Estimating base temperatures in econometric models that include degree days

Heating and cooling degree days are in common use in conditional demand models, billing analysis, and large-scale energy forecasts. The implications of either choosing an ex-ante base temperature, or scanning over the base temperatures, as suggested in Fels (1986) and recommended by some evaluation protocols, are infrequently considered. These procedures result in biased estimates of weather-driven loads because of correlated errors-in-variables, and impart a downward bias to the variance of those estimates by a factor of two. A non-linear estimation procedure that corrects for these biases and an ex-ante correction factor for evaluating prior evaluations are offered.