2 BULLETIN 1069, IT. S. DEPARTMENT OF AGRICULTURE. 
to ascertain the best methods of dairy-herd management, Some of 
the 96 associations furnished records for more than one year. In such 
cases the records for each year were tabulated as though they were 
from different associations. On that basis the 96 associations fur- 
nished 120 sets of records. Therefore, to determine from these 
figures the average number of cows on test in an association we must 
consider each set of records as a separate association. 
In the 120 sets of records there were 2,939 herds and 41,990 cows 
on test. The study covered the period 1910 to 1920, inclusive; but 
every cow record was for a period of one year, the records of less 
than one year being included only when making herd and associa- 
tion summaries. In all other tabulations 12 months' records only 
were used. 
/S * 1 \ " 
/ /""" — "~-L_ * /* 
1? 
1 \ * r \^ \ • *n ^— -*3^L/ 
I L L^ X*^ JZ\ 
Pig. 1. — Location of the 96 cow-testing associations whose records were tabulated. 
It is well known that cows having high-production records are gen- 
erally more profitable than those whose production records are not 
so high, and this bulletin shows the rate at which income over feed 
cost advances as average production increases. 
CHARACTER OF DATA USED. 
Figure 1 shows the distribution of the cow-testing associations 
whose records were tabulated. 
So far as practicable the records of each association were studied 
independently of the others before being grouped into general tables. 
In this way any differences clue to geographical location were noted 
and interpreted. 
The first cow-testing association in the United States began work 
in Newaygo County, Mich., in 1906. On July 1, 1921, there were 
