Randomized Response Approach to Criminological Surveys (From Methods in Quantitative Criminology, P 41-57, 1981, James Alan Fox, ed. - See NCJ-85331)

• NCJ Number: 85334
• Author(s): P E Tracy; J A Fox
• Date Published: 1981
• Length: 17 pages
• Grant Number(s): 78-NI-0123
• Sponsoring Agency: National Institute of Justice (NIJ); US Dept of Justice NIJ Pub
• Sale Source: Academic Press, Inc
• Language: English
• Country: United States of America
• Note: Adapted from Evaluation Review, V 4, N 5 (October, 1980).

Annotation

The randomized response approach to self-report surveys yields aggregate estimates and prevents the derivation of specific information about particular respondents, so it tends to reduce the bias of evasive answers and deals with ethical concerns about the collection of sensitive information.

Abstract

Randomized response is a survey technique for reducing response bias arising from respondent concern over revealing sensitive information. The randomized response method uses indeterminate questions (the question answered by the respondent is unknown to the researcher) and thus maintains the anonymity of the responses. The interviewer merely records the response to a random question. Based on various stochastic relations between the questions and the observed responses, it is possible to obtain estimates of parameters in the aggregate. Because only aggregate estimates are possible, not only are respondents protected, but many ethical concerns surrounding the solicitation of sensitive information are nullified. Critics of the randomized response approach have contended that the reduction in bias earned through the randomized response may not be sufficient to outweigh the severe loss in efficiency. While this argument is valid for many applications involving mildly sensitive information where the potential bias with direct questions might be relatively small, in criminology, not only is the information solicited quite sensitive, but at times it is potentially incriminating. Mathematical equations are provided for randomized response models, and 33 references are provided.