This paper examines how unemployment can be measured in normative fashion - taking into account the mean and inequality of spell lengths - and how the extent of unemployment can be estimated from cross section data of the type found in labour force surveys. The issue is not straightforward since in these surveys completed durations of unemployed individuals are not observed yet they constitute the basis for calculating the kind of index that has been proposed to measure the extent of unemployment in a way that goes beyond the unemployment rate. The index proposed by Shorrocks has robust normative foundations and has an equivalent representation in terms of average complete duration and the density of completed durations. Building upon earlier work applied in the United States for estimating the first of these, we present a method that enables the index to be calculated based on an estimate of the density of completed durations. The approach is illustrated in the context of comparing male-female unemployment differences in France, where historically female unemployment has been higher than that of males.