Further Index-Number Problems

Abstract

There are many problems which have been only partially solved in the theory and in the practice of index numbers and some of them can be conveniently considered in this last chapter. The first problem is that of the influence of sampling and other errors, first raised in 1.7 and in need of further examination. As an opening gambit we can quote Hofsten taking a ‘hard line’ on the British retail price index:

Like any other statistics, the accuracy of an index number should be given in terms of its variance… So far nobody seems to have attempted to produce any theory for index numbers in such terms, and it seems doubtful whether any such theory could be of any practical use… The concept of an accurate index should be given up. Statisticians should not be forced to behave as if it were possible to make the index very accurate, if only much money is spent on its construction. If government policy requires great confidence in the index, the statisticians should not permit themselves to be used as hostage. Hofsten (1956), pp. 8, 14