respond to the census despite numerous attempts to 

 contact them, and not all farm operators are requested to 

 provide the sample data items. All farms were surveyed for 

 all data items so only nonresponse estimation was neces- 

 sary. 



Whole Farm Nonresponse Estimation 



A statistical estimation procedure was used to account 

 for the census farms among mail list nonrespondents that 

 were not designated for telephone followup. A stratified 

 systematic sample of eligible census nonrespondents were 

 mailed a simplified report form. Five sample strata were 

 defined based on form type, expected value of sales, and 

 previous census status. The report form was designed to 

 provide sufficient information to determine farm status. 

 Additional mail and telephone contacts were made to 

 survey nonrespondents to obtain sufficient response for 

 survey estimates. 



Estimates of the proportion of census nonrespondents 

 that operated farms were made for each stratum in the 

 State using survey results and applied to the total number 

 of census nonrespondents in that stratum. A synthetic 

 estimation procedure was used to estimate the number of 

 census nonrespondents that operated farms for each 

 county by stratum. This estimation procedure is based on 

 the assumption that the distribution of farms in a stratum 

 by county is the same for census nonrespondents as for 

 census respondents. 



Within each stratum in a county, a noninteger nonre- 

 sponse weight was calculated and assigned to each 

 eligible respondent farm record. The procedure used for 

 calculating the nonresponse weight assumed the eligible 

 census respondents and the nonrespondent farm opera- 

 tions in a county had similar characteristics within each 

 stratum. The noninteger nonresponse weight was the ratio 

 of the sum of the estimated number of nonrespondent 

 farms (using nonresponse survey results) and the number 

 of eligible census respondent farms to the number of 

 eligible census respondent farms. Stratum controls were 

 established to ensure that this weight was never greater 

 than 2.0. The noninteger nonresponse weight was used in 

 the estimation of the final weight for the sample items. It 

 was randomly rounded to an integer weight of either 1 or 2 

 for each record for tabulating the complete count items. 



The procedure assumed that we obtain complete response 

 from large and unique farm operations because these 

 cases received intensive telephone followup during cen- 

 sus processing. In situations where addressees could not 

 be contacted by telephone or refused to cooperate, sec- 

 ondary sources such as Agricultural Stabilization and 

 Conservation Service offices or county extension agents 

 were asked to provide information as to whether or not the 

 addressee had agricultural activities. Data from previous 

 census reports for the specific addressee, in conjunction 

 with other information, were used to complete the census 

 report form. 



Table A quantifies the effect of the nonresponse esti- 

 mation procedure on selected census data items. The 

 percentage of the census value contributed by nonre- 

 sponse estimation as provided in this table indicates the 

 potential for bias in published figures resulting from this 

 procedure. The estimates provided in these tables do not 

 reflect the effect of nonresponse to individual data items 

 on respondents' census report forms. The effect of this 

 item nonresponse is discussed further under Census Non- 

 sampling Error. 



Table A. Percent of State Totals Contributed by 



Whole Farm Nonresponse Estimation: 1987 



Item 



Farms number- 

 Land in farms acres- 

 Value of land and buildings $1 ,000- 



Market value of agricultural products sold --$1,000- 

 Harvested cropland acres- 

 Corn for grain or seed acres- 

 Wheat for grain acres- 

 Livestock and poultry inventory: 

 Cattle and calves number- 

 Hogs and pigs number- 

 Hens and pullets of laying age number- 



Percent of total 



16.3 

 1.1 

 4.9 

 1.6 

 2.1 



2.6 



11.3 

 .6 



CENSUS SAIVIPLING ERROR 



Sampling error in the census data results from the 

 nonresponse sample and the census sample data collec- 

 tion. Census items were classified as either complete 

 count or sample data items. In Hawaii, all farms were 

 asked the complete count and sample items. The com- 

 plete count data items included land in farms, harvested 

 cropland, livestock inventory and sales, crop acreages, 

 quantities harvested and crop sales, land use, irrigation, 

 government loans and payments, conservation acreage, 

 type of organization, and operator characteristics (sections 

 1 through 22 of the census report form). The sample data 

 items included farm production expenditures, fertilizer and 

 chemical usage, farm machinery and equipment, value of 

 land and buildings, and farm-related income (sections 23 

 through 28 of the census report form). Variability in the 

 estimates of complete count and sample items is due to 

 the nonresponse sample estimation procedure that is 

 based on a sample of nonrespondents. 



This sample for the 1 987 Census of Agriculture is one of 

 a large number of possible samples of the same size that 

 could have been selected using the same sample design. 

 Estimates derived from the different samples would differ 

 from each other. The difference between a sample esti- 

 mate and the average of all possible sample estimates is 

 called the sampling deviation. The standard error or sam- 

 pling error of a survey estimate is a measure of the 

 variation among the estimates from all possible samples, 

 and thus is a measure of the precision with which an 

 estimate from a particular sample approximates the aver- 

 age result of all possible samples. The percent relative 



C-2 APPENDIX C 



1987 CENSUS OF AGRICULTURE 



