Four alternative measures of the average duration of unemployment are examined with the intention of illustrating: (1) the biases inherent in the average incomplete duration of unemployment, a statistic that is often the only one reported by many statistical agencies; and (2) the robustness of the average complete duration of unemployment to a host of assumptions underlying its derivation by non‐parametric methods. Canadian data are employed, but the results offer a guide to the construction of the average complete duration of unemployment that may have broader application