A0917
Title: Spatial-temporal modeling of branch-level deposits
Authors: Shi Qi - William & Mary (United States) [presenting]
Abstract: A spatial-temporal autoregressive panel model is developed to analyze branch-level deposit dynamics and uncover localized interdependencies in the U.S. banking system. Using a dataset of commercial bank branches and annual deposit balances spanning multiple decades, it is estimated how deposit growth at one branch responds to both its own historical values and the deposit behavior of spatially proximate branches. The framework incorporates spatial lags and time lags simultaneously, allowing for the joint modeling of geographic spillovers and persistence in deposit trends. Branch-specific spatial weight matrices are constructed based on geographic proximity and bank ownership, enabling differentiation between intra-bank and inter-bank spatial effects. By comparing model specifications with and without spatial terms, the extent to which traditional models understate inter-branch dependencies is quantified. Findings demonstrate that spatial autocorrelation in deposit growth is economically and statistically significant, particularly in urban markets with dense branch networks. This methodology provides a foundation for measuring the stability of deposit funding across space and time, offering a scalable approach for studying liquidity dynamics at fine geographic resolution. The contribution is to empirical banking and spatial econometrics literature by adapting spatiotemporal modeling tools to branch-level financial data.
