Guido Imbens MA ’89, PhD ’91, applied econometrics professor and professor of economics at the Stanford Graduate School of Business, sits in front of two square images of his children playing chess.
Both of his sons started playing the game more during the COVID-19 pandemic, he said. He wanted to memorialize the first time that they beat him. He himself played chess in the past, a pursuit that he says has given him a good idea of how to focus on one problem for many hours.
This is a type of focus that is essential for Imbens’s work in economics. He recently received the Nobel Memorial Prize in Economic Sciences along with David Card and Joshua Angrist for his work on causal inference. Imbens has worked on a variety of projects, including education research and work on inequality and mobility.
Imbens’s career started with studying mathematics, but he wanted to find a way to apply his technical understanding. In the Netherlands, students apply for a specific undergraduate major instead of an entire school, Imbens said. He enjoyed mathematics, but “wanted to do something reasonably technical.” He appreciated that economics could be applied to society and addressed practical problems, but also used enough theoretical mathematical concepts to cater to his interests.
One of those interests was combining different fields, similar to the economic scientist Jan Tinbergen, Imbens said, whom he described as personifying the ideal of connecting economic science and government advising through projects such as the Central Planning Bureau. Tinbergen lent scientific authority to policy decisions and inspired Imbens’s journey through his undergraduate and graduate careers.
Having landed on economics, Imbens has been working to figure out how to draw credible causal inferences in non-experimental settings since the early 1990s. His work is therefore of particular importance to economics, because it can provide frameworks for organizing data.
Imbens attended Erasmus University in the Netherlands for his undergraduate education, and then met his mentor, Anthony Lancaster, at The University of Hull in the United Kingdom. He then followed Lancaster, professor emeritus of economics, to Brown to obtain his MA and PhD.
“We worked together on problems of homelessness in New York City,” Lancaster said. “The idea is that we could apply up-to-date econometric methods to understand more clearly the causes of homelessness and the difficulties that homeless people found.” Normally, when economists or econometricians collect data, they sample a population manually, Lancaster said — an approach that encounters challenges when looking to sample people experiencing homelessness. Lancaster and Imbens gathered data from homeless shelters and published a series of papers on the subject.
Lancaster said that the precise topic of causal inference arose from Imbens’s interaction with economists such as Donald Rubin, a professor emeritus of statistics at Harvard who has done profound work on the use of statistical data to make causal inferences. Their research in turn built on the work of Jerzy Neymann, a mathematician and statistician from Poland who published papers in the 1930s.
“Economists have a lot of data, and increasingly with the Internet, they have vastly more,” Lancaster said. “And so the question is, what to do with this to cast light onto the solutions to economic problems.” Econometricians such as Lancaster and Imbens are interested in finding the underlying events that cause relationships like the connection between unemployment and the level of unemployment compensation.
“If we can understand the physiology of some correlation in economics, then we can start to claim that we really do understand the phenomenon, and can use it to guide policy,” Lancaster said. Imbens, Angrist and Card’s research is an example of viewing the structure of economic relations to gain a deeper understanding of worldly phenomena.
“Much research in empirical economics makes use of what are called instrumental variables — variables that influence one economic variable (say, a person’s years of schooling) without directly influencing an outcome (say, the person’s earnings),” Jesse Shapiro, professor of political economy, wrote in an email to The Herald. “Angrist and Imbens taught us new ways to understand what we can learn using instrumental variables.”
In many cases, a randomized experiment is not possible due to ethical or financial considerations, Imbens said. The inability to perform direct experiments makes it difficult for economists to study some of their biggest questions.
“We are not going to take a bunch of individuals, and 100 of them will go to college and the others do not get to go to college,” Imbens said. Instead, economists look for a naturally occurring situation where the individuals only differ by one major variable. In one such study, Angrist and Alan Krueger, former professor of political economics at Princeton, exploited the fact that compulsory schooling laws affected people differently depending on when they were born. People born on Sept. 30 vs. Oct. 1 are not different in their ability, but they will be sorted into different school years by compulsory schooling laws. The researchers leveraged this variation to investigate causal effects, namely how schooling affects earnings.
In another case, Imbens’s colleague, Wilbert Van der Klaauw, researched the college admissions process. In one process, college admissions officers sorted through applications and scored them. The people in a narrow band on either side of the margin do not differ significantly, but their outcomes are markedly different. “You can learn something meaningful from comparing individuals close to these thresholds,” Imbens said.
In the case of comparing levels of education,“we are looking for ways of modeling and making credible assumptions about why some people went to school and some people didn’t go to school, in a way that still allows us to draw credible inferences,” Imbens said.
Imbens’s work on causal inference can be applied to different fields. He worked with Donald Rubin to write “Causal Inference for Statistics, Social and Biomedical Sciences.” Imbens has since coauthored papers touching on biomedical and psychological fields.
Imbens said he is not sure where the Nobel Prize will take him in the future.
“It has been a very strange, busy two and a half weeks,” Imbens said. But, in the long run, he believes it will give him opportunities to get attention for parts of econometrics. Within economics as well, it helps to make students excited about the field and gives him opportunities to teach a new generation.
For his future work, Imbens is interested in two major topics. He wants to investigate the idea that experiments can be used along with insights from economic theory to determine mechanisms of economic behavior. Additionally, he wants to use experiments not only to answer specific questions but also to complement observational studies. A combination of these approaches can be integrated to improve the credibility of economic conclusions.
It has always been useful to speak to individuals about their empirical research, Imbens said. He hopes that younger econometricians see the value in collaborating with other economists instead of staying insular.
But at the base of his research, Imbens loves puzzling over problems. Although he says he does not have much time now, he has found it helpful to work on a problem alone and “think until your head hurts.”
“The ability to sink into a problem and think deeply about it I find very useful, and I still find it very comfortable to do,” he said.
