Advanced measurement and sampling for marketing research

This doctoral thesis investigates how to (i) elicit truthful answers on sensitive survey questions using indirect questioning techniques and (ii) identify optimal interventions to stimulate desirable behavior. The first chapter develops four statistical models for the Crosswise Model, a survey technique to infer the prevalence of sensitive characteristics and their potential correlates while hiding the responses of survey participants. The new models introduced here improve accuracy and validity over common applications of the Crosswise Model. The second chapter develops the Paired Response Technique. This technique is more general, being suitable for categorical sensitive survey questions. Empirically, this chapter estimates using a large scale survey study applying the Paired Response Technique that 17.7 %of the Dutch population has purchased medications to enhance sexual performance at least once. In contrast, with direct questioning, only 5.1 % of the sample admits committing these sensitive purchases. The third chapter develops an adaptive sampling technique to identify the most effective intervention out of a potentially large set of available interventions. The technique allows researcher to minimise required sample size and stop data collection once sufficient statistical evidence has been collected.