Measuring leadership and productivity in an organisational structure

TL;DR

This paper introduces a new framework using cooperative game theory to measure leadership potential and productivity in organizational structures modeled as directed graphs, with practical algorithms for large networks.

Contribution

It develops the Average Forest measure based on maximal spanning forests, providing a novel way to assess leadership and productivity in hierarchical networks.

Findings

The AF measure captures agents' expected contributions across configurations.

Theoretical properties of the AF measure are established.

A Monte Carlo algorithm is proposed for large network estimation.

Abstract

This paper develops a novel methodological framework for assessing leadership potential and productivity within organisational structure represented by directed graphs. In this setting, individuals are modeled as nodes and asymmetric supervisory or reporting relationships as directed edges. Leveraging the theory of transferable utility cooperative games, we introduce the Average Forest (AF) measure, a marginalist leadership measure grounded in the enumeration of maximal spanning forests, where teams are hierarchically structured as arborescences. The AF measure captures each agent`s expected contribution across all feasible team configurations under the assumption of superadditivity of the underlying game. We further define a measure of organisational productivity as the expected aggregate value derived from these configurations. The paper investigates key theoretical properties of the AF measure -- such as linearity, component feasibility, and monotonicity -- and analyzes its sensitivity to structural modifications in the underlying digraph. To address computational challenges in large networks, a Monte Carlo simulation algorithm is proposed for practical estimation. This framework enables the identification of structurally optimal leaders and enhances understanding of how network design impacts collective performance.