Liquidity and Capital Thresholds as Buffers against Macroeconomic Uncertainty: Evidence from Panel Threshold and Causal Machine Learning in ASEAN

Abstract

This study examines the nonlinear and heterogeneous effects of macroeconomic uncertainty on bank stability in ASEAN emerging markets, with particular emphasis on the moderating roles of capital and liquidity buffers. Using panel data of 62 banks over 2010‒2023, the analysis integrates panel threshold regression (PTR) with double machine learning (DML) and causal forests. PTR results reveal a statistically significant capital adequacy threshold of approximately 8%, above which the adverse effect of uncertainty is significantly attenuated, while liquidity shows consistent mitigating effects in machine learning (ML) estimates but lacks robust threshold evidence. In contrast, ML evidence uncovers conditional heterogeneity: higher liquidity systematically reduces the destabilising impact of uncertainty, whereas elevated capital ratios in some cases amplify vulnerability, reflecting structural frictions and risk taking incentives in emerging markets. Robustness checks across alternative stability measures, subsamples (pre- vs. post-COVID, small vs. large banks), and uncertainty regimes confirm that uncertainty shocks are most damaging during high-uncertainty episodes and for smaller banks. The findings highlight that while both buffers matter, capital provides strong protection only once a critical threshold is reached, whereas liquidity consistently supports resilience. These insights underscore the need for tailored macroprudential strategies that integrate buffer design with institutional quality.