Assessing Nonresponse Bias in the Green Technologies and Practices Survey

• Activity Number: 606
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Brian Meekins*+ and Michael Sverchkov and Sharon Stang
• Companies: U S Bureau of Labor Statistics and US Bureau of Labor Statistics and Bureau of Labor Statistics
• Keywords: Nonresponse ; Propensity ; MAR ; NMAR

Abstract

In the fall of 2011 the Bureau of Labor Statistics fielded a survey to assess the number and nature of green technology jobs in the U.S. This mixed-mode (mail, web, phone, fax) survey collected information on occupation and wages of workers who spent more than half of their time using green technologies or practices (GTP). Approximately 35,000 establishments were selected from the population stratified by geography, industry, and ownership. About 30 percent of sampled establishments either did not respond to the survey or did not provide sufficient information. We seek to improve the current practice to account for nonresponse where estimates are adjusted to employment (from frame) within cells based on sample design factors. This work uses logistic regression to identify the "best" response propensity model, which modestly predicts response propensity. It is difficult to determine whether response is MAR, with additional adjustment unnecessary, or NMAR, where the auxiliary information fails to account for bias in the outcome estimate. We use methodology based on Sverchkov (2008 & 2011) in order to determine that nature of the nonresponse.