POIDS which recognized farms by operator/operation. Following this code, each distinct
agricultural operation (e.g., a farm, ranch, feedlot, or greenhouse) was considered a
separate farm record. In other words, an operator received as many forms as the number
of different businesses s/he ran. Consequently, the matching of farms between the 1997
and 2002 censuses was not as precise as between the 1992 and 1997 censuses. The
matching mechanism was a correspondence rather than a function and resulted in a bias
in the absolute number of entering and exiting farms in 2002. Because the bias affected
all size cohorts, we limit our analysis of the cohorts to relative changes.
Summary statistics are computed for each cohort in each census to determine changes
in size distribution characteristics of dairy farms over time. They include: (1) number of
farms, (2) mean size, (3) median size, (4) size range, (5) size standard deviation, (6) size
skewness, (7) size kurtosis, (8) number of exiting firms, and (9) portion of farms in each
of the four sales categories (i.e., 90% or more, 75-94.9%, 50-74.9%, or less than 50%
percent of all agricultural sales from milk and dairy products). Incumbent farms in
subsequent censuses do not change their cohort assignment. Therefore, size ranges of
cohorts in the 1997 and 2002 censuses overlap due to growth or decline in size of
individual farms but represent all surviving farms in each cohort.
For the 1997 and 2002 censuses, an additional cohort of new entrants is created for
farms entering the dairy business since the previous census. The same statistical
information is recorded for each cohort of new entrants. In addition, the number of new
entrants that are in the size range of each 1992 cohort is also recorded.
To permit valid calculations of firm growth between censuses, agricultural receipts in
each census are deflated by the index of prices received. Milk and dairy product sales for
10