The direction of globalization and the integration of the financial system continues to increase in line with the trend of increasing capital flows which is the focus of discussion in this research. This study applies panel data analysis to analyze banking behavior to improve its performance. The analysis uses a linear model and a threshold model—panel data from 1991 to 2020 in 39 countries. Threshold panel regression is a non-linear model applied in this study to prove a change in the impact of independent variables that affect banking performance in specific regimes. In general, the linear model Coef.icients are as expected according to the theory. Analysis using threshold panel regression will give more profound results and higher intuition. Threshold panel regression has a smaller SSR value than the linear model. This study applies one threshold value to produce two different regimes. Changes in the impact occur at a certain threshold. Conclusively, this study finds threshold values ??for GDP growth, Concentration, Inflation, Leverage, Efficiency, and Credit Allocation. The GDP growth rate as a threshold is the most efficient model. The Coef.icients on the threshold panel regression model generally are in line with theoretical expectations. Still, some differences become the advantages of this non-linear model, revealing different economic conditions.