This study analyzes the mobility between three labor market states: working in low paid jobs, working in higher paid jobs and not working. Using German panel data I estimate dynamic multinomial logit panel data models with random effects taking the initial conditions problem and potential endogeneity of panel attrition into account. In line with results from other countries, this first study on Germany finds true state dependence in low pay jobs and confirms previous results of state dependence in non-employment. Moreover, I find evidence for a \low pay no pay cycle", i.e. being low paid or not employed itself increases the probability of being in one of these states in the next year. However, compared to non working, being low paid does not have adverse effects on future employment prospects: the employment probability increases with low pay employment and the probability of being high paid seems to be higher for previously low paid workers. I find no evidence for endogenous panel attrition.