A method of estimating the distribution of uninterrupted unemploymentspells from aggregate unemployment data is outlined. The methodology does not i mpose constant individual hazard rates and allows for heterogeneity. Characteris tics of the distribution of unemployment spells are then considered including ha zard rates, measures of inequality such as Gini coefficients, and the fraction o f total unemployment associated with S weeks of unemploy-ment. The relationship between these char-acteristics and labor demand variables (GNP) and supply-side variables (transfer payments, unemployment insurance) is considered in or-der to better understand the relative role of short- and long-run unemployment.