Purchase Print Copy

 FormatList Price Price
Add to Cart Paperback41 pages $23.00 $18.40 20% Web Discount

Summarizes, for potential users of survey data, methods and findings from a full study of survey response errors to sensitive topics. An analysis of published measurement studies suggests that the average measurement bias centers on zero for sensitive topic surveys but that responses are very unreliable or noisy. The biasing effects of these response errors for analysis (e.g., on estimates of correlations, regression, transition probabilities, and means) are examined both theoretically and using computer simulation. Simple and complex statistics that describe relationships or change are biased by unreliability as are the statistical inferences. Several strategies for neutralizing the effects of response errors on statistics and inferences are examined. Record checks and reinterviews yield satisfactory corrections in many situations, while internal consistency and instrumental variable approaches were less robust. Randomized response and multiplicity techniques were not effective because they do not address response unreliability.

This report is part of the RAND Corporation report series. The report was a product of the RAND Corporation from 1948 to 1993 that represented the principal publication documenting and transmitting RAND's major research findings and final research.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.