Abstract
This paper explores the liability of newness in the population of national labor unions in the United States from 1836–1985. It compares the fits of a variety of stochastic models that express different forms of time dependence. Both with respect to the overall hazard of mortality and the three specific types of mortality, disbanding, absorption and merger, there is evidence of a strong liability of newness. Moreover, the estimates reveal that this dependence is monotonic negative. Finally the Gompertz model that has formed the basis of recent work on this problem fits less well than a number of other models used.