Back to overview

Social Interactions, Endogenous Networks, and Production

English title Social Interactions, Endogenous Networks, and Production
Applicant Pellizzari Michele
Number 165618
Funding scheme Project funding (Div. I-III)
Research institution Institute of Economics and Econometrics Geneva School of Economics and Management Université de Genève
Institution of higher education University of Geneva - GE
Main discipline Economics
Start/End 01.10.2016 - 30.09.2020
Approved amount 250'823.00
Show all

All Disciplines (3)

Education and learning sciences, subject-specific education

Keywords (3)

Networks; Education; Social interactions

Lay Summary (Italian)

Questo progetto si propone di studiare il ruolo dei contatti sociali (amici, parenti, conoscenti) nell’influenzare una varietà di comportamenti, quali ad esempio il rendimento scolastico, la riceca di lavoro ma anche comportamenti anti-sociali o addictive (crimine, fumo, et.). L’innovazione più importante di questo progetto consiste nell’incorporare nell’analisi anche il processo di formazione dei contatti sociali, permettendo quindi di comprendere non solo le conseguenze delle reti sociali ma anche i motivi e le intenzioni che guidano la formazione di tali reti.
Lay summary

È ormai noto che la interazioni tra amici, parenti e conoscenti rivestono un’importanza fondamentale nell’influenzare una grande varietà di comportamenti economici e sociali. Per esempio, moltissime persone trovano lavoro attraverso i propri contatti personali, gli studenti studiano spesso in gruppo influenzando in modo importante i propri rendimenti, numerosi comportamenti anti-sociali o che generano dipendenza nascono dalle interazioni con altre persone (crimine, fumo, alcolismo, et.). L’importanza delle reti sociali è stata da tempo riconosciuta da moltissime discipline scientifiche, l’economia, la sociologia, la psicologia. Ciò nonostante, la stragrande maggioranza degli studi in questo campo si concentra o sull’analisi dell’effetto delle reti sociali su uno specifico o comportamento (e.g. qual’è l’effetto dei compagni di studio sul proprio rendimento scolastico?) oppure sul processo di formazione di tali reti. In realtà questi due processi sono strettamente legati tra di loro perché le persone scelgono di formare relazioni per un motivo, con l’intenzione di svolgere delle attività in collaborazione.

Questo progetto si propone di studiare l’importanza delle reti sociali nel determinare i comportamenti delle persone tenendo in corretta considerazione il processo attraverso il quale tali reti si formano. Il nostro lavoro permetterà di rispondere, per esempio, alle seguenti questioni: qual’è il modo migliore di assegnare le persone a gruppi come le classi scolastiche o i gruppi di lavoro all’interno di un’azienda? È possibile (e come) migliorare la produttività di un gruppo di lavoro o di studio scegliendo in modo ottimale la composizione dei membri in termini di età, sesso e altre caratteristiche? Si tratta di questioni cruciali per qualsiasi istituzione di formazione (scuole, università, et.) e qualsiasi azienda.

Direct link to Lay Summary Last update: 09.09.2016

Responsible applicant and co-applicants



An Estimable Model of Production Interactions in Endogenous Networks
PellizzariMichele, De GiorgiGiacomo, TomásRodríguez Barraquer (2020), An Estimable Model of Production Interactions in Endogenous Networks, CEPR, London.


Networks of study partners at Bocconi University

Author Pellizzari, Michele; De Giorgi, Giacomo
Persistent Identifier (PID) not available
Repository Bocconi universty
The dataset contains information collected via online surveys from the students who entered the BA in Management at Bocconi university in September 2011. The surveys take place at the end of each academic year and collect information on the names of classmates with whom the respondents collaborated during the year. The names of both respondents and nominated students are anonymously linked to the administrative archives of the university with information on background socio-economic characteristics and academic performance.The data are used for the paper:De Giorgi, G, Pellizzari, M. and Rodríguez Barraquer, T. "An Estimable Model of Production Interactions in Endogenous Networks", 2020, CEPR Discussion Paper.The paper is currently being submitted for publication in scientific journals.The dataset is property of Bocconi university but a fully anonymised version with all the information required to replicate the results of the paper can be obtained free of charge from the university upon request and for the only purpose of replication


Group / person Country
Types of collaboration
NY Fed - Prof. Giacomo De Giorgi United States of America (North America)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel
Universidad de los Andes Colombia (South America)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel
Universitat Autònoma de Barcelona - Tomás Rodríguez Barraquer Spain (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Research Infrastructure
- Exchange of personnel

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
IEE Research Day Talk given at a conference Production interactions in endogenous networks 07.02.2020 Geneva, Switzerland Pellizzari Michele;
ASSA Annual Meeting 2021 Talk given at a conference Production Interaction in Endogenous Networks 04.01.2020 Virtual conference, United States of America Pellizzari Michele;
Causal Learning with Interactions Talk given at a conference Production interactions in endogenous networks 11.12.2019 London, Great Britain and Northern Ireland Pellizzari Michele;


Title Date Place
Topics in Structural Econometrics 01.11.2017 Geneva, Switzerland


This project seeks to simultaneously study the dynamic process of social network formation and the significance of peer effects in the formation of human capital. It is well known that interactions among friends, relatives and other acquaintances are of primary importance in numerous aspects of social and economic behavior. For example, very many people find work through social contacts, students often study in groups crucially affecting their academic achievement, vary many addictive or anti-social behaviors arise out of social interactions (crime, smoking, drinking, etc.). The importance of social networks has been acknowledged by many disciplines, from economics to sociology and others. However, the vast majority of studies in this area either look at the effect of social interactions on a specific phenomenon (e.g. what is the effect of peers on own achievement in school) or investigate the process of network formation (i.e. how people make links with each other). In reality, of course, the two processes are closely linked as people connect to one another with the purpose of engaging in some sort of interaction once the connection is established. In fact, it is impossible to formulate policy prescriptions without a structural model of behavior which endogenizes the network formation stage and recognizes the potential feedback between production and connection processes. This project aims at developing such a structural model and taking it to the data in a specific application (students in higher education). Ultimately, our work will allow answering the following set of questions: how can individuals be optimally allocated to classrooms, teams, or any other type of productive groups? Can productivity be improved by choosing team members optimally? By how much? These policy questions are crucial for educational institutions as well as firms in general.The problem is also important from a methodological perspective, as there do not exist general purpose estimable models of social dynamics which treat the network formation process as endogenous and simultaneously contemplate peer effects which alter behavior and feed-back into the network formation process. The paucity of the literature in this respect is due to two crucial aspects: (i) first, it is theoretically quite difficult to develop such a complex interaction model while at the same time maintaining it sufficiently parsimonious to be implemented empirically; and, second and perhaps more importantly, (ii) the data requirements for a convincing estimation strategy of such a model are extremely stringent. In fact, one needs information on true social networks, outcomes of a production or learning process, as well as some sources of exogenous variation in the linking probabilities. Essentially a large amount of information is needed to credibly disentangle the three key mechanisms behind the co-determination of network structure and behavior: (i) potential biases in the preferences of individuals over the types of their friends, (ii) biases in the opportunities they have to meet people of different types and (iii) the fact that via peer effects, as determined by their social networks, the characteristics (types) of agents change over time. Putting it simply, in the absence of longitudinal data and exogenous variation in the probability a link is formed, it is usually impossible to tell whether a given set of observable characteristics is relevant for the formation of links, or whether certain agents are linked because they share some of those characteristics.The current project possesses all the features required to overcome these difficulties, both from the point of view of the theoretical modeling and from the point of view of the availability of data. We will thus be able to develop and estimate a structural model in which agents decide to link with other agents while taking into account the type of interactions they will engage in. Using the resulting estimates it will be possible to conduct counterfactual experiments varying the allocation of agents across groups or varying the composition of the original allocations, thus allowing us to formulate policies prescriptions. For example, we will be able to propose alternative policies of allocating students to classes which would cost nothing and potentially improve achievement, similarly we can modify the characteristics of the entry cohorts in terms of their immutable attributes such as gender.