Evaluation of Nonresponse in the Nationwide Food Consumption Survey, 1987-88.Nationwide Food Consumption Survey

摘要

Response rates for the 1987-88 Nationwide Food Consumption Survey (NFCS 1987-88)were very low--38 percent at the household level and 31 percent at the individual level. The publication provides information on the procedures used in NFCS 1987-88, the response rates, and the characteristics of the unweighted sample compared with population estimates from the U.S. Bureau of the Census. The regression weighting approach used to adjust for nonresponse is described. Analysis done to evaluate the effect of nonresponse on estimates of food and nutrient intakes in NFCS 1987-88 are presented. Results of a study of attrition suggested that the regression weighting may correct nonresponse bias. The study showed that differences between respondents and nonrespondents in eating behavior were predictable because they were caused by known socioeconomic variables, which can be adjusted for by weighting, and were not caused by some other unknown and nonrandom, and thus unpredictable, response propensity. Users must balance their need for the data and their tolerance for error against the limitations of the data.