r

OH. 8 COMPARISONS OF PURCHASING POWER 119

a war—intervenes which, materially changes relativeprices and the character of consumption, but thatafter an interval equilibrium is restored and bothprices and consumption return to exactly what theywere before. Now it is obvious that in such circum-stances the purchasing power of money at the end isexactly the same as it was at the beginning. But ifthe Chain Method has been employed, there is noguarantee that the index-number will return to itsoriginal position—indeed one may be pretty sure thatit will not. 1

Finally, the Chain Method is statistically laboriousand inconvenient to apply in practice—so much sothat it has been very seldom employed, in spite of themany years which have passed by since it was firstrecommended and the general approval (more than itdeserves in my opinion) which it has received fromtheorists.

I conclude, therefore, that it would be betterin comparing, let us say, the purchasing power ofmoney to-day and fifty years ago to compare theprice of that part of expenditure (say 50-70 per cent—I have no idea as to the true proportion) which ismore or less unchanged in character, supplementingthis by a list of the expenditures discarded and added(so as to enable a general judgment to be made as tothe extent of the improved opportunities), rather thanto compare the price-levels by means of the ChainMethod applied year by year over the interveningperiod.

We are left, therefore, with the Highest CommonFactor Method supplemented by the Method of

1 An interesting analysis bearing on the above, though primarily directedto an examination of the conditions in which a Chain Index diverges fromthe corresponding Fixed Base Index, has been published by ProfessorPersons (Review of Economic Statistics, May 1928, “ The Effect of Correlationbetween Weights and Relatives in the Construction of Index Numbers,”pp. 100-105).