Measuring Loss on Defaulted Bank Loans: A 24-Year Study

Article excerpt

In trying to understand the investment characteristics of any credit risk-sensitive instrument, one of the most important measures is expected loss, that is, the average default loss anticipated over a defined time horizon. With respect to loans to medium- and large-size corporations, banks have traditionally been pure buy-and-hold investors in what are generally floating rate instruments. Given this approach, the main concern is gauging the potential reductions in contractual cash flows paid directly by the borrower (issuer) to the holder of the loan agreement (the bank/investor). In other words, the concern is exclusively about potential defaults. As such, expected loss is clearly a critical measure.

Investors who actively rebalance their portfolios using secondary markets are concerned about price risk as well as default risk: If they sell an investment before maturity, they may sell at some price other than par. For floating rate, credit risk-sensitive instruments, price risk is primarily a function of changes in issuer credit quality and general changes in the market level of credit spreads. Even in this context, expected loss remains a critical measure.

Expected loss can be expressed as follows:

Expected Loss = Probability of Default x Loss in the Event of Default (LIED)

The probability of default represents the expectation that investment performance will become impaired over a defined time frame, such as one year. In public bond market studies, default is generally defined as a missed principal or interest payment, a bankruptcy filing, or a troubled debt restructuring. Technical covenant violations do not qualify as defaults.

LIED is an estimate of the full economic (present-value) cost if there is a default.

Bond Default Studies

A significant amount of published research has been devoted to studying bond default rates, while less research has been devoted to measuring LIED for bonds. Even less research has been published on default rates or LIED for corporate loans. Fortunately, bond default studies can be used as benchmarks in assessing the likelihood of medium and large corporate borrowers defaulting, which constitutes a major advantage of breaking down expected loss into its two component parts. However, applying bond LIED rates to loans is a dubious practice. It is reasonable to expect that bank loans and bonds could have different LIED rates, since loans have many features not found in bonds.

Furthermore, just as LIED rates vary among different classes of bonds, such as senior secured, senior unsecured, and subordinated, LIED rates vary among different classes of loans. For example, a borrower may have two bank loans, one unsecured and one secured by its customer receivables. A default by this borrower would most likely lead to a different loss experience on the two loans, but the before-the-fact likelihood of default is identical.

For lenders, understanding the potential economic impact of a loan default is at least as important as the need for corporate bond investors to know the potential recovery rate on defaulted bonds. A lender's estimation of the cost of a default affects its pricing decisions, economic capital allocation, loan loss forecasts, and valuation of the existing loan portfolio. Therefore, it stands to reason that developing good LIED statistics for loans meets an important practical need.

In this article, we present the data, methodology, and empirical results of a study of LIED performed on a historical population of Citibank's U.S. defaulted borrowers. We also discuss some of the practical conclusions regarding the interpretation and use of the results.

Our study covers two groups of loans. One loan group is composed of general commercial and industrial (C&I) corporate loans. The other group is composed of structured loans. The structured loan group has the following characteristics:

* The loans are very closely monitored - the bank directly controls the company's cash receipts and disbursements. …