Comparison of forecasts from aggregate and disaggregate models for personal vehicle energy consumption and CO2 emissions

Disaggregate models are generally more policy sensitive than aggregate models. However, disaggregate models may not track aggregate totals, such as total U.S. fuel consumption, as well as aggregate models. We obtain forecasts of aggregate fuel consumption from the DRI aggregate model and from Train's disaggregate model, using the same set of inputs for the two models. We find that the two models produce fairly similar forecasts. This result suggests that taking advantage of the greater policy sensitivity of the disaggregate model does not, in this case, entail a loss of accuracy in forecasting compared to the aggregate model.