The paper discusses the composition and dynamics of low paid workers whose hourly wages do not exceed two thirds of the median value. Using RLMS-HSE data for 2002–2016, we analyze how the size and the composition of low-paid employment, and the likelihood of being low-paid have changed over time. Our findings suggest that the size of the group has decreased from 30 до 24% of the total employment over the study period, while the chances of being low paid are significantly higher for those who have lower level of education, who are employed in low skilled jobs, and who reside outside large cities. However, the main focus of the study is on the earnings mobility of low paid workers. The key question is whether the low paid state is a dead-end and a long-term trap or is it just a stepping stone towards the high pay group? In order to answer this question, we apply a dynamic random effects multinomial logit-model, which allows to control for unobserved heterogeneity and to account for initial conditions problem. The study demonstrates high degree of structural dependence: two out of three low paid workers cannot exit this state over the year. This trap effect is stronger for women than for men. Though the stepping stone effect is also present, it is much weaker and relates to only one out of four low paid workers.