Peer-reviewed research

• • Brewer, Dylan. (2020). “Do preferred thermostat settings differ by sex?” Energy and Buildings,  vol. 224. Abstract.
    This paper examines the extent to which males and females report different thermostat settings using data from single-occupant homes in the United States. Reported thermostat settings for heating and cooling at night, at home, and when away from home are rigorously tested for behavioral differences between the sexes. In general, single-occupant males and females report the same thermostat settings on average. The lack of differences between sexes is robust to controlling for demographic characteristics and environmental variables that differ systematically between the samples using both regression and decomposition approaches. These findings have important implications for thermostat settings and thermal comfort in shared spaces.
  • Brewer, Dylan. (2022). “Equilibrium Sorting and Moral Hazard in Residential Energy Contracts.” Journal of Urban Economics, vol. 129. Abstract.
    This paper explores tenant behavior in rental housing where the landlord pays for heating. I develop a model in which renters have heterogeneous preferences for home size and indoor temperature. When energy is costly, renters choose smaller apartments and turn down the heat—or sort into apartments with landlord-pay energy bills. I estimate the model using exogenous variation in energy prices and use a machine-learning algorithm to explore preference heterogeneity. I find no evidence to support sorting into landlord-pay apartments based on temperature preferences. Eliminating moral hazard by forcing all renters to pay their own bill reduces energy consumption by 25% due to renters turning down the heat (22%) and choosing smaller units (3%). Incentive problems in residential energy contracts cost the United States $839 million per year in welfare losses including $238 million from damages due to increased carbon emissions.
  • Brewer, Dylan. (2023). “Changes in electricity use following COVID-19 stay-at-home behavior.” Economics of Energy & Environmental Policy, vol. 12, no. 1. Abstract.
    Using hourly electricity consumption data from the PJM Interconnection in the United States in a difference-in-predicted-differences strategy, this article shows that while in the first months of the COVID-19 pandemic total electricity consumption declined by 3.7-5.4% relative to a predicted counterfactual, in July and August 2020 electricity consumption was 2.9-4.6% higher than the predicted counterfactual. In addition, higher temperatures had an increased effect on electricity consumption in 2020 relative to previous years. Nationwide monthly data on electricity consumption by load class reveals that commercial and industrial consumption was below its expected baseline from March-November 2020, while residential consumption was above its expected baseline, peaking in July. This suggests that increased demand for residential cooling offset declines in commercial and industrial demand for electricity. Estimates of the total effect of the pandemic on electricity consumption from March through December 2020 suggest that early reductions in electricity use were almost perfectly offset by later increases, implying that any expected “silver lining” of decreased emissions from electricity production may be smaller than previously thought.
  • Brewer, Dylan, Dench, Daniel, and Taylor, Laura.  (2023).  “Advances in Causal Inference at the Intersection of Air Pollution and Health Outcomes Data.” Annual Reviews of Resource Economics, vol. 15. Abstract.
    This article provides an overview of the recent economics literature analyzing the effect of air pollution on health outcomes.   We review the common approaches to measuring and modeling air pollution exposures and the scientific literature on health outcomes that undergirds federal air regulations in the United States.  We contrast this scientific evidence to the evidence provided by economists that employ a causal inference framework.  In particular, we review the common sources of estimation bias in epidemiological approaches that the economics literature has sought to overcome with research designs that take advantage of natural experiments.  We review new promising research designs for estimating concentration-response functions and identify areas for further research.
  • Brewer, Dylan. (2023). “Household Responses to Winter Heating Costs: Implications for Energy Pricing Policies and Demand-Side Alternatives.”  Energy Policy, vol. 177, 113550. Abstract.
    I conduct a survey that presents research subjects with hypothetical costs to adjust their thermostats. I estimate responses to the cost of heating and analyze the causes for heterogeneity in household demand for energy services using the survey results as a complete-information baseline. I find that even at the highest price level, half of the participants exhibit zero response to price. On average, a 100 percent increase in the cost of heating the home induces a 0.31 to 0.97 degree Fahrenheit (0.17 to 0.51 degrees Celsius) reduction in the winter heating level, corresponding to a -0.005 to -0.014 elasticity. Further, I find that participants’ experimental behavior with complete information can explain observed real-world temperature settings, suggesting a limited role for informational barriers or salience issues in energy-service demand heterogeneity. Inelastic demand suggests that energy efficiency policies may have high returns and that centralized demand-response policies may be required to address winter energy emergencies. Further, individuals with higher temperature preferences are more price responsive, suggesting that increasing block pricing policies for energy may reduce energy consumption while minimizing the regressivity of energy pricing.
  • “Addressing Sample Selection Bias for Machine Learning Methods.” Joint with Alyssa Carlson.  Supplemental appendix. Accepted at Journal of Applied Econometrics. Abstract.
    We study approaches for adjusting machine learning methods when the training sample differs from the prediction sample on unobserved dimensions. The machine learning literature predominately assumes selection only on observed dimensions. Common suggestions are to re-weight or control for variables that influence selection as solutions to selection on observables. Simulation results indicate that common machine learning practices such as re-weighting or controlling for variables that influence selection into the training or testing sample often worsens sample selection bias. We suggest two control-function approaches that remove the effects of selection bias before training and find that they reduce mean-squared prediction error in simulations with a high degree of selection. We apply these approaches to predicting the vote share of the incumbent in gubernatorial elections using previously observed re-election bids. We find that ignoring selection on unobservables leads to substantially higher predicted vote shares for the incumbent than when the control function approach is used.
  • “Habit and skill retention in recycling.”  Joint with Samantha Cameron. Accepted at Journal of Policy Analysis and Management. Abstract.
    From 2002-2004, New York City ceased collecting residential glass and plastic recycling due to city budgetary pressure. We use data on recycling rates in New York City, New Jersey, and Massachusetts in a synthetic difference-in-differences (DID) approach to determine whether this exogenous pause led to a depreciation of habitual capital in recycling. Despite a 50% decline in the overall recycling rate in 2003, by 2005 the overall recycling rate had fully recovered. Our results suggest that recycling habits are persistent in the short term and that habitual capital depreciation may not be a large cost when pausing unprofitable recycling programs. We show that re-estimating synthetic DID time weights for each post-treatment period in an event-study context increases the precision of the treatment effect estimates.

Grant-sponsored research

• • “Monitoring Air Pollution in Underserved South Atlanta (MAP-USA).” United States Environmental Protection Agency, 2022. $498,401. PIs: Garry Harris (Center for Sustainable Communities) and Dylan Brewer.
  • “The Effects of Internal Combustion Engine Transit Systems on Health: An Interdisciplinary Research Program Linking Transit-Related Pollution to Birth Outcomes.” The Strategic Energy Institute, 2022. $119,855. PIs: Dylan Brewer and Randall Guensler.
  • “The Health Effects of Air Pollution: An Interdisciplinary Research Program.” Georgia Tech Executive Vice Provost for Research, 2022. PIs: Laura Taylor and Randall Guensler.

Working papers

• • “Who heeds the call to conserve in an energy emergency?  Evidence from smart thermostat data.” Joint with Jim Crozier. Revisions requested at the Journal of the Association of Environmental and Resource Economists. Abstract.
    In 2019, a fire at a natural gas plant and historically low temperatures caused an emergency shortage of natural gas in Michigan. To avoid an outage, the Governor issued a request via statewide text alert to turn thermostats down to 65 degrees F. We analyze the effectiveness of this request using high-frequency smart-thermostat data from Michigan and four neighboring states. Using a difference-in-differences research design, we find that Michigan households reduced thermostat settings by 1.1 degrees on average following the Governor’s request. Households that were previously above 65 degrees F responded strongly, while households that were below did not respond at all or increased their thermostat settings. Meanwhile, households in districts that voted for the Governor in 2018 were more likely to comply. Our results suggest that unrealistic compliance goals and political polarization reduce the effectiveness of emergency calls to conserve energy.
  • “The effect of extreme temperatures on evictions.” Joint with Sarah Goldgar. Revisions requested at the Journal of Environmental Economics and Management. Abstract.
    Using data on evictions in the United States, we estimate the relationship between temperature and evictions. We find that extreme heat days result in a statistically significant increase in evictions while extreme cold days do not have the same relationship. To explain these findings, we show that energy expenditures are more sensitive to extreme heat than extreme cold and that energy assistance programs in the United States prioritize funding for heating rather than cooling. These findings suggest that future climate change scenarios with more hot days and fewer cold days will result in an increase in evictions without other policy or private adaptation.
  • “The pandemic, work from home, and new residential construction.” Joint with Graham Lewis. Under review. Abstract.
    This article examines the effect of the COVID-19 pandemic and work-from-home trends on residential construction. We document a decline in building permits filed early in the pandemic, followed by a substantial increase in permits filed during late 2020 and 2021. Pre-pandemic trends of building in below-median population density and metropolitan areas continued, while substantially more permits were filed in Republican-voting counties and counties in states with Republican governors. Our results suggest political re-sorting during the pandemic that may continue in a work-from-home environment.
  • “Benefits to Agriculture from an Afforestation Program: Evidence from India.” Joint with Vikrant Kamble and Matthew E. Oliver. Under review. Abstract.
    Afforestation is a popular strategy to mitigate climate change. When successful, afforestation programs can produce important co-benefits beyond carbon sequestration, which have significant implications for the net social benefit of carbon abatement through afforestation. In 2003, one of the largest afforestation programs in India was implemented in Rajasthan state. Using a yearly, district-level panel from 1997 to 2017, we estimate the effects of this program on the agricultural sector using two-way fixed effects and synthetic difference-in-differences approaches. Our findings suggest that the afforestation program led to robust, statistically significant increases in rainfall and agricultural production, area, and yield. We discuss the implications of our findings for afforestation as a climate mitigation strategy.
  • “The effect of mass layoffs on related industries: Evidence from a mining ban in India” Joint with Vikrant Kamble. Under review. Abstract.
    In 2011, the Supreme Court of India banned iron ore mining in three districts of Karnataka state to curb illegal mining. This decision had unintended consequences on the market for unskilled labor as displaced miners shifted to other fields such as agriculture. In districts where mining was banned, we find that male agricultural field labor wages declined by 24%, household consumption declined by about 20%, and demand for guaranteed government work programs increased by about five percentage points. These findings have important implications for sunset industries in transitioning economies that are experiencing structural transformation.
  • “Of presidents and prices at the pump: Do economists have a blind spot?” Joint with Matthew E. Oliver. Under review. Abstract.
    Many people believe political executives (e.g., the US President) can influence retail gasoline prices. We conduct a survey of US-based energy market experts and non-experts and find that non-experts are significantly more likely to believe the US President plays a major role in determining retail fuel prices. Next, we analyze weekly retail gasoline and diesel price data from the US, UK, France, and Canada under different heads of state. After controlling for fluctuations in the wholesale crude oil price, country-specific long-run averages in prices, and common time shocks, our results suggest that non-expert beliefs are more consistent with reality–fuel prices differ significantly across leaders. We argue that this miscalibration of expert perceptions with reality has led to a ‘blind spot’ in economic research on the influence of executive policy on retail fuel prices and make a call to economists to look for other potential blind spots by paying closer attention to similarly prominent economic narratives.

Works in progress

• • “Re-Lighting Detroit’s Streets: Impacts on Energy Use, Crime, and Property Values.” Joint with Soren Anderson, Prabhat Barnwal, Joe Herriges, Anna Terkelsen, Alex Tybl, and Asa Watten. Abstract.
    At the beginning of 2014 more than 40% of Detroit’s street lights were burned out or broken. During 2014-2016 the city completely overhauled its street light system, installing 65,000 new LED lights and removing 23,000 old lights to become the largest U.S. city lit entirely using LEDs. The program was phased in gradually across the city—with many broken lights replaced and some working lights removed—leading to large variation in both the quantity and quality of street light over time and space. Using administrative data from the Detroit Public Lighting Authority, we calculate the program’s effects on energy use. We then use the program’s phased implementation across the city to estimate its non-energy impacts on crime, safety, and property values.
  • “Estimating energy efficiency using smart thermostat data.” Joint with R. Jim Crozier.

Other publications

• • Brewer, Dylan. “The Economic Costs of Forecasting Errors in the PJM Interconnection Due to the COVID-19 Quarantine.” IAEE Energy Forum, Covid-19 Special Issue 2020: 116-119. Abstract.
    Using hourly metered load and weather data, I show that PJM electricity consumption during the COVID-19 period declined 10.6%, leading to poor-performance of load forecasts. The costs of over-purchasing day-ahead generation were likely low in March 2020 due to mild temperatures; however, the costs may increase as summer approaches.