Benefit transfer (BT) has evolved as the dominant valuation method for environmental benefit-cost analyses, including those required of US federal agencies. Yet even best-practice approaches for BT based on meta-regression models (MRMs) typically exhibit poor predictive fit and out-of-sample precision. This article introduces random forests (RFs) for nonparametric estimation of MRMs and construction of BT predictions. We compare the performance of different RF models to current best-practice approaches for BT. We find that forest-based models substantially improve the out-of-sample accuracy of welfare predictions and tighten confidence intervals of predicted benefits for stipulated policy scenarios. The best performers reside within the family of local linear forests (LLFs), a hybrid approach that combines elements of RFs and locally weighted regression. Results suggest that this new approach has the potential to substantially improve BT accuracy for environmental policymaking without sacrificing theoretical properties, while simultaneously reducing econometric and computational difficulties relative to leading alternatives.
Despite high noise pollution, empirical research on the impact of noise on learning in developing economies has been sparse. We investigate this question by analyzing spatial-and-temporal variation in noise pollution recorded by monitoring stations in India and academic performance in high-stakes Class 12 examinations across schools. We find that a ten-percentage-point increase in noisy days during the exam quarter is associated with a 2.3 percent increase in the failure rate of Class 12 boys. This pattern is not gender-neutral, as no association between noise pollution and failure rate is observed for girls. We do not find any effect of noise pollution recorded during nonexam months. By documenting how noise pollution hinders learning in India and shedding light on cognitive burden explanation, our study highlights an overlooked threat to human capital accumulation in developing economies and underscores the importance of integrating noise pollution mitigation strategies into education and urban policies.
College education, homeownership, and automobiles are all examples of purchases where consumers sink a large up-front investment expecting to enjoy a stream of future net benefits. Often those benefits are uncertain at the time of purchase. We consider how uncertainty in future operating costs impacts the choice of vehicle fuel economy. Using measures of future gasoline price uncertainty from real options theory and comprehensive data on sales of automobiles in the United States, we find that future gasoline price uncertainty has economically meaningful impacts on vehicle demand, with a one standard deviation increase in the variance of the future price distribution increasing willingness to pay for fuel economy by 6.7%, slightly improving mean fuel economy but substantially reducing short-run vehicle sales. This new finding has implications for transportation policies that may introduce new uncertainties in future vehicle operating costs.
Emerging datasets capture rich temporal variation in recreation behavior, but recreation demand analyses have traditionally used variation across sites, rather than across time, to value environmental amenities. We introduce a model and estimation procedure designed to exploit panel variation in recreation demand analyses by embedding panel data causal inference techniques within a travel cost random utility model. To demonstrate the method, we use "structural synthetic controls" to value the welfare losses caused by water-quality-induced beach closures in southeast Michigan. Losses tend to be larger on weekends and hotter days, and our results suggest that a stigma effect reduces visitation even after the beach reopens. Our method is particularly useful for valuing the recreational impacts of resource shocks, like harmful algal blooms or wildfires, and it can be applied broadly given the increasing availability of high-frequency recreation data.
This study extends the monocentric city model to incorporate endogenous household automobile fuel choice and dwelling energy consumption. Electric vehicle (EV) tax credits lower direct energy consumption and emissions via commuting but also cause important urban general equilibrium effects, including sprawl. This creates a substantial emissions rebound effect via larger homes, longer commutes, and greater consumption of the numeraire good. Nevertheless, EV tax credits are welfare enhancing unless electricity production is heavily tilted toward coal, with total household energy consumption and carbon emissions falling substantially.
This study examines a Pennsylvania policy reform that increased bonded liability for coal mines with reclamation obligations. A 2001 change for surface mines was largely priced into existing reclamation insurance mechanisms, which exist to prevent mines from shirking their liabilities. We find that the changes increased mine-level insured liability and idling (approximate to 15% for median mines), led to declines in production, and did not change environmental performance directly for treated mines. However, we do find that these effects were driven largely by higher polluting mines. Major results are robust to both comparing bituminous and anthracite mines within Pennsylvania as well as bituminous mines between Pennsylvania and Virginia.
Researchers deploying stated preference surveys to value public goods commonly use techniques designed to reduce bias in hypothetical choice settings. This practice is at odds with evidence that most survey respondents perceive that their decisions are not hypothetical but instead have economic consequences. We examine three bias reduction procedures in both hypothetical and incentive-compatible, real payment settings: cheap talk, solemn oath, and certainty adjustment. We confirm that adjusting hypothetical choices based on response certainty or using a solemn oath can reduce hypothetical bias. In the incentive-compatible decision setting, the oath increases willingness to pay (WTP), and certainty adjustment can lead to serious distortions in demand estimates. Cheap talk does not alter mean WTP but leads to a stark difference in WTP across sexes. To minimize unintended consequences, our results suggest that survey researchers should deploy screening questions to better target these hypothetical bias reduction techniques.
Utilities have invested billions of dollars in advanced metering infrastructure (AMI) but proceeded slowly to deploy AMI-enabled programs that benefit consumers. High bill alert (HBA) programs, which inform consumers of unusually high usage patterns, offer an avenue for tapping AMI-enabled benefits. We evaluate an HBA program and find that the program reduced mean electricity and natural gas consumption by about 0.5%. The effects were largest at the top of the usage distribution, especially when normalized by preprogram usage, indicating that households experienced fewer expenditure shocks. Estimates of the program's benefits and costs illustrate the importance of considering the forgone value of conserved energy when evaluating demand-side management programs in the energy sector.
This study examines the effects of droughts on economic activity, proxied by nighttime lights. Using two different indices of drought severity, one remotely sensed and one from ground-sensed meteorological data, we provide some of the first global estimates of the economic effects of drought, as opposed to temperature and precipitation. Economic impacts depend on drought severity, with moderate-or-worse droughts reducing luminosity by about 2%. Water storage in the form of accessible groundwater mitigates drought impacts. Dams moderate the effects of all but the most extreme droughts. For moderate or worse droughts, negative effects are fully offset by the presence of dams, unless some dams are hydroelectric. Upstream dams do not appear to create harmful spillovers for drought resilience downstream. Results are robust to allowing separate mediating impacts from irrigation and suggest that the resilience benefits of dams and groundwater extend beyond agriculture.
This study investigates the impact of glyphosate, the most widely used herbicide, on birth outcomes in the US Corn Belt. Using water flow mechanisms to identify causal effects, the study shows that glyphosate affects populations far from application sites through waterborne transmission. The results suggest that a 10 kg/km2 increase in upstream glyphosate use led to a 4.6% rise in neonatal deaths in lower-income areas, with no observed effects in higher-income regions. The research design also incorporates variations in spatial distances, seasonal exposure patterns, and rainfall data to ensure that the observed health impacts are attributable to glyphosate. Evidence suggests that avoidance behaviors and water treatment are potential mechanisms of the heterogeneous effects.
Policy makers frequently champion information provision about carbon impact on the premise that consumers are willing to mitigate their emissions but are poorly informed about how to do so. We empirically test this argument and reject it. We collect an extensive new dataset and find both large misperceptions of the carbon impact of different consumption behaviors and clear preferences for mitigation. Yet, in two separate experiments, we show that correcting beliefs has no effect on consumption in large representative samples. Our null results are well-powered and informative, as we target information for maximal impact. These results call into question the potential of correcting carbon footprint misperceptions as a tool to fight climate change.
Scheduling electric vehicle (EV) charging behavior in the context of large-scale EV development is vital for building low-carbon societies. Using 2020 public charging data from Beijing, China, this study applies a difference-in-differences (DID) model to examine the impact of flexible charging price subsidies on drivers' behavior, distinguishing between business and private drivers to capture heterogeneity. The results show that a subsidy of CNY 0.4/kWh significantly increased the average daily charging volume by 25.29%. Specifically, the daily charging volume increased by 59.68% for business drivers and 12.22% for private drivers. The price elasticity of charging demand for all EV drivers was estimated at -0.95, with business and private drivers showing elasticities of -2.24 and -0.46, respectively. The economic and environmental benefits resulting from these behavioral changes after policy optimization are further discussed. These findings highlight the importance of tailoring price incentive policies to the varying sensitivities of different driver types.
This study explores the path-dependent nature of cheating and investment in innovation as strategies for automobile manufacturers subject to environmental regulations. Firms first decide on their investments in innovation to develop compliant technologies. Then, knowing the outcome of innovation, they may opt to activate a cheating device. Successful innovation achieves compliance at reduced costs, while undetected cheating creates the appearance of compliance while eliminating all compliance costs. We show that higher investment in innovation discourages cheating. In contrast, the availability of cheating devices, used either systematically or as a fallback when innovation fails, leads to lower investment levels. We derive policy recommendations relying on comparative statics with respect to compliance costs, the likelihood of detecting cheating, the penalty imposed on cheating firms, and market competition.
