Optimal net interest margin and bank stability in emerging markets: evidence from dual banking Indonesia during the COVID-19 pandemic – a game theory perspective

Purpose This study aims to estimate the optimal net interest margins (NIMs) of large and small conventional banks in Indonesia before and during the COVID-19 pandemic. Indonesia’s banking market is oligopolistic, with a few dominant banks shaping system-wide pricing. By focusing on strategic leader–follower dynamics and uncertainty, the study evaluates whether banks’ observed NIMs align with model-implied optimal levels. The research also situates the findings in the context of Indonesia’s dual-banking environment (conventional and Islamic banks) and draws implications for profitability, intermediation efficiency and regulatory monitoring of pricing behavior during periods of systemic stress. Design/methodology/approach The authors extend the risk-averse dealer model (RADM) by embedding a dynamic Stackelberg game under incomplete information to capture leader–follower strategic behavior. Using quarterly data for 20 conventional banks from 2018Q1 to 2021Q4, the authors classify large banks as leaders and small banks as followers. The sample is divided into prepandemic (2018–2019) and pandemic (2020–2021) periods. Model parameters are estimated using bank-level panel regressions, and optimal NIMs are derived from the first-order conditions of the game. Observed and optimal NIMs are compared across periods, with robustness checks using subsamples of the five largest and five smallest banks. Findings The authors extend the RADM by embedding a dynamic Stackelberg game under incomplete information to capture leader–follower strategic behavior. Using quarterly data for 20 conventional banks from 2018Q1 to 2021Q4, the authors classify large banks as leaders and small banks as followers. The sample is divided into prepandemic (2018–2019) and pandemic (2020–2021) periods. Model parameters are estimated using bank-level panel regressions, and optimal NIMs are derived from the first-order conditions of the game. Observed and optimal NIMs are compared across periods, with robustness checks using subsamples of the five largest and five smallest banks. Research limitations/implications The study focuses on group-level leader–follower dynamics among conventional banks and does not capture within-group heterogeneity. Alternative game-theoretic structures, such as static or complete-information games, are not modeled. The analysis is limited to 20 conventional banks covering approximately 85% of system assets, and thus does not directly estimate Islamic bank margins despite discussing the dual-banking context. The pandemic dummy cannot fully isolate COVID-19 effects from other contemporaneous shocks. Future research could extend the model to include bank-level risk management variables, richer regulatory measures and adapt it to Islamic profit-sharing structures for broader applicability in dual-banking economies. Practical implications The study focuses on group-level leader–follower dynamics among conventional banks and does not capture within-group heterogeneity. Alternative game-theoretic structures, such as static or complete-information games, are not modeled. The analysis is limited to 20 conventional banks covering approximately 85% of system assets, and thus does not directly estimate Islamic bank margins despite discussing the dual-banking context. The pandemic dummy cannot fully isolate COVID-19 effects from other contemporaneous shocks. Future research could extend the model to include bank-level risk management variables, richer regulatory measures and adapt it to Islamic profit-sharing structures for broader applicability in dual-banking economies. Social implications The study provides system-level insights into how leader–follower dynamics influence margin setting among Indonesia’s conventional banks. While the model captures strategic interactions between large and small banks, it does not account for within-group heterogeneity or alternative game structures such as static or complete-information settings. The sample, covering 20 conventional banks that represent about 85% of system assets, excludes Islamic institutions; hence, optimal margins apply only to conventional banking. The pandemic dummy captures major disruptions but may also reflect concurrent policy and liquidity shocks. Future research should incorporate bank-specific risk management indicators, dynamic regulatory variables, and uncertainty measures to refine model precision. Extending the framework to Islamic banks through profit-sharing structures would enhance its policy applicability, enabling regulators to benchmark intermediation efficiency and pricing behavior across both conventional and Sharia-compliant segments of Indonesia’s dual-banking system. Originality/value This study is among the first to extend the RADM with a dynamic Stackelberg game and apply it to Indonesia’s oligopolistic banking sector. It provides bank-group-specific optimal NIM estimates and quantifies how these align with observed margins before and during COVID-19. By explicitly modeling leader–follower dynamics and macro uncertainty, the framework offers a richer understanding of pricing efficiency than traditional RADM applications. The analysis yields actionable benchmarks for regulators and banks to diagnose misalignment and refine pricing strategies. The framework can be readily adapted to other emerging markets and to Islamic profit-sharing margins in dual-banking economies.