End-use metered data collected for five years from 350 California residences are used to compare three types of models for allocating estimates of annual residential central air conditioning energy use to hours of the year. We assess how well the model fits the data for daily energy, peak demand, and demand coincident with system peak. A model which couples regression-based functions for daily load estimation with hourly estimation according to a library of load profiles is judged to have a slightly better fit to the data than a model that estimates hourly loads directly from hourly functions derived from linear regressions. Concerns regarding the applicability of end-use metered data for long-term resource planning are described.