Building vintage and electricity use: Old homes use less electricity in hot weather

This paper studies whether electricity use in newer or older residential buildings rises more in response to high temperature in a region of Southern California. Peak electricity demand occurs at the highest temperatures which are predicted to increase due to climate change. Understanding how newer buildings differ from older buildings improves forecasts of how peak electricity use will grow over time. Newer buildings are subject to stricter building energy codes, but are larger and more likely to have air conditioning; hence, the cumulative effect is ambiguous. This paper combines four large datasets of building and household characteristics, weather data, and utility data to estimate the electricity–temperature response of different building vintages. Estimation results show that new buildings (1970–2000) have a statistically significantly higher temperature response (i.e., use more electricity) than old buildings (pre1970). Auxiliary regressions with controls for tiered electricity prices, number of bedrooms, income, square footage, central air conditioning, ownership, and type of residential structure partially decompose the effect. Though California has had extensive energy efficiency building standards that by themselves would lower temperature response for new buildings, the cumulative effect of new buildings is an increase in temperature response. As new buildings are added, aggregate temperature response is predicted to increase. ► This paper looks at electricity use in buildings of different ages in a region of Southern California, US. ► Temperature response (electricity as a function of temperature) for air conditioning varies by vintage. ► New buildings have higher temperature response; technological improvements have been overwhelmed by other changes. ► This find means that adding new buildings has, in this area, increased aggregate temperature response. ► This paper estimates temperature response using monthly billing data, rather than utility/state-level aggregates.