The Impact of Climate Change: An Empirical Analysis of Smart Thermostat Data

Extreme weather from climate change creates unprecedented fluctuations in residential heating and cooling demand. Understanding how households use thermostats and react to ambient weather is key to achieving demand reductions and avoiding power crises. To this end, we analyze high-frequency microlevel time-series data from smart thermostat users to examine how households adjust their thermostat operations in response to outdoor temperature. Comparing households within a city, we find that households that automate their HVAC mode decisions exhibit reduced responsiveness to temperature shocks. In particular, their heating energy consumption remains elevated, even in the face of a changing climate, compared with households that rely on less automation. With respect to temperature setpoints, households that frequently override preprogrammed schedules show inertia, remaining at the overridden setpoints for long periods, and only partially revert to their preprogrammed setpoints. This can lead to higher energy use compared with those who stay in their preprogrammed schedules. To quantify the impact of these results, we conduct a set of simulations under hypothetical climate scenarios. These simulations reveal that by 2050, total cooling energy usage could rise by at least 67% and as high as 300%, depending on the climate scenario and geographical location, with peak usage and demand variability increasing by over 50%. Whereas heating consumption is reduced, this decrease is dominated by the increase in cooling energy usage. As a case study of the operational implications, we project total hourly loads and generation costs for Texas’ electricity grid in 2050. The increased heating and cooling demand alone could raise annual grid operating costs by 20%–25% and require 21,700 megawatts of additional capacity to maintain grid reliability.

This paper was accepted by Jayashankar Swaminathan, operations management.

Supplemental Material: The e-companion and data files are available at https://doi.org/10.1287/mnsc.2024.06952.