Why David Card, Joshua Angrist and Guido Imbens won the Nobel Prize in economics
This year, Sveriges Riksbank Prize in the Economic Sciences in the memory of Alfred Nobel — more popularly known as the economics Nobel — has been awarded to David Card “for his empirical contributions to labour economics” and Joshua Angrist and Guido Imbens “for their methodological contributions to the analysis of causal relationships”.
The prize this year is unique in the sense that it is awarded both for “methodological contributions” and applied work. The broad theme that unites their work is the use of “natural experiments” to answer important economic questions.
Economists are often interested in the causal effects of policies. A “naive” way of finding the causal effects is just looking at the association between two phenomena. This could be very misleading, however, due to multiple reasons. These reasons — collectively known as endogeneity problems in econometrics — include omitted variable bias, reverse causality etc.
Let us take some examples. Assume a researcher analyses the data on the deployment of police force and the incidence of crime in a particular area and gets a positive association, i.e. areas with more police force also have higher crime rates? Does it mean that the police deployment causes a higher incidence of crime? Not at all. The reality is probably the other way around: It’s the high crime rate that results in higher deployment of police forces.
Similarly, if one looks at the association between hair length and wage rate and finds a negative correlation, one should not jump to the conclusion that hair length itself causally affects wage rate. More likely, the hair length is a marker of some other variable such as gender that affects the wages negatively. In other words, this association is driven by an “omitted variable”.
It’s easy to think of ways in which associational studies can be horrendously wrong. The bigger question is how to learn about causal relations despite such challenges? One way to do is through the so-called Randomised Control Trials (RCTs), which are actual experiments done in the field, for which the Nobel prize was awarded in 2019. Despite their popularity, RCTs have severely limited applicability in the social science due to ethical and practical reasons.
Think of the impact of historical institutions on growth for example. Unless one designs a time machine, how can one go back in history and randomly change the historical institutions? Or think about the causal effects of education or nutrition. In order to perform an RCT, one has to deprive a randomly selected group of students from education or nutrition. That would be unacceptable on ethical grounds. Even if feasible, RCTs are expensive to carry out.
Is it possible to establish causal effects without running into these practical and ethical problems? The answer is provided by “natural experiments”, which are random events or policy changes that were not experimentally assigned, but can be used to establish causal effects. This year’s laureates have contributed to the establishment of a toolkit for economists to use such natural experiments to test existing theories or formulate new ones.
The potential for empirical work to challenge theoretical orthodoxy is perhaps best demonstrated by a paper that David Card wrote with his long term collaborator Alan B Krueger (who would perhaps have shared the prize if he was still alive) in the early 1990s, showing that an increase in minimum wages did not lead to loss of employment. They compared restaurants in New Jersey and Pennsylvania, showing that they had very similar employment trajectories historically. Minimum wages increases in New Jersey were shown to not impact employment at all in comparison to Pennsylvania where there had been no change in the minimum wage.
With this, they not only challenged the orthodoxy that maintained that increasing state-mandated minimum wages will lead to business employing fewer people, but they also demonstrated a novel way of using a quasi-natural experiment to establish the causal effect of a policy, which has become widely used in many different contexts since then.
Other scholars continued to build on the work of Card and Krueger, and one can draw a line from their work to the political campaigns that are trying and succeeding in increasing minimum wages in the United States. David Card also contributed similarly to empirically understanding the effects of school education and of immigration.
Joshua Angrist is the pioneer of one of the most widely used techniques of identifying causal effect, called the instrumental variable technique. Many of the most important empirical questions that economists want to answer are not amenable to experiments. For example, to find the effect of schooling on wages, one would need to randomly allocate some children to go to school and keep some children out of school — which is morally indefensible and practically infeasible.
In such cases, one can use some other randomly occurring event, called the instrumental variable, that influences the probability of acquiring education to estimate the effect of education. For example, Angrist, along with Alan Krueger, used the fact that laws regarding compulsory schooling in parts of the United States were based on the age of the student and not the class that they studied in. Hence, students whose birth-months were such that they would be the older ones in their cohort, would reach the school leaving age earlier and hence were less likely to remain in school than their younger counterparts. Here, the random event of being born in a certain month affects the probability of staying in school longer, and hence was used by Angrist and Krueger as an instrumental variable to find what the benefit, in terms of higher wages, was of staying in school.
Angrist then joined hands with Guido Imbens, the third of this year’s awardees, and went on to refine the instrumental variable technique. Together, they provided a framework for thinking about issues of causal identification and the conditions that such a “natural experimental” must satisfy for the valid estimation.
Imbens made important contributions to developing two other techniques for causal identification using non-experimental data called regression discontinuity and matching. Regression discontinuity uses an arbitrary discontinuity in the implementation of a policy — say an age limit, or a geographical boundary, and compares the outcomes on either side of the discontinuity to establish the effect of the policy. Matching techniques are a set of methods where one tries to find individuals, households or firms who were not affected by the policy but were very similar to those who were affected. The difference in outcome between these two matched sets can be a valid estimate of the effect of the policy and Imbens established useful ways of creating such valid matches, which have been widely used by other researchers since.
Angrist is one of those rare Nobel laureates whose contribution to the teaching of his field is arguably as valuable as his research. He, along with Jorn-Steffen Pischke, has authored two textbooks — one at the graduate level (Mostly Harmless Econometrics) and one at the undergraduate level (Mastering Metrics). These have revolutionised the teaching of econometrics (the branch of economics that deals with data analysis and statistics), rescuing it from being a forlorn study in proving well-proven theorems and turning it into a way to actively engage in the empirical turn that the discipline of economics has taken in the last two decades.
It is this turn that the Nobel committee has started acknowledging in the last few years and the trend has continued with the well-deserved prizes this year.
Anand Shrivastava and Avinash Tripathi teach economics in Azim Premji University, Bengaluru
The views expressed are personal