Applying discrete choice experiments to recent topics in health economics
This thesis comprises four studies in which health related preferences are elicited from the German public. All studies have in common that DCE data are analyzed while they are related to different research questions. In “Rescuing Schelling’s girl” it is analyzed whether the identifiable victim effect can be observed for a sample of the German public. Generally, it is widely believed that in modern societies identified lives are valued more highly than statistical lives. However, only a limited number of empirical studies finds evidence for the existence of this so-called identifiable victim effect. In this chapter, the results of a labeled DCE are reported, which was designed to overcome some of the limitations of the methods that have been applied to date to investigate preferences in the context of identified and statistical lives. It is found that the respondents prefer reducing the mortality risk of an identified life over the one of a statistical life on an aggregate level. When expressed in monetary terms, the respondents attach an additional value of approximately USD 3.8 million to saving an identified life over an otherwise comparable statistical life. However, an analysis of subgroups suggests that the older part of the respondents in the sample drives this effect. In “The dead-anyway effect from a societal perspective” we draw on the same data set used in “Rescuing Schelling’s girl” and use more sophisticated MIXL models to investigate the relationship between willingness to pay (WTP) for a given level of risk reduction and the initial mortality risk of the beneficiary when society spends the resources. Pratt & Zeckhauser (1996) assert that this relationship should be positive on the individual level and termed it the dead-anyway effect (DAE). We find evidence in favor of this assertion on the societal level. Furthermore, our results suggest that the WTP for a fixed risk reduction does not seem to increase incrementally with initial risk. In “Evaluating the consequences of rheumatoid arthritis” we empirically investigate whether patients and non-patients tend to attach different utility values to the state of suffering from a specific illness. If that would be the case, we were naturally led to the question whose utility values should be used as the basis in cost-effectiveness analysis (CEA). Intuitively, one would presume that patients are better informed about the consequences of their illness and public authorities should therefore use the patients' utility values in CEA. Contrary to this presumption, it has been argued that society at large should determine which values are to be used and not the patients because, in the end, it is societal resources that are to be allocated. Against this background, we use data from a DCE that was completed by patients suffering from rheumatoid arthritis (RA) and non-patients to explore the discrepancies between the two groups’ utility estimates for typical consequences of RA. Our results indicate that both groups attach remarkably similar part-worth utilities to the symptoms pain, fatigue and functional limitations. However, non-patients significantly undervalue the ability to work when compared to patients. In “A comparison of discrete choice and best-worst scaling” we supplement the DCE described in “Evaluating the consequences of rheumatoid arthritis” by a BWS survey with the identical underlying experimental design to shed more light on potential differences between the two methods. In particular, we estimate models capable of accounting for taste and scale heterogeneity in order to investigate if there are method induced differences in this regard. We find that the DCE and BWS lead to considerably different coefficient estimates and that the BWS responses exhibit a larger degree of taste heterogeneity.