Mapping Organizational Design as NK Graphs: A Neural Network Approach

Abstract

The concept of interdependence has been widely theorized in organization theory and strategy using NK model, yet rarely has it been empirically explored. In this research we develop a novel way, building upon the advancements in Artificial Intelligence, to measure the interdependencies (K) among the firm activities (N). The method enables reconstructing the NK configurations of individual firms and represent them as graphs providing detailed insights into the whole structure of interdependencies. We use a sample of 2298 SMEs, 6 months of financial transactions per firm. In the obtained NK graphs, the bilateral relationships, i.e. interdependencies, and the unilateral relationships, i.e. dependencies, can exist simultaneously. We find that while interdependencies (compared to dependencies) are less prevalent than previously assumed, they have a more positive relationship with firm performance. We further observe that the centralization of the (inter)dependency pattern, instead of the bilateral or unilateral relationships, drives the firm performance differences. Finally, we observe substantial heterogeneity in terms of how firms in the same industry enact the K, and find that firm-level rather than industry-level interdependence variation drives the firm performance differential. We find that while interdependencies (compared to dependencies) are less prevalent than previously assumed, they have a more positive relationship with firm performance. We further observe that the centralization of the (inter)dependency pattern, instead of the bilateral or unilateral relationships, drives the firm performance differences. Finally, we observe substantial heterogeneity in terms of how firms in the same industry enact the K, and find that firm-level rather than industry-level interdependence variation drives the firm performance differences.