Discrete choice experiments (DCEs) can be used to obtain monetary measures of benefit for use in economic evaluation. However, concerns exist that DCE predictions may have poor external validity because they are subject to hypothetical bias. Hypothetical bias exists when respondents to a stated preference survey makes choices that do not reflect their true preferences. The implication is often over-prediction of service uptake and over-estimation of willingness to pay, potentially leading to biased policy recommendations. This thesis aims to determine the best approaches to mitigate hypothetical bias and generate externally valid DCE predictions that can be used to inform robust resource allocation decisions in health. Hypothetical bias mitigation techniques can be described as ex-ante (before the task) interventions or ex-post (after the task) corrections. The thesis investigates the impact of three ex-ante interventions (a cheap talk script, a consequentiality script and an honesty oath) and two ex-post correction tools (quantitative and qualitative certainty scales) on DCE data quality, willingness to pay, and the external validity of service uptake predictions in. To determine the impact on external validity, stated and revealed preferences are compared in two DCE case studies eliciting general population preferences for preventative dental care services. The overall findings of the thesis suggest that preventing hypothetical bias a priori with ex ante interventions is preferable to trying to remedy it after the fact with ex-post corrections.