The dynamics of perception: modelling subjective well-being in a short panel

We consider the neglected issue of the dynamics of perceptions, as expressed in responses to survey questions on subjective well-being. We develop a simulated ML method for estimation for dynamic linear models, where the dependent variable is partially observed through ordinal scales. This latent autoregression (LAR) model is often more appropriate than the usual state-dependence (SD) model for attitudinal and interval variables. The paper contains an application to a model of households perceptions of their financial well-being, demonstrating the superior fit of the LAR model to both the usual static model and the SD model.