40 BULLETIN 1277, U. S. DEPARTMENT OF AGRICULTURE 
various products, and undoubtedly as a result make wiser decisions 
than they are now making and probably could make on the basis of 
cost data now available. In fact, it is likely that this is the form in 
which aid will have to be brought to the large majority of farmers 
for a long time to come. Such indices should be constructed as soon 
as possible for all the important systems of farming in the United 
States. 
ELASTICITY OF SUPPLY 
To what extent and how can input data be used in forecasting the 
supply which will be forthcoming at any given price? As already 
suggested, cost indices may indicate necessary relative changes in 
price. But for tariff making and other purposes it is sometimes 
desired to have absolute data. It has been assumed at times that one 
can determine from an array of costs what amount will be produced 
at any particular price. Unfortunately, the problem is not so simple 
as this. The new supplies forthcoming after a rise in price come 
more from farmers already producing at a profit than from the 
marginal producers. If input data are to be used in this way, there 
must be an analysis of cost rates and profits to go along with them 
that will show how cost rates and profits vary with the proportions 
of enterprises in the enterprise combination. 
STATISTICAL METHODS 
PROPER THEORETICAL ANALYSIS ESSENTIAL 
The statistical analysis of any given problem can be no better 
than the theoretical considerations, upon which it is based. An 
economic basis for the analysis of farm production has been pre- 
sented ; but before proceeding to the statistical analysis, it is neces- 
sary so to outline the procedure as to take account of the biological 
relations involved. The production of agricultural commodities in- 
volves the life-process of plants and animals; variations in input 
affect output only through physiological reactions. For this reason 
the theoretical framework of the analysis should be based on all 
available technical and experimental knowledge of the productive 
process, as well as upon the economic considerations. Statistical 
method is but a tool; it enables one to test or measure any specific 
relation or set of relations, but it can not indicate the direction or 
meaning of the relations. 
THE NATURE OF THE PROBLEM 
The methods employed in the analysis of farm data to determine 
the relations between the various input factors and the resulting 
output must be such as to show the net contribution to the product 
of each different input. Since the average unit contribution of a 
given input tends to be different for different total quantities of 
input, that is, the increases in output tend to follow the law of 
diminishing returns, the method must farther show the contribution 
of additional units of each input as well as the total or average 
contribution. 
The first of these conditions requires the use of multiple correla- 
tion, whereas the second requires some method of measuring curvi- 
linear relations both between pairs of variables and among many 
variables. 
