Two Arms of One Curve: Cooling and Heating Degree Days and Monthly Residential Electricity Sales in North Carolina, 1990–2026

Studies of weather and electricity demand in the Southeast tend to focus on summer cooling, but many North Carolina homes also heat with electricity. I used 436 months of public data (January 1990 to April 2026) from NOAA's nClimDiv dataset and Form EIA-861M to estimate how statewide residential electricity sales respond to both cooling and heating degree days. Because sales nearly doubled over the period, I modeled the log of sales, so that the weather effect grows with the size of the system. A model with cooling degree days alone explained R² = .13 of the variation in log sales. Adding heating degree days raised this to .45, and adding a linear trend raised it to .85. In that model, 100 more cooling degree days in a month went with 18.7% higher sales, 95% CI [18.0, 19.5], and 100 more heating degree days with 8.7% higher sales, 95% CI [8.3, 9.1]. Heating and cooling each accounted for a similar share of annual sales. The heating-to-cooling ratio ranged from 0.87 to 1.24 across specifications, so neither can be called the larger one. Between 2008 and 2026, each degree day was tied to less electricity per customer, by 15% to 27% for cooling. Most of that drop tracks a general decline in use per customer: in proportional terms the cooling response fell by only 6% to 12%, and that change was not robust. Code and data: https://github.com/Shubham6883/nc-degree-day-electricity