Validity of Forced Responses in a Randomized Response Model

Abstract

An empirical test was conducted to determine the degree to which respondents would follow the directions of the randomizing device in the directed response model of the randomized response technique (RRT). The results showed that a sizable percentage of the respondents would not give the directed response to a question if that response was a socially undesirable one and the question was sufficiently sensitive.

Do randomized‐response designs eliminate response biases? An empirical study of non‐compliance behavior

“…For example, in an FC study reported by Edgell et al (1982) respondents were asked to say 'Yes' when the outcome of a randomizing device is 0 and 1, 'No' when the outcome is 8 and 9, and to answer honestly for outcomes between 2 and 7. By fixing outcomes of the randomizing design a priori, the investigators found that about 25% of the respondents did not follow the instructions when answering a question on homosexual experiences: They answered 'No' although they should have responded 'Yes' according to the randomizing device.…”

E-research – selected issues concerning the Internet attitudes and opinions research

“…This design was created to address the problem that under the other randomized response designs some respondents may still feel uncomfortable providing a particular response (e.g., answering "yes") even when interviewers do not know whether they are answering the sensitive question. For example, Edgell, Himmelfarb, and Duchan (1982) used the forced response design to study college students' experiences with homosexuality. By fixing the outcome of the randomization device unbeknownst to the respondents, the researchers found that 25% of the respondents who were forced to reply "yes" by design did not do so.…”

“…Under the standard forced response design, we assume that both the probability of answering the sensitive question p and that of a forced "yes" ("no") p 1 (p 0 ) are known. However, respondents may not reply "yes" even when forced to do so (Edgell, Himmelfarb, and Duchan 1982). The modified forced response design addresses such noncompliance behavior.…”

“…Indeed, our extensive search yields only a handful of published studies that use the randomized response method to answer substantive questions (Madigan et al 1976;Chaloupka 1985;Wimbush and Dalton 1997;Donovan, Dwight, and Hurtz 2003;St John et al 2012). In contrast, a vast majority of existing studies apply the randomized response method to empirically illustrate its methodological properties by including some substantive examples (e.g., Abernathy, Greenberg, and Horvitz 1970;Chi, Chow, and Rider 1972;Goodstadt and Gruson 1975;Reinmuth and Geurts 1975;Locander, Sudman, and Bradburn 1976;Fidler and Kleinknecht 1977;Lamb and Stem 1978;Tezcan and Omran 1981;Tracy and Fox 1981;Edgell, Himmelfarb, and Duchan 1982;Volicer and Volicer 1982;van der Heijden and van Gils 1996;van der Heijden et al 2000;Elffers, Van Der Heijden, and Hezemans 2003;Lensvelt-Mulders, Hox, and Van Der Heijden 2005a;Lara et al 2006;Cruyff et al 2007;Himmelfarb 2008;De Jong, Pieters, and Fox 2010;Gingerich 2010;Krumpal 2012). This finding is consistent with previous reviews of the literature.…”

Bayesian randomized item response modeling for sensitive measurements

“…The RD is specified in such a way that the sensitive question is selected with probability ϕ 1 and the unrelated question with probability 1 − ϕ 1 . This RR method is extended to a forced response method (Edgell, Himmelfarb, & Duchan, 1982), where the unrelated question is not specified but an additional RD is used to generate a forced answer. Each observed individual answer is protected, since it cannot be retrieved whether it is a true answer to the sensitive question or a forced answer generated by the RD.…”