The Survey of Income and Program Participation is a major vehicle for producing information about the economic situation of people and families in the United States. Recent research, however, has indicated the presence of important amounts of measurement (or response) error (Marquis and Moore, 1990a) and that the measurement error can result in substantial distortions in subject-matter estimates from the survey (Bollinger and David, 1993, Marquis and Moore, 1990a).

As a result, the Census Bureau has designed a new, cognitive interview approach for SIPP that may reduce measurement errors importantly. In this paper we mention the primary features of the new interview, discuss the design of the pretest, describe the design of the associated check of administrative records and present some results concerning measurement errors, response rates and costs. We conclude by raising some issues to be addressed in later research.