52 BULLETIN 1440, U. S. DEPAKTMENT OF AGRICULTURE 
FORECASTING THE PRICE OF HOGS FROM THE FUTURES PRICES FOR HOG 
PRODUCTS 
The price of live hogs is very closely correlated with the price of 
hog products at the same time. Thus, live hog prices may be rather 
closely estimated from the price of three products, as shown in Figure 
5, the correlation being 0.955. The multiple regression equation by 
which this chart is drawn, in terms of prices per 100 pounds for each 
product, is: Heavy hogs=0.2450 lard+0.0479 mess pork+0.6504 
short ribs— 0.1645. 
There is an active futures market for lard and short ribs, with 
deliveries as far as five months in the future. Prices are quoted for 
delivery only in four months, January, May, July, and September for 
lard. That gives some basis for making at least a quarterly pre- 
diction of the price. 
Although there is a high correlation between future quotations 
and the final price for each product, much of that correlation is due 
to general trends. Thus for the period 1896 to 1915 the correlation 
between lard futures for delivery in five months and the average 
price for the delivery month was +0.893. Taking first differences, 
however, for the change in price in five months, the correlation was 
only 0.290. 
For the period 1896 to 1915 the future prices of short ribs and 
lard for delivery in five months were correlated with the average 
price of heavy hogs for the delivery month, giving a multiple corre- 
lation of R = 0.879. For the period 1916 to 1924 the same factors 
gave a multiple correlation of R = 0.886. In both cases, especially 
the latter, the largest part of this correlation was due to the inclusion 
of the similar movements in trends in both actual and estimated 
values. It is probable that the correlation for 1916 to 1924 would 
drop to an insignificant value if computed as first differences instead. 
The net regression equation (from the pre-war solution) used in 
constructing Figure 23, is as follows: 
Letting Y= quotation for short ribs for delivery in five months, 
Z= quotation for lard for delivery in five months, and 
P=forecasted price of heavy hogs for five months later. 
P=0.0439Y-f0.6256Z + 0.668. 
The relative accuracy of these five-month forecasts, as compared 
with the six-month forecasts from the correlation with corn-hog 
differential, etc., on page 48, may be judged from the square of the 
standard errors of estimate, 21 which are 0.215 for the former, and 
0.131 for the latter. Since the accuracy of estimate varies inversely 
with this coefficient, it is evident that the latter method was nearly 
twice as accurate as the former for the period under consideration. 22 
2i The standard error of estimate, e, may be computed by the equation 
where x is the dependent variable. In percentage terms, the square of the standard error represents the 
part of the variation still unaccounted for — the difference beteen 100 per cent and the total determination 
by all factors measured. 
" For a method of estimating hog receipts directly from prices, weather conditions, and similar factors, 
see Elliott, F. F., adjusting hog production to market demand. [Unpublished manuscript in the 
files of the Bureau of Agricultural Economics.] 
