Policy impact of key-player targeting

Yves Zenou 08 January 2015



In her article in The New York Times on 11 June 2013, the journalist Monica Davey documents that Chicago has witnessed an important drop in shootings and crime.

  • Chicago killings in 2013 have dipped to a level not seen since the early 1960s – down by 34% from 2012 to 2013.

Her explanation of this important decrease in crime is that, in recent months, as many as 400 officers a day, working overtime, have been dispatched to just 20 small zones deemed the city’s most dangerous.

The police say they are tamping down retaliatory shootings between gang factions by using a comprehensive analysis of the city’s tens of thousands of suspected gang members, the turf they claim, and their rivalries. The police are also focusing on more than 400 people they have identified as having associations that make them the most likely to be involved in a murder – as a victim or an offender.

New research: Focusing on key players in social networks

My recent research with several coauthors (see below) aims to investigate how targeting key players embedded in a network can have important effects on total activity. This is particularly true for crime – the success in reducing crime in Chicago was due to the targeting of 400 key players rather than spending resources on more general targets.

Doing this confronts two important questions concerning the role of social interactions and social networks in different activities such as crime, education, R&D, finance, and the diffusion of microfinance.

  • First is the role of peer effects in these activities.

To what extent do the people in your social network influence you to commit more or less crime, to have more or less R&D collaborations, or to give more or less bank loans?

  • Second, what is the empirical relevance of the ‘key player’ policy in real-world settings, where the aim is to target the agent who, once removed from the network, reduces total activity the most.

We also want to compare the predictive performance of the key player policy to the performance of other policies based on more traditional measures of network centrality and to other reasonable policies such as targeting the most active agents. In this article, we give an overview of these policies for crime, R&D, education, finance, and microfinance.


For crime, consider the study of Lindquist and Zenou (2014), in which we look at individuals in Sweden who are over 16 years old and who have been suspected of at least one crime. We have access to the official police register of all persons who are suspected of committing a crime in Sweden. In this register, the police keeps records of who is suspected of committing a crime with whom. In this context, a (criminal) link exists between two individuals if they are suspected of committing a crime together (and are then convicted). Both the convictions data and suspects data include crime type, crime date, and sanction received. We look at the relative effect of removing the key player in those cases in which the key player is no longer a part of the active network. To do this, we create an indicator variable for each person indicating whether or not they have died during the relevant time period and if they have been placed in prison. Our results indicate that, in the real world, the key player policy outperforms the random player policy by 9.58%. The key player policy also outperforms the policy of removing the most active player by 3.16% and the policy of removing the player with the highest eigenvector and betweenness centrality by 8.12% and 2.09%, respectively.

R&D networks

For R&D networks, we follow the study of König et al. (2014), in which we determine the key players – or more exactly the key firms – in R&D networks. In other words, our aim is to determine which firms are crucial for an industry in the sense that, if they exit the market (i.e. go bankrupt), the cost in terms of total activity or welfare for the remaining firms and consumers will be the highest possible. Using the MERIT-CATI dataset, we determine the key firms over a period of more than 40 years, and show that the key firms are usually not the ones with the largest number of R&D collaborations (degree), number of patents, nor the highest eigenvector, betweenness, or closeness centrality and, more importantly, not the firm with the highest market share in its sector. Interestingly, General Motors, which was bailed out in 2008 by President-elected Obama, was among the key firms. We show that, if General Motors were to have been removed from the market in 1990, then total welfare would have been reduced by 8.37%, while total output would have been decreased by 2.14%.


For education, we consider the recent study by Hahn et al. (2014), in which we conducted a controlled experiment in Bangladesh for fourth graders in rural primary schools. After conducting a network survey asking each student to nominate his/her best friends, we generated random groups of four in each school and asked each group to perform different maths and knowledge tests. The key question we want to answer is: If, by chance, a student ends up in a group with a high average centrality, does this positively impact on his/her test score as compared to someone who finds him/herself in a group with a lower average centrality? We also test the role of leadership in a group on grade outcomes by looking at the impact of the individual with the highest centrality in the group on outcomes. We find that, among the six centralities considered, the key-player and the Katz-Bonacich centrality are the ones that matter the most in explaining the performance of the students both in groups and individually. We find that, for both tests, it is the Katz-Bonacich centrality that performs best. These results indicate that the composition of a group in terms of centrality matters for both individual and group outcomes, and that key-player and Katz-Bonacich centralities are key for these outcomes.

Financial networks

We also look at financial networks using the study of Denbee et al. (2014). Using daily data from January 2006 to September 2010, the authors identify risk key players, that is, the banks that contribute the most to aggregate liquidity risk. They show that the risks that key players took during these periods vary a lot. They also find that the key players in the network are not necessarily the largest borrowers. In fact, during the credit boom, large lenders and borrowers are equally likely to be key players. This set of findings is of policy relevance, and gives guidance on how to effectively inject liquidity, to reduce the network risk, if the government decides to intervene.

Finally, we examine the role of key players in the diffusion of microfinance loans in India using the study of Banerjee et al. (2013). Their key question is: How do the network positions of the first individuals in a village to receive information about a new product affect its eventual diffusion? These authors show that the communication centrality of the injection points is a strong predictor of eventual participation in microfinance, and should therefore provide guidance to anyone trying to spread the news about microfinance in similar villages.


Banerjee, A, A G Chandrasekhar, E Duflo, and M O Jackson (2013), “The diffusion of microfinance”, Science 341(6144). 

Denbee, E, C Julliard, Y Li, and K Yuan (2014), “Network risk and key players: A structural analysis of interbank liquidity”, unpublished manuscript, London School of Economics and Political Science.

Hahn, Y, A Islam, E Patacchini, and Y Zenou (2014), “Network and peer effects in education: Evidence from a field experiment in Bangladesh”, unpublished manuscript, Stockholm University.

König, M D, X Liu, and Y Zenou (2014), “R&D networks: Theory, empirics and policy implications”, CEPR Discussion Paper 9872. 

Lindquist, M J and Y Zenou (2014), “Key players in co-offending networks”, CEPR Discussion Paper 9889. 



Topics:  Education Financial markets Frontiers of economic research

Tags:  networks, Peer Effects, crime, R&D, financial markets, microfinance

Professor of Economics at Stockholm University and CEPR Research Fellow