Analysis of the phenomenon of unemployment is an important issue in socio economic research. According to the literature on the subject, analyses concerning the labour market are focused on reducing the resources of the unemployed by taking up employment. However, there are more reasons for the outflow of the unemployed from labour office registers. The proposed job offers often do not suit the unemployed. They refuse to take up the offered employment, do not show up at the office within the set deadlines, interrupt participation in activisation programmes for no reason. As a result, they are removed from the register. In order to pursue an effective labour market policy, it is important to identify such groups of people. The aim of the study is to compare the duration in unemployment of three groups of de-registered people. These are de-registrations due to taking up work, removal and other causes (resignation from the mediation of the labour office). In the empirical example, a cohort of the unemployed, de-registered persons in 2017 from the Poviat Labour Office in Szczecin was used. As a research tool, a non-parametric model of the duration of a specific phenomenon was used duration tables. Cohort tables for particular types of exiting from unemployment were built on the basis of individual data of 19,931 people (SYRIUSZ IT system). In the construction of the tables the following features of the unemployed were taken into account: gender, age, education, seniority and number of subsequent registrations.