Multiple Matrix Sampling: Approaches to Redesigning the U.S. Consumer Expenditure Survey

Jeff Gonzales

  Bureau of Labor Statistics

Abstract

The U.S. Consumer Expenditure Survey (CE) is a large-scale ongoing panel survey of American households conducted to collect information on consumer expenditures at a relatively fine level of detail. Data are currently collected through quarterly interviews (the Consumer Expenditure Interview Survey) and weekly diaries (the Consumer Expenditure Diary Survey) and are used in calculating the cost weight revisions for the Consumer Price Index (CPI), one of the nation’s leading economic indicators. Due to the interest in the reduction of respondent burden, improvement of data quality and/or the possible reduction of data collection costs, researchers at the Bureau of Labor Statistics (BLS) are investigating survey (re)design approaches to address these concerns. One proposed method is multiple matrix sampling, a technique for dividing a questionnaire into subsets of questions and then administering them to subsamples of the initial sample. This presentation will begin with a brief overview of the CE and multiple matrix sampling. It will include an illustration of a simple multiple matrix sampling design and a discussion of the implications and considerations for different phases of the survey process (e.g., including data collection and estimation procedures). The talk will also identify possible modified multiple matrix sampling designs that incorporate adaptive survey procedures and/or mixed mode surveys. The presentation will conclude with a discussion on how these designs will be evaluated and plans for future research.