This paper uses a difference-in-differences model with staggered treatment timing to estimate the causal effect of utility-scale renewable energy projects (wind turbines and solar facilities) on out-migration at the county level for the United States. We find that counties that receive wind turbines experience decrease in population of 2%–3% while solar facilities do not cause out-migration. We also find evidence of similarly-sized spillover effects in adjacent counties. Given the literature on the rural resentment towards renewable energy project in their communities, we split our sample into rural and urban counties using Rural–Urban Continuum Codes and re-run our analysis. We find that rural and urban counties exhibit nearly identical out-migration responses to wind projects, indicating that, in practice, their attitudes towards renewable energy projects are more similar. Combined with the previous literature on the externalities of utility-scale renewable energy projects, our results suggest that solar energy is the better option for generating renewable energy when the concerns of — and impacts on — communities are considered by policy makers.
The previous literature on site selection for utility-scale photovoltaic (PV) and agrivoltaic (AgriPV) development has considered broad geographies with the intention to identify the best regions of countries for PV development. This paper considers Central Virginia, a region with a rich agricultural history that is experiencing increased PV development, and uses parcel-level data from local governments and a Weighted Product Model multi-criteria decision making process to identify suitable parcels of land for utility-scale PV and AgriPV development. We consider energy infrastructure proximity, land suitability, road access, and local zoning policies at the parcel level as necessary criteria. We identify parcels with optimal characteristics for both PV development and AgriPV development, as well as those parcels that are suitable for PV development yet not situated on prime agricultural land, a land use that these communities want to preserve. We also perform a sensitivity analysis that shows that zoning is the most significant bottleneck for one of the counties in our study.Cary, M. (2025). Post-Kyoto emissions in the United States. Renewable and Sustainable Energy Reviews, 209, 115068.
In this study I show that even though the United States (US) did not ratify the Kyoto Protocol, it still largely behaved like a nation with a binding emissions target under the Kyoto Protocol. This is determined by running two sets of synthetic controls models — one using a sample comprised of nations that set a binding emissions target under the Kyoto Protocol, and one using a sample of nations that did not set a binding emissions target. With the exception of methane emissions, the emissions profile of the US resembles its counterparts who did set a binding emissions target. Thus, the US effectively reduced greenhouse emissions similarly to nations that set binding emissions targets under the Kyoto Protocol, but, by opting for natural gas, the US would not experience the same level of public health gains associated with reducing emissions. Given this, the primary implication for future climate policy is that, on the margin, ratifying a binding emissions target is the better choice for the potential signatory as it leads to a more fully internalized externality.
The Kyoto Protocol is an international agreement signed in 1997 to reduce emissions of carbon dioxide and mitigate the consequences of global climate change. Early studies of the effectiveness of the Kyoto Protocol have attributed it to approximately a 7%–10% reduction in global carbon dioxide emissions. Given the carbon intensity of cement manufacturing, we examine the impact of the Kyoto Protocol on cement manufacturing output and carbon dioxide emissions. We are especially concerned about whether the Kyoto Protocol may have led cement manufacturing to move to places without binding emissions targets, thus resulting in carbon leakage. We find that nations with binding emissions targets under the Kyoto Protocol saw a 5% reduction in both cement manufacturing and carbon dioxide emissions from cement manufacturing compared to other nations, suggesting that the Kyoto Protocol may have led to technological innovation in cement manufacturing. Using data on carbon intensities for cement manufacturing, we support this notion and find further evidence that it fostered diffusion of existing, cleaner technologies from relatively more developed nations during the first phase, and technical innovation by these nations during the second phase, resulting in significant reductions in carbon dioxide emissions.
Climate policy affects the economy. Economists typically consider the economic costs associated with climate policy, but not all economic consequences of climate policy are necessarily negative. This paper considers the impact of the Kyoto Protocol on the international trade network for timber products and the competition graph derived therefrom. Implementing an instrumental variables difference-in-differences approach which uses a panel of centrality and clustering measures describing the relative importance of individual nations in the international trade network for timber products, this paper finds that nations that set a binding emissions target under the Kyoto Protocol saw the centrality of their position within this network decrease. Since the nations which set binding emissions targets were also the wealthiest nations, an additional consequence is that less wealthy economies became relative more important members of this international trade network. This is likely due to nations with a binding emissions target placing an increased importance on forested land for carbon sequestration. These results are extended to the competition graph of the international trade network for timber products where it is found that nations with binding emissions targets became marginally less important. However, increased clustering among nations with binding emissions targets in the competition graph occurred as a consequence of the Kyoto Protocol. Collectively, these results indicate that the competitive playing field was leveled in the international timber trade network while helping make progress towards multiple economically and environmentally oriented United Nations Sustainable Development Goals.
Core-periphery models are a commonly used and powerful tool in network analysis which allow researchers to identify a set of highly influential members (the core) based on the structure of the network. However, there is no standard method for core identification. Worse yet, for researchers in fields such as economics, existing methods often do not allow for researchers to make meaningful inferences from core-periphery analyses. This paper introduces a new method using integer programming to identify the core of a weighted network with either symmetric or asymmetric relationships and without the aid of any a priori assumptions. The integer program maximizes the weighted influence of the members of the core while imposing a set of user-parameterized connectivity constraints, both within the core and between the core and the periphery. As a result, this model generalizes previous concepts of the core of a network, i.e., an idealized core, through the flexibility created by the two connectivity parameters. These parameters give researchers the freedom to apply discipline specific theory to core-periphery analyses. This approach also offers drastically improved handling of directed data compared to existing methods of core-periphery analysis. An algorithm which converts the initially non-linear integer program into a linear integer programming approach is presented, and several examples are studied using this method to demonstrate its efficacy including a data set of international collaborations in academic publications and a data set of US railway shipments.
Introduction: The use of telemedicine increased during the global Coronavirus disease 2019 (COVID-19) pandemic. Rural populations often struggle with adequate access to care while simultaneously experiencing multiple health disparities. Yet, telemedicine use during the COVID-19 pandemic has been understudied on its effect on visit completion in rural populations. The primary purpose of this study is to understand how telemedicine delivery of family medicine care affects patient access and visit completion rates in a rural primary care setting. Methods: We performed a retrospective cohort study on primary care patient visits at an academic family medicine clinic that serves a largely rural population. We gathered patient demographic and visit type and completion data on all patients seen in the West Virginia University Department of Family Medicine between January 2019 and November 2020. Results: The final sample included 110,999 patient visits, including 13,013 telemedicine visit types. Our results show that telemedicine can increase completion rates by about 20% among a sample of all ages and a sample of adults only. Working-aged persons are more likely to complete telemedicine visits. Older persons with higher risk scores are more likely to complete their visits if they use telemedicine. Conclusions: Telemedicine can be a tool to improve patient access to primary care in rural populations. Our findings suggest that telemedicine may facilitate access to care for difficult-to-reach patients, such as those in rural areas, as well as those who have rigid work schedules, live longer distances from the clinic, have complex health problems, and are from areas of higher poverty and/or lower education.
Working Papers
Local Labor Market and Land Use Effects from Deforestation Caused by an Invasive Species (*Under Review at the American Journal of Agricultural Economics*)
The spotted lanternfly (SLF) is an invasive species that first appeared in the United States in 2014 and is rapidly spreading across the country. The SLF feeds on the sap from many economically important plants including grapevine, apple, and maple. Using a staggered difference-in-differences model, this paper produces the first causal estimates of the labor market effects induced by the spread of the SLF and also uncovers the mechanisms driving this result. I find that average wages decrease, the unemployment rate decreases, and the number of jobs increases in counties with established colonies of SLF. These new jobs come from lower paying sectors, specifically from construction and from low-paying urban amenities jobs. The decrease in average wages is driven largely by an industry mix effect, although productivity effects likely also contribute to the observed decrease in wages. To explain these economic outcomes, I show that the SLF causes a reduction in forested land and an increase in developed land. These land use changes appear to drive the increase in urban amenities and thus the effects on local economies. Interestingly, these communities appear to view the land use changes and increase in urban amenities as a net amenity gain as I show that rents increase using a county-level house price index.
Job Growth and Environmental Degradation (*Revise & Resubmit at Ecological Economics*)
In this paper I characterize how job growth affects land use. Specifically, I characterize how county level job growth affects each of the categories of land cover included in the National Land Cover Database (NLCD) for the United States between 2011 and 2021. To obtain the causal effect of job growth on land cover, I instrument county level job growth with a Bartik shift-share instrument. I find that job growth decreases the proportion of wetlands, agricultural lands, and barren land, while job growth increases the proportion of land covered by water. However, these results vary across the urban-rural gradient, the wealth gradient, the growth-rate gradient, and the east-west geographic gradient. I find that job growth comes at the expense of agricultural land and the wetlands in both rural and slower growing counties, but not in urban or faster growing counties. I also find that job growth causes an increase in the proportion of surface area covered by water, with this result driven exclusively by counties in the more arid, western US, implying that job growth is not environmentally sustainable in its current form.
Data Centers Reduce Residential Property Values (with Le Wang) (*Suibmitted to the Journal of Urban Economics*)
Data centers have expanded rapidly with the growth of cloud computing and artificial intelligence, raising concerns about their local impacts. Using a repeat-sales spatial difference-in-differences model, we estimate how data center openings affect nearby residential property values. We find that homes within two kilometers decrease in value by about 6–7%, with losses exceeding 15% for homes within two-thirds of a kilometer. These results suggest that data centers generate substantial localized externalities that are capitalized into housing markets.
How and Where the Kyoto Protocol Affected Forest Cover: Evidence from Satellite Data (with Piper Zimmerman) (*Under Review at the Annals of the American Association of Geographers*)
The Clean Development Mechanisms of the Kyoto Protocol incentivize the (developed) nations that set binding emissions targets to invest in sustainable forest projects in developing nations. Since we know that the Kyoto Protocol was largely successful at reducing emissions, and also that several nations needed to exploit these Clean Development Mechanisms in order to meet their emissions targets, a reasonable question to ask is if the Kyoto Protocol had a measurable effect on forest cover, and if so, where? To answer this question, we use satellite forest cover data and an instrumental variables difference-in-differences approach to estimate the effect of setting a binding emissions target under the Kyoto Protocol on annual percentage changes in forest cover. Our combined results suggest that the Kyoto Protocol did have a positive effect on the annual percentage change in forest cover in the poorest of developing nations. Combined with previous literature, this suggests that the Kyoto Protocol promoted environmentally and economically sustainable growth in global forest cover.
Road Network Structure and Air Pollution: Moving Beyond the Fundamental Law of Road Congestion (with Heather Stephens)
Transportation is one of the primary contributors to local pollution stocks and flows. This paper considers how the structure of local road networks and the accompanying vehicular emissions might affect pollution stocks and flows. A pollution stock and flow model building on the Fundamental Law of Road Congestion that considers the impact of road network structure is presented and used to generate hypotheses for how the structure of road networks should affect pollution stocks and flows. The main avenues for these effects are via traffic congestion and the opportunity cost of driving. Using topological indices to describe the structure of road networks, these hypotheses are tested using a Hausman-Taylor approach using a measure of urban form as an instrument to address the endogeneity of the network structure. Evidence is found supporting the hypotheses that better connected road networks, i.e., those with fewer bottlenecks and which generally allow for more efficient traversal, lead to lower levels of pollution stocks and flows. Evidence is also found that drivers adapt to more circuitous road networks with lower levels of driving. These mechanisms are confirmed by regressing measures of congestion and the opportunity cost of driving against the topological indices.
Air Pollution and Physical Productivity: Evidence From Ultramarathons (with Heather Stephens)
Air pollution is known to lower worker productivity in myriad settings. However, causal inference studies with reliable air pollution data studying the effect of air pollution on physical productivity in outdoor settings over periods of time comparable to a workday are sparse. Furthermore, identification of the specific pollutant(s) responsible for this effect is even less common. By using ultramarathon race performances as a measure of physical productivity and spatiotemporal variation in ambient pollution levels, it is determined that particulate matter (PM10) is the key pollutant responsible for decreasing consistent, metered physical productivity in outdoor settings, with a 1% increase in ambient levels of particulate matter leading to a 0.071% decrease in physical productivity as measured by ultramarathon race performance. In addition to this, this paper also extends the existing literature by suggesting a distinction in which pollutants affect physical productivity along an aerobic/anaerobic divide, thereby explaining part of the existing ozone-particulate matter divide in the air pollution and physical productivity literature, as well as by providing the first evidence of gender based differences in the effects of air pollution on physical productivity.