Consumer price indices and the identification problem

Conventionally, consumer price indices are constructed on the assumption that we are observing a stable system of consumer demand, and that all price movements are, therefore, the result of supply-side changes. This often leads to an emphasis on consumer price substitution and to a recommendation that we should allow for it by using the geometric mean for first-stage aggregation. This paper argues, on the basis of economic theory and from observations on the UK clothing sub-index, that demand-side changes are also important in generating price movements. For most items we are unable to solve the resulting identification problem of whether supply-side or demand-side influences predominate: in these circumstances, the appropriate formula to use for first-stage aggregation is one that makes no assumptions about the cause of price changes – i.e. one that uses an arithmetic rather than a geometric average. Allowing for both sources of price movements also affects the way in which elementary aggregates should be defined: this should be on the basis of both demand and supply characteristics, in order to minimise problems that arise when aggregating disparate products.