The Distribution of the Unemployment Burden: Do the Last Hired Leave First?

Abstract

R ECENT theoretical and empirical analysis of unemployment has emphasized its dynamic character: flows into and out of unemployment are very substantial in relation to the stock of unemployed individuals,' and groups with high unemployment rates tend to be those whose members experience short but frequent spells of joblessness (Hall, 1970). This view of unemployment has afforded important new insights into the question of why average unemployment rates differ so markedly across certain labor force groups, and has led to major shifts in the focus of labor market policy. For example, the discovery that the higher unemployment rates of blacks are due almost entirely to the higher frequency of jobless spells they experience (Perry, 1972) has shifted emphasis away from policies that stimulate demand toward policies that will help to reduce job turnover. Existing studies have focused almost exclusively on differences in the average values of the duration and frequency of spells of unemployment between certain labor market groups; none has examined systematically the variation in unemployment spell lengths and frequencies across individuals. This paper seeks to fill this gap by examining the relative contributions of unemployment frequency and unemployment duration to the distribution of total hours of unemployment across individuals within each of several important labor force groups. Dispersion in the distribution of unemployment across individuals results when either the length or frequency of unemployment spells is unevenly distributed across individuals. This dispersion is reinforced when the length and frequency of spells are positively correlated, i.e., when those individuals who have the greatest difficulty finding jobs (the last hired) tend to be the same individuals who experience the greatest difficulty keeping them (the first fired). The principal contribution of our study is that it enables the variations in individual unemployment experience to be linked explicitly to individual variations in the length and frequency of unemployment spells. Section II below outlines a probabilistic model of individual labor market transitions that serves as the basis for our empirical estimation procedures. In section III we then describe the data employed in our study (the National Longitudinal Survey) and present our estimates of the distributions of individual transition probabilities for each of four demographic groups. Section IV examines the effect of a business cycle downturn on the unemployment experienced by individuals in these groups.