Applied Econometrics

Empirical analysis of social interactions and knowledge spillover in which the 'invisible hand' does not work

Published online 30 October 2018

Economics is a dynamic and multifaceted area of research. ©2018 RTCorp

Why study economics?

"I grew up with a keen interest in nature and wildlife in the greenery-rich suburbs of Tokyo," says Ryo Nakajima, Professor at the Department of Economics of Keio University. "At high school I read a book that led me to think not only about nature itself but about the coexistence of human society with nature, especially in rural society. This was the reason I decided to major in agricultural economics at university."

Nakajima became interested in the pragmatic aspects of economics in validating theories using data while taking a course on economics at the Department of Agricultural Economics at Kyoto University. "But I found that in other courses, such as rural sociology and farm management science, teaching tended to emphasize ideology based on hypotheses that were impossible to disprove; although in retrospect I realize that this impression was a product of inexperience," recalls Nakajima. "I remember rebelling against these courses and decided to focus only on economics."

After completing his undergraduate studies at Kyoto University, Nakajima enrolled on a doctorate course in labor economics at the Faculty of Economics at New York University (NYU), specializing in methodology of applied economic analysis.

Features of economics courses in the US and Japan

In the mid-1990s when Nakajima was a student at NYU, the curricula taught at faculties of economics in Japan were different from those in the United States. "The biggest difference was content on Marxian economics," says Nakajima. "In Japan there were lectures on neoclassical economics, with Marxian economics also designated as a compulsory subject. Meanwhile, on curricula at American universities the focus was on microeconomics and macroeconomics; and this remains the case today." Nakajima recalls studying Marx's capital theory at the Department of Agricultural Economics at Kyoto University. "I remember being very surprised that there were opportunities to discuss this with American classmates and international students while I was studying in the USA."

Nakajima has seen firsthand how the collapse of socialism led to the decline in the teaching of Marxian economics in Japan, with the result that there is very little difference in curricula taught in Japan and the US today. "I think that economics undergraduates, not only in Japan and the United States but also in other countries, use classic economics textbooks. Economics involves acquiring knowledge by means of step-by-step methodologies using standard textbooks. So the main difference in undergraduate lectures on economics given at Keio and at American universities is the language in which they are taught, rather than the content."

Nakajima adds that lectures and teaching content vary greatly for graduate schools, with the differences reflecting the characteristics of each graduate school, or the personality of supervising teachers, rather than differences between Japan and the United States.

Research on empirical analysis of social interactions

In the traditional economic paradigm, people's economic behavior is analyzed based on their interactions via results of market transactions. As far as relying on market transactions, it has been shown that voluntary and self-interested behavior by individuals will lead to a desirable outcome without external intervention. Adam Smith expressed this as persons being controlled by the "invisible hand".

In contrast, economic analysis of social interactions focuses on direct interactions occurring independently of transactions involving goods and services. For example, it is known that people's health and longevity are greatly influenced by the people they interact with on a daily basis, such as family members, friends and acquaintances. This is because behavior that is harmful to health, such as smoking and excessive eating and drinking is influenced by interactions with people who are directly implicated in such habits. This behavior may be indirectly influenced by friends of friends and even the actions of strangers connected to them through social networks. The important point to draw from this is that even though decisions may seem to be the result of free will, people are influenced by their surroundings.

In circumstances in which individual actions are influenced by factors other than market transactions, the "invisible hand" does not work, so the overall results may not be desirable. Since it is not possible to predict the overall macroscopic outcomes from a simple summation of individual micro-actions, it is necessary to consider the influence of individuals on other individuals as part of a system of global social interaction.

"I wanted to find out more about the relationship between micro- and the overall macro-behavior so I decided to focus my research activities on the empirical analysis of social interaction," says Nakajima. "In my NYU doctoral thesis I described this by means of research on how youth smoking is influenced by peers and how that influence spreads. I published the results in The Review of Economic Studies in 20071.

"It may be of interest to note that my choice of this research theme was in fact influenced by social interactions with faculty members who were themselves studying social interaction at the Faculty of Economics at NYU at the time. They were not my academic advisors, but I cannot deny that contact with them affected my choice of research theme. I think this further goes to show that the range of voluntary decision-making by people is unexpectedly limited, and I conclude that decision-making depends on where you are and who you are with."

Approach to research

Nakajima examines social interactions using large-scale micro data on individuals employing two main methods. The first is experimental research design. Since it is difficult to intervene in the decision-making of actual subjects by carrying out randomized control experiments, Nakajima uses situations that can be regarded as natural experiments, where random intervention occurs, formulates a hypothesis about social interactions, and quantifies the magnitude of interactions.

The second approach is an analytical method called structural estimation, consisting of the task of constructing an economic theory model that explains people's behaviors and estimating the parameters of the model that might best describe the observed action pattern. "Based on the estimated model, we carry out predictive simulation about people's behavior and verify the validity of the theoretical model by examining whether that prediction can explain the behavior of real people consistently," explains Nakajima. "Predictive simulation by modelling is also used for policy evaluation. It is possible to quantitatively calculate the effect of policies that have already been implemented and to compare the impact of alternate policies that may be executed in the future and to clarify in advance which is the best policy."

Recent research highlights

Nakajima is looking into the social interactions of experts who create new knowledge and have carried out empirical research. Researchers and inventors are considered to "produce" new knowledge, but most of the knowledge "produced" is based on existing knowledge. Previous research shows that researchers and inventors incorporate knowledge through direct contact with their colleagues. In other words, most of the expert knowledge is transmitted and spread through social networks of peers, not by market transactions.

In a paper published in 2014, Nakajima and colleagues traced the flow of knowledge of inventors using US patent data and quantified the physical distances over which knowledge spreads2.

In more recent collaborative research Nakajima and his collaborator analyzed the transmission of scientific knowledge across generations3. "We quantitatively measured the degree of succession of scientific knowledge through guidance at the graduate school - in other words a guided relationship," says Nakajima. "To that end, we introduced a value-added model that correlates the research results of graduate students with the "quality" of the supervisor and estimated the "quality" of research guidance. To identify the teaching guidance effect of teachers, we focused on the separation of faculty members due to mandatory retirement, transfer, death, and so on."

When a professor leaves their job, the research guidance of students is handed over to another professor, so the approach to the supervision of master's and doctoral students in the laboratory will differ. Meanwhile, in the laboratory of the generation before the professor left, the same person conducts research guidance for graduate students. By comparing differences between generations of such research guidance environments as natural experiments, this research identified the impact of changes in the quality of research guidance due to changes in supervisors on the growth of research findings by students. In empirical analysis for physics researchers who obtained degrees at the Graduate School of Science of the University of Tokyo, the effect of supervisor turnover was measured. The results show that research guidance of university supervisors has a significant influence on the research achievements of students.

Future research

Nakajima is currently researching social interactions to measure the impact of generational changes in researchers on the inheritance of knowledge in Japanese manufacturing companies and the connection to the company' productivity.

"We are also working on a project to study how companies can incorporate new knowledge and outside ideas, and what kind of organizational structure is most effective to utilize such knowledge."

About the researcher

Ryo Nakajima― Professor

Faculty and Graduate School of Economics

After receiving his PhD in 2004 from New York University, Nakajima worked as a postdoctoral research fellow at Osaka University and Duke University. In 2006, he returned to Japan and started working at Tsukuba University as an assistant professor. From 2009 to 2012, he was an associate professor at Yokohama National University. In 2012 he joined Keio University as an associate professor and was promoted to a full professorship in 2016. He was awarded the Ishikawa Prize of the Japanese Economic Association in 2018.



  1. Ryo Nakajima, Measuring Peer Effects on Youth Smoking Behaviour, The Review of Economic Studies, 74, p. 897-935, July 2007. | article
  2. Yasusada Murata, Ryo Nakajima, Ryosuke Okamoto and Ryuichi Tamura, Localized Knowledge Spillovers and Patent Citations: A Distance-Based Approach, Review of Economics and Statistics, 96, p.967-985, December 2014. | article
  3. Yuta Kikuchi and Ryo Nakajima, Evaluating Professor Value-added: Evidence from Professor and Student Matching in Physics, Institute for Economic Studies, Keio University, Keio-IES Discussion Paper Series, 24 March 2017. | article