| Abstract: | Panel surveys, such as the Survey of Income and Program Participationand the Consumer Expenditure Survey, interview respondents every3 or 4 months, but ask the respondents for monthly data. A typicalfinding in such surveys is that changes in responses to a questionare relatively small for adjacent months within a referenceperiod but much more abrupt for adjacent months across referenceperiods. Previous studies have attributed this "seam effect"either to underreporting of changes within the periods or tooverreporting of changes across them. In the present studies,we attempt to distinguish these possibilities, using an experimentalmethod that allows us to gauge respondents' accuracy as wellas the number of times they change their answers. The studiesproduced seam effects and accompanying evidence for forgettingof queried information and bias toward constant responses withinthe reference period. In general, seam effects appear to increaseas a function of the demands on memory. We also find that separatingquestions with the same content in the survey instrument decreasesthe seam effect. To account for these data, we propose a modelin which respondents' answers are initially based on attemptedmemory retrieval. Inability to recall leads to (possibly biased)guessing or subsequent repetition of an earlier answer. |
