Mapping scientists' career trajectories in the survey of doctorate recipients using three statistical methods.

This paper investigates to what extent there is a 'traditional' career among individuals with a Ph.D. in a science, technology, engineering, or math (STEM) discipline. We use longitudinal data that follows the first 7-9 years of post-conferral employment among scientists who attained their degree in the U.S. between 2000 and 2008. We use three methods to identify a traditional career. The first two emphasize those most commonly observed, with two notions of commonality; the third compares the observed careers with archetypes defined by the academic pipeline. Our analysis includes the use of machine-learning methods to find patterns in careers; this paper is the first to use such methods in this setting. We find that if there is a modal, or traditional, science career, it is in non-academic employment. However, given the diversity of pathways observed, we offer the observation that traditional is a poor descriptor of science careers.