A STATISTICAL METHOD FOR MEASURING ‘MARGINAL UTILITY’ 193 
Evidently a demand schedule is not the simple thing it seems 
and I distrust statistically calculated demand curves except as 
they represent the temporary situation in the market as con- 
cerns substitutes, complements and money valuation. But a sta- 
tistical determination of the subjective value of income as a whole 
and of “physical” indexes of its constituent groups, such as food, 
clothing, housing, etec., seems a practical possibility. 
a 
Conclusion 
I have emphasized the fact throughout that I am here offering 
no statistics but only a statistical method. I have, however, 
applied the method to certain available statistics of the U. 8S. 
Bureau of Labor Statistics. The results confirm the common 
idea that progressive rather than regressive taxation of incomes 
is justified. 
I do not give these statistical figures here because the data need 
to be “smoothed” and subjected to critical analysis for varying 
size of family and other complications before the results can be 
considered even roughly accurate. Thus, the actual application 
of the formule here given to statistical use is deferred for 
another paper. If someone else than I will perform this arduous 
task I shall be more than pleased. 
The only important point which is made in the present paper 
is that if we have given budget tables for two different places (or 
times) relating to families presumably very similar in all essential 
particulars and have given also the relative price levels for both 
places or times for each of the budget groups (such as food, cloth- 
ing, or rent), and if these relative prices diverge sufficiently from 
each other, our formul® ought to give results of at least some 
statistical value.