In the past 30 years, a revolution in causal inference has erupted in modern economics research.

If I read one more year, will my income rise? How much? Will raising the minimum wage lead to an increase in the unemployment rate? Will the increase in immigration affect the employment and salary levels of locals? This year’s Nobel Prize in Economics was awarded to three American economists who gave rigorous causal answers through various measurement tools. The Nobel Prize recognizes them for solving one after another historical, social and economic mystery.

On October 11th, Beijing time, the Nobel Prize in Economics was announced as the finale. Half of the awards were awarded to David Card “because of his “Empirical contribution to science”, the other half was jointly picked by Joshua D. Angrist and GuidoW Imbens, “in recognition of their methodological contributions to causality analysis” .

The award speech wrote: They provided new insights on the labor market and showed that conclusions about causality can be drawn from natural experiments. Their methods have been extended to other fields and revolutionized empirical research.

Hongliang Zhang, Associate Professor of Economics at the School of Business Administration of Hong Kong Baptist University, said in an interview with News ( that everyone expects them to win the prize, the only one is uncertain When will the award be given. Before accepting an interview with a news reporter, Zhang Hongliang just sent a congratulatory email to his mentor Angrist, speaking of his joy and excitement at the relief of this uncertainty, and received Angrist’s letter on October 13th. Thanks Reply. Photo taken by Angrist’s wife when he learned that he won the prize.

Angrist’s photo taken by his wife when he learned that he won the prize.

Looking for causality in real-world economics

From 2003 to 2009, Zhang Hongliang studied for a PhD in economics at the Massachusetts Institute of Technology. During his studies, his advisor was Angrist and Esther Duflo, one of the 2019 Nobel Prize winners in economics. Duflo is the youngest Nobel economy in historyThe winner of the academic prize (47 years old at the time of the award) is also the second woman to receive the prize; in 2019, Duflo and her husband Abhijit Banerjee (Abhijit Banerjee) contributed to the “experimental method proposed in reducing global poverty” , Won the Nobel Prize in Economics with Michael Kremer.

Zhang Hongliang told the news that there was a time when labor economics was generally optimistic a few years ago. Except for the three popular candidates who won this year, Allen, who died in 2019 · Kruger is also a popular candidate. This year, after Gary Becker was the first labor economist to win the prize in 1992, the Nobel Prize was once again awarded to the labor economist. This year’s award was given to empirical research and methods based on natural experiments. The rise of empirical research based on natural experiments has a revolutionary impact on the development of economics, and has had extensive and profound impacts on economists who have grown up in the past three decades. In the past, economics basically required a theoretical model to be determined first, and related analysis was based on the assumption that the model was correct. Empirical research based on natural experiments emphasizes “let the data speak for themselves” and directly investigates the causal relationship between explanatory variables (explanatory variables) and outcome variables (outcome variables).

Zhang Hongliang introduced that this year’s Nobel Prize in Economics can be seen in conjunction with the random experiment in 2019. Why does the Nobel Prize want to emphasize the importance of “natural experiment” this time, because many scientific experiments can be done in natural sciences. The Nobel Prize in 2019 commends the application of randomized control trials to development economics research middle. Random trials have many applications in medicine. For example, when testing new drugs, patients can be randomly assigned to the control group (using the commonly used drugs) and the experimental group (using new drugs), and then see if the cure rate of the experimental group is higher than And to what extent is higher than the control group. Economists apply random experiments to the study of economic and social issues, such as providing subsidies and incentives for certain regions for vaccination, and comparing them with those regions that maintain the status quo to see the impact of related mechanisms on vaccination rates. But on many economic issues, it is not always possible to do a complete random experiment. At this time, it is necessary to use the existing data to evaluate what has happened, and construct a natural experiment-to make causal inferences through the observed data instead of the data collected by field experiments. After learning of the award, Inbens (middle) took a photo with his family. Angrist and Inbens met at the beginning of their careers, and they went to Harvard University after graduation in 1989 and 1990 respectively.  Teaching in the department, it hits it off quickly, and I am happy to do research together.

After learning that he won the award, Imbens (middle) A photo with his family. Angrist and Inbens met at the beginning of their careers. After graduating in 1989 and 1990, the two came to teach in the Department of Economics of Harvard University. They hit it off quickly and were happy to do research together.

In the mid-1990s, Angrist and Inbens challenged how to assess the average intervention effect. They used the instrumental variable (IV) framework commonly used in economics and the “inverse of causal reasoning” commonly used in statistics. Combining the framework of “counterfactual outcome” (counterfactual outcome), it analyzes how to identify and evaluate intervention effects that can be clearly explained in real scenarios, that is, the local average treatment effect (LATE). Angrist and Yinben Si’s research shows that the use of instrumental variables, that is, variables that are related to the intervention or not but have nothing to do with other factors that affect the outcome, can identify the local average intervention effect for individuals who comply with the experimental or quasi-experimental assignment (compliers to the assigned treatment) In the above-mentioned experiment of testing new drugs as an example, the instrumental variable is the allocation of the experimental group/control group, and the individuals who follow the allocation refer to the patients who will follow the doctor’s order no matter they are allocated to the experimental group or the control group, that is, allocation New drugs will be taken in the experimental group and old drugs will be taken in the control group. The Royal Swedish Academy of Sciences commented that their LATE framework has become the leading framework for quasi-experimental and experimental research in economics or other social sciences. The LATE framework provided by Angrist and Inbens is also used to test the conditions of other causal inference methods including breakpoint regression design and double difference recognition.

The Royal Swedish Academy of Sciences believes that the contributions of the three winners this year are complementary to each other, and the combined overall contribution is greater than the sum of their respective contributions. Card’s research in the early 1990s showed how to use natural experiments to reveal causal effects, using observations The focus of empirical research on data has shifted to relying on quasi-experimental changes to establish causal effects. Angrist and Inbens have greatly changed the use of data generated by natural experiments or random experiments with imperfect compliance. Empirical questions. Therefore, the contributions of all the winners are added together, and they played a central role in establishing an empirical method based on research design in economics. Their method aims to simulate random experiments to answer causality through observed data. Problem bigIt greatly improves the ability of researchers to use observable data to answer causal questions that are of great significance to economic and social policy.

In 1994, Angrist and Inbens jointly published “Identification and Estimation of Local Average Treatment Effects” (Identification and Estimation of Local Average Treatment Effects, Econometrica , March 1994.) The LATE, a causal inference framework model, was officially published, which is also the Nobel Prize commending their main contribution this time.

Zhang Hongliang introduced that Angrist’s research and contributions to natural experiments are embodied in his two masterpieces. The first can be traced back to his doctoral thesis “Lifetime Earnings and the Vietnam Era Draft Lottery: Evidence from Social Security Administrative Records, The American Economic Review,” June 1990.), he studied the impact of military service on future income during the Vietnam War due to a random lottery based on birthday. The second is the paper “Identification and Estimation of Local Average Intervention Effect” published by him and Inbens in 1994. This paper derives the LATE theoretical model through rigorous mathematics. Zhang Hongliang pointed out that “based on the LATE causal inference model of economics empirical papers, the biggest advantage of the research design is clean and neat. The causal inference method, the factual data on which it is based, and what conclusions can be inferred from it can be clearly explained. For even non-professionals without a background in economics.”LATE's mathematical derivation

LATE’s mathematical derivation

MIT’s economics department’s most acute-questioning teacher

According to the institution at the time of the award, the University of Chicago has the most Nobel Prize winners in economics There are 13 universities and colleges, and this year MIT (8) won the award because of Angrist’sAwarded and surpassed Harvard University (7 places).

Zhang Hongliang recalled his school days and talked about the tradition of the Department of Economics at MIT: Every year before the Nobel Prize is issued, teachers and students will guess who will get it. Each person can guess multiple candidates, and each person can vote $1 into the guessing prize pool. In the end, the money in the prize pool is divided equally between the guessers. During the years when he was studying, although the teachers of the MIT Department of Economics “gathered together”, no one has ever won the prize, and only 1-2 people can guess it every year. However, since 2010, five teachers have won awards (Peter Diamond 08′; Bengt Holstrom, 16′; Ahhijit Banerjee and Esther Duflo, 19′; Joshua Angrist, 21′). shared.

MIT implemented a dual tutor system. Zhang Hongliang’s research fields at the time were public economics and labor economics, and the two tutors chosen were Duflo and Angris special. The two tutors also have different styles-Duflo is more open to research questions and responds positively to students’ topic selection as much as possible; while Angrist is very strict with the criteria for topic selection.

What impressed Zhang Hongliang most was the weekly lunch seminar. “MIT has a very good atmosphere. There are more interactions between students and teachers. According to the research field, there is a lunch seminar every week. All doctoral students in this field have a meal with the teacher. All the teachers comment on your own research topics.”

Zhang Hongliang said when introducing MIT’s luncheon that Angrist was famous for asking questions at the lunch seminar. The sharpest.

“He will always ask you’so what’? What is your research contribution: What is the contribution in the literature? What is the significance of the empirical? Everyone thinks The most difficult thing is that in the first 5-10 minutes, if you can successfully pass Angrist’s questioning, it often means you have passed the test of topic selection.”

Zhang Hongliang also I think I have benefited most from Angrist’s guidance on topic selection and writing. Angrist’s requirements for writing are also very strict. He gave each student a writing guide (“The Elements of Style” by William Strunk Jr.), and personally taught the students how to write the opening paragraph of the article.

“His own articles are very well written. He has very high demands on himself, and high demands on others.” Zhang Hongliang commented on his teacher like this. Angrist’s

Angrist’s “Basically Harmless Econometrics” as the first author was also The domestic academic circles recommend it as a must-read for empirical research in economics.

Angrist’s design of educational mechanisms may be of reference to China

In fact, Angrist’s role in the field of educational economics Many articles have been regarded as classics of economics.

This time the Royal Swedish Academy of Sciences also mentioned that Angrist and Alan Kruger published a report in the top magazine “Economic Quarterly” in 1991. This article discusses the issue of education time and future wage levels.

The United States promulgated the “Compulsory Education Act” very early. The compulsory education law in the United States stipulates that as long as children who have reached the age of 6 in that year, they need to enter school in September of that year. In other words, if a child’s birthday falls on January 1, his age at school is 6 years and 8 months, while for a child born on December 31 of the same year, the age at school is 5 years and 8 months, which is higher than that of those born on January 1. ‘S child is 1 year younger. The compulsory education law in the United States also stipulates that you must pass the 16th birthday before you can legally leave the school. In this way, if you also leave school after the 16th birthday, the child born on January 1 will be one year later than the school age, and the number of years of education will be reduced by one year. The compulsory education law in the United States thus created A natural experimental environment-children who dropped out of school at the age of 16 can be divided into two groups: the control group has a birthday earlier in the year; the intervention group has a birthday later in the year. Angrist and Kruger collected income information for children born in the 1920s, 1930s, 1940s, and 1950s in the United States in 1970 and 1980, respectively, and found that for children born in the 1920s, People born in the first quarter are better than the other threePeople who were born in this quarter missed 0.126 years of schooling, and the rate of return to education was 0.7% lower. For children born in the 1940s, those born in the first quarter went to school 0.109 less than those born in the other three quarters, and their income was 1.11% lower. This inferred the rate of return for an extra year of schooling. It is 10.2% (1.11%/0.109).

Angrist’s research effectively proves that the extension of education time can significantly increase future wages, and thus refutes that the compulsory education law is of little economic significance to students. the opinion of.

Zhang Hongliang introduced that Angrist’s recent research is also about education issues-school-student matching, student-teacher matching mechanism design, and How to evaluate the impact of schools on student learning. He founded Blueprint Lab at MIT to promote the mechanism design of education, medical and employment policies.

“There is an organization in the United States called’Teach for America’. This organization does something similar to teaching support in China. When it comes to teaching support, teachers will have some preferences , Especially in the United States, different individuals have different preferences for religion and race; students will also have many preferences when choosing a school. In this case, how to do a good job of matching students, teachers and schools, Assign teachers and students to different schools to maximize their preferences and grades. Angrist is designing an algorithm to help this organization.”

Zhang Hongliang believes that Angrist’s current research is also of great reference to China. For example, if China wants to achieve de-schooling and the transfer of teachers, there will definitely be some difficulties, because teachers will have some preferences. How to respect the preferences of teachers to a certain extent and realize the transfer mechanism of teachers to make the distribution of educational resources more equitable is exactly what Angrist is currently doing.