A social network’s changing statistical properties and the quality of human innovation

We examined the entire network of creative artists that made Broadway musicals, in the post-War period, a collaboration network of international acclaim and influence, with an eye to investigating how the network’s structural features condition the relationship between individual artistic talent and the success of their musicals. Our findings show that some of the evolving topographical qualities of degree distributions, path lengths and assortativity are relatively stable with time even as collaboration patterns shift, which suggests their changes are only minimally associated with the ebb and flux of the success of new productions. In contrast, the clustering coefficient changed substantially over time and we found that it had a nonlinear association with the production of financially and artistically successful shows. When the clustering coefficient ratio is low or high, the financial and artistic success of the industry is low, while an intermediate level of clustering is associated with successful shows. We supported these findings with sociological theory on the relationship between social structure and collaboration and with tests of statistical inference. Our discussion focuses on connecting the statistical properties of social networks to their performance and the performance of the actors embedded within them. PACS number: 89.75.−k “ . . . the stage is ‘The Mirror of Nature,’ and the actors are ‘The Abstract, and brief Chronicles of the Time;’ – and pray what can a man of sense study better?” (The Critic 1779, in Sheridan 1962). What drives outstanding human achievement? Many have believed that great minds work in isolation and find singular inspiration. Yet, recent research has shown that this stereotype only rarely fits the reality. Following the careers of all the notable scientists, artists and philosophers since recorded time in the Eastern and Western civilizations, Collins (1999) found that the most creative geniuses, from Freud and Darwin to Beethoven and Curie, were embedded in 1751-8113/08/224023+12$30.00 © 2008 IOP Publishing Ltd Printed in the UK 1 J. Phys. A: Math. Theor. 41 (2008) 224023 B Uzzi networks of other scientists, researchers and artists who shared ideas through competition and collaboration. This trend is apparently intensifying in contemporary science. Teams on average now produce the most highly cited work, inverting the advantage that individuals once possessed (Wuchty et al 2007). At the same time, methodological developments have made the mysteries of large network structures reckonable, enabling new analyses of how network topology may affect human dynamics (Granovetter 1973, Barabási et al 2000, Burt 2004, Guimera et al 2005, Colizza et al 2006, Caldarelli 2007). In this paper, we build on previous results to analyze the statistical properties of a large collaboration network over a long time period, with a focus on how change in the network’s topology affects the performance of the system (Uzzi and Spiro 2005). We examined whether different topologies could shape the organization of the creative talent, amplifying or stultifying its innovativeness. Our analysis investigates the statistical properties of the collaboration network of all the creative artists that made Broadway musicals from 1945 to 1989. Our results indicate that the topography of the network may substantially affect the performance of the actors within it, suggesting that the arrangements of talent, not just the presence or absence of talent, underlies successful creative enterprises. Broadway musical creative artist network, 1945–1990 The Broadway musical industry (BMI) network of creative artists includes the artists responsible for the creation of Broadway musicals, an internationally recognized performance art that blends music, lyrics, dance, stage design and story into a single seamless artistic production. It takes its name from the locale in the New York City along Broadway Street where it developed in the late 1890s. Typically, six specialists—choreographer, librettist, composer, lyricist, producer and director—team up to create a musical. In this network, artists are nodes and collaborations between two artists are undirected edges. Our analysis examines all Broadway musicals made from 1945 to 1989. After 1945, an influx of new talent created a distinct post-1945 period. Since virtually all productions are done by a team of approximately six or seven artists, our network is a bipartite network with artists clustered within teams of about the same size. Artists who worked on the same show are the members of a fully linked team-clique (e.g., the team that made ‘Evita’). Links form between artists on different teams when an artist works on multiple teams. Our data include all 474 musicals of new material released between 1945 and 1989 as well as 49 musicals that closed in preproduction during the 1945–1989 period. These data allow us to capture all the professional links among artists and not just those due to finished products, which avoids network sampling bias. We have information on a musical’s opening date, theatre, creative artist team, financial success and critical success. These data are recorded in Playbills (Simas 1987, Green 1996). By adding new shows to the network and removing inactive artists, the network structure can change with time. Based on interviews with contemporary artists, we determined that artists ‘drop out’ of this network after seven years of no new productions. (Failure to drop out inactive artists gives a biased view of the real network by maintaining impossible links such as a link between Andrew Lloyd Weber (b. Mar 22, 1948) and George Cohan (b. Jul 4, 1878—d. Nov 5, 1942). Consequently, we added nodes and ties each year with the founding of new productions and deleted nodes and their ties that were inactive for seven years. The seven-year decay rule was confirmed statistically. There is 5% chance that an artist would make another show once they have been inactive for seven years. Our first year of observation was constructed using the artists who had been active within the prior seven years and each year after that new artists were added to it and inactive artists dropped from it. To link the network’s structure to performance,