When can micro properties be used to predict aggregate demand?

Heterogeneity in consumer behaviour creates differences in demand responses, which may create problems with aggregation across consumers. If aggregation problems exist, results from estimation based on aggregate data may prove difficult to interpret. Using estimation results from micro data to predict aggregate demand responses may also create disaggregation bias (the reverse aggregation problem). The aim of this paper is to discuss potential problems of using micro data to predict aggregate demand, and how such problems relate to the linear and non-linear aggregation problem. We also expand the theories of linear and non-linear aggregation to the case in which prices vary across agents. We formulate and test criteria for aggregation by using data on Norwegian household electricity consumption. We find clear evidence of aggregation problems, as heterogeneity in both price and income derivatives are significant. We thus expect to experience problems with aggregation when analysing Norwegian household electricity consumption.