FACTORS AFFECTING THE PRICE OF HOGS 51 
dependent, R IlFGH , was 0.878. The relative importance as determ- 
ining factors was as follows : 
Per cent 
F Corn-hog differential, 18-month lag 65. 9 
G Corn-hog differential, 6-month lag 1. 
H Industrial stocks, 6-month lag 10. 1 
Total determination 77. 
Besides I, above, a new statement of the dependent variable was 
computed as follows: 
12-month moving average of (price of hogs -*- seasonal variation-f- 
Bureau of Labor Index) 
-k- = Same average, lagged 18 months 
This gave a correlation with the corn-hog differential of r K F = + 
0.895, and a multiple correlation of R k .fgh = 0.937. The determina- 
tion coefficients were as follows: 
Per cent 
F 76. 5 
G 1. 7 
H 9. 5 
Total 87. 7 
That the corn-hog ratio lagged 18 months was the dominant cause 
of variations in both I and K is evident. This ratio is an expression 
of factors affecting the number of sows that farmers bred. The 
index of industrial stocks was next in importance, probably because 
it gave some slight forecast of changes in consumers' purchasing 
power and demand. The corn-hog differential with a six-month lag, 
introduced to measure the effect of corn and hog prices in changing 
the weight to which the hogs were fattened, was of only negligible 
importance. 
For the period January, 1920, to June, 1924. the following results 
were obtained: 
I^i-fgh = 0- 83 
and 
Rk-fgh = 0' 79 
In neither case were these high enough to use for forecasting 
purposes. 
One difficulty with this approach is that before the results can 
be reduced to actual prices, either the level of prices must be fore- 
casted or it must be assumed to remain unchanged. The seriousness 
of this limitation is shown by the fact that when prices were esti- 
mated for the pre-war period by the correlations with I, and the 
actual index of prices was used (that is, assuming the level of prices 
could be predicted six months in advance with perfect accuracy), 
the correlation between the actual and the predicted prices was 
+ 0.91; whereas when prices were estimated using the price index 
six months earlier (when the prediction was theoretically being 
made), the correlation was only +0.88. In either case it is evident 
that the forecasts for the pre-war period were not nearly so accurate 
as were the forecasts based on deviations from the trends. With 
further refinement, however, it may ultimately prove the most 
satisfactory method. 
