This paper considers the dynamic specification of the relationship bet
ween electricity load and air temperature. It is shown that this relat
ionship has an important dynamic component and that ignoring this appe
ars to bias the estimated effects of temperature on load. It is also s
hown that forms of autoregressive specification may give a good explan
ation of present load even if there are no dynamics in the causal rela
tion. The assumptions underlying effective temperature and its empiric
al validity are investigated. Dynamic specification of the temperature
response is shown to be conditional on the time of day, which brings
into question the use of the effective temperature concept.