Some additional comments
Hi Berk, we used a list randomization in our last survey to elicit condom use of sex workers in Senegal. We found that condom use was over-reported by 20 pp (condom use was 97% when self-reported and 77% with the list experiment). Overall, I think this is a good and inexpensive method to elicit sensitive behaviours in LMICs and it does allow sub-group analysis, which can provide important information. In our case, we find that it was mainly HIV positive and sex workers at high risk of HIV who were not using condoms, as one may expect.
I have a couple of additional comments on the method. Firstly, while in theory, the number and choices of non-key items should not affect the results, I strongly believe that it does and that non-key items should be related to the sensitive behaviours. There is some evidence that non-key items that are unrelated to sexual behaviors were better to elicit unbiased HIV related behaviors. For instance, in Droitcour (1991), they found that the use of non-key items that were unrelated to the key content make participants suspicious about the survey, and therefore reduces the success of the list randomization method. Despite this, most list randomizations I am aware of are designed by using non sensitive items that have nothing to do with the sensitive behaviour. Secondly, the results obtained from the list randomization are to some extent imprecise, and given the implementation challenges when doing a list randomization, the method is often applied to small samples. In this case, using a double list randomization where each group serves once as the control group and once as the treated group can increase precision. Thirdly, I still believe that the condom use elicited with our list experiment is still over-estimated, hence it would be useful to test the validity of the results obtained with the list randomization to other indirect elicitation methods or other methods (get condom use from clients?). Finally, I think that more methodological research should be undertaken to be able to use the results from the list randomization at the individual level, which is the main drawback of the method. We were thinking of using information from the sub-group analysis results to estimate the probability of condom use from the list experiment at the participant level but of course this is a bit tricky to implement.
Droitcour, J., Caspar, R. A., Hubbard, M. L., Parsley, T. L., Visscher, W., & Ezzati, T. M. (1991). The item count technique as a method of indirect questioning: A review of its development and a case study application. Measurement errors in surveys, 185-210.