This paper investigates the reliability of self-assessed measures of health using panel data collected in Albania by the World Bank from 2002 to 2004 through the Living Standard Measurement Study project. As the survey includes questions on the usual self-assessed measure of health and on more objective health problems, we combine both types of information to better understand how respondents revise their answers to the self-reported measures over time. The estimation of random effects ordered Probit equations provides two main results. First, differences in self-reported subjective health between individuals are much more important than those over time, suggesting a strong state dependence in subjective health status. Second, our empirical analysis sheds light on the coherence of respondents, both from a subjective and an objective viewpoint. Health evolution is influenced by more permanent shocks in health than by more transitory illness or injury.