UC Berkeley

Volumes have been written on the subject of innovation, in academic literature as well as the popular press. Conventional wisdom would have one believe that innovation is the purview of the lone inventor at the workbench or the solo scientist huddled over the laboratory microscope. Yet a growing body of literature points to how relationships among actors - be they individuals, teams or entire companies - form the foundations of new knowledge (Hargadon & Sutton 1997; Powell, Koput & Smith-Doerr 1996; Uzzi & Spiro 2005). The most effective networks of relations are those that facilitate the exchange of unique information, often from distinct domains (Fleming 2002; Hargadon 2003). Within those networks, the people best situated to conduct such exchanges possess either large numbers of connections or, more typically, occupy positions as bridges between otherwise disconnected third parties (Burt 1992; Galaskiewicz 1979; Gould & Fernandez 1989).

Despite the role social networks may play in fostering innovation, we do not well understand the processes by which actors come to occupy these positions. In this dissertation, I make use of a novel dataset to provide insight into how some people come to occupy certain preferential network position while others do not. Specifically, I study the work-flow network among software engineers at a large, West Coast technology company as defined by the writing and reviewing of lines of code. By following a sample of 804 new entrants to the firm over a period of three years, I am able to trace the evolution of their positions within the larger network and understand the influence of factors present at time of entry.

The results of the analyses paint a consistent picture as to the determinants of position in the code review network for software engineers: one's team and one's job at time of entry play the strongest roles in determining one's subsequent centrality and autonomy. I find little to no evidence in support of a role for either intellectual endowment (educational attainment, years of previous work experience, coding ability) or social endowment (pre-employment relationships with other employees present at time of entry). Indeed, while non-significant, results were frequently opposite the hypothesized direction for individual attributes. The strong evidence on the influence of job- and team-related factors suggests that individuals have the greatest opportunity to shape their network positions when they choose their jobs and their managers, and thus their teams. One additional finding suggests that network activity also matters. That is, one can increase both centrality and autonomy through the brute force of writing a great deal of code.