect on employment that theory predicted. Card and Krueger contributed to the probably most cited application of the difference method in economics.
Today many debate whether or not Card’s contributions contradicted the most orthodox economic theory, which indeed they did, but the Nobel Prize has been given to him for his methodological contribution.
The quantitative results obtained in these studies have been reviewed and corrected by many economists, but the methodological contribution survives, and today it is not possible to open an issue of a scientific journal on economics without seeing some application.
The other two winners, Angrist and Imbens, have contributed, among other contributions, to the study of causal relationships through the use of “instrumental variables”. Education constitutes an investment in human capital that should have a positive impact on the salary gains of the most educated. Angrist considered that a study on the salary earnings of Vietnam War veterans could contribute to the study of the returns to education, since war veterans have fewer years of study.
In his study, he argues that the lottery to determine whether or not a man was going to serve in the US military could be used as an “instrumental variable”, similar to random assignment in experiments. However, some of the men “luck” assigned not to serve in the military decided to volunteer and others who had been assigned to military service filed claims to be exempted from service.
In experimental terms, we would say that some of those assigned to the treatment group were untreated, and that some assigned to the control group eventually received the treatment.
The existence of cause and effect relationships
This is a case of “interference” that Angrist and Imbens showed can be analyzed by focusing on those subjects who followed the protocol, that is, who served in the army when they were assigned to do so and who did not serve in the army when they were not. they were assigned to it. This was an improvement over the “intention to treat” method that was traditionally used in the literature.
These examples and many other “natural experiments” are used today by researchers in economics and many other social sciences to determine the existence of cause and effect relationships.
Logically, these methods that imitate or recreate situations and conditions similar to the experimental ones result in inference whose power is lower than that obtained by means of the traditional experimental procedure, but they allow to have an inference based on the scientific procedure.
Javier Gardeazabal, Professor of Economic Analysis, University of the Basque Country / University of the Basque Country