In the context of income dynamics, we investigate whether aspects of agents' superior information relative to the econometrician's limited information are captured in subjective expectations data. It is natural, for instance, to assume that the econometrician cannot observe idiosyncratic shocks to both permanent and transitory components of income. In this case perceptions of risk differ between agents and the econometrician. Our tests are based on panel data elicited yearly from British households. We find evidence of superior information consistent with standard income modelling.