Recent research that looked into the dispositional base of job satisfaction focused on relating observed job satisfaction to core self-evaluations (CSE). This study was concerned with (a) the relation between the trait variance of job satisfaction and CSE and (b) the structure of the CSE-variables. Using a longitudinal measurement model in a secondary analysis of four waves of a longitudinal study we first tested whether CSE are sufficiently stable over time. Results indicate a high stability of CSE (.87 across 2 years). We then performed a state-trait decomposition of job satisfaction in order to separate trait variance of job satisfaction from changing variance. The stable job satisfaction factor was regressed on CSE-variables, using different models of CSE (a collective set, a latent factor, or an aggregate concept). Results were in favor of treating the CSE-variables as a collective set, and this set explained almost all stable variance of job satisfaction (84%). Moreover, only negative affectivity and internal locus of control had a significant impact, whereas self-esteem and self-efficacy had not. It is concluded that current conceptualisations of CSE as a superordinate concept underlying its four dimensions is possible but overly broad in job satisfaction research; collective consideration of LOC and KA is better and sufficient.