Improving Comprehension and Recall in the Consumer Expenditure Interview Survey: Discrepancies in Comprehension and Recall as a Source of Nonsampling Error *

The current paper describes research focusing on the Consumer Expenditure Interview Survey. Over the years, data analysis has indicated that this survey may be underestimating expenditure levels (Gieseman,1987). This, combined with the fact that researchers increasingly depend on surveys as a fundamental source of data (Anderson, Kasper, and Frankel, 1979)leads us to investigate ways to improve our data collection methods. The Bureau of Labor Statistics (BLS) has begun research to develop and test improvements in questionnaire wording and design for the Consumer Expenditure Interview Survey Questionnaire. Specifically, the research discussed below was designed to employ both cognitive laboratory research and survey methodology to evaluate and recommend a new collection format for the Health and Medical Expenditures (Section 15) of the Consumer Expenditure Interview Survey Questionnaire. We begin with a discussion of the Consumer Expenditure Survey, including the current format of the Health and Medical Expenditures Section. Baseline research in comprehension and recall is outlined, followed by the goals of the current research. The current research includes two phases: laboratory research on the cognitive approaches respondents use in answering questions, and a feasibility field test of the revised instrument to be conducted in fall of 1989. The results of the laboratory research are presented and a discussion of how these results and cognitive theory proved useful and supportive in redesigning the instrument follows.