Abstract:

In the next decades a dramatic change in the age composition of the populations of western societies will occur. The proportion of older people will increase due to declining birth rates after a baby boom. For products whose consumption is contingent upon age an important question is how the demographic changes will affect consumption. Two models can be used to predict the net effect of the changing age distribution on the consumption of a product. Model 1 weights todays penetration per age class with the expected size of the age classes. Model 2 weights todays penetration per birth cohort with the expected size of the cohorts. Whereas model 1 assumes that consumption behavior changes when people grow older, model 2 assumes that consumption behavior may differ between generations but remains constant when a generation ages. Developmental psychologists have shown that one cannot know which assumption is correct, if only cross sectional data at a single point in time are available. An example (oven baked french fries) demonstrates that the forecasts may be very different according to which model is used. Whereas model 1 predicts a decline in the total consumption of oven baked french fries between 1983 and 1993 model 2 predicts a strong increase. The question, which model is true, can only be answered if data from several cohorts for at least two points in time are analyzed. But since three interfering effects, age, cohort, and period may be at work, data interpretation of cohort designs has to be done carefully. By comparing data from 1983 and 1987 it can be shown that in the example of oven baked french fries cohort effects exist whereas age effects are implausible. Therefore model 1 is incorrect.