Table 3 gives mean values for the key variables. Table 4 shows changes in costs at sample mean
values as effort devoted to performance of sanitation and process control tasks and the use of
food safety technology changes from the 95th to the 5th percentiles. The table shows that there
are no differences in costs due to differences in performance of sanitation and process control
tasks for plants above the 75th percentile. Costs rise only modestly for plants in meat slaughter
with performance in the 75th to the 5th percentiles and then jump dramatically at the 5th
percentile. In meat processing and poultry, cost rises from their 75th percentiles to the mean
performance of sanitation and process control tasks is followed by no change in costs for meat
processing and a small cost increase in poultry slaughter before both rise sharply between the
25th and 5th percentile.
The results for the performance of sanitation and process control tasks are surprising.
Intuitively, greater effort devoted to the performance of sanitation and process control control
tasks should lead to higher costs because more labor must be expended to complete a higher
percentage of tasks. Lower costs with better performance of sanitation and process control tasks
implies that (1) higher task performance of sanitation and process control leads to higher
productivity or (2) the sample is biased.
To see if the sample is biased, a model containing the tasks variable but not the
technology index was examined. This model permitted a near doubling of the sample sizes for
each industry since observations without the technology index did not have to be dropped. We
do not report coefficients but do provide the mean values (table 3b) and show how costs change
with changes in the percentile of task performance (table 4, bottom panel). Results are consistent
with the findings using the smaller set of data, except that the rise in costs over the 5th to 25th
15