procedures could have over- adjusted or under- 
adjusted for eommodity produetion. To address this, 
a seeond set of variables, known as eommodity 
targets, was added to the ealibration algorithm. 
These targets were eommodity totals from 
administrative sourees or from NASS surveys of 
nonfarm populations (e.g. USD A Farm Serviee 
Ageney program data, Agrieultural Marketing 
Serviee market orders, livestoek slaughter data, 
cotton ginning data). The introduetion of these 
eommodity eoverage targets strengthened the overall 
adjustment proeedure by ensuring that major 
eommodity totals remained within reasonable 
bounds of established benehmarks. Commodity 
eoverage targets with aceeptable ranges were 
established by subjeet-matter experts for eaeh State, 
with New England treated as a State. 
Eaeh State was ealibrated separately. The ealibration 
algorithm addressed commodity coverage. The 
algorithm was eontrolled by the 65 State farm 
operation eoverage targets and the State commodity 
coverage targets. To ensure that the ealibration 
proeess eonverged with so many eonstraints, it was 
desirable to provide some toleranee ranges for eaeh 
target. Although full ealibration to a single point 
estimate would assure that the weighted total among 
eensus respondents equaled its target for eaeh 
ealibration variable in either set, it was not always 
possible to ealibrate to sueh a large number of target 
values while ensuring that farm weights were within 
a reasonable range and not less than one. Beeause of 
this and beeause ealibration targets are estimates 
themselves subjeet to uneertainty, NASS allowed 
some toleranee in the determination of the adjusted 
weights. Rather than foreing the total for eaeh 
ealibration variable eomputed using the adjusted 
weights to equal a speeifie amount, NASS allowed 
the estimated total to fall within a toleranee range. 
This toleranee strategy made it possible for the 
ealibration algorithm to produee a set of satisfaetory, 
adjusted weights. 
Ranges for the farm operation eoverage targets were 
determined differently from the eommodity targets. 
The State target for number of farms had no 
toleranee range. The toleranee range for the 64 other 
State farm operation eoverage targets was the 
estimated smoothed State total for the variable plus 
or minus one-half of the standard error of the 
eapture-reeapture estimate. This ehoiee limited the 
2012 Census of Agriculture 
USDA, National Agricultural Statistics Service 
eumulative deviation from the estimated total for a 
variable when State totals were summed to a U.S. 
level total. The eommodity target toleranee ranges 
were determined by subjeet-matter experts, based on 
the amount of eonfidenee in the souree, and usually 
were less than plus or minus two pereent of the 
target. Ranges were not neeessarily symmetrie 
around the target value. 
Census data eolleetion was assumed to be eomplete 
for very large and unique farms with their weight 
being eontrolled to 1 during the ealibration 
adjustment process. For all other farms, adjustment 
weights were obtained using truneated linear 
ealibration whieh foreed the final census reeord 
weights to fall in the interval [1,6]. Adjustments 
began with the nonresponse and miselassifieation 
adjusted weights. Through ealibration, a seeond 
stage weight that simultaneously satisfied all farm 
operation eoverage and commodity coverage 
ealibration targets was obtained. Calibration was 
seldom able to adjust weights so that all State targets 
were met. Within the ealibration proeess, the highest 
priority for meeting a target was given to the number 
of farms, total land in farms, and top eash-reeeipt 
eommodities aeeounting for 80 percent of the State’s 
produetion. All remaining targets assoeiated with 
eommodities and eharaeteristies of farms and farm 
operators had equal priority. If a value within the 
toleranee range of any variable eould not be 
aehieved in a given State, the variable was removed 
as a target in that State and the ealibration algorithm 
was rerun. 
Weight eomputations in the final algorithms were 
performed to several deeimals. Thus, the fully- 
adjusted weights were non-integer numbers. To 
ensure that all subdomains for whieh NASS 
publishes summed to their grand total, fully-adjusted 
weights were integerized. This eliminated the need 
for rounding individual eell values and ensured that 
marginal totals always added eorrectly to the grand 
total. As an example of how the integerization 
proeess worked, assume there were five eensus 
records in a county with final noninteger coverage 
weights of 2.2, for a total of 1 1 . The integerization 
proeess randomly seleeted four of these reeords and 
rounded their final weight down to 2.0 and rounded 
the fifth reeord up to 3.0, for a total of 1 1 . 
The proportions of seleeted eensus data items that 
APPENDIX A A- 13 
