The validity of anchoring vignettes: Testing response consistency with an experiment

Anchoring vignettes are increasingly used to correct reporting heterogeneity of self-assessment sin (cross)national surveys and clinical research with the aim of improving interpersonal comparability. The validity of the method has however not been sufficiently tested. This thesis aims to test response consistency, one of the assumptions of anchoring vignettes, by means of an experiment. A sample of 48 students participated in this research. Respondents were asked for a self-assessment on three health domains: sleep, drinking behaviour and affect. Additionally participants were asked three detailed question on their health for these domains. This information was used to build a personal ‘replica’ vignette describing an hypothetical individual with exactly the same health problems as themselves for every domain. Respondents were then asked to rate the replica vignette on the same scale as the self-assessment. If response consistency holds, people should rate their own health the same as the replica vignette on these three domains. Results indicate that the participants in this research assessed their own health worse than that of the hypothetical individual in the vignette on average on each domain. Different statistical techniques and econometric models consistently show that this difference is (marginally) significant. Although these results point to a rejection of the response consistency assumption other explanations are considered.