Age effects in demand models: What about cohort and period effects?

摘要

Age effects in demand models are commonly estimated by using one year of microdata. However, what is usually forgotten is that there is a linear dependency between age, period, and cohort. Hence, the age effect may be biased when cohort is excluded. In this paper, a mixed-effects-demand model for milk is estimated, and it is shown how it might go wrong when cohort is not accounted for. In addition, lower intercepts in younger cohorts than in older cohorts indicate that over time, milk purchase will continue to decrease as younger cohorts replace older cohorts.