Average duration of unemployment is one of the main indicators of labor market condition along with the rate of unemployment. There are different approaches to the estimation of unemployment duration, each having its advantages and disadvantages. As these approaches were developed concerning certain type of data, they can lead to a serious bias when applied to data that have another structure. In this article we propose a method of measuring average unemployment duration which can be used with RLMS data. Also we show that the results of previous studies of unemployment duration based on RLMS suffer from selectivity bias. Estimates that we obtain agree with the measures of official statistics and supplement them providing the information about effectiveness of job search.