Économie et Statistique n° 403-404 - Health, Ageing and Retirement in Europe
Are International Comparisons of Subjective States of Health Relevant? An Evaluation using Standard Vignette Methods
When the response modalities for a question which is evaluated subjectively are used in different ways by respondents, it is said that the modalities are affected by an «item bias» or by a DIF (Differential Item Functioning)-type effect. The responses must in this case be corrected before they can be used to formulate comparative hypotheses on the subjective state of health of sub-populations. We present a detection and correction method for DIF effects in self-assessments of subjective health in this paper. This method is based on individuals' assessments of states of health which correspond to imaginary situations, called standard vignettes. It entails «calibrating» an individual's response to a subjective question using the assessments he or she makes of situations described in different vignettes. It supposes that all the individuals refer to a single aspect of health when responding to subjective questions, that they assess the vignettes and their own situation in comparable ways, and that the vignettes can be a priori and unambiguously ordered by the analyst, according to the aspect being analysed. A potential DIF bias can be identified by comparing the distribution of the assessments made of the vignettes by two sub-groups of individuals. It can be corrected by positioning each individual's response to the subjectively assessed question in relation to his or her assessments of situations described in the vignettes. Taking into account coherence problems leads to the establishment of uncertainty intervals for the values of the variable which is produced in this way. The proposed method is illustrated by highlighting a possible DIF effect in the self-assessment of physical pain, according to the geographical origin of samples from the Share 2004 survey, which would particularly lead us to modify our comparison of subjective states of health in the Swedish and Dutch samples.