Past research shows that individuals with a common mental illness are more likely to change residence than those without such illnesses, but the spatial aspect of these moves is rarely considered, and models used imply that migration is consistent across areas. We address these issues with data from the original British Household Panel Survey cohort and the 1999 booster sample, followed through releases of the British Household Panel Survey and Understanding Society (1991-2013). Longitudinal probit models predict the probability of moving between waves of the survey(s). We test whether individuals with a common mental illness (measured using the 12 item General Health Questionnaire) are more likely to move between survey waves than those without a common mental illness. We use a novel cross-classified approach, in which the structure of the BHPS/US allows us to control for individual and time differences in migration behaviour. Additional parameters are estimated for the effect of common mental illness in each local authority to test whether certain areas of Great Britain ‘push’ those with mental illness away, and if they are ‘pulled’ towards other areas. We find that common mental illness is associated with greater probability of moving between survey waves (coefficient = 0.17, CI 0.13 – 0.19), so the predicted probability of moving for the population with a CMI (11.3%) is higher than for population without a common mental illness (8.5%). In addition, those with a common mental illness are particularly more likely to move to local authorities where migration is otherwise relatively rare.