We introduce an R&D network formation model where firms choose both R&D efforts and collaboration partners. Neighbors in the network benefit from each other's R&D through technology spillovers, and there exists competition effects reflecting strategic substitutability in R&D. We provide a complete equilibrium characterization, and develop an estimation method that is computationally feasible for large networks. We then conduct an analysis of R&D collaboration subsidies, and find that a subsidy scheme targeting specific R&D collaborations can be more effective than a uniform subsidy, with a welfare gain up to five times larger than the cost of the subsidy.

This paper studies gender and racial disparities in altruism among social network members who are endogenously linked. We specify group (gender or race) specific altruistic interactions models, as well as intra- and inter-group altruistic interactions models, to capture the heterogeneous patterns of altruism associated with the characteristics of two individuals in pairs. We apply the models to the Add Health data to identify altruism and social interaction effects on academic achievement and smoking behaviors among adolescents. The results indicate that females are generally more altruistic than males, and whites appear to be the most altruistic racial group. We also find that blacks exhibit spiteful effects toward other black students who are considered to “act white.”

This paper analyzes statistical issues arising from non-representative samples of a network. Sampled network data could systematically bias the network properties and generate non-classical measurement error problems. Apart from the sampling rate and the elicitation procedure, the biases on network structural measures depend non-trivially on which subpopulations of nodes are missing with higher probability. We propose a methodology, adapting weighted estimators to networked contexts, which enables researchers to recover several network-level statistics and reduce the biases in the estimated network effects. The proposed weighted estimators are consistent and asymptotically normally distributed and have good performance in finite samples. Notably, our approach does not require users to assume any network formation model and is straightforward to implement.

AbstractThis paper introduces a structural model for the coevolution of networks and behavior. We characterize the equilibrium of the underlying game and adopt the Bayesian Double Metropolis-Hastings algorithm to estimate the model. We further extend the model to incorporate unobserved heterogeneity and show that ignoring this heterogeneity can lead to biased estimates in simulation experiments. We apply the model to study R&D investment and collaboration decisions in the chemical and pharmaceutical industry and find a positive knowledge spillover effect. Our model also provides a tractable framework for a long-run key player analysis.

This paper studies the impact of collaboration on research output. First, we build a micro-founded model for scientific knowledge production, where collaboration between researchers is represented by a bipartite network. The Nash equilibrium of the game incorporates both the complementarity effect between collaborating researchers and the substitutability effect between concurrent projects of the same researcher. Next, we propose a Bayesian MCMC procedure to estimate the structural parameters, taking into account the endogenous participation of researchers in projects. Finally, we illustrate the empirical relevance of the model by analyzing the coauthorship network of economists registered in the RePEc Author Service. The estimated complementarity and substitutability effects are both positive and significant when the endogenous matching between researchers and projects is controlled for, and are downward biased otherwise. To show the importance of correctly estimating the structural model in policy evaluation, we conduct a counterfactual analysis of research incentives. We find that the effectiveness of research incentives tends to be understated when the complementarity effect is ignored and overstated when the substitutability effect is ignored.

This paper studies cross-border networks formed in firms' prior destinations in the context of cross-border mergers and acquisitions (M&A). An acquirer owns a subsidiary in a prior destination, and the subsidiary's domestic network forms the cross-border network of the acquirer due to their ownership link. We focus on the subsidiary's network of neighboring firms in a location within the prior destination. Using rich data on cross-border M&A, we show that knowledge spillovers from this cross-border network have a positive impact on the acquirer's likelihood of entering a new destination. Our results highlight the global externalities of multinationals' international economic activities.

Abstract Empirical evidence suggests that close to 100 million women are “missing” worldwide. We revisit the empirical evidence for China, the country with the most missing women. Nearly ten million girls born in the 1980s and 1990s who were “missing” according to earlier census data can be found again in the 2010 population census. We discuss two possible explanations for the re-emergence of these formerly missing girls: the delayed registration of girls owing to economic reasons, and the response to amendments to the Chinese Statistics Law in 2009 and policy changes in the 2010 population census. Using the most recent statistics, we document patterns of the underreporting of women over time and across regions as well as explore the basic determinants of underreporting of women. Important policy challenges remain. For the unregistered children, the lack of access to public services will increase their vulnerability and adversely affect their quality of life.

This paper is concerned with methods for analyzing social interaction effects. The attention is focused on how to estimate endogenous effects, where an individual's choice may depend on those of his/her contacts about the same activity. The analysis is guided by the data structure that is available to measure social interactions, an intuitive aspect that allows empirical researchers to understand whether and how they could study social interaction effects in their own data. First, the case where the information on social interaction patterns is limited to membership to a given group is considered, then the discussion moves to the case where the data contain information on specific relationships among pairs of individuals within each group, and the availability of data on the co-evolution of social structures and outcomes. This paper also discusses some basic methods to deal with online social network data, and the novel literature estimating social interaction effects relying only on outcome data. For each data structure, the challenges and the main methods proposed in the literature to tackle them are reviewed.

We model network formation and interactions under a unified framework by considering that individuals anticipate the effect of network structure on the utility of network interactions when choosing links. There are two advantages of this modeling approach: first, we can evaluate whether network interactions drive friendship formation or not. Second, we can control for the friendship selection bias on estimated interaction effects. We provide microfoundations of this statistical model based on the subgame perfect equilibrium of a two‐stage game and propose a Bayesian MCMC approach for estimating the model. We apply the model to study American high school students' friendship networks using the Add Health dataset. From two interaction variables, GPA and smoking frequency, we find that the utility of interactions in academic learning is important for friendship formation, whereas the utility of interactions in smoking is not. However, both GPA and smoking frequency are subject to significant peer effects.

This paper studies the impact of collaboration on research output. First, we build a micro founded model for scientific knowledge production, where collaboration between researchers is represented by a bipartite network. The equilibrium of the game incorporates both the complementarity effect between collaborating researchers and the substitutability effect between concurrent projects of the same researcher. Next, we develop a Bayesian MCMC procedure to estimate the structural parameters, taking into account the endogenous matching of researchers and projects. Finally, we illustrate the empirical relevance of the model by analyzing the co-authorship network of economists registered in the RePEc Author Service.

SummaryWe study social interactions when individuals hold altruistic preferences in social networks. Rich network features can be captured in the resulting best response function. The inward network links provide unique information for identifying the altruism effect. We demonstrate that the often ignored altruism is another serious confounding factor of peer effects. Specifically, the estimates of peer effects are approximately 36% smaller after taking into account social preferences. Furthermore, we could identify two types of effects caused by peers' outcomes: the spillover effects and the externality effects, which is impossible in a conventional social interactions model based on the self‐interest hypothesis.

Abstract This study primarily seeks to answer the following question: How do social networks evolve over time and affect individual economic activity? To provide an adequate empirical tool to answer this question, we propose a new modeling approach for longitudinal data of networks and activity outcomes. The key features of our model are the inclusion of dynamic effects and the use of time-varying latent variables to determine unobserved individual traits in network formation and activity interactions. The proposed model combines two well-known models in the field: latent space model for dynamic network formation and spatial dynamic panel data model for network interactions. This combination reflects real situations, where network links and activity outcomes are interdependent and jointly influenced by unobserved individual traits. Moreover, this combination enables us to (1) manage the endogenous selection issue inherited in network interaction studies, and (2) investigate the effect of homophily and individual heterogeneity in network formation. We develop a Bayesian Markov chain Monte Carlo sampling approach to estimate the model. We also provide a Monte Carlo experiment to analyze the performance of our estimation method and apply the model to a longitudinal student network data in Taiwan to study the friendship network formation and peer effect on academic performance. Supplementary materials for this article are available online.

This paper introduces a structural model for the coevolution of networks and behavior. The microfoundation of our model is a network game where agents adjust actions and network links in a stochastic best-response dynamics with a utility function allowing for both strategic externalities and unobserved heterogeneity. We show the network game admits a potential function and the coevolution process converges to a unique stationary distribution characterized by a Gibbs measure. To bypass the evaluation of the intractable normalizing constant in the Gibbs measure, we adopt the Double Metropolis-Hastings algorithm to sample from the posterior distribution of the structural parameters. To illustrate the empirical relevance of our structural model, we apply it to study R&D investment and collaboration decisions in the chemicals and pharmaceutical industry and find a positive knowledge spillover effect. Finally, our structural model provides a tractable framework for a long-run key player analysis.

We study the impact of research collaborations in coauthorship networks on research output and how optimal funding can maximize it. Through the links in the collaboration network, researchers create spillovers not only to their direct coauthors but also to researchers indirectly linked to them. We characterize the equilibrium when agents collaborate in multiple and possibly overlapping projects. We bring our model to the data by analyzing the coauthorship network of economists registered in the RePEc Author Service. We rank the authors and research institutions according to their contribution to the aggregate research output and thus provide a novel ranking measure that explicitly takes into account the spillover effect generated in the coauthorship network. Moreover, we analyze funding instruments for individual researchers as well as research institutions and compare them with the economics funding program of the National Science Foundation. Our results indicate that, because current funding schemes do not take into account the availability of coauthorship network data, they are ill-designed to take advantage of the spillover effects generated in scientific knowledge production networks.

Social interactions are widely recognized to play an important role in smoking initiation among adolescents. In this paper we hypothesize that individual with `stronger' personalities (i.e. emotionally stable, conscientious individuals) are better able to resist peer pressure in the uptake of smoking. We exploit detailed friendship nominations in the US Add Health data, and extend the Spatial Autoregressive Model (SAR) model to deal with (i) endogenous peer selection, and (ii) unobserved contextual effects, in order to identify heterogeneity in peer effects with respect to personality. The results indicate that peer effects in the uptake of smoking are predominantly affecting individuals who are emotionally unstable. That is, individuals with `weaker' personalities are more vulnerable to peer pressure. This finding not only helps understanding heterogeneity in peer effects, but additionally provides a promising mechanism through which personality affects later life health and socioeconomic outcomes.

This paper analyzes statistical issues arising from networks based on non-representative samples of the population. We first characterize the biases in both network statistics and estimates of network effects under non-random sampling theoretically and numerically. Sampled network data systematically bias the properties of observed networks and suffer from non-classical measurement-error problems when applied as regressors. Apart from the sampling rate and the elicitation procedure, these biases depend in a non-trivial way on which subpopulations are missing with higher probability. We propose a methodology, adapting post-stratification weighting approaches to networked contexts, which enables researchers to recover several network-level statistics and reduce the biases in the estimated network effects. The advantages of the proposed methodology are that it can be applied to network data collected via both designed and non-designed sampling procedures, does not require one to assume any network formation model, and is straightforward to implement. We apply our approach to two widely used network data sets and show that accounting for the non-representativeness of the sample dramatically changes the results of regression analysis.

We apply a high order spatial autoregressive (SAR) model to simultaneously capture'heterogeneous peer effects from multiple gender and racial groups, as well as endogenous network formation. In students' GPA and smoking behaviors, we find that within-gender endogenous effects are stronger than cross-gender effects. Females and whites are more sensitive to peer influences and more influential than other students. Intra-race spillover effects are stronger than inter-race effects for whites, but not for non-whites. Homophily on observed and unobserved characteristics are important for friendship formation. However, the formation of friendship is not necessary motivated by common interest in outcomes such as smoking. Our findings suggest that coeducational or desegregated schooling may help increase academic achievement, but not reduce smoking frequency.

In this paper we introduce a stochastic network formation model where agents choose both actions and links. Neighbors in the network benefit from each other’s action levels through local complementarities and there exists a global interaction effect reflecting a strategic substitutability in actions. The tractability of the model allows us to provide a complete equilibrium characterization in the form of a Gibbs measure, and we show that the structural features of equilibrium networks are consistent with empirically observed networks. We then use our equilibrium characterization to show that the model can be conveniently estimated even for large networks. The policy relevance is demonstrated with examples of firm exit, mergers and acquisitions and subsidies in the context of R&D collaboration networks.