Valuation of Non-Performing Loans: Calibration of unsecured recovery curves



The statistical valuation of granular loan portfolios requires investors to determine the most suitable recovery curves.


In this article we discuss the pitfalls of calibrating recovery curves and propose a simple constant hazard competing risk model. A recovery curve is a model for the amount and timing of expected collection and expense cash flows used in a discounted cash flow analysis. Calibration pitfalls include potential selection biases where the available historical data are not representative for the portfolio to be valued especially when the loans sold are unresolved positions several years after default whereas the available data are based on resolved cases only. We describe some key model decisions that investors or banks must take when valuing granular portfolios of unsecured non-performing loans.

We propose a tractable competing risk Markov chain model to capture some of the observed dynamics of recovery cash flows. The model provides a simple closed formula for the investor’s expected gross cash flow expressed as a multiple of the last 12 months collections and the net present value of recovery cash. We offer several examples of recovery curves of European NPL. In Italy after 2020, we find that special loan servicers increased the weighted average life of the expected remaining unsecured recoveries by around 1.5 years. Some NPL servicers do not adjust their recovery expectation for elapsed time after default in stark contrast to the empirical evidence and theoretical prediction.

You can access the full technical paper by clicking here:

NPL Markets – Unsecured recovery curves