Estimating and Pricing Credit Risk: An Overview

In the past five years, many sophisticated models for pricing credit risk have been developed. The rapid progress in this area is primarily a result of the growth of credit derivatives, securitized asset pools, and other structured products. Factors such as regulatory concerns and the availability of empirical data on default, rating changes, and asset recovery have also sparked interest in credit-risk-pricing models. This article reviews the development of modeling and pricing credit risk during the last three decades. It starts with a discussion of the statistical properties of credit spread behavior over time. It then reviews various quantitative models for assessing a company's creditworthiness and default probabilities. Next, it focuses on the ultimate objective of credit-risk assessment—how credit risk should be priced. First, the basic building blocks of a credit-risk-valuation model—interest rates, default/rating migration, and recovery rates—are discussed. Then, based on these building blocks, the article compares two primary credit-risk-pricing approaches—the structural (firm-value) and the reduced-form models—and reviews several other simple but popular pricing approaches. The article concludes with a brief discussion of credit-risk-pricing applications and possible future research directions. Traditionally, fundamental credit research is the main method practitioners have used to assess a company's credit risk. The pricing of credit risk, however, has been somewhat disconnected from such practitioner credit research; instead, it has relied mostly on trending relationships or anecdotal evidence about credit spreads. Estimating and pricing credit risk via quantitative methods has been largely ignored. In the past five years, however, U.S. market participants have witnessed several credit/liquidity crises, the rapid development of credit derivatives and securitized assets, the increasing availability of credit data, and increasing regulatory scrutiny. Against this backdrop, investors and academic researchers have shown a renewed interest in credit-risk pricing and have developed several sophisticated quantitative models.To construct a sensible and practical credit-risk-pricing model, one must understand the properties of credit-risk behavior over time. I examine several significant factors underlying changes in credit spread properties: the level and slope of the U.S. Treasury yield curve, interest rate option volatility, LIBOR, swap spreads, equity returns, and equity return volatility. Multivariate relationships seem to describe the dynamic nature of credit-risk changes better than a single factor.Quantitative credit-risk models are often used to estimate a company's statistical likelihood of bankruptcy or financial distress over time. These models use various statistical techniques, such as multiple discriminant analysis, survival analysis, neural networks, and recursive partitioning. Input variables are normally derived from balance sheets, income statements, equity information, and macroeconomic data. Quantitative credit-risk models as complements to traditional fundamental credit research have been empirically proven to effectively assess a company's creditworthiness in the United States and abroad.Recently developed credit-risk-pricing methodologies can be categorized as “structural” or “reduced form.” In the structural (or firm-value) approach, debt is treated as a contingent claim on firm value. Default risk estimation and credit pricing rely on examining the relationship between firm value and debt value over time. The reduced-form approach bypasses company valuation and extracts the value of credit risk directly from market information (e.g., credit spreads and rating transitions). For each approach, the models can be compared on the basis of each of the following three modeling components: interest rate process, default process, and asset-recovery process. Variations of the models can be viewed according to how the default risk is determined, how default intensity and the recovery rate are defined, how the pricing process is presented, and how the elements of the process are implemented.Following the model descriptions, I review the relevant published empirical results and provide comments about the strengths and weaknesses of the model variations. In addition, I briefly discuss other approaches popular with practitioners.I conclude that the recently developed credit-risk-pricing models still find their primary applications in valuing credit derivatives; they remain theoretical as far as other applications are concerned. To extend these modeling approaches to pricing a broader array of instruments, as well as to gain wider acceptance of the models by practitioners, more empirical studies are needed. Increased empirical research would require a database, however, with detailed and frequently updated information about individual corporate issues and the issuers' capital structures, which is not easily obtainable. Moreover, a credit-risk-pricing model that would provide practitioners with an intuitively appealing solution would have to link the basic characteristics of an issuer and debenture to the model's parameters.