Economists increasingly focus on peer effects to explain behaviour, i.e. situations where one person's outcomes are partially determined by the characteristics of the people around him. Applications address diverse questions:
- How did Silicon Valley become a centre of innovation?
- Why do some people pick up on new ideas so quickly, while other people wait to adopt them?
- How can localities capitalise on high skill immigration and take advantage of the increased human capital stock, creating a cycle of ever-greater returns?
In all these, it seems reasonable that people affect each others’ behaviour and, thus, overall outcomes.
These common sense ideas, however, mask a deep imprecision: Who exactly are our peers?
- Are they the people who live and work near us, that we see and interact with every day, even though we may have little else in common (Waldinger 2012)?
- Are they the people most similar to us in terms of interests and training, even though we may never have met (Borjas and Doran 2012)?
- Or are the true peers the ones who directly interact with us in the pursuit of a shared goal (Waldinger 2009, Azoulay et al. 2010)?
Obviously, all of these alternative definitions of a ‘peer group’ have merit, and we refer to them as geographic, intellectual, and collaborative peers, respectively.
In Borjas and Doran (2014), we try to disentangle the three kinds of peer effects on productivity, specifically academic productivity. Other research has reached disparate conclusions regarding the effect of gaining or losing productive colleagues on one's own output. Now, using the conceptual distinction outlined above, we are able to clarify which kinds of peers, if any, have strong effects on a person's output.
Academics are a useful population in which to study peer effects for two reasons:
- There are many ways to define and track individual productivity over time, using a plethora of information about publications and citations; and
- Academics face all three varieties of peers at once (geographic peers work at the same institution, intellectual peers publish papers in the same field, and collaborative peers coauthor papers together).
Specifically, we looked at mathematicians in the former USSR who lost peers (in all three ‘spaces’) when the Soviet Union collapsed and many academics left for Western institutions.
After 1992, about 10% of the mathematical scientists in the Soviet Union emigrated to other countries. Among those who remained, some lost many peers in the first sense – colleagues who worked at their home institutions. Others lost many peers given by the second definition – intellectually similar scientists who worked on the same topics. If not collaborators, these peers were often competitors. Finally, some mathematicians lost peers from the third space – collaborators with whom they had worked closely together in the joint creation of knowledge.
Of course, the fraction of a mathematician’s peer group who emigrated was endogenously determined along with the changing productivity and behaviour of that mathematician over time. So, we make use of instruments that predict the likelihood that a mathematician’s peers would emigrate but that themselves are unlikely to independently cause sudden changes in that mathematician’s productivity around 1992. Specifically, we observe the proportion of a mathematician’s peers from each category who have names that indicate a high likelihood of a Jewish descent (see Waldinger 2010 for the first use of such an instrument). Because of historical and continuing anti-Semitism, mathematicians of Jewish descent were significantly more likely to emigrate following the opening of the borders during the collapse of Soviet communism than were non-Jewish mathematicians. Mathematicians who had many Jewish peers in their institution were likely to lose many geographic peers; mathematicians who worked in topics that also interested many Jewish mathematicians were likely to lose many intellectual peers; and mathematicians who had coauthored with Jewish mathematicians were likely to lose many collaborators.
Our results suggest that losing geographic peers does not really affect a person’s productivity. But losing intellectual competitors matters quite a bit – it results in an increase in output. Finally, losing collaborators can dramatically reduce a person’s productivity if those couthors were of extremely high quality.
In particular, we find that a 10 percentage point increase in the outmigration rate of intellectual competitors increased the number of papers that a mathematician published in any given year by about 7%. On the other hand, authors with a very high quality coauthors suffered a 8% decline in their publication rate for every 10% of their collaborators who emigrated.
These results suggest that there are two competing forces at work when peers interact with each other.
- On the one hand, a peer is using resources that you might otherwise use yourself. The outmigration of such a peer would increase your productivity.
- On the other hand, the peer is producing knowledge that may spill over to you, is setting an example that may inspire you, and is attracting an inflow of new resources that you can use as well. The exit of such a peer would decrease your productivity.
Our results suggest that competition for scarce resources is the dominant effect among intellectual rivals, while positive peer effects are the dominant effect from high quality coauthors. If this finding holds in general, not only in the context of Soviet mathematicians, it could be a powerful explanation of otherwise disparate results in the existing literature. In particular, Waldinger (2012) finds that losing institutional colleagues has no effect on academic productivity just as we do, while Waldinger (2010) and Azoulay et al. (2010) find that losing brilliant mentors or superstar coauthors has a negative effect on academic productivity. Finally, Borjas and Doran (2012) find that gaining intellectual competitors through the entry of many high-skill immigrants has a negative effect on productivity. The disparate results in the existing literature may thus be artefacts of different definitions of who is a peer.
Taken together, the conceptual distinction among the different types of peers suggests that the policy debate on high-skill immigration should concentrate more heavily on what makes people collaborate and what keeps people from collaborating. The most important positive spillovers from high-skill immigration are likely to emerge from direct collaborations between high skill immigrants and the pre-existing workforce. Papers such as Kerr and Lincoln (2010) and Freeman and Huang (2014) show that much of this collaboration has tended to occur along ethnic lines, but Boudreau et al. (2014) suggests that simple interventions may greatly increase the likelihood of collaboration. If the key to positive peer effects in productivity and innovation lies in collaboration, then both the policy debate and the academic literature should focus on collaboration as the lynchpin that could turn competition for scarce resources into a larger economic pie for society.
Azoulay, Pierre, Joshua S. Graff Zivin, and Jialan Wang (2010), “Superstar Extinction,” Quarterly Journal of Economics 125:2, 549-589.
Borjas, George J., and Kirk B. Doran (2012), “The Collapse of the Soviet Union and the Productivity of American Mathematicians,” Quarterly Journal of Economics 127:3, 1143-1203.
Borjas, George J., and Kirk B. Doran (2014), “Which Peers Matter? The Relative Impacts of Collaborators, Colleagues, and Competitors,” National Bureau of Economic Research Working Paper # 20026, March.
Boudreau, Kevin, Ina Ganguli, Patrick Gaule, Eva C. Guinan, and Karim Lakhani (2014), “The Formation of Scientific Collaborations: Field Experimental Evidence on Search Frictions in Collaborator Matching”, unpublished manuscript, Harvard University.
Freeman, Richard B. and Wei Huang (2014), “Collaborating With People Like Me: Ethnic Co-authorship within the US”, unpublished manuscript, Harvard University.
Kerr, William R. and William F. Lincoln (2010), “The supply side of innovation: H-1B visa reforms and U.S. ethnic invention”, Journal of Labour Economics 28, no. 3: 473-508.
Waldinger, Fabian (2010), “Quality Matters: The Expulsion of Professors and the Consequences for Ph.D. Student Outcomes in Nazi Germany,” Journal of Political Economy 118:4, 787-831.
Waldinger, Fabian (2012), “Peer Effects in Science: Evidence from the Dismissal of Scientists in Nazi Germany,” Review of Economic Studies 79:2, 838-861.