Social Interactions with Endogeneity

We identify peer and contextual effects in social interactions models with endogenous covariates (e.g., self-selected treatments). Unobserved individual characteristics are correlated with these endogenous covariates and affect peer outcomes in the reduced form. Our method uses instruments for endogenous covariates, but does not require additional instruments for simultaneity in outcomes, which are often hard to find in linear-in-means models. Nor does it require any exclusion restriction that some covariates have no contextual effects. The method can be applied to relax the Stable Unit Treatment Value Assumption (SUTVA) in the program evaluation literature, allowing individual treatments to influence the outcomes of other group members through both peer and contextual effects. We apply our method to estimate social effects in Grade 3 math scores of elementary school students in the State of Tennessee. Using lagged class sizes, teacher qualifications and students' self-reported motivation scores as instruments for Grade 2 scores, we find significant evidence for positive peer effects and path dependence on G2 scores.

This paper provides a method to study quantile effects in discrete choice with social interactions. The method is based on a behavioral social interactions model from quantile preference in decision making and demonstrates peer effects on different quantiles of discrete outcomes. The peer effects parameters are estimated by a nested pseudo score (NPS) approach, which is developed to tackle the computational burden pertaining to the social interactions model. Consistency and asymptotic normality are established for the proposed NPS estimator. We illustrate the finite sample performance of the model and the estimator by Monte Carlo experiments and an application of peer effects among students on exercise decisions, using the National Longitudinal Study of Adolescent Health dataset.

Social interaction model enhanced with speculation stage for human trajectory prediction

This dissertation focuses on analyzing peer effects in household decisions and the diffusion of renewable energy. The first chapter investigates peer effects in two family planning decisions among Chinese households – having a second child and having a son. The second chapter focuses on evaluating environmental policies (tax credit and Corporate Average Fuel Economy standard) on the diffusion of electric vehicles in the US. And the third chapter analyzes peer effects in residential solar panels adoption with a geographic focus on California. In summary, the first and third chapter adopt two structural peer effects models to analyze household behavior under two distinct decision-making context – family planning and solar panels adoption. And the second and third chapter focuses the diffusion of two renewable energy powered products – electric vehicle and solar panels.Peer effect measures how much the decision made by an agent (usually refers to an house-hold in this dissertation) is influenced by peers’ decisions under the same decision context. Manski (1993) summarizes the obstacles in identifying peer effects. The first is to separate peer effects with contextual effects, or how much the observed similarity in decision making among peer group members are attributed to similar backgrounds between peers due to endogenous group formation. The second is to separate peer effects with correlated effects, which refers to unobserved household characteristics and are believed to be correlated with each other. We use static and dynamic structural peer effects models to analyze family planning decisions and solar panels adoption decision separately, and these models are capable of disentangling the difficulties mentioned above. For the demand estimation of electric vehicle, we use a random coefficient model which has been broadly used in industrial organization.The first chapter is motivated by the increasingly unbalanced sex ratio in China. This phenomenon and associated social challenges have been widely documented, though few studies have rigorously investigated the role that peer effects have played in this unbalanced sex ratio. This paper fills this gap by focusing on peer effects in the decision to have a second child, and to have a son. The data we use comes from the 2016 data of China Family Panel Studies, and is a ten-year cohort of women aged 45-54 by 2016; we use a structural discrete choice model to estimate the peer effects. We find that peer choices significantly influence the probability that a family has a second child, but not the probability of having a son. Instead, having a son is largely driven by contextual effects, and in particular, by the education level of one’s peer group.The second chapter uses the random coefficient model with post-estimation counterfactual analysis to answer two research questions: (1) How much the tax credit has facilitated the diffusion of EV; (2) How much the CAFE standard and penalty level have facilitated the diffusion of EV. We obtain the data from Wards Auto with a years range from 2012 to 2019.We find that the EV market share will decrease by 35.82% if there is no tax credit. CAFE marks down the price of EV in average by 3.4 percent but marks up the price of other types of vehicles by 3.26 percent, whose absolute value far exceeds the CAFE penalty itself. We also find that increasing the penalty level from$55 to$140 per vehicle per mpg below the standard will only increase the EV market share by 0.23% and decrease the non-EV market share by 0.12%.The third chapter applies a utility-based structural optimal stopping time model developed by de Paula (2009) to analyze solar PV adoption. We use both econometrics model and nonparametric test to support the evidence of peer effects, using public solar PV data obstained from CaliforniaDGStats. And we apply the optimal stopping time model with a confidential data set obtained from PG&E. We find significant peer effects and correlated effects in a case study which contains 20 non-adjacent communities in the suburb of San Jose. And we predicted the adoption rate in this area will increase from 19.77% in 2019 to39.65% in 2029.

In this paper we analyze the impact of classroom peers on individual student performance with a unique longitudinal data set covering all Florida public school students in grades 3-10 over a five-year period. Unlike many previous data sets used to study peer effects in education, our data set allow us to identify each member of a given student's classroom peer group in elementary, middle, and high school as well as the classroom teacher responsible for instruction. As a result, we can control for individual student fixed effects simultaneously with individual teacher fixed effects, thereby alleviating biases due to endogenous assignment of both peers and teachers, including some dynamic aspects of such assignments. Our estimation strategy, which focuses on the influence of peers' fixed characteristics - both observed and unobserved - on individual test score gains, also alleviates potential biases due to error in measuring peer quality, simultaneity of peer outcomes, and mean reversion. Under linear-inmeans specifications, estimated peer effects are small to non-existent, but we find some sizable and significant peer effects within non-linear models. For example, we find that peer effects depend on an individual student's own ability and on the ability level of the peers under consideration, results that suggest Pareto-improving redistributions of students across classrooms and/or schools. Estimated peer effects tend to be smaller when teacher fixed effects are included than when they are omitted, a result that suggests co-movement of peer and teacher quality effects within a student over time. We also find that peer effects tend to be stronger at the classroom level than at the grade level.

: Many previous peer effects in higher education studies have assumed that peer groups form at the roommate, dorm floor, or dorm-level. Random assignment of students into squadrons at the US Air Force Academy allows us to identify the peer group with which students spend a majority of their time interacting. Using the squadron as the peer group, we find peer effects of much larger magnitude than those found in the previous literature. In separate estimations, we find for freshman students, a 100-point increase in the peer group average SAT verbal score increases individual GPA by 0.45 grade points and a 1-point increase in peer group GPA increases individual GPA by 0.65 grade points. Our results demonstrate the critical importance of properly identifying the relevant peer group when estimating peer effects. As evidence of this, we find that geographic proximity of students in dorm halls alone, as in Foster \201forthcoming\202, does not generate measurable peer effects. We also find smaller peer effects at the roommate level, which virtually disappear once we control for the squadron-level peer effects. Our models correct for the endogeneity of individual and peer outcomes and rule out common shocks as the mechanism driving the peer effects.

: Many previous peer effects in higher education studies have assumed that peer groups form at the roommate, dorm floor, or dorm-level. Random assignment of students into squadrons at the US Air Force Academy allows us to identify the peer group with which students spend a majority of their time interacting. Using the squadron as the peer group, we find peer effects of much larger magnitude than those found in the previous literature. In separate estimations, we find for freshman students, a 100-point increase in the peer group average SAT verbal score increases individual GPA by 0.45 grade points and a 1-point increase in peer group GPA increases individual GPA by 0.65 grade points. Our results demonstrate the critical importance of properly identifying the relevant peer group when estimating peer effects. As evidence of this, we find that geographic proximity of students in dorm halls alone, as in Foster \201forthcoming\202, does not generate measurable peer effects. We also find smaller peer effects at the roommate level, which virtually disappear once we control for the squadron-level peer effects. Our models correct for the endogeneity of individual and peer outcomes and rule out common shocks as the mechanism driving the peer effects.

In literature, the analysis of the influence of the environmental context on individual choices and behaviors, with particular reference to areas related to school and academic education, is usually performed through the methodology of peer effects. In this study, after presenting a brief overview on this approach, we propose a procedure for the analysis of the dependence of students’ academic performances on the contextual effects, determined by the sociometric ties observed between them, and by the subsequent division of the network into groups, using the linear-in-mean model for social interactions. This procedure is then applied to real data, collected in a second level degree course of the university of Rome Tor Vergata, for the construction of some models on student academic achievement. Empirical evidence suggests that peer effects are a significant determinant of performance, and that they identify explanatory aspects of individual achievement, that usual regressors are not able to catch.

We construct a peer effects model where mean expenditures of consumers in one's peer group affect utility through perceived consumption needs. We provide a novel method for obtaining identification in social interactions models like ours, using ordinary survey data, where very few members of each peer group are observed. We implement the model using standard household‐level consumer expenditure survey microdata from India. We find that each additional rupee spent by one's peers increases perceived needs, and thereby reduces one's utility, by the equivalent of a 0.25 rupee decrease in one's own expenditures. These peer costs may be larger for richer households, meaning transfers from rich to poor could improve even inequality‐neutral social welfare, by reducing peer consumption externalities. We show welfare gains of billions of dollars per year might be possible by replacing government transfers of private goods to households with providing public goods or services, to reduce peer effects.

Skin damage is a trait of economic and welfare importance that results from social interactions between animals. These interactions may produce wound signs on the gilt's skin as a result of damage behavior (i.e., fighting), biting syndromes (i.e., tail, vulva, or ear biting), and swine inflammation and necrosis syndrome. Although current selection for traits that are affected by social interactions primarily focuses on improving direct genetic effects, combined selection on direct and social genetic effects could increase genetic gain and avoid a negative response to selection in cases of competitive behavior. The objectives of this study were to (1) estimate variance components for combined skin damage (CSD), with or without accounting for social genetic effects, (2) investigate the impact of including genomic information on the prediction accuracy, bias, and dispersion of CSD estimated breeding values, and (3) perform a single-step genome-wide association study (ssGWAS) of CSD under a classical and a social interaction model. Our results show that CSD is heritable and affected by social genetic effects. Modeling CSD with social interaction models increased the total heritable variance relative to the phenotypic variance by three-fold compared to the classical model. Including genomic information increased the prediction accuracy of direct, social, and total estimated breeding values for purebred sires by at least 21.2%. Bias and dispersion of estimated breeding values were reduced by including genomic information in classical and social interaction models but remained present. The ssGWAS did not identify any single nucleotide polymorphism that was significantly associated with social or direct genetic effects for CSD. Combined skin damage is heritable, and genetic selection against this trait will increase the welfare of animals in the long term. Combined skin damage is affected by social genetic effects, and modeling this trait with a social interaction model increases the potential for genetic improvement. Including genomic information increases the prediction accuracy of estimated breeding values and reduces their bias and dispersion, although some biases persist. The results of the genome-wide association study indicate that CSD has a polygenic architecture and no major quantitative trait locus was detected.

We point-identify and estimate linear social network models without observing any network links. The required data consist of many small networks of individuals, such as classrooms or villages, with individuals who are each observed only once. We apply our estimator to data from Tennessee’s Project STAR (Student-Teacher Achievement Ratio). Without observing the latent network in each classroom, we identify and estimate peer and contextual effects on students’ performance in mathematics. We find that peer effects tend to be larger in bigger classes and that increasing peer effects would significantly improve students’ average test scores in some classes.

Summary In this paper, we introduce a social interaction econometric model with an extreme order statistic to model peer effects. We show that the model is a well-defined system of equations and that it is a static game with complete information. The social interaction model can include exogenous regressors and group effects. Instrumental variables estimators are proposed for the general model that includes exogenous regressors. We also consider distribution-free methods that use recurrence relations to generate moment conditions for estimation. For a model without exogenous regressors, the maximum likelihood approach is computationally feasible.

Cross-section and panel estimates of peer effects in early adolescent cannabis use: With a little help from my ‘friends once removed’

Neural correlates of rapid reversal learning in a simple model of human social interaction

This paper aims at opening the black box of peer effects in adolescent weight gain. Using Add Health data on secondary schools in the U.S., we investigate whether these effects partly flow through the eating habits channel. Adolescents are assumed to interact through a friendship social network. We first propose a social interaction model of fast food consumption using a generalized spatial autoregressive approach. We exploit results by Bramoulle, Djebbari and Fortin (2009) which show that intransitive links within a network (i.e., a friend of one of my friends is not my friend) help identify peer effects. The model is estimated using maximum likelihood and generalized 2SLS strategies. We also estimate a panel dynamic weight gain production function relating an adolescent’s Body Mass Index (BMI) to his current fast food consumption and his lagged BMI level. Results show that there are positive significant peer effects in fast food consumption among adolescents belonging to a same friendship school network. The estimated social multiplier is 1.59. Our results also suggest that, at the network level, an extra day of weekly fast food restaurant visits increases BMI by 2.4%, when peer effects are taken into account.