Errors in Survey Reports of Consumption Expenditures

This paper considers data quality issues to analyze the pattern of con-sumption inequality in the 1990s exploiting two complementary datasets from the US Consumer Expenditure Survey. The Interview sample fol-lows survey households over four calendar quarters and consists of retro-spectively asked information about monthly expenditures on durable and non-durable goods. The Diary sample interviews household for two con-secutive weeks, and it includes detailed information about frequently pur-chased items (food, personal cares and household supplies). Each survey has its own questionnaire and sample. We exploit information from one sample as an instrument for the other to derive a correction for the mea-surement error affecting observed measures of consumption. We produce some evidence of non-classical measurement error affecting the aggregate measure of consumption both for diary and recall based data; we also show the implications of our findings to test for the Permanent income hypothesis.