It explores the implication of one central assumption: when firms receive multiple acceptable applications, they hire the worker who has been unemployed for the least amount of time. This assumption ("ranking") is contrasted with the assumption of random hiring ("no-ranking"). By embodying this assumption in a model of the labour market with job creation/destruction and matching, the joint behaviour of unemployment, the distribution of unemployment durations, and wages are characterized. The implication that the re-employment prospects of employed workers, were they to become unemployed, are better than those of the currently unemployed appears to have been an important feature of European unemployment experience in the 1980's. This paper examines the effects of the composition of unemployment on wage determination. It explores one central assumption: when firms receive multiple acceptable applications, they hire the worker who has been unemployed for the least amount of time. We refer to this assumption as "ranking" and contrast it throughout to the assumption of random hiring, or "no-ranking". By embodying this assumption in a model of the labour market with job creation/destruction and matching, we characterize the joint behaviour of unemployment, the distribution of unemployment durations, and wages. "Ranking" has a number of important implications. First, in addition to the obvious result that the exit rate from unemployment is a decreasing function of duration, we show that the effect of duration is stronger the higher the rate of unemployment. Second, with the wage depending on the distributions of prospective unemployment durations, the presence of long-term unemployed, per se, has little effect on wages. Third, looking at the relation between the wage and the aggregate level of unemployment, and comparing ranking to the case of no-ranking, the wage moves less with the level of unemployment, but more with the change in unemployment. Put simply, ranking and no-ranking models show exactly the same aggregate unemployment, slightly different long-run responses of wages to unemployment, and sharply different wage dynamics. The paper is organized as follows. Section I characterizes the matching process and derives a matching function based on an urn-ball process, with vacancies as urns and applications by unemployed workers as balls. Thus the distribution of applications per vacancy depends on aggregate unemployment and vacancies. Vacancies with at least one acceptable application are filled. When a vacancy receives more than one acceptable application, firms use a rule to choose among applicants. Thus, the rule determines who is hired but does not affect the number of hires. Section II uses this matching function in a model of the labour market with continual job creation and destruction and characterizes the equilibrium values of unemployment, vacancies and hires.