Testing the vignettes method. Correcting self-reported health measures

The vignettes method is increasingly used to correct for reporting heterogeneity present in self-reported health measures. The general idea is that the reporting behaviour of a specific respondent can be anchored using the way in which that respondent evaluates the health status of hypothetical vignette persons. After anchoring the reporting behaviour of each individual, reporting heterogeneity is corrected for and thus cleansed (self-reported) health measures can be obtained. However, the validity of this vignettes method rests on the two crucial assumptions of vignette equivalence and response consistency. Violation of these assumptions may reduce the power of the vignettes method as a correction tool. In this thesis we test for the validity of this method by running regressions in which an outcome measure that is expected to be affected by true health in the end is regressed on a number of individual characteristics, the corrected health measures obtained from the vignettes method and the uncorrected observed self-reports. Evaluating the relative significance of the corrected and uncorrected health measures then allows one to assess the quality of using anchoring vignettes as a tool to correct for reporting heterogeneity. We do not find (strong) positive results on the validity of the vignettes method. While a large number of our test regressions are inconclusive, we do find negative results in a considerable amount of test regressions. The most positive result that we find is both corrected and uncorrected health measures having a similar significance in the test regressions. Our results do not reject the vignettes method as a correction tool per se, but it seems to be that one should be cautious in using the method. Further research is needed to improve the vignettes method thereby enhancing health measurement.