Graduate education and long‐term inventive performance: Evidence from undergraduates' choices during recessions

• Author: Koichiro Onishi,Sadao Nagaoka
• DOI: http://doi.org/10.1111/jems.12382
• Published date: 01 July 2020
• Date: 01 July 2020

J Econ Manage Strat. 2020;29:465–491. wileyonlinelibrary.com/journal/jems © 2020 Wiley Periodicals LLC

|

465

Received: 12 February 2019

|

Revised: 18 May 2020

|

Accepted: 20 May 2020

DOI: 10.1111/jems.12382

ORIGINAL ARTICLE

Graduate education and long‐term inventive performance:

Evidence from undergraduates' choices during recessions

Koichiro Onishi

1

|Sadao Nagaoka

2,3

1

Faculty of Education and Integrated Arts

and Science, Waseda University,

Shinjyuku‐ku, Tokyo, Japan

2

Faculty of Economics, Tokyo Keizai

University, Kokubunji‐shi, Tokyo, Japan

3

Research Institute of Economy, Trade,

and Industry, Chiyoda‐ku, Tokyo, Japan

Correspondence

Koichiro Onishi, Faculty of Education and

Integrated Arts and Science, Waseda

University, 1‐6‐1 Nishiwaseda,

Shinjyuku‐ku, Tokyo 169‐8050, Japan.

Email: onishi@waseda.jp

Funding information

Japan Society for the Promotion of

Science, Grant/Award Numbers:

JP15K03486, JP26285055

Abstract

Using individuals' life‐cycle invention data, we investigate how graduate

education affects inventive performance and inventors' abilities to absorb and

combine diverse knowledge sources. To control for the endogeneity of edu-

cational choice, we use the status of college labor markets as an instrumental

variable (IV), specifically the difference between the unemployment rate and

its long‐run average rate by academic field. We find that graduate education,

induced by the IV, significantly enhances inventive performance and the scope

of exploited knowledge, exceeding the levels implied by ordinary least squares.

Graduate education can have a significant causal effect on inventive capability

and performance.

1|INTRODUCTION

Human resources are often seen as central to creating inventions and, therefore, to knowledge‐based economic growth.

On the theoretical front, endogenous growth theory identifies the number of highly educated persons in a society as a

key determinant of the growth rate (as pioneered by Romer, 1990). Both emerging and developed countries have

expanded their graduate education programs (at the master's and doctoral levels) based on this view (OECD, 2016,

p. 146). Despite the widespread belief that graduate education enhances inventive performance, there is surprisingly

little empirical support for this conjecture.

1

To fill this important gap, we identify the effect of graduate education on the number and quality of inventions

made by inventors because of this education. We suggest that these additional inventions reflect how well graduate

education nurtures inventors' knowledge absorption capabilities and their inventive productivity. The ability to absorb

scientific advances is a particularly important consideration in the assessment of graduate education's contribution to

inventor performance, given that the absorptive capability is an important determinant of the innovative capability of a

firm (Cohen & Levinthal, 1989; March, 1991). Further, since innovation is often created by recombining existing

knowledge (Schumpeter, 1977; Weitzman, 1998), one can expect individuals with higher education to be able to

combine existing knowledge more effectively. On the other hand, successful technologies and ideas have often arisen

without the influence of existing scientific knowledge (Price, 1965).

To assess the contribution of graduate education, we focused on more than 2,200 Japanese corporate inventors and

derived detailed personal information from the Research Institute of Economy, Trade and Industry (RIETI) Inventor

Survey. To capture these people's long‐term inventive activities, we gathered all the life‐cycle patents applications they

had submitted to the Japan Patent Office (JPO).

2

Then, we matched the two databases using the inventors' names,

addresses, and other details. We thus obtained data on the inventors' final educational degrees and their long‐term

inventive activities.

Obtaining causal evidence of the effects of education is challenging because an individual's level of education is

endogenous to their unobserved characteristics (superior students would go to graduate school). To control for this

possible self‐selection problem, we use the unemployment rate of college graduates in academic fields in the year

preceding the focal inventor's graduation as an instrumental variable (IV). Graduate school seems to be an important

alternative for students facing a negative labor market at the time of college graduation. They may spend their time in

graduate school until the labor market recovers, and a recession can decrease the opportunity costs for attending

graduate school (Charles, Hurst, & Notowidigdo, 2018; Laeven & Popov, 2016). Prior studies have found that labor

market conditions affect enrollment for graduate school and doctoral programs in the United States (Bedard &

Herman, 2008; Johnson, 2013; Shu et al., 2012). Further, Kondo (2007) found a negative correlation between a lower

local job opening ratio and the propensity of high school students to go to college in Japan. Thus, unemployment is a

promising candidate for an instrument.

The unemployment rate of college graduates works as a valid instrument if it affects long‐term inventive activities

only through its effect on the choice for graduate education (the exclusion restriction). To control for this possibility

through the changes in technological opportunities and demand‐side opportunities, we introduce a full set of cross

terms between invention year dummies and technological area dummies. In addition, we addressed the following two

potentially important confounding factors. The first is that college students' unemployment rate could be correlated

with long‐term technological opportunities for inventions across academic fields. For example, an academic subject

experiencing rapid scientific progress provides abundant future inventive opportunities and may attract many capable

students; however, unemployment may also be high due to the lag between scientific advances and industrial in-

vestment. To control for this possibility, we focus on the variation within academic fields rather than the level

differences across fields because our instrument is the difference between the unemployment rate and its average rate

in each academic field divided by the standard deviation. Furthermore, we introduce a time trend for each academic

field in the estimations to control for the variations of trends across academic fields. The second confounding factor is

that labor market conditions may directly affect the quality of inventors, because hiring is restricted during a recession,

and therefore, their future inventive performance. However, we will show that a recession in the year preceding

graduation does not affect students' long‐run probability of a college graduate ultimately becoming an inventor in

Japan. This is perhaps because students can postpone entering the labor market by choosing to go to a graduate school

when hiring is limited. Furthermore, they can change their jobs over time to realize their aspirations for becoming

inventors. Because there is no reason for us to expect that the students who chose graduate education only due to poor

labor market situation had better inventiveness (e.g., ability, persistence, interest, etc.) than those who could find a job

immediately after the college graduation, we assume that the effect of unemployment fluctuation on the decision to

undertake further graduate education is unrelated to the factors that affect inventiveness.

In this paper, we aim to identify the effect of graduate education on students who would have chosen to become

inventors irrespective of their taking graduate education. That is, we seek to uncover the additional effect of acquiring a

graduate degree, conditional on inventor status (intensive margin). Graduate education can also affect a nation's

inventive performance by enhancing college graduates' propensities to become inventors, as some of them may have

become inventors only due to attending graduate school (extensive margin). This is beyond the focus of our paper due to

data limitations. However, assessing the intensive margin of graduate education itself would be an important con-

tribution to understanding its role, especially because some policy and management decisions depend only on intensive

margins, such as the choice between on‐the‐job training and graduate education.

While our IV (unemployment) can induce effects due to both the intensive and extensive margin, our IV estimation

generates an estimate equivalent to the intensive margin as long as additional inventors who become inventors only

because of their graduate educations have the same capabilities on average as those of the other inventors. Given this

assumption, the extensive margin does not affect our estimate for the intensive margin (proof given in Appendix A).

Even if such an assumption does not hold, it is likely that the additional inventors, due to the extensive margin, have

lower capabilities and smaller graduate education effects than the other inventors. In this case, our IV estimation gives

the lower bound of the intensive margin of graduate education.

In summary, we find that the inventive productivities of inventors with graduate education, as revealed by an IV, are

about 45–47% higher than those with less education. These graduate‐educated inventors also cite more scientific

literature and build on more diverse fields of prior patent literature. Further, the results in the IV estimation are larger

than those in ordinary least squares (OLS), showing that a significant negative selection exists in college students'

466

|

ONISHI AND NAGAOKA