http://publications.iadb.org/bitstream/handle/11319/3312/How%20accurate%20is%20our%20missinformation%3f.pdf?sequence=2

The study reports on a randomized trial of 1,200 young adults enrolled in an employment training program, to determine the most cost-effective and appropriate interview mode for measuring youth risk behaviors. Four different survey administration modes -two interviewer-assisted (FTFI and CATI) and two self-administered modes (SAI and ACASI)-were randomly assigned. The authors have centered the study on the question of cost-effectiveness -actual implementation costs and estimates of measurement bias- and the randomization of interviewer gender in order to assess the interaction between gender and data quality. The research shows that the target population is likely to underreport sensitive questions in self-administered surveys, and thus the degree to which a mode improves self-reporting of a particular risk behavior or set of behaviors is likely to be context specific.