International Financial Reporting Standard 17 (IFRS17) requires the risk-free discount rate to have “no or only negligible credit risk.”¹ Most (if not all) South African insurers treat the government bond curve (whether internally derived or specified by the Prudential Authority) as risk-free. Is this assumption reasonable?

Sovereign default can be a significant risk, as highlighted in the book This Time Is Different.² The authors emphasise the history of frequent sovereign defaults and currency crises. Our collective tendency is to expect better outcomes in the future, despite evidence to the contrary. In a 2023 analysis,³ the World Economic Forum ranks sovereign default as a major risk for emerging market and developing economies.

This short paper will look at a few different ways of estimating credit risk with the aim of concluding whether or not South African government bonds have only negligible credit risk.

As a quick refresher on credit risk before we delve into the next sections, credit risk arises in three ways:

• Default risk: The failure of a borrower to meet their contractual obligations, such as repaying principal or interest.
• Credit migration risk: The upgrade or downgrade of a borrower's creditworthiness, causing a shift in the credit spread and the instrument's value.
• Spread risk: Fluctuations in the difference between the yield of a debt instrument and a risk-free benchmark, affecting the instrument's market value.

Default risk implied from historical default frequency

South African government bonds have a BB/BB- rating by S&P⁴ (local currency / foreign currency as of 2024), which is considered "sub-investment grade" or "junk." Fitch rates South Africa as BB- even for local currency denominated instruments,⁵ Historical default experiences of BB rated and other sub-investment grade government bonds can provide insights into the credit risk associated with these bonds.

Several countries, including Argentina, Belarus, Ecuador, Ghana, Lebanon, Sri Lanka, Venezuela and Zambia have all defaulted in last five years. Greece spent much of the 20th century in default and missed a payment to the International Monetary Fund (IMF) in 2015 again.

Most of these countries had credit ratings not very different from South Africa’s current BB.

Sources^6,7 suggest that, on average across the last 30 to 40 years, 1-year default rates for BB rated government bonds have been around 0.4%. On average, 5-year default rates are between 2% and 5% depending on the source. For BB rated government bonds over the last 50 years, the 5-year rate of default has varied, including up to 9% at times.

By definition, bonds rated below BBB are considered “junk” or more politely “sub-investment grade” or “speculative” given their default risk. Historical experience has shown that this can be significant and would not be a “negligible” consideration for an outside investor looking at South African bonds with an impartial eye.

In contrast, AAA and AA rated government bonds typically experience a 0% default rate in the short term, as shown in a recent S&P’s global default and rating transition study.⁸

To add some further background, under the Solvency II Standard Formula capital rules applicable in the European Union, sovereign debt with a AAA or AA rating has no capital charge, whilst all other ratings have a charge equivalent to a corporate bond of a letter better rating (e.g., a BB rated sovereign debt enjoys the same capital charge as a BBB rated corporate bond).⁹

Credit migration risk from historical experience

Credit migration risk, the potential for a borrower's creditworthiness to be upgraded or downgraded, can significantly impact the value of debt instruments. This risk is particularly relevant for government bonds, where changes in sovereign credit ratings can lead to substantial shifts in yield spreads and market prices.

South Africa's credit rating history

South Africa's credit rating history provides a clear example of credit migration risk in action. Over the past two decades, the country has experienced a series of downgrades:

• 2000: BBB- (investment grade)
• 2003: BBB (upgrade)
• 2005: BBB+ (upgrade)
• 2012: BBB (downgrade)
• 2014: BBB- (downgrade)
• 2017: BB+ then to BB (downgrade to sub-investment grade, two downgrades within the year)
• 2020: BB- (further downgrade)

This progression illustrates how a country can transition from investment grade to sub-investment grade ("junk") status over time, with each step potentially impacting the cost of borrowing and investor perception.

Spreads vary by credit rating

Credit rating changes can have a material impact on bond spreads. Figure 2 is based on the analysis by Damordan¹⁰ looking at traded bonds in the US. The spreads will vary over time, and between corporates and sovereign exposures. The chart in Figure 2 just highlights the significant differences between typical spreads for different credit ratings. Not all ratings were available in the source (e.g., BB- is missing).

Credit ratings as predictors of spreads

The relationship between credit ratings and spreads is complex (and it usually takes the form of positive correlation, usually easily confirmed by historical data analyses)

1. Leading indicator: Rating agencies generally aim to review and adjust ratings based on forward-looking assessments, which can precede market spread movements.
2. Lagging indicator: In some cases, market spreads may widen before a rating downgrade, as traders and investors react to deteriorating fundamentals more quickly than rating agencies. This is particularly true for sovereign risks where market participants typically have access to much the same information as credit rating agencies.
3. Sometimes, especially when an instrument or issuer is downgraded from investment grade to junk, the change in credit rating can have a causal impact on spreads. The mandate of certain funds or investors may exclude junk bonds. A downgrade can therefore result in significant selling from these investors, depressing prices and increasing yields and therefore spreads.

Research suggests that credit ratings are often a mixture of leading and lagging indicators, with their predictive power varying across different market conditions and time horizons. However, the announcement of rating changes, particularly downgrades, often leads to immediate market reactions, underscoring their importance in credit risk assessment.

Rating transition matrices

Credit rating agencies like S&P, Moody's and Fitch publish rating transition matrices that show the probability of rating changes over various time horizons. However, the number of sovereigns is small relative to corporate bonds, and the number in each credit rating is even smaller. Therefore, these results should be understood in the context of that imperfect credibility. It is worth noting that the transition probabilities are fairly consistent with corporate bond transitions, where the exposure is much higher.

According to S&P's recent global sovereign average transition rates (1975-2023)¹¹ and Moody’s similar 2023 report¹² based on 1983 to 2022:

• A BB rated bond will remain BB with a 13% to 14% chance of being B or BBB rated a year later. The probability of moving a single notch will be higher.
• Over a five-year period, this probably rises to about 50%. I.e., it’s about as likely that a BB bond will be rated differently by a full credit step as it is that it will remain the same within five years. Given the term of many life insurance obligations, it seems quite likely that during the life of the contract a credit rating change (upwards or downwards) will take place.

These matrices demonstrate that credit migrations are not uncommon, especially for sub-investment grade issuers. South Africa has been rerated several times in recent history and these different ratings will typically result in a materially different spread, which would affect the value of assets or liabilities valued using a discount rate incorporating a credit spread.

Credit risk implied from credit default swaps

Another way to evaluate credit risk is to consider implied risks using available market information on the credit risk of South African government bonds.

Introduction to credit default swaps

A credit default swap (CDS) insures the risk of default by paying out to the purchaser on a specified credit event of the reference entity. The CDS spread is the annual cost (as a percentage of nominal) that the purchaser must pay for this protection. In the event of a covered credit event, the CDS typically pays out the difference between the face value of the bonds and their post-default market value. This excludes the risk of currency depreciation or inflation as the CDS only pays out on default.

Credit risk can be measured by the absolute level of the CDS spread and the variability in the spread over time.

While government bonds are issued across a wide range of maturities, CDS contracts are less frequently issued and have a more limited set of maturities. This analysis focuses on 5-year CDS spreads, as they are more commonly traded and provide more consistent and reliable pricing data.

Local vs. foreign currency CDS spreads and risk

CDS spreads on USD-denominated South African government bonds are available, but ZAR-denominated bond CDS spreads are not readily available. They might be expected to be cheaper, however, given the better credit rating and lower subjective assessment of risk of default.

It is often casually stated that a South African government default on ZAR-denominated debt would be far less likely to occur than on USD-denominated debt, given:

• The ability to use inflation as an alternative to outright default for ZAR denominated debt
• Selective default is common, and USD-denominated debt is much smaller than ZAR-denominated debt
• The more disruptive impact of defaulting on ZAR debt for the local economy and financial services sector

This logic is sound in general, but the Bank of England has also stated that “Sovereign defaults on local currency debt are more common than sometimes assumed. Since 1960, 36 sovereigns have defaulted on local currency debt.”¹³ South Africa’s local currency credit rating is only one modifier notch above the foreign currency rating.

Levels and variability of CDS spreads on government bonds

The 5-year CDS spreads on South African USD-denominated government bonds are currently (12 September 2024) around 185 basis points (bps).¹⁴

For comparison, Türkiye rated at B+ is around 280 bps.⁵ Türkiye has different problems from South Africa, including out-of-control inflation and monetary policy, but its economy is about twice the size and has experienced higher economic growth than South Africa.

As another comparison, Brazil is rated BB and has current 5-year CDS spreads of 157 bps.¹⁶

South Africa's credit spread has been mostly between 200 bps and 300 bps for the last decade. Thus, the cost of insuring against default is typically around 2% to 3% per year. This 100 bps difference is significant for a 5- or 10-year bond, or a portfolio of life insurance obligations with an average effective term of 10 years.

This estimated credit spread is also consistent with spreads between South Africa government bonds and US government bonds (of around 550 bps) relative to expected inflation differentials of about 2%. This simple analysis gives a high-level indication of a 3.5% spread for credit and liquidity risk, broadly consistent with a 2% to 3% range for credit risk allowance from the CDS market.

Deriving probability of default from CDS spreads

A 40% recovery rate (RR) is a standard assumption and is reasonable compared to the last 40 years.^17,18 It varies significantly based on the circumstances of the particular sovereign and the nature of the default. There are relatively few sovereign defaults (at least compared to corporate defaults) therefore estimating this accurately from past data is more difficult. Government bonds are typically thought to have better RRs than corporate bonds, although a Moody’s study using data from 1983 to 2008 shows a value-weighted RR of only 31%.¹⁹

Assuming a recovery rate of 40% in the event of default, the implied annual probability of default based on current CDS spreads is approximately 3.1%. Using the spreads from the last decade, the implied annual probability of default ranges from 3.3% to 5.0%. This prospective view reflects higher default experience than recent historical default rates.

These figures translate to a market-consistent 5-year probability of default between 16% and 23%, assuming independence. While these probabilities are likely overstated, given that they include a margin for profit, risk, or return on capital for CDS sellers, it is difficult to argue that someone paying around 2% per annum for credit protection considers the credit risk to be negligible.

Conclusion

Based on the historical default frequencies and the implied probabilities of default from CDS spreads, it seems inescapable to conclude that South African government debt does not have only negligible credit risk.

The exact quantification of the risk is difficult. Different methods and sources will derive different results. These estimates can also change rapidly as local and global economic and political conditions change. This uncertainty and volatility arguably support the conclusion that there is credit risk within a government bond-derived yield curve.

Consequently, any yield curve constructed from unadjusted government bond yields implicitly includes some allowance for nonnegligible credit risk. From a South African perspective, it is common to treat government bonds as “risk-free” on a day-to-day basis. Our solvency regulations reinforce this because no solvency capital is held against government default risks. This will be true for many other African countries and emerging markets in general.

Accepting that a risk-free yield curve requires a downwards adjustment for sovereign risk would have implications for IFRS17 liability valuations and the onerosity of insurance contracts. The impact would depend on interest rate sensitivity, the timing of in-flows and out-flows, average profitability, and the distribution of profitability across contracts.

¹ IFRS17, B79: “For cash flows of insurance contracts that do not vary based on the returns on underlying items, the discount rate reflects the yield curve in the appropriate currency for instruments that expose the holder to no or negligible credit risk, adjusted to reflect the liquidity characteristics of the group of insurance contracts.”

² Carmen M. Reinhart & Kenneth S. Rogoff (2009). This Time Is Different: Eight Centuries of Financial Folly.

³ Pallan, H. (13 October 2023). The impacts of sovereign debt defaults explained in four charts. World Economic Forum. Retrieved 3 October 2024 from https://www.weforum.org/agenda/2023/10/sovereign-debt-default-charts/.

⁴ Gupta, Z.S. et al. (17 November 2023). Research Update: South Africa ”BB-/B” Foreign Currency and ”BB/B” Local Currency Ratings Affirmed; Outlook Stable. S&P Global. Retrieved 3 October 2024 from https://www.spglobal.com/ratings/en/research/articles/231117-research-update-south-africa-bb-b-foreign-currency-and-bb-b-local-currency-ratings-affirmed-outlook-sta-12915727.

⁵ Fitch Ratings (13 September 2024). Rating Action Commentary Fitch Affirms South Africa at ”BB-“; Outlook Stable. Retrieved 3 October 2024 from https://www.fitchratings.com/research/sovereigns/fitch-affirms-south-africa-at-bb-outlook-stable-13-09-2024.

⁶ Tolstova, E. et al. (24 June 2024). Default, Transition, and Recovery: 2023 Annual Global Sovereign Default and Rating Transition Study. S&P Global. Retrieved 3 October 2024 from https://www.spglobal.com/ratings/en/research/articles/240624-default-transition-and-recovery-2023-annual-global-financial-services-default-and-rating-transition-study-13137806#:~:text=Defaults%20And%20Downgrades%20Rise%20But%20Remain%20Historically%20Low,-In%202023%2C%20eight&text=Because%20of%20the%20rise%20in,a%20similar%20number%20of%20defaults.

⁷ Moody’s (2023). Sovereign Default and Recovery Rates, 1983-2022.

⁸ Tolstova, E. et al., op cit.

⁹ See the Solvency II Delegated Regulations, Article 176: Spread Risk on Bonds and Loans, and Article 180: Specific Exposures.

¹⁰ Damordan (2024). Ratings, Interest Coverage Ratios and Default Spread. Retrieved 3 October 2024 from https://pages.stern.nyu.edu/~adamodar/New_Home_Page/datafile/ratings.html.

¹¹ Tolstova, E. et al., op cit.

¹² Moody’s (2023), op cit.

¹³ Bank of England (2022). BoC–BoE Sovereign Default Database: What’s New in 2022?

¹⁴ World Government Bonds. South Africa 5 Years CDS – Historical Data, as at 12 September 2024. Retrieved 3 October 2024 from https://www.worldgovernmentbonds.com/cds-historical-data/south-africa/5-years/.

¹⁵ World Government Bonds. Turkey 5 Years CDS – Historical Data, as at 12 September 2024. Retrieved 3 October 2024 from https://www.worldgovernmentbonds.com/cds-historical-data/turkey/5-years/.

¹⁶ World Government Bonds. Brazil 5 Years CDS – Historical Data, as at 12 September 2024. Retrieved 3 October 2024 from https://www.worldgovernmentbonds.com/cds-historical-data/brazil/5-years/.

¹⁷ Moody’s (2023), op cit.

¹⁸ Luckner, Meyer, Reinhart, & Trebesch (2003). Sovereign Debt: 200 Years of Creditor Losses.

¹⁹ Moody’s (2008). Sovereign Default and Recovery Rates, 1983-2007.