Nonresponse Bias Study for the Annual Capital Expenditures Survey

• Activity Number: 477
• Type: Contributed
• Date/Time: Wednesday, August 5, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Justin Z. Smith*+ and Katherine J. Thompson
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: nonresponse bias analysis ; response rates ; propensity model ; prediction model

Abstract

This paper presents a nonresponse bias analysis conducted for the Annual Capital Expenditures Survey (ACES). The ACES is a stratified SRS of approximately 45,000 companies with employees and 15,000 companies without employees. The ACES uses different weight adjustment procedures to compensate for unit nonresponse: the employer component uses an auxiliary variable; and the nonemployer component uses sample counts. We examine five years' of response rate measures then focus on the employer component of the survey. Here, we assess the validity of the assumed response propensity model, later studying the assumed prediction model. We conclude with recommendations for a slight modification of the existing weighting procedures; this modification addresses a systematic deficiency in the existing procedure that has very little impact on the estimate levels.