Real national income comparison is one of the most frequently performed exercises in empirical economics. While the welfare implications of such comparisons are often not spelt out, there is little doubt that the significance that is attached to comparisons of real national income depend greatly on their implicit welfare content. This also influences the statistical procedures that are chosen, and as has been observed, the basic conventions that have been adopted by the statisticians of most countries for purposes of GNP-measurement are still founded on some notion of what measure of economic activity can best represent the contribution of that activity to welfare (Beckerman [2], p. 80). The welfare theory of real national income comparisons is, however, incomplete in several important ways, despite outstanding contributions by several distinguished economists, including Hicks [17], [18], Scitovsky [40], Kuznets [24], Samuelson [38], [39], Little [26] and Graaff [12], among others. Perhaps the most serious difficulty is with the treatment of income distribution. The starting point of thispaper lies in treating the same commodity going to two different persons as two different goods. The weighting of goods , thus defined, will incorporate distributional judgements. (The approach presented in this paper is best seen in the context of Graaff's [12] analysis of the need to dispense with the time-honoured device of drawing a distinction between the size and the of the national income and saying that welfare upon them both , and his penetrating observation that welfare depends (if we must use the term) on size only-and we do not know what the size is until we know the distribution ([12], p. 92).) This will permit a welfare interpretation of real income comparisons without the usual restrictive assumptions, e.g. leaving out distributional considerations explicitly (cf. Hicks [17], [18]), or making the peculiarly unrealistic assumption that is made by lump-sum transfers (cf. Samuelson [38], pp. 28-29, [39]). The important point to note is that typically social welfare can be seen to be a function of the vector of goods as defined here, but not as a function of the vector of commodities in the usual sense (i.e. irrespective of who gets them). Another departure lies in the explicit recognition of the fact that real national income comparisons involve different groups of people.' In this, these comparisons differ from traditional welfare economics (e.g. the use of Pareto Optimality or of Compensation Tests ), collective choice theory, (e.g. the use of Arrowian social welfare functions ), and the standard theory of national planning (e.g. cost-benefit analysis , or optimal growth theory ), which are concerned with comparing alternative positions of the same group of people. One way of avoiding the complex problems of inter-group contrasts is to confine real income comparisons to contrasting the actual position of a group with what its position would have been it were plhced in the position of another group. (Cf. Pigou: If the German population with German tastes were given the national dividend of England, . [34], pp. 52-53.) But this problem is ill-defined. For two groups of n people each, there are n! different ways of placing one group in the position of another. And, for two groups of different sizes, the interpretation of such as if comparisons is totally ambiguous.
Read full abstract