Stated preferences in health economics

Choice experiments (CEs) are commonly used in applied economics to value non-market goods or to overcome market imperfections. In a CE, participants are asked to choose between two or more multi-attribute hypothetical descriptions of the good. These stated preferences are then used to estimate the marginal utility of changes in the composition of the good. This PhD thesis aims to contribute to the base of knowledge on the applications of stated preference methods in health economics. The thesis consists of three independent papers. All the studies have in common that they feature a choice experiment. However, regarding content, various policy-relevant questions in health economics are addressed. The first paper investigates heterogeneity in patients’ willingness to wait (WTW) for changes in time and risk attributes of kidney transplantation and examined how heterogeneity in WTW can be mapped with observable characteristics of the patients. Using mixed logit models in WTW-space, we find evidence of heterogeneity in WTW for attributes of kidney transplantation. We demonstrate that younger patients are willing to wait longer for a transplant with the better-expected outcome. Moreover, patients with longer duration of dialysis are willing to wait longer for a better-quality organ. The implication for transplant practice is that accounting patients' preferences in kidney allocation algorithm may improve patients’ satisfaction and the donor-receiver matching process. The second paper explores whether there is a link between cognitive ability, choice consistency, and WTW, using heteroskedastic multinomial logit, generalised multinomial logit models, and the same data set as in the first paper. A higher cognitive ability tended to result in more consistent choices, and consistency resulted in a lower WTW for changes in the multi-attribute content of kidney transplantation. The paper highlighted the importance of incorporating a cognitive ability test in CEs to determine the consistency of choice responses. The third paper investigates whether individuals aggregate multi-attribute information when completing choice tasks in CEs. An existing CE survey concerned with preferences for personalisation of chronic pain self-management programmes in the UK is used to explore attributes aggregation (AA) in multi-attribute choices. We develop a framework in which individuals restructure the multi-attribute information into a meta-attribute (e.g., convert non-monetary attributes into a single quality dimension) before making their decisions. We find evidence of AA when responding to CEs, with the probability of adopting AA greater for homogenous information. AA is more prevalent amongst participants who adopted a quick and click strategy (shorter response time), more likely to occur for later positioned choice tasks (potentially due to fatigue effect), leads to improvements in model fit and has implications for welfare estimates. Our results underline the importance of accounting individuals’ information processing rules when modelling multi-attribute choices.