What is credit risk?
Credit risk has four broad elements:
1. Probability of default;
2. Market risk and return on capital;
3. Concentration risk; and
4. Correlation risk
Probability of default: When a portfolio holds credit instruments it is exposed to the risk of losses from defaults and from credit deterioration, which is otherwise known as downgrade risk. Rating agencies’ credit ratings are summary measures of credit quality and encapsulate both an estimate of the probability of default, and an assessment of the stability of that rating.
A common criticism of credit ratings from agencies is that they may not be accurate and that they should not be relied on in isolation. However, we believe that it is good practice to use a measure from an independent source. This is not to argue that that a credit rating from Moody’s or S&P, is better than our own, but rather to say that consideration of credit risk should be separated from the investment decision. The portfolio holds a credit instrument because, presumably, it is believed that the pricing of the instrument compensates for the risks. In this context, the independent source’s credit view is a challenge to the views of the investment manager. We can therefore use the credit rating to determine a measure of the probability of default. The average probability of default for the portfolio is a measure of the percentage of the portfolio expected to suffer a default. Thus, a portfolio consisting entirely of BBB Non-Financial issuers should expect 0.19% of the assets to be subject to default each year (See Table 1). This number is determined only by the credit rating and is not affected by the number of instruments held – it cannot be reduced by diversification. Also, as it measures the probability of default in a single year, it is independent of the average maturity and duration of the credit exposure. Finally, it is not a measure of market risk as it is unaffected by the level or changes in credit spreads.
Market risk and return on capital: An alternative measure of credit risk is provided by the capital charge under the solvency II spread risk calculation. The intention of this calculation is to estimate the losses suffered by the portfolio if spreads were to widen by an amount expected to happen only once every 200 years; it is sensitive both to the rating and duration of the bond. The capital charge is therefore a measure of credit risk.
Conventionally, one expects to earn a return on capital and so the capital charge provides a means of assessing whether one is sufficiently compensated for taking credit risk. This will depend on:
If market risk is a concern, then the capital charge is a suitable measure of credit risk. The capital charge for spread risk is much reduced under the assumptions used in the matching adjustment, but nevertheless it exists and can be used to set limits on credit risk appetite. Finally, the amount of credit risk that one wishes to take may be influenced by the expected return. Under Solvency II the yield on an instrument is reduced on the basis of a formula that takes in an average of long run spreads and losses from defaults and downgrades. This is the impaired spread to which we refer in the Non-linked asset reports. We can use the impaired spread as a measure of the expected return in credit. In this case, the amount of credit risk capital can respond to the level of spreads, with exposure being reduced if the expected return from credit falls below requirements.
Concentration Risk: Although a portfolio consisting entirely of BBB Non-Financial issuers should expect 0.19% of the assets to be subject to default each year (See Table 1), the actual losses in any one year will be very different. Indeed, the experience is generally of several years of no losses from defaults, and then a loss in a single year that is a multiple of the expected annual average. This situation is analogous to the experience of throwing a dice and hoping not to get a six. On each throw there is a 1 in 6 chance of getting a six, but the long run pattern of losses is that we have five throws with no 6 followed by one throw with a 6. The average default loss disguises the true pattern of defaults. This experience motivates the discussion of concentration risk.
If we have a portfolio of 100 BBB Non-financial bonds, then we should expect no defaults in at least 8 years in 10 and precisely 1 default in no more than 2 years in 10. This calculation is based on Table 1 and assumes that defaults are uncorrelated. If instead, as history suggests, defaults are bunched, then, somewhat counter-intuitively, we should expect years with a single default to be less common and years with either no defaults or two or more to be more common. While we may be happy to accept losses of 0.19% per anum, losses of 1% once every five years may be unacceptable. This effect can be mitigated by increasing the number of holdings or setting maximum holding limits based on the credit rating of the bond. To illustrate, if we have a portfolio of 200 BBB non-financial bonds, then we are less likely to have a year without defaults and more likely to have at least one default, but the losses from that default are less because the holding size is smaller. This makes a large loss less likely, but increases the likelihood of suffering a smaller loss: the average loss from default is unchanged.
Correlation Risk: Default and downgrade risk can arise from the choices made by, or failing to be made by the issuer, but they can also arise from shocks to economies, or economic or market sectors; these are so-called factor exposures. These last can give rise to a bunching of defaults, or, in the jargon, correlated defaults. In order to reduce the risk of exposure to such an event, it is good practice to ensure that issuers are diversified by the economies and sectors to which they are exposed and by their ratings.
Incorporating into guidelines
One approach to controlling credit risk in investment guidelines is discussed below.
Probability of default: Define the one-year probability of default to be that provided by EIOPA for the calculation of the Fundamental Spread. These are shown in Table 1 and are applied to the assets based on their sector and credit rating. The guideline could require that the overall portfolio’s annual probability of default will be no more than for example 15bp, which is roughly equivalent to a 1 in 200 chance of a loss of GBP 500,000 on a GBP100m portfolio.
Market risk and return on capital: We can use the Solvency II definition to calculate the capital requirements for credit bonds in the portfolio. We then assume that the spread, after impairments, over gilts is the return on investment in the credit bonds. The guidelines could require, for example, that the capital charges applicable to the portfolio cannot exceed 10% and the return on capital is maintained at 5%.
Concentration Risk: Seek to control concentration risk by using a method borrowed from Solvency II. A threshold is set for the holding size of bonds based on their credit rating, there is then a notional charge that is proportional to the amount of the portfolio in excess of these thresholds. The total charge attaching to issuer exposures in excess of the specified thresholds is required to be no more than GBP 50,000 for example.
This guideline interacts with other guidelines in the portfolio. For example, in isolation this guideline would permit a portfolio of 67 BBB bonds, but as a portfolio consisting entirely of BBB bonds would have a capital charge of approximately 20%, which is materially in excess of the 10% ceiling, it is not possible in practice. This example highlights the difficulty of focusing on isolated constraints and is one reason why stress tests can complement guideline limits and help refine risk appetite statements.
Correlation risk: We could seek to control this risk by limiting the duration and default risk contributions from any sector other than Treasuries to 40% for example. Together these measures encourage a diversity of exposures by sector and rating.