In this paper, we propose a model of income dynamics which takes account of mobility both within and between jobs. The model is a hybrid of the mover-stayer model of income dynamics and a geometric random walk. In any period, individuals face a discrete probability of moving, in which case their income is a random drawn from a stationary recurrent distribution. Otherwise, they stay and incomes follow a geometric random walk. The model is estimated on income transition data for the United Kingdom from the British Household Panel Survey (BHPS) and provides a good explanation of observed non-linearities in income dynamics. The steady-state distribution of the model provides a good fit for the observed, cross-sectional distribution of earnings. We also evaluate the impact of tertiary education on income transitions and on the long-run distribution of incomes.