The effect of leadership style on return on equity classification: An artificial neural network approach

Abstract

This study investigates whether specific board-level leadership behaviors, particularly decision-rigour and strategic monitoring, can predict Return on Equity (ROE) tiers in Ghanaian banks. Addressing a gap in emerging market governance research, it integrates Upper-Echelon Theory and the Resource-Based View to explore behavioral influences on financial performance. Using a mixed-methods approach, data were collected via a 15-item Board-Efficacy Scorecard (BES) across ten banks and combined with 2023 ROE figures, categorized into Low, Medium, and High tiers. An Artificial Neural Network (ANN) model with two hidden layers (15 neurons each) was trained and benchmarked against multinomial logistic regression and random forest classifiers. The ANN achieved superior predictive accuracy (AUC = 0.94), significantly outperforming traditional models. Sensitivity analysis revealed that decision-rigour and strategic monitoring collectively increased the probability of High-ROE classification by 22 percentage points. These findings validate both the BES tool and the use of ANN in leadership-performance research. The study proposes the Board-Analytics and Disclosure Directive (BADD), advocating for BES publication, AI-powered board dashboards, and performance-linked compensation for directors. These measures aim to enhance governance transparency and financial outcomes in emerging markets. This is the first empirical application of ANN for ROE classification in Ghana's banking sector. By integrating advanced machine learning with context-sensitive behavioral metrics, the study demonstrates how board dynamics can be transformed into strategic tools for improving firm performance.