Cognitive research; Measurement error; Questionnaire design; Record use; Seam bias
This paper describes the development and initial testing of experimental data collection procedures for the Survey of Income and Program Participation (SIPP). The new procedures derive from prior research which has revealed serious levels of measurement error in some of SIPP’s basic statistics, the important implications of the errors for standard analytical uses of the data, and which has suggested the cognitive bases of the errors. The key features of the redesigned procedures are a clear and consistent message to all participants that accuracy is the primary goal, and an emphasis on the use of records to assist income reporting. Initial results from small-scale tests of the new procedures indicate a high rate of record use to report income flows, and decreased response error (as indicated by a reduction in underreport errors and in the “seam bias”); on the negative side, the initial tests have suffered substantially greater nonresponse than does standard SIPP, and possibly increased per-case costs.
Moore, Jeffrey C., Karen Bogen, and Kent H. Marquis. (2010). A ‘Cognitive’ Interviewing Approach for the Survey of Income and Program Participation: Development of Procedures and Initial Test Results. Statistical Research Division Research Report Series (Survey Methodology #2010-07). U.S. Census Bureau. Available online at <http://www.census.gov/srd/papers/pdf/rsm2010-07.pdf>.