Quality with Non-Probability Samples, Administrative Records, and 'Found' Data

• Activity Number: 491 - Sirken Award Session
• Type: Invited
• Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Sirken Award
• Author(s): J. Michael Brick*
• Companies: Westat
• Keywords: model assumptions ; Big Data

Abstract

For many years, probability sample surveys have been the accepted norm to produce statistical information of high quality for policy makers and official statistics. New approaches such as obtaining data from non-probability samples, administrative records, and other existing sources (Big Data) are alternatives that have become much more popular in recent years. In this environment, claims about the quality of information are difficult to judge and the consequences for policy making may be severe. We examine the struggle to define or re-define the quality of the information under all of these approaches. The process-oriented total survey error approach that is widely used for traditional surveys is not applicable for these alternatives. Outcome measures of quality are scarce and incomplete, but they may be required in this new era.