FARMERS' EARXIXGS IX SOUTHEASTERN PENNSYLVANIA 45 
is one reason why the " operator's earnings'' figure used in this "bul- 
letin as a measure of farmers' success varies widely from the " labor 
income" figure which has been used in other studies. The fact, 
brought out elsewhere (7), that the figure for the operator's earnings 
does tally with variations in farm efficiency much more closely than 
does the labor income figure, would seem to offer at least some 
measure of empirical proof that the reasoning upon which the new 
measure is based is sound and useful. 
A correlation study of the relation of farm buildings to the value 
of the farm was made for the specific purpose of determining what 
part of the " value" of the farm dwelling, as estimated by the farmers 
in the way described, was represented in the value of the farm. In 
a region where land was being brought under cultivation and now 
houses were being constantly erected, a $6,000 house should add 
86,000 or even more to the value of the farm upon which it was built. 
In a region where farms were being abandoned and the population 
was diminishing rapidly, a SO, 000 house might add nothing to the 
value of the farm upon which it was built. Chester County lies 
between these two extremes; the problem is to find just what part 
of the dwelling values (on the average) are added to the value oi the 
land on which they stand. 
The simplest way to approach this problem would be to pick out 
farms of the same size, fertility, and with farm buildings other than 
the dwelling of the same value, but with dwellings of different values 
upon them. Then, if enough such very similar farms could be found, 
comparing the differences in the values of the farms with the differ- 
ences in the estimated values of the dwellings would show the relation 
of the latter to the former, for the particular size and type of farm 
considered. This would be a fairly accurate result, because the 
method of grouping would hold constant the other factors such as 
size of farm, fertility of soil, and value of the dwelling, which otherwise 
might influence the farm value and so obscure the relation of dwel- 
ling value to farm value. 
In practice, however, there are so many different causes affecting 
farm value that it is impracticable to use this method. With the 
limited number of records it would be almost impossible to pick out 
enough farms exactly or even approximately alike in all particulars 
except value of the dwelling and value of the farm to give reliable 
averages. And even could such a group be chosen, the results from 
that group would be applicable only to farms of the particular size, 
fertility, etc., included in the group, and would not apply to other 
farms in the area. 
The statistical method of multiple correlation furnishes another 
way of getting the same result, not merely for a particular group of 
farms but as an average for all the farms in the area. By using this 
method it is possible to determine exactly the average relation be- 
tween the estimated value of the dwelling and the value of the 
farm, holding constant (by mathematical means) all the other factors 
included which may influence the farm value. This mathematical 
process may be thought of as equivalent to making all possible 
groupings of the kind suggested in the two previous paragraphs, 
noting the results of each grouping, and then making a weighted 
average of the relation between dwelling value and farm value as 
found for each grouping. 
