Abstract We study opinion dynamics in a social network consisting of two groups. Agents update their opinions by conforming to members of their own group while rejecting the views of the opposing group (affective polarization), and by listening to a media outlet that may provide biased information. We characterize the long-run opinions and identify when affective polarization and media bias lead to ideological polarization, persistent disagreement, or failures of learning. We also derive when information interventions or censorship improve the accuracy of average opinions and reduce disagreement, and when they backfire: better information helps only under specific media bias configurations and when directed to the agents we identify as most effective at propagating it through the network.

This paper studies the behavior of Katz–Bonacich centrality and key-player intercentrality in linear–quadratic network games when spillovers are strong. Understanding this regime is both theoretically important and empirically relevant in environments where interactions are intense. We show that as spillovers become large, Katz–Bonacich centrality converges (up to scale) to the Perron eigenvector, while intercentrality diverges at the same rate for all nodes and becomes dominated by a common leading component. As a result, first-order proportional differences in intercentrality—normalized by the common divergent scale—vanish near the critical threshold. Rankings based on lower-order differences in intercentrality, or on post-removal activity and welfare levels, need not collapse and may remain stable even arbitrarily close to the boundary.

We examine how social interactions and friendships shape students’ political opinions in a natural experiment at Sciences Po, a leading French university specializing in social and political sciences. The quasi-random assignment of students into short-term integration groups before their academic curriculum reduces political opinion gaps and fosters friendship formation. Using same-group membership as an instrumental variable for friendship, we find that friendship reduces opinion differences by 44 percent of the mean opinion gap. Our evidence supports a homophily-enforced mechanism: Friendships form among initially politically similar students, leading them to join political associations together, reinforcing their similarity. (JEL C93, D72, Z13)

This paper studies how natural disasters spread conflicts within a network. We first construct a new panel data set that combines geo-referenced information about conflict events and natural disasters, for 5,944 districts in 53 African countries, over the period 1989-2020. Considering natural disasters as exogenous shocks that affect the combatants' activity in a locality, we find that natural disasters decrease conflict incidence in the affected locality, increase conflict incidence in neighbouring localities, and lead to an overall net increase in conflict incidence. The spatial dispersion of conflict varies by the level of local rent-seeking opportunities and the level of international, post-disaster aid. We then provide a simple theoretical framework that may explain this conflict dispersion pattern. Findings provide important implications for implementing local and aggregate level conflict mitigation policies.

We study how the degree of zero-sumness in a social environment shapes cooperation and network formation. When individual gains partly come at others' expense, agents face a trade-off between contributing to a public good and extracting rents from their network position. We model this as a public good game on a directed, weighted network with a parameter capturing the intensity of zero-sum conflict. Our Zero-Sum Theorem shows that greater zero-sumness increases contributions while reducing social connectedness: agents sever links to limit exposure to extraction and increase self-provision to compensate for the diminished free-riding opportunities associated with sparser networks. The unique Nash equilibrium has contributions equal to weighted sign-alternating Bonacich centralities, reflecting how network position shapes behavior through alternating spillovers while incorporating extraction incentives. With endogenous links, equilibrium total connectivity is uniform across homogeneous agents, and heterogeneity in extraction incentives generates an inverse relationship between rent-extraction benefits and network centrality.

We develop a dynamic model in which individuals allocate time between work and religious activities, and parents invest in their children's human capital and religious belief. The model delivers, in a unified framework, the two empirical regularities documented by Barro and McCleary (2003) and McCleary and Barro (2019): controlling for religious activities, stronger belief raises economic growth because it raises human capital investment; controlling for belief, more time spent on religious activities lowers growth by crowding out labor supply. While the labor-supply margin is individually optimal, the human-capital margin is not: parents do not internalize that greater human-capital investment crowds out future religious transmission through the socialization channel, leading to inefficiently high human capital in equilibrium under strong socialization externality. We extend this baseline framework in three directions. First, introducing a complementarity between religious belief and human capital-capturing the Protestant-ethic channel of Weber (1930)-we show that the efficiency of equilibrium depends non-monotonically on the strength of this complementarity. Second, allowing for cultural conflict between two religious groups à la Bisin and Verdier (2000), we show that cultural intolerance depresses human capital investment and, if human capital raises labor productivity, reduces economic growth. Third, embedding the model in a system of cities, we show that larger, more productive cities endogenously attract workers who invest more in human capital and spend less time on religious activities, generating a negative cross-city relationship between city size and religiosity that is consistent with the empirical evidence in McCleary and Barro (2019).

We develop a simple multilayer network model in which agents allocate effort across layers with heterogeneous structures, subject to an aggregate effort constraint. Incentives are shaped by agents' network positions within each layer, and equilibrium behavior reflects both within-and cross-layer interactions. We analyze how shocks propagate through the network and characterize optimal targeting interventions. Our results show that effective policy design must account for effort allocation across layers. We also demonstrate that predictions from monolayer models can diverge sharply from those of multilayer models, underscoring the importance of accounting for network complexity in both empirical and policy analyses.

We study an Arrow-Debreu economy with externalities generated by multiplex networks. Market equilibrium prices reflect both the preferences and scarcity of goods, consumers' network centralities arising from goods' externalities, as well as linkages across goods (layers) through the budget constraint. Despite the presence of externalities, competitive markets can still be efficient: the First and Second Welfare Theorems hold if either all networks are regular or all layers share the same network structure. When markets allocate goods inefficiently, a Lindahl equilibrium-implemented through personalized prices-can restore efficiency, but may leave some consumers worse off.

This article investigates whether virtual contact, initiated through a documentary film, can promote interethnic harmony. We carried out a cluster-randomized field experiment involving over 3,300 households across 121 multiethnic villages in Bangladesh. We find that a documentary film, designed to humanize the ethnic-minority Santals and evoke empathy among the ethnic majority Bengalis, increased the ethnic majority's prosociality toward minorities, though the strength of the evidence varies by treatment arm and outcome. Using emotion-detecting software to analyze facial expressions during the film viewing suggests that the documentary elicited emotional responses consistent with empathy. We do not find evidence that the intervention reduced the prevalence of negative stereotypes and discriminatory opinions toward minorities. In villages assigned to target network-central people, we find positive behavioral effects on untreated individuals, including Santals, and village-level administrative data suggest a reduction in police complaints in those villages. About five months after the intervention, we conducted a casual-work field experiment involving 720 participants from the main intervention. In this task, pairs of ethnic-majority and minority participants jointly produced paper bags for a local supplier under a piece-rate compensation scheme. We find positive treatment effects on productivity for both ethnic groups, with effects concentrated in villages where network-central people were treated. For the ethnic majority, increased prosociality, and for the ethnic minority, reciprocity or peer pressure may have contributed to increased productivity. Overall, our findings suggest that virtual contact and social networks may help promote harmony in multiethnic communities.

We study the optimal joint intervention of a planner who can influence both the standalone marginal utilities of agents in a network and the weights of the links connecting them. The welfare-maximizing intervention displays two key features. First, when the planner's budget is moderate (yielding interior solutions), the optimal change in link weight between any pair of agents is proportional to the product of their eigen-centralities. Second, when the budget is sufficiently large, the optimal network converges to a simple structure: a complete network under strategic complements, or a complete balanced bipartite network under strategic substitutes. We show that welfare effects are governed by the principal eigenvalue of the network, while distributional outcomes are driven by the dispersion of the corresponding eigen-centralities. Comparing joint interventions to single interventions targeting only standalone marginal utilities, we find that joint interventions consistently generate higher aggregate welfare, but may also increase inequality, revealing a potential trade-off between efficiency and equity.

In settings where resistance and rampant misinformation against vaccines exist, the prospect of containing infectious diseases remains a challenge. Can delivery of information regarding the benefits of vaccination through personal home visits by local ambassadors increase vaccine uptake? We conduct a door-to-door randomized information campaign targeted towards COVID-19 unvaccinated individuals in rural Indonesia. We recruited ambassadors from local villages tasked to deliver information about COVID-19 vaccines and promote vaccination through one-on-one meetings, using an interpersonal behavioral change communication approach. To investigate which type of ambassador—health cadres, influential individuals, and laypersons—is the most effective, we randomly vary the type of ambassador that delivers the information at the village level. We find that the overall vaccination take-up is quite moderate and that there are no differences in vaccination outcomes across the treatment groups. These results highlight the challenge of boosting vaccine uptake in late stages of a pandemic.

This paper analyzes the role of networks in the spatial diffusion of local economic shocks in Africa. We show that geographic connectivity, along with road and ethnic connectivity, are important for diffusing economic spillovers over longer distances. We then determine the key players, that is, which districts are key in propagating local economic shocks across Africa. Using these results, we conduct counterfactual policy exercises to evaluate the potential gains from policies that increase economic activity in specific districts or improve road connectivity between districts.

We study the dynamics between local segregation, partisanship, and political polarization. We exploit large-scale, exogenous and high-stakes peer assignment due to universal conscription of soldiers assigned from each of 34,947 municipalities to French infantry regiments during WWI. We find that municipalities with soldiers serving with the same line regiment converge in their post-war voting behaviors. Soldiers from rural municipalities exposed to more leftist regimental peers become more leftist for the first time after the war, while adjacent municipalities assigned to the right are inoculated against the left. We provide evidence that these differences reflect persuasive information exchanged among peers when the stakes for cooperation and trust are high rather than group conformity. These differences further lead to the emergence of sharp and enduring post-war discontinuities across 435 regimental boundaries that are reflected, not only in voting, but also inviolent civil conflicts between Collaborators and Resistants during WWII.

We consider a model in which each individual belonging to an ethnic minority group is embedded in a network of relationships and decides whether she wants to be integrated in the society. Each individual wants her behavior to agree with her personal ideal preference for integration but also wants her behavior to be as close as possible to the average integration behavior of her peers. We show that there is always convergence to a steady-state and characterize it. We also show that different preferences for integration may emerge in steady state depending on the structure of the network. Then, we consider an optimal tax/subsidy policy which aim is to reach a certain level of integration in the population.

We develop a multi-agent model of the education production function where investments of students, parents, and teachers are linked to the presence of minorities in the classroom. We then test the key implications of this model using rich survey data and a mandate to randomly assign students to classrooms. Consistent with our model, we show that exposure to minority peers decreases student effort, parental investments, and teacher engagement and it results in lower student test scores. Observables correlated with minority status explain less than a third of the reduced-form test score effect while over a third can be descriptively attributed to endogenous responses of the agents.