Cognitive Aspects of Dependent Verification in Survey Operations

• Activity Number: 476
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Lars Lyberg*+ and Stephanie Eckman and Magnus Roos and Frauke Kreuter
• Companies: Stockholm University and Institute for Employment Research and Gothenburg School of Economics and Joint Program in Survey Methodology
• Keywords: quality control ; CASM ; survey error rates ; data processing

Abstract

Dependent verification is a commonly used approach to check the validity of survey operations such as coding, listing, and translation. In this context, dependent means that the verifier has access to the initial outcome of the operation, which he/she then reviews for errors. There is ample evidence that this method tends to discover only a fraction of the errors in the initial operation. The authors' experiments with dependent listing and coding verification have shown that listers add only 81% of the housing units that are missing from the initial frame and control coders miss at least 50% of the errors. Survey methodologists have dealt with this shortcoming in dependent verification by turning to independent verification, where the verifier has no access to the initial outcome. This approach reduces the problem of under detection of errors, but is more expensive and more complex to administer than dependent verification. In addition, independent verification is not itself error free. By shedding more light on the cognitive mechanisms at work in the verification task, we hope to be able to design more effective dependent verification procedures that neutralize its drawbacks.