Is it time to re-think how we model credit risk?

Timing matters

Despite occasional downgrades, investment-grade bonds have shown remarkable stability over recent decades. Analysing S&P data, default rates remained notably low during certain crises. 

Over a period of 25 years (1998-2022), analysis of S&P data reveals that twelve out of 25 years exhibited a 0% default rate in investment-grade bonds. Notably, the highest annual default rate of 0.42% was recorded in both 2002, following the dot-com bubble burst, and 2008, amidst the global financial crash. Focusing on the most recent ten-year period showcases a significant upsurge, with 75% of years exhibiting a 0% default rate in investment-grade bonds. Whether this is attributable to volatility, or a changing economic dynamic remains to be proven.

What is apparent in the data is that choosing the right time frame when assessing default risk is important, as it can have a significant impact on credit risk capital. Even when considering the leading economic scenario generators (ESGs), it is paramount to consider timing, time horizon and relevance of the chosen indices for the specific use. 

Given the (relatively) low level of default for investment-grade bonds, it should be noted that the risk of a downgrade is principally what impacts a firm’s portfolio value rather than the risk of an outright default. Whether to contemplate similarly for government bonds is an interesting thought, albeit beyond the scope of this article. 

Is the traditional definition of a default still relevant?  

Recent evidence from Chinese bonds suggest that the classical definition of a default event can and has evolved. In some cases, defaults are overlooked as a timing issue, with a commensurate response from investors. This raises some interesting questions - for example, how should the capital model treat situations where missed interest payments are rectified afterwards? Should the capital model incorporate a transition state between defaulted and not defaulted positions? Should collateral be allowed for in the capital model, and if so to what degree?

Diversification as a mitigant

The historical data provides useful insights when considering the likelihood of at least one default, but it also highlights that the number of investment grade defaults remain low even in a crisis hit year. Considering the risk of concentration, a diversified portfolio could expect to mitigate the potential financial loss should some investments fail to uphold their financial obligations. In this sense, the problem of credit default on bonds could be considered a numbers game, to which the risk can be mitigated by design. 

Conversely, it is worth noting that there is a limit to how much credit default risk can be diversified. Some elements of default experience will be correlated across bonds, so even if you were to hold an enormous number of bonds, the portfolio default experience will inevitably still vary over time.

What it means for capital models

This solvency article raises some interesting capital modelling design questions:

1. Should you model each individual bond or investment, or group them by investment type, currency, duration and/or credit rating? 
2. To what level of complexity should the model allow for correlations between the underlying investments defaulting, in both a normal and stressed environment? 
3. In the event of a credit default, is the loss amount known or would there exist a range of potential credit default severities? Would this be similar for all the underlying investments, or differ based on the underlying characteristics of the different investments?
4. Should you model separately the impact of credit spread changes, credit rating transitions and credit defaults separately, or as a single total return?
5. Do you need a stronger tie in between liquidity risk with credit risk?

Regardless of your design choices, it is important that the business ensure that controls and feedback loops are in place to appropriately understand, monitor and communicate changes in the landscape as they may arise. 

It can be tough to justify when a complex natural catastrophe has not been fully captured by the capital model, so it might be even more difficult to justify if the model has not fully captured a simple credit default on bonds.