We analyze how mobile internet access affects desire and plans to emigrate. Our empirical analysis combines survey data on 617,402 individuals with data on worldwide 3G mobile internet rollout from 2008 to 2018. Exploiting temporal variation in 3G rollout from 2,120 subnational districts in 112 countries, we show that an increase in mobile internet access increases the desire and plans to emigrate. Using lightning incidence as an instrument provides additional evidence that the effects are causal. The effect on the desire to emi- grate is particularly strong for those with secondary education. In line with our theory, an important mechanism appears to be that access to the mobile internet lowers the cost of acquiring information on potential destinations. In addition to this, increased internet access reduces perceived material well-being and trust in government. Municipal-level data from Spain shows that 3G rollout also increased actual emigration flows.

Abstract: Understanding behavior in digital settings is important for researchers, policymakers, and companies. But measuring behavior online and conducting experiments is difficult for independent researchers, who do not have access to the user bases or technologies of large companies.  To address the challenge of conducting large-scale empirical studies for policy-relevant research online, we have developed Webmunk, an open-source tool designed to make research studies much easier. The user facing side of Webmunk is a browser extension that can track consumer browsing behavior and experimentally modify their experience as they browse the Internet.  We demonstrate the power of Webmunk in a project that explores self-preferencing and vertical integration by Amazon, where third party products are sold in direct competition with Amazon brands, such as Amazon Essentials and Amazon Basics. This talk will present the first results from a recently completed field experiment. 

Abstract: This paper investigates the impact of broadband internet penetration on economic growth using China’s city-level data during 2005 to 2019. The baseline OLS results from a structural model suggest that on average a 10% increase in broadband internet penetration growth is associated with a 0.4% increase in GDP per capita growth. A cost-benefit analysis translates this estimate into an average rate of return around 40% for investment in broadband internet infrastructure in China. To address the potential endogeneity, we exploit historical and topographical instrumental variables for broadband internet penetration for causal inference. Furthermore, utilizing the temporal and spatial variations from the Broadband China Pilot City Program, our staggered difference-in-differences analysis reveals innovation and entrepreneurship as two potential mechanisms underlying the growth effect of broadband internet penetration. Such mechanisms are consistent with the evidence that firms located in cities that participated in this program and especially at an earlier stage exhibit a notably higher level of digital adoption.

Abstract: In this paper, we propose a new approach to estimate short panel regression models. The model considered is general enough: the model can be either static or dynamic, the time-varying regressor can be either endogenous, predetermined or strictly exogenous variable, and time invariant regressors can also be included. The errors can contain flexible form of observable fixed effects and/or unknown interactive fixed effects. We propose the minimum distance(MD) estimator to estimate these models in a unified way by utilizing the covariance structure analysis. The distinctive feature of our approach is that we do not need to use instrumental variables even in the presence of endogenous regressors. Theoretical investigation demonstrates that an identification problem arises when there are endogenous regressors, and a simple solution to address that problem is proposed. Test for endogeneity is also discussed. Monte Carlo simulation results reveals that the proposed MD estimator outperforms most of the existing estimators such as GMM estimators. 

Abstract: We study differences across demographic groups in their exposure to high-income individuals in shared spaces using smartphone movement data. Black, Hispanic, and lower-income individuals are less exposed to high-income people. To distinguish preferences over the demographics of co-patrons from preferences for venue attributes and differences in physical proximity to venues, we study choices of establishments within business chains. Social preferences are similar across demographic groups:

people prefer high-income and own-race co-patrons. Black and Hispanic individuals, however, face a trade-off between income and racial exposure. More broadly, individuals' preferences over co-patron demographics align with the demographics of their neighborhood of residence and can predict the neighborhood choices of movers.

Abstract: Utilizing a novel database including nearly 2.2 billion booking records in China, we examine whether people escape from pollution by traveling to “cleaner” places. Combining an instrumental variable approach with high-dimensional fixed effects, we find a 50-unit increase in the AQI gap between a city pair leads to a 1.30% (1.33%) increase in train and airline ticket bookings from the origin to the destination city departing within one day (2-7 days). In addition, the destination of such pollution-induced trips is more likely to be an intra-province city with more tourist attractions. We also measure willingness to pay for clean air.

Abstract: This study explores the effectiveness of Payments for Environmental Services (PES) in addressing both climate change and biodiversity loss within the framework of bioeconomic supply chain management. Leveraging variational inequality methods within a multi-criteria decision-making framework, we substantiate our theoretical claims through numerical simulations executed via an optimized machine learning algorithm. Our findings indicate that reductions exceeding 50% in both greenhouse gas emissions and biodiversity loss are achievable. While PES are integral to this success, they are insufficient on their own. A synergistic approach that combines a moderate decrease in production levels due to the economic decoupling effect, heightened environmental awareness among stakeholders, and targeted monetary incentives is essential to realize such substantial reductions. Thus, PES are necessary but not sufficient to achieve a meaningful reduction in ecological footprint. The adoption of sustainable practices and the improvement of resource efficiency are equally crucial.

Abstract: While recommendation algorithms have been increasingly used in daily life, little has been done to investigate their effect on decision making in terms of decision quality and preferences. Here we examine this question in an experimental setting whereby subjects from a representative US sample are randomly assigned to five conditions and make sets of binary choices between two lotteries. The two control conditions provide either no recommendation or recommendations based on a randomization device. The three treatment conditions provide recommendations developed by algorithms: one is based on the choice of the majority, and the other two use AI-based recommenders including content-based filtering and user-based collaborative filtering. We find that subjects tend to follow recommended choices and are willing to pay a small fee to receive recommendations for their subsequent decisions. Compared to control conditions, recommendations help subjects make better and faster decisions and behave in accordance with the independence axiom and the betweenness axiom. These results can be explained by some classes of stochastic choice models. Our work adds to the growing literature on the behavioral underpinnings of algorithms including AI and sheds light on the design of choice architecture for decision making under risk.

Abstract: We examine the impact of asymmetric data access within a vertically integrated e-commerce platform. Using a unique daily panel, we find that both the platform owner and third-party sellers base pricing on past sales, but only the owner exploits competitors' sales data. We estimate a price competition model with heterogeneous seller learning. We find that (1) eliminating the owner's access to competitors' data increases social welfare by 0.33%, negatively impacting the owner through reduced first-party sales and (2) providing third-party sellers with equal access to competitors' data as the owner increases social welfare by 1.65%, benefiting the owner through increased referral fees and yielding a Pareto improvement.

Abstract: This paper investigates how carbon tax aversion impacts the macroeconomy. Through an analysis of carbon tax announcements in Alberta, we first provide empirical evidence that carbon taxes result in the formation of pessimistic macroeconomic beliefs, characterised by higher inflation expectations and lower expected real income.  We term these beliefs as carbon tax aversion in the macroeconomy and incorporate them into a two-sector New Keynesian model that departs from the standard assumption of rational expectations.  From our benchmark model, we find that carbon tax aversion exacerbates output losses but stabilise aggregate prices after carbon tax announcements.  With carbon tax aversion, the announcement of a 1% increase in carbon taxes results in larger output losses by 0.3% but smaller decrease in aggregate prices by 0.02pp, relative to the rational expectations model. Furthermore,  earlier  announcement results in lower output losses.   Our findings highlight the importance of managing pessimistic beliefs against carbon taxes and the role of central banks in supporting the green transition.

Abstract: We propose an adjusted 2SLS estimator for social network models when existing network links are missing from the sample at random (due, e.g., to recall errors by survey respondents, or lapses in data input). In the feasible structural form, missing links make all covariates endogenous and add a new source of correlation between the errors and endogenous peer outcomes (in addition to simultaneity), thus invalidating conventional estimators used in the literature. We resolve these issues by rescaling peer outcomes with estimates of missing rates and constructing instruments that exploit properties of the noisy network measures. We apply our method to study peer effects in household decisions to participate in a microfinance program in Indian villages. We find that ignoring missing links and applying conventional instruments would result in a sizeable upward bias in peer effect estimates.

Abstract: We study the interaction between horizontal (peer) and vertical (manager) social factors in workers' motivation. In our setting, individuals work using open-plan desks. Using a natural experiment, we identify a sharp increase in workers' productivity following the occupation of adjacent desks. We link this peer pressure effect to two key aspects of the worker-manager relation. First, we find stronger peer pressure when managers monitor workers less. Second, we find stronger peer pressure among workers performance-evaluated by the same manager. In a set of counterfactual exercises, we illustrate how organisations could take advantage of these interdependencies to increase worker productivity.

Abstract: Forests accompany the cities we build. Globally, about 25 percent of urban land is covered by tree canopy. This study examines urban forests as a policy tool for air pollution mitigation. We study an afforestation program in the city of Beijing, which planted a total of 2 million mu of greenery – roughly the size of Los Angeles – across the city over a decade. We conduct a remote-sensing audit of the program, finding that it contributes to a substantial greening up of the city. This causes significant downwind air quality improvement, reducing average PM2.5 concentration at city population hubs by 4.2 percent. Rapid vegetation growth, however, led to a 7.4 percent increase in pollen exposure. Analysis of medical claims data shows aeroallergens triggered emergency room visits, mirroring well-documented industrial pollution effects though less severe. We offer insight on managing urban forests’ health risks, identifying harmful pollen species and susceptible population subgroups.

Abstract: This  paper  investigates  how  carbon  tax  aversion  impacts  the  macroeconomy. Through  an  analysis  of  carbon  tax  announcements  in  Alberta,  we  first  provide empirical evidence that carbon taxes result in the formation of pessimistic macroeconomic beliefs, characterised by higher inflation expectations and lower expected real income.  We term these beliefs as carbon tax aversion in the macroeconomy and incorporate them into a two-sector New Keynesian model that departs from the standard assumption of rational expectations.  From our benchmark model, we find that carbon  tax aversion exacerbates  output losses  but stabilise  aggregate  prices after carbon tax announcements.  With carbon tax aversion, the announcement of a 1% increase in carbon taxes results in larger output losses by 0.3% but smaller decrease in aggregate prices by 0.02pp, relative to the rational expectations model. Furthermore,  earlier  announcement  results  in  lower  output  losses.   Our  findings highlight the importance of managing pessimistic beliefs against carbon taxes and the role of central banks in supporting the green transition.

Abstract: In this paper, we study how dormant historical legacies can become active and resurface to influence individual behaviors. The historical  backdrop is the establishment of the Chinese Soviet Republic during the 1930s, which yielded two historical legacies: an underdeveloped economic environment and a diminished level of social capital. Using a novel dataset covering 10,000 couriers and 4 million daily delivery records spanning the periods before, during, and after the 2020 COVID-19 outbreak, we show that in response to the sudden escalation of risks, migrant couriers originating from former Soviet regions curtailed their delivery volumes to a greater extent compared to their peers.  Combining a large-scale survey of 2,800 couriers within the same company and other social value surveys, we show that the heightened panic-like behavioral response of Soviet base couriers can be partially attributed to a lack of trust in local government, rather than lower levels of altruism or economic circumstances of their upbringing. Furthermore, lower levels of political trust correlate with an increased perception of infection risk, lowering labor supply.

Abstract: In recent years, there has been an interest in improving people’s happiness and life satisfaction. Despite this interest, the benefits of an improved state of well-being on economic behavior have yet to be fully explored. In this paper, I use subjective well-being measures and examine their influence on economic decision-making, particularly on savings. I use data from all 7 waves of the World Values Survey, from 1981 to 2022. To tackle the potential endogeneity in the model, I use an instrumental variable identification strategy and employ a 2SLS estimation model, using the average annual temperature of each country as an instrument. I find that both happiness and life satisfaction have a positive and significant impact on savings behavior. This suggests that a person’s well-being plays an important role in savings behavior. Through heterogeneity analysis, I find that this impact varies depending on a person’s age, gender, social status, and employment.

Abstract: Social norms that are costly for individuals can remain in place for a long time because of social pressure to conform. Examples include duelling, footbinding, female genital cutting, and norms of conspicuous consumption. These and other costly norms are rarely “all-or-nothing”; instead, they take on many alternative forms. We develop a theory of norm dynamics that allows for such intermediate forms. The theory predicts rich dynamics that are consistent with cases where norm shifts have been documented.

Abstract: Hydropower dams along international rivers often spark disputes due to cross-border spillover effects. We examine the impact of upstream dams which do not internalize downstream costs, by looking at Chinese dams on the Mekong River located near the Chinese border with downstream countries. Using machine learning and difference-in-difference methods, we find that, on average, dams smooth downstream water levels across seasons and improve agricultural productivity. However, when facing extreme weather events, dams’ operational priorities can conflict with downstream interests, exacerbating adverse weather shock damages. Our findings shed light on the conditions under which international spillover effects of dams turn negative.

Abstract: We provide new evidence on the inflationary effects of Quantitative Easing (QE) using Swedish administrative data at the bank, firm, and product level. For identification, we rely on bank-firm lending relationships and the heterogeneous participation rates of banks in the government bond purchase program by the Swedish central bank. Our results show that the bond purchase program led to a significant and persistent increase in producer prices. Importantly, we find that the degree of financial frictions considerably influences firms' price response: low leverage firms do not change their prices, whereas high leverage firms raise their prices significantly. This divergent pricing behaviour can be rationalized by a significant increase in long-term borrowing and interest rate expenses among high leverage firms. The difference in price responses across high and low leverage firms is less pronounced for exogenous changes in the repo rate implying that the transmission mechanism of QE differs from the one of conventional interest rate policy.

Abstract: The nonlinear and endogenous effect is common in empirical studies with observational data. This paper proposes new estimation and inference procedures for nonparametric treatment effect functions with endogeneity and potentially high- dimensional covariates. One innovation of this paper is the uniform confidence band of the marginal effect function, defined as the derivative of the nonparametric treatment function, which is essential in policy-making. The asymptotic honesty of the confidence band is verified in theory. Simulation studies and an empirical study of air pollution and migration show the validity of our procedures.

Abstract: Do safety interventions create a false sense of security? We investigated whether people adjust their behavior optimally when there are exogenous changes to risk. In a laboratory experiment, subjects played an insurance-buying game under various risks of losing endowment. We found strong evidence of overcompensation - subjects purchased more (less) insurance than optimum in response to an increase (decrease) in risk. The degree of overcompensation was larger when risk increased than when it decreased. We exposed the same subjects to changes in safety conditions in real life and found little evidence that lab behaviors predict behaviors in the field.

Abstract: In many workplace settings, collaborative work is prevalent. While members of both genders arguably contribute equally to outcomes, previous research has found that women tend to receive less credit for their contributions in group settings compared to men. In this study, we aim to examine the extent to which specialisation and gender stereotypes contributes to such phenomena. We investigate this by designing a laboratory experiment where men and women work together in separate tasks to produce a joint output, subsequently bargaining over its distribution. Importantly, the tasks which they work could either be i) their specialisation or not, and/or ii) gender stereotypical or not. We discuss potential hypotheses about the treatment effects and mechanisms which could contribute to such patterns. 

Abstract: Lying and deception are common in economic interactions and have important strategic implications. While related, they are distinct phenomena that may have different effects on communication outcomes. In this paper, we study repeated communication with a reputation concern in a two-dimensional belief domain, and identify two environments where lying and deception are completely separated. In one environment, the sender must tell the truth to conceal a bad intention, while in the other, the sender must lie to reveal a good intention. Our experimental data show that the proportion of senders who successfully build reputations is lower than predicted in both environments. Furthermore, the deviation from theory is greater when reputation-building requires lying rather than deception. Finally, we observe that receivers punish senders for lying, even when the intention behind it is good. Our findings suggest that different communication mechanisms may perform differently depending on their reliance on lying or deception, highlighting the need to distinguish between these two concepts investigating organizational and political phenomena.

Abstract: This paper studies when and how intertemporal bundling arises in optimal dynamic selling mechanisms in a two-period setting, where two objects are sequentially available for sale. When the buyer's values of the two objects are independent, in contrast to conventional insight from simultaneous sales, we find in our setting that there is no loss in relying on separate sales for optimal design. This result extends under widely adopted regularity conditions when the values are positively correlated. However, with a negative correlation, the optimal integrated design features intertemporal bundling. The optimal bundle in general is contingent on the first-stage type; it consists of the whole second object and a probabilistic proportion of the first object, which makes the interim expected value of the bundle independent of the first-stage type. At optimum, the allocation of the first object is necessarily stochastic, and the second object is always sold.

Abstract: The central government in China is keen on catching up with the World Bank's doing business survey (renamed as the business readiness survey since 2023), and business environment assessment and improvement in China have been developing rapidly in the past decade. Close and clean state-business relations are crucial to doing business in transition economies like China, and it is meaningful to measure and examine state-business relations. In this talk I will share our research about the ratings and rankings of state-business relations of Chinese cities, and explore how they can be used in empirical research. I will also talk about the game of business environment rankings and how the theory of performance feedback helps to understand government behaviors and business responses.

Abstract: This  paper  formalizes  the  principle  that  persecution  power  of  government  may  generate  violent contests over it. We show that this principle yields a large set of theoretical insights on different separation-of-powers institutions  that  can  help  to  preempt  such  contests  under  different  socio-economic  conditions.  When socio-economic  cohesion  is  low,  the  risk  of  contests  can  be eliminated only by individual veto against persecution. Moreover, such unanimity rule is resilient to autocratic shocks only when the chief executive does not control the legislative agenda, i.e., the executive and legislative branches are separate. When socio-economic cohesion is high, the risk of violent contests can be eliminated without individual veto, but only by a persecution-reviewing judiciary whose members cannot join the executive branch in the future, i.e., when the executive and judicial branches are separate. Our results shed light on the evolution of separation of powers in European history.

Abstract: We use a threshold spatial dynamic panel data model to study the selection of target locations for large-scale interventions in the presence of multilayer networks, regional differences, and feedback effects. Using the US COVID-19 vaccination rollout as a case study, we find peer effects of vaccination through a friendship network, COVID-19 transmission through a travel network, and various vaccination effect channels. Simulations show that targeting the most populated or spatially connected states is the most clinically effective whereas targeting socially influential states is the most cost-effectively. The framework can address societal challenges where social and spatial spillover effects are critical.

Abstract: This paper quantifies the impacts of transportation infrastructure on welfare gains and structural change across space. By modeling transportation service as a perfect complementary input in production, improvement in transportation technology reduces within-region trade costs, thus acts as sector productivity progress and reduces transportation service share in production. Our quantitative application to China shows that increases in truck load capacity reallocates economic activity from agriculture and manufacturing to services by 0.84 percentage points in terms of value added, and generates an average welfare gain of 5.6% across Chinese provinces between 2000 and 2013.