Too big to fail? : an analysis of the Colombian banking system through compositional data

Although still incipient in economics and finance, compositional data analysis (in which relative information is more important than absolute values are) has become more relevant in statistical analysis in recent years. This article constructs a concentration index for financial/banking systems via compositional analysis to establish the potential existence of "too big to fail" financial entities. The intention is to provide an early warning tool for regulators about this type of institution, so they can define thresholds and measures depending on their risk appetite and the systems' specificities. The index has been applied to the Colombian banking system and assessed over time with a forecast to determine whether the system is becoming more concentrated. Results found that the concentration index has been decreasing in recent years and the model predicts this trend will continue. Regarding the methodology used, compositional models were shown to be more stable and to lead to better prediction of the index compared to the classical multivariate methodologies.