Correlation in credit risk is a well-recognized phenomenon, and understanding the sources of correlated credit losses is crucial for many purposes, such as setting capital requirements for banks and pricing of structured credit products that are heavily exposed to correlations in credit risk (for example, the collateralized debt obligations (CDOs)). The issue becomes particularly important given the rapid growth of structured credit products in the financial markets in recent years. In spite of much research on this subject, researchers still do not understand many aspects of the correlation in credit risk, and this paper aims to move this literature forward.
The first question we explore in this paper is how much of the correlation in credit risk is driven by observable factors? This is an open empirical question. Most credit models are based on the doubly stochastic assumption that, conditional on observable risk factors, defaults are independent. This assumption is widely accepted and implemented in the banking practice when determining capital requirements. However, this assumption is challenged by Das, Duffie, Kapadia, and Saita (2007), and their findings are echoed by Duffie, Eckner, Horel, and Saita (2008). Further, several studies have documented that “contagion” has prominent impact on the credit risk of other firms (i.e., Jorion and Zhang (2007)), raising the possibility that “contagion” may play a more important role in the correlation in credit risk than people have realized. Based on this empirical phenomenon, some researchers have tried to model “contagion” in credit models (e.g., Giesecke (2004), Jarrow and Yu (2001), and Schönbucher and Schubert (2001)). However, it is not clear from the literature how much of the correlation is due to observable factors and how much is due to unobservable factors, such as “contagion,” and whether “contagion” is a general phenomenon in the economy.