192 ECONOMIC ESSAYS IN HONOR OF JOHN BATES CLARK 
EAN 
nr a 
The curve could, of course, be extended to other points cor- 
responding to Cases 3, 5, 7, ete., and could be drawn on “doubly 
logarithmic” paper and treated as we have indicated for the want- 
of-income curve. 
Similarly the want-for-one-more “sq. ft.” of rent or shelter may 
be worked out as follows: 
Sip1 1000 X 24 
Si re RE 80.00 
Wilk, == .75X 3.= 2.25 
i i 
Rola 
giving the point in the curve corresponding to Case 1; and, for 
Case 3: 
Sus 1440 X 25 
3 =120.00 
pen 
HL a 
from which we see that an increase from 80 to 120 “sq. ft.” 
diminishes the marginal wantability of shelter from 2.25 to 1.00 
wantabs. 
According to these figures the food curve descends faster than 
the rent curve, this being due in the calculations to the more rapid 
change of the percentage (4) spent on food with a given change 
of income as compared with the corresponding change in the 
percentage (p) for rent. Thus by means of our formule we 
extract from “Engel’s law” its true significance psychologically. 
In the same way we may calculate the curves for clothing or 
any other consumption group, provided it is reasonably inde- 
pendent of the other groups. It is not feasible to construct any 
curve for bread, or butter, potatoes, or other items, the substitutes 
and complements of which have an important influence on their 
wantabilities. The reason is that a curve can only represent a 
variable as dependent on one other variable. When, as in the case 
of, say, bread or butter, its wantability depends on many vari- 
ables (e.g., on the quantities of bread, butter, potatoes), we need 
something more than a curve. A surface can show one variable 
dependent on two others. Beyond that no purely geometric repre- 
sentation will suffice, although a set of numerical schedules might 
conceivably be made out. 
Of course, these want curves or want schedules, when taken in 
conjunction with the want curve for income, first discussed, 
underlie demand curves and schedules.