The role of heterogeneous demand for temporal and structural aggregation bias

Differences in estimated parameters depending on the frequency of aggregate data have been reported in several fields of economic research. Some differences are due to seasonal variations in demand, but temporal aggregation bias is reported even in seasonally adjusted models. These biases have been explained by time-nonseparable preferences and excluded dynamic components. We show that it is possible to observe temporal aggregation bias in a seasonally adjusted static model even when preferences are time-separable. This is because of changes in the distribution of exogenous factors describing the variation in seasonal demand across consumers. To show this, we develop a method for aggregation based on an Almost Ideal Demand System, where demand response varies across both consumers and time