46 BULLETIN 1400, U. S. DEPARTMENT OF AGRICULTURE 
As a by-product of this analysis, too, we have not only the average 
relation of dwelling value to farm value, but also the average relation 
of every other factor considered to the value of the farm. 
As the first step toward the multiple correlation analysis, the 
different factors to be considered must be decided upon. Planning 
such an analysis after the basic records have been obtained, instead 
of before, the analysis is necessarily limited to the data already 
obtained. Within this limit, the following items were selected as 
likely to affect the farm value : 
Value of the dwelling (and tenant houses, if any). 
Value of the dairy barn, silo, and milk house. 
Value of other buildings. 
Fertility of the soil (as shown by a crop index). 
Percentage of the farm area tillable. 
Percentage of the farm area level. 
Type of road upon which the farm fronts. 
Distance to the nearest town. 
The crop index was computed by the usual methods, using only 
the yields of the most significant crops — corn, wheat, clover, and 
alfalfa — in making the computation. 
The type of road was designated according to a "code," as follows: 
(1) " State" road, concrete or asphalt; (2) macadam road; (3) broken 
stone, gravel, or slag, water-bound; and (4) dirt road. 
The correlation might have been made by adding the number of 
acres of land as a factor and then correlating all these factors directly 
with the total value of the farm. This would necessarily have given 
a very high correlation, since obviously the larger the farm the 
greater the value. By reducing all the items to terms of value per 
acre this difficulty can be avoided. Correlating the value per acre 
of the different classes of buildings, and the other items, with the 
value per acre of the farm should give a much more delicate measure- 
ment of the effect of the different factors upon farm values, since, 
other factors being equal, the value of the land should not vary with 
the size of the farm. 18 This procedure had been found to give good 
results in similar previous work (11) and was therefore adopted for 
this analysis. 
The factors used in the initial correlation study were therefore 
stated as follows: 
A. Value per acre of the dwelling (and tenant house). 
B. Value per acre of the dairy buildings. 
C. Value per acre of other buildings. 
D. Crop index. 
E. Percentage of the farm area tillable. 
F. Percentage of the farm area level. 
G. Code for type of road. 
H. Distance to town. 
I. Value of the farm per acre. 
Correlating the eight other factors with land value, 19 a multiple 
correlation of i?i. a b..h = 0.98 was obtained, using all of the 422 
> k This last assumption may be open to quastion. As will be pointed out later, the analysis could be so 
organized as to tesl it out and determine whether it is correct, instead of making it to a large extent implicit 
in all the subsequent analyses. 
19 Correlation analyses such as this, involving many variables, maybe accomplished most easily by 
using the methods outlined in (19). 
