The Demand for Health and the Contingent Valuation Method

Sammanfattning: The theoretical part develops Michael Grossman’s dynamic demand-for-health model by (a) letting the depreciation rate depend upon the level of health, (b) allowing a continuous set of health states, (c) introducing uncertainty (by letting health be a stochastic variable), (d) introducing social and private insurance and (e) releasing the assumption of an isoperimetric budget constraint. Beside the theoretical results, there are also results with important policy implications. When conducting empirical willingness-to-pay (WTP) studies, one must acknowledge whether the individual regards the hypothetical scenario as uncertain or not and whether insurance exists for the relevant good. The empirical part first investigates whether it is possible to apply the contingent valuation (CV) method to operation queues and waiting lists. Due to the exploratory nature and to the poor significance of the statistical model, the results are tentative at most. The dichotomous-choice (DC) WTP questions worked better than the open-ended questions and choosing the bid-vector and not having a too small population are important issues. The impact of ‘objective’ risk information on patients’ WTP for autologous blood donation (ABD) was then estimated. This information reduced the variance and magnitude of the WTP, which showed that the patients initially overestimated the risks. The WTP was significantly related to dread, perceived transfusion risk and income, indicating that ABD provides substantial benefits in the form of ‘peace of mind’. The experimental part presents the results of two experiments comparing the DC CV approach with ‘real’ purchase decisions for a consumer good. In addition, the hypothesis that a more conservative interpretation of the DC CV approach (where only absolutely sure yes-responses are counted as yes-responses) correctly predicted real purchase decisions was tested. Both experiments showed that the hypothetical yes-responses overestimated the real yes-responses. In the first experiment, the hypothetical absolutely sure yes-responses underestimated the real yes-responses, but in the second experiment the null hypothesis that the conservative DC CV approach corresponded to the real yes-responses could not be rejected. This suggests that it may be possible to sort out the real yes-responses from the false yes-responses by adding a question about the certainty of the yes-response.