Measuring Indecision in Happiness Studies

The main objective of this paper is to evaluate the degree of uncertainty in self-reported happiness responses by means of a statistical model able to detect the relevant features of the expressed ratings. We consider a mixture model to address a twofold research question: how can we measure the indecision in expressed well-being; how to assess if this latent trait varies depending on the covariates of those surveyed? The selected modelling approach investigates the feeling/agreement component, making the underlying indecision explicit without imposing extra constraints to the model. Furthermore, our proposal allows to enhance the presence of a “refuge” option in the response patterns. The effects of individual characteristics may be highlighted, when significant. Results are presented stemming from an observational study showing that responses are characterized by a large variability among subjects. The methodology here experimented may be considered a general one since it can be exploited both in observational and in experimental surveys.