Structural diversity is a poor proxy for information diversity: Evidence from 25 scientific fields

Abstract

Many empirical applications of structural holes theory rest on the assumption that there is a close relationship between network position, or ‘structural diversity,’ and access to complementary information, or ‘information diversity.’ Constraint, and to a lesser extent efficiency, are two measures that are typically used as measures of structural diversity, which is in turn used as a proxy for information diversity. In this article, we question the assumed relationship between structural and information diversity at the core of brokerage theory. We collected metadata for 864,450 journal articles from the Web of Science database, identified 1,192,488 authors, and constructed 13 different measures of information diversity that are based on the text data and co-authorship network structure. We correlated each of these measures with Burt’s constraint and efficiency as measures of structural diversity, finding only weak correlations and low R2 values. Efficiency performs slightly better than constraint, however the strongest correlation was 0.56, corresponding to an R2 of 0.3163. Furthermore, the variance in this relationship is non-uniform and indicates that, aside from the poor correlation, other factors are likely important mediators to the relationship between efficiency and information diversity. We argue that this shows that, even in the best cases, neither constraint nor efficiency should be used as proxies for information diversity and make suggestions for more diligent treatment of information in structural holes and brokerage research.