On the Modeling of Prepayments for Variable Rate Institutional Loans - Ascertaining the Inference of Bank Internal Default Probabilities on Subsequent Prepayments

This paper aims to evaluate an inference of bank internal PDs (Default Probabilities) on subsequent prepayments of variable rate institutional loans. Since variable rate loans hardly present an economic motivation for early prepayments in that they would not o er a cheaper re nancing alternative, we test the conjecture of a correlation between improvements in obligors´creditworthiness (as re ected by negative changes in Bank Internal PDs) and subsequent loan prepayments, as obligors might be tempted to renegotiate more advantageous terms (lower credit spreads) with their lenders. The analysis is purported to serve as an early warning mechanism for banks pursuing the Basel III IRB (internal rating based) approach for unexpected in ows of liquidity in the near future. We use Machine Learning (ML) ensemble methods to forecast potential prepayments and perform a conditional prepayment analysis to make an inference on the prepayment amounts and the prepayment timing distributions while controlling for macroeconomic and corporate idiosyncratic characteristics.