Tracks through time and continuous processes: transitions, sequences, and social structure

1 The richness of longitudinal data In the social and economic sciences it appears that there was a pioneering enthusiasm for longitudinal data in the late 1960s and early 1970s, resulting inter alia in the US Panel Study on Income Dynamics (1968) and lifehistory studies such as Natalie Rogoff Ramsoy’s Norwegian Life History Study (Rogoff Ramsoy, 1975). While the continuity of the PSID helped a lot in the development of techniques for the use of individually longitudinal data, it has been a slow development, but has been nonetheless relentless and cumulative. By now we have a very substantial array of panel and life-history studies, longitudinal elements incorporated in exercises such as the Labour Force Survey, and a growing battery of longitudinal data sets based on official data collection. Commensurate with the rich data is a growing body of high-quality research that takes full advantage of its longitudinality. The argument no longer needs to be made for the greater richness, power and sheer interest of longitudinal data. Technical development has also been slow, but has been cumulative and is by now very substantial.1 1984 saw sociologists become aware of what has become the iconic technique for the analysis of longitudinal data, certainly on the sociological side, hazard rate modelling or ‘Event History ∗Paper prepared for the Conference, ‘Frontiers in Social and Economic Mobility’, Cornell University, March 27–29 2003 (tracks.tex,v 1.4 2003/04/04 11:30:18 brendan Exp) 1By development here I mean not simply the invention of new techniques, but also their diffusion into the imaginations of researchers, and researchers’ overcoming of the often substantial practical difficulties of handling longitudinal data