The paper analyzes effects of randomized response with respect to some binary dependent variable on the estimation of the probit model. This approach is used in interviews when asking sensitive questions. Alternatively randomization can be considered as a means of statistical disclosure control which has been termed post randomization method (PRAM). The paper shows that all properties concerning parameter estimation are maintained although there is a loss in (asymptotic) efficiency.
Asymptotic Efficiency Maximum Likelihood Post Randomisation Statistical Disclosure.