Abstract Book of the 14th International Conference on New Ideas in Management, Economics, and Accounting
Year: 2025
DOI:
[PDF]
Early Warning Systems for Banking Crises: Does Regulatory and Institutional Quality Matter?
Laura Muncey
ABSTRACT:
This thesis proposes and estimates an early warning system (EWS) model for banking crises, with particular focus on the role of governance quality, banking regulation, and market power. A sample of 130 countries over a 25-year period (1980-2012) was analysed. The EWS consists of a baseline model followed by an extended model to ensure the incorporation of all relevant variables.
The baseline EWS model used is a hazard model that considers the time to one-yearforward and two-year-forward crisis event, modelled as a function of banking, financial, and macroeconomic indicators. Model fit statistics across six estimators (probit, logit, and cloglog applied to pooled data, and xtprobit, xtlogit, and xtcloglog that take account of the panel structure in the data) indicate that the probit and xtprobit estimators are preferable compared to other estimators. Estimation results indicate that bank deposits to gross domestic product (GDP) and GDP growth tend to reduce the probability of a banking crisis whereas the ratio of domestic credit to GDP, the ratio of broad money to total reserves, and the deposit rate tend to increase the probability of banking crises.
Results from the extended EWS model using a probit estimator show that absence of an explicit insurance scheme and existence independent supervisory authority reduce the probability of a banking crisis, while the Clientelism Index, lower levels of competition as measured by the Boone Indicator, and the Central Bank Independence Index all increase the probability of banking crises. These findings are consistent with the theoretical perspective underpinning the EWS model and provide policymakers with evidence to make changes where appropriate to deter a banking crisis.
The extended model also considers the nonlinear relationship between the variables investigated and the occurrence of banking crises in the sample countries. Findings are consistent with the literature, which suggests that there is a non-linear relationship between banking competition and the probability of a banking crisis. However, investigation as to whether Central Bank independence differs between developed and developing countries suggests that there is no difference.
This research also investigates the issue of controlling for potential endogeneity that may arise from the correlation between frailty (i.e. unobserved country heterogeneity in vulnerability to financial crises) and regressors. The findings remain robust in the face of different financial distress events for both the baseline model and the extended model.
In summation, an EWS hazard model for banking crises with the inclusion of regulation, competition, and governance quality has been developed. This model produced statistically significant results that have been verified using the Mundalk Correction and Mixed Random Effects models in combination with linearity and sensitivity robustness checks.
This thesis has incorporated several innovations in comparison to exiting literature. Controlling for duration within the hazard model, whereby this variable shows that if a country does not experience a crisis until a certain year, the probability of experiencing a crisis after that year is smaller. This is key information for guiding monetary policy. The innovation of using other categories of variables enriches the model, thus giving policymakers and other financial forecasters the ability to predict a banking crisis. These other categories incorporate banking competition, regulation and governance quality, in which no other paper has used these additional categories of variables within a hazard model along with the traditional categories. Thus, the EWS produces significant results of variables to use within the EWS.
keywords: Banking crises, Early warning systems, Hazard Model