Bank Risk Measurement

Eric Falkenstein

April, 1999

The Objective of Credit Modeling

The primary objective of credit modeling is not to estimate a one-year mark-to-market probability distribution function (PDF) of a portfolio, but to simply estimate the true expected loss for various relevant product type and along various risk dimensions.  Not that PDFs are unimportant, just that one naturally precedes the other and is much more amenable to discussion and testing.  This assertion is in contrast to the findings of the Fed Task Force on Internal Credit Models, as well as many thoughtful commentators on this matter.  This revelation  comes from a year in the business actually trying to estimate capital used in pricing and internal corporate performance evaluation.  Instead of the PDF priority, the most important objective of credit modeling is to calibrate and validate a models that relate a measure of risk with future net losses.  The extent a risk measure can reliably create different groupings that experience different "bad" rates, and these bad rates map into cardinal measures of net charge offs, is the main payoff.  Expected loss is the number one focal point of any lending activity.  This often goes unnoticed because for publicly traded companies S&P and Moody's data exist, and this leaves expected losses somewhat less controvertible (if it's a BB+, it has an expected loss of 0.15%); however, evaluating the risks  of agency rated debt is the purview of fixed income mutual funds, not banks.

Move to next section (empirical review)

Return to Outline