The purpose of this paper is to present a comprehensive framework for assisting lending banks in their current expected credit losses (CECL) forthcoming computations.
The bottom-up approach requires multiple steps including the spline method for identifying optimal segments in the lifetimes of loans, Poisson regressions for evaluating the explanatory variables and hazard rate probes for gaining inferences toward the expected credit losses and their projected schedule.
The CECL paradigm has both advantages and disadvantages, as discussed hereafter.
The model is practical, accurate in the sense that provisions are properly and timely allocated, it can be programmed and it relies on merely a few mild assumptions, thus it can be conveniently calibrated to fit broad macroeconomic scenarios.
This study provides background on the subject, motivate each module, construct the advised model, assemble a pseudo-database, demonstrate the functionality of the procedures and further draw conclusions on the effectiveness of the current strategy.
