The main goal of the present work is to split effective unemployment into two components, one dealing with the natural rate of unemployment, and another with cyclical unemployment. With this purpose in mind, an estimation of stochastic cost frontiers is performed where natural unemployment is identified as a lower limit and cyclical unemployment as the deviation of effective unemployment with regard to that limit. To achieve this purpose, information is used from the 17 autonomous communities in Spain over the period spanning 1982 to 2013. Results evidence a greater importance of the natural component as the principal determinant of effective unemployment at a regional scale. The latter part of the work compares stochastic frontier estimations to those obtained when applying univariate filters, which are in widespread use in economic literature. The main conclusion to emerge is that the proposed decomposition modifies the weight distribution amongst the various types of unemployment, increasing the importance of cyclical unemployment. This finding has significant implications for economic policy, such as the existence of a greater margin for aggregate demand policies in order to reduce cyclical unemployment, particularly during growth periods.