South Africa
Financial Sector Assessment Program-Stress Testing the Financial System-Technical Note

This Technical Note discusses stress testing (ST) results for the financial system of South Africa. The bank STs suggest that banks have adequate capital to withstand severe shocks, but need larger liquidity capacity to meet regulatory requirements. Even in the severe scenario in which GDP falls for three consecutive years, banks’ capital buffers seem sufficient, although the impact of a large default could be significant. Banks also appear resilient to market risks in both the trading and banking books. Some banks, however, would have difficulty meeting the Liquidity Coverage Ratio without the Committed Liquidity Facility of the South African Reserve Bank.

Abstract

This Technical Note discusses stress testing (ST) results for the financial system of South Africa. The bank STs suggest that banks have adequate capital to withstand severe shocks, but need larger liquidity capacity to meet regulatory requirements. Even in the severe scenario in which GDP falls for three consecutive years, banks’ capital buffers seem sufficient, although the impact of a large default could be significant. Banks also appear resilient to market risks in both the trading and banking books. Some banks, however, would have difficulty meeting the Liquidity Coverage Ratio without the Committed Liquidity Facility of the South African Reserve Bank.

Glossary

BU

Bottom-up (stress test)

CAR

Capital Adequacy Requirement

CAT

Catastrophic event

CCAR

Comprehensive Capital Analysis and Review

CET1

Common Equity Tier One

CLF

Committed Liquidity Facility

CVA

Credit Valuation Adjustment

D-SIB

Domestic Systemically Important Bank

FSAP

Financial Sector Assessment Program

FSB

Financial Stability Board

GDP

Gross Domestic Product

HQLA

High Quality Liquid Assets

IDR

Incremental Default Risk

IMF

International Monetary Fund

IRB

Internal Ratings-Based

JIBAR

Johannesburg Interbank Agreed Rate

LCR

Liquidity Coverage Ratio

LGD

Loss Given Default

LTV

Loan-to-Value

NBFI

Nonbank Financial Institution

NII

Net Interest Income

NRR

Negative Rand Reserves

NSFR

Net Stable Funding Ratio

ORSA

Own Risk and Solvency Assessment

OTC

Over the Counter

P&L

Profit and Loss

PD

Probability of Default

PIT

Point-in-Time

PSE

Public Sector Entities

RAM

Risk Assessment Matrix

RoE

Return on Equity

RWAs

Risk-Weighted Assets

SAM

Solvency Assessment and Management

SARB

South African Reserve Bank

STeM

Stress Test Matrix (for FSAP stress tests)

STs

Stress Tests

TD

Top-down (stress test)

TTC

Through-the-Cycle

ZAR

South African Rand

Executive Summary

The mission conducted stress tests (STs) for the banking and insurance sectors. The bank ST includes both top-down (TD) and bottom-up (BU) components, which cover 86 and 94 percent of banking assets in the system, respectively. The BU insurance STs cover the major life and non-life insurers.

The bank STs suggest that banks have adequate capital to withstand severe shocks, but need larger liquidity capacity to meet regulatory requirements. Even in the severe scenario in which GDP falls for three consecutive years, banks’ capital buffers seem sufficient, although the impact of a large default could be significant. Banks also appear resilient to market risks in both the trading and banking books. Some banks, however, would have difficulty meeting the Liquidity Coverage Ratio (LCR) without the Committed Liquidity Facility (CLF) of the South African Reserve Bank (SARB), and face even bigger challenges meeting the Net Stable Funding Ratio (NSFR) without reducing the maturity profile of assets.

The insurance ST indicates that, despite market value losses, the sector, on aggregate, remains solvent and profitable. The ST pointed to a high sensitivity of insurers to market risks, especially equity prices and the default of the largest banking counterparty. Life insurers’ solvency ratios declined sharply under the shocks, although the decline was mitigated by passing on lower investment returns to policyholders. Non-life insurers were more resilient with their more conservative investment strategies.

On a conglomerate level, financial institutions could weather the combined losses from their banking and insurance operations. The banking operations would be able to withstand the severe shocks; the recapitalization need, at ZAR 72 bn or 1.6 percent of 2018 nominal GDP at the maximum, is manageable, and the capital shortfall for the insurance companies, at 0.1 percent of nominal GDP, is even smaller. Recapitalizing the few insurance companies would be within the capacity of the corresponding conglomerates, owing to the small size of the capital required as compared to the level of aggregate capital in the group.

Table 1.

Summary of Recommendations

article image

“NT-near-term” is 1–3 years; “MT-medium-term” is 3–5 years.

Introduction

1. The financial sector in South Africa is large and sophisticated. Total financial sector assets of about 298 percent of GDP exceed those of most other emerging market economies. Commercial banks’ assets total 112 percent of GDP, and their share in total financial assets has been declining in recent years with the rapid growth of the nonbank financial sector. The insurance sector’s gross assets1 account for 67 percent of GDP.

2. The financial sector emerged relatively unscathed from the global financial crisis. Banks remained profitable during the crisis, with the return on equity of the four largest banks remaining above 20 percent. The non-performing loans declined from 6 percent in late 2009 to 3.6 percent in 2013, and the regulatory tier one ratio of 13.5 percent in 2013 also compares favorably with banks in other countries. The nonbank financial sector also continued to thrive, as seen by the record increase in assets under management. The life insurance sector is overall well capitalized and both life and nonlife companies are, on aggregate, highly profitable.

3. Nevertheless, the financial sector operates in a challenging economic environment and macroeconomic and credit risks are building up. A combination of slow growth, high unemployment, low savings, and high public investment is sustaining large current account and fiscal deficits. In June, Standard and Poor’s downgraded the sovereign foreign and local currency rating by one notch to BBB- and BBB+, respectively. Although unlikely, further rating actions could potentially threaten the inclusion of South African debt in global benchmark indices. An exclusion from the benchmark would significantly weaken the demand for sovereign debt from foreign investors, pushing funding costs notably higher and adding further strain on the fiscal outlook from higher contingent liabilities.

4. Against this background, the mission assessed risks to financial stability in four broad areas:2

  • Systemic liquidity risk: The need to finance large current account and fiscal deficits leaves financial markets potentially vulnerable to a re-pricing of risk and a sudden stop of capital inflows. The second source of potential systemic liquidity risk stems from banks’ very high dependence on short-term wholesale funding. Moreover, domestic participants active in over-the-counter (OTC) derivatives trading are exposed to counterparty credit risk and potentially to global liquidity shocks through increased trading with foreign counterparties.

  • Household and corporate indebtedness: Household debt may become a risk in an environment of rising interest rates. The expansionary fiscal policy, combined with loose monetary conditions after the global financial crisis, led to a record increase in unsecured lending to low-income households, which grew by 50 percent year-on-year in 2011. As growth slows, the already stretched repayment capacity of low-income households could be further undermined. In addition, floating-rate mortgages account for a large proportion of household debt. Under these circumstances, rising interest rates would make it difficult for households to service their debt. On the corporate side, private sector balance sheets appear to be strong, but rising debt levels leave public sector corporations vulnerable to interest rate and exchange rate shocks.

  • Concentration and interconnectedness within the financial system: The financial sector has a high degree of concentration and interconnectedness. Five large banks together account for 90.5 percent of total banking assets. As for the nonbank financial institutions (NBFIs), the top five insurance companies account for 74 percent of the market for long-term insurance and 44 percent of the market for short-term insurance; while the seven largest fund managers control 60 percent of the unit trust collective investment schemes industry. All the major banks are affiliated with insurance companies through either a holding company structure or direct ownership. In addition to shareholdings, these entities are connected through balance sheet transactions. For instance, NBFIs hold substantial amounts of assets in the largest four banks’ deposits, which expose them to counterparty risks and banks to liquidity risks.

  • Cross-border expansion of the banking sector: Claims of South African banks on other African countries, mostly in Sub-Saharan Africa and mostly through subsidiaries, have more than tripled over the last five years to 2 percent of total assets.

5. The impacts of these risks on the financial system are assessed through comprehensive stress tests (STs) of banks and insurance companies. The bank STs include both top-down (TD) and bottom-up (BU) components, covering four banks and six banks, corresponding to 86 and 94 percent of banking system assets, respectively. The BU STs cover both local and foreign operations to adequately assess the risks associated with the rapid expansion in cross-border business.3 The insurance ST includes only a BU component and covers five major life insurers with 70 percent of the sector’s assets, and four non-life insurers with 50 percent of the sector’s premiums. To address the risks identified above, a comprehensive set of risk factors are considered, including credit risks, market risks, liquidity risks, and contagion risks. In particular, one sensitivity analysis assesses the credit risks of household debt to large interest rate hikes, and a number of sensitivity analyses are introduced to assess counterparty risks due to the interconnected structure of the financial system, including the failure of a large bank for insurance companies. Insurance underwriting risks, like the impact of a catastrophic event, have been added as a sensitivity analysis for the insurance sector.

Figure 1.
Figure 1.

Testing Program under the 2014 FSAP Update

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: IMF staff.

Solvency Risk

A. Methodologies

Macro-financial Scenarios

6. The macro-financial scenarios for STs are aligned with the IMF’s Global Risk Assessment Matrix (G-RAM), and generated using a model-based approach. The G-RAM identified as major global shocks a surge in global financial market volatility and a protracted period of slower growth in advanced and emerging economies.4 The external shocks were used as inputs to the SARB’s Global Projection Model and Core Model to generate three global and domestically consistent macro-financial scenarios for South Africa and its main trading partners:5

  • Baseline scenario: modest GDP growth, persistent high unemployment level, and the gradual upward normalization of interest rates. The projections are aligned with the IMF’s economic projections for South Africa as of April 2014, which were subsequently adjusted downward in August 2014. Since the stress tests were conducted, the baseline macroeconomic scenario has become considerably worse.

  • Adverse scenario: mild decline in GDP growth because of a disorderly exit from unconventional monetary policy by the advanced economies, resulting in capital outflows, increased market volatility, and higher-than-expected increases in interest rates.

  • Severely adverse scenario: a “perfect storm,” where GDP growth falls sharply and stays in negative territory for three consecutive years due to the confluence of several negative factors including a recession in advanced economies; large capital outflow; and substantially higher domestic interest rates. In terms of the GDP path, the scenario would be equivalent to a cumulative eight standard deviation event from the baseline.6

Figure 2.
Figure 2.

Macro-financial Scenarios (Key Variables)

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: Haver and IMF staff projection.

Bank stress test

7. The BU and TD STs assess banks’ capital adequacy under the above scenarios.7 The TD ST relies on supervisory data, which span a relatively short horizon, and high level assumptions, while the BU STs, following broadly the guidance note issued by the FSAP team, are based on banks’ proprietary data, own estimated models, and expert judgment (Box 1). In terms of their scope, the BU STs cover six banks and both local and foreign operations to adequately assess the risks associated with the rapid expansion in cross-border business. Due to data limitations, the TD ST is based on the four largest banks and local operations.

Modeling Approaches and Assumptions in the TD and BU STs

Probabilities of default (PDs)

In the TD ST, panel data models for PDs are estimated for the 10 asset classes.1 The panel data model is chosen to address the issue of short time series.2

  • We perform logit transformation to the PDs and take the first difference of the new series as the dependent variables.

  • The explanatory variables for each model are chosen from the following list of variables: GDP growth rate, policy rate, 10-year government bond yield, property price growth, household debt-to-income ratio, government debt to GDP ratio, official unemployment rate, domestic credit extension, M3 money supply growth, nominal effective exchange rate, Rand/US Dollar exchange rate, commodity price in US$, stock market index, crude oil price, capital ratio and exposure growth. For each variable, the current value and up to four lags are considered.

  • The best model is selected based on the Bayesian information criteria.

In the BU STs, to project PDs, banks use a combination of modeling and expert judgment, reflecting the heterogeneities in customer base and risk appetite across banks.

Loss given default (LGD)

In the TD ST, for each asset class and each bank, the downturn LGD is computed using linear mapping functions (denoted “LGD_TTC,” the through-the-cycle (TTC) LGD as of Dec 2013 in the supervisory data):

Baseline scenario: LGD=LGD_TTC

Adverse scenario: LGD=0.08+0.92*LGD_TTC

Severe scenario: LGD=0.10 + 0.92*LGD_TTC

Note that the above gives a conservative assumption on LGDs. For instance, for the residential mortgage portfolio, the LGD under the severely adverse scenario would imply, for a loan with an LTV of 80 percent, a decline of the collateral price by 40 percent since the origination point.

In the BU STs, banks use their internal models to estimate the LGD.

Balance sheet growth

  • In both the TD and BU STs, the relative proportion of the balance sheet items and the trading positions remain the same under all scenarios.

In both tests, banks with capital ratios above the minimum common equity tier one (CET1) capital requirements increase their estimated balance sheet at an own estimated rate or at the same pace as nominal GDP, if the latter is positive.3 If the nominal GDP growth rate is negative, balance sheet growth will be capped at zero. Banks that fail to meet the Tier 1 capital requirements will keep their balance sheet even under positive nominal GDP growth in the TD ST, while some banks would deleverage in the BU STs.

Loan loss provision

  • The provision is equal to the expected loss, which is the product of PD, LGD and the amount of exposure. We assume that the expected loss is equal to the realized loss.

Net income growth

  • In the TD ST, the net income for each bank is assumed to grow at the same rate as the nominal GDP. The net income is defined as income before deducting the loan loss provision.

  • In the BU STs, to reflect the impact from banks’ asset liability structure, income projections under each scenario are disaggregated into its main components: interest income, interest expenses, trading income, and operating expenses. Both pre-tax and after-tax projections are reported.

Risk weighted assets (RWAs)

  • In the TD approach, the changes in RWAs are calculated following an approximate Basel II internal ratings-based (IRB) approach. The IRB formula is used to determine regulatory RWAs for the loan book and their changes over the stress horizon. For calculating the regulatory RWAs, the point-in-time (PIT) PD and the above downturn LGDs are assumed and the exposures are net of provisions. Percentage changes in the regulatory RWAs from the base year (i.e., year 2013) are applied to the base year realized RWAs4 to generate the final RWAs.

  • In the BU STs, banks follow the regulatory approach to formulate RWAs with the one exception that PIT PDs, rather than TTC PDs, are used in the exercise. Using TTC PDs would be consistent with the regulatory requirement, which aims to avoid the procyclicality issue. However, PIT PDs are preferred in STs since they fully incorporate the unexpected loss, measured by RWAs, which could materialize under stress conditions. Nevertheless, some banks noted that their system does not have the capacity to generate PIT equivalent results for certain loan portfolios.

Dividend

  • The dividend payout depends on a bank’s buffer over the minimum capital ratio, following the maximum payout ratio in the table below. In the BU STs, banks pay out less than in the maximum, which is assumed in the TD ST.

article image
Source: IMF Staff.

Hurdle rates

  • The minimum capital ratio consists of Basel III minimum CET1 requirements, and Pillar 2A and Pillar 2B charges. It reflects the phase-in of the Domestic Systemically Important Bank (D-SIB) charge and the capital conservation buffer.5

article image
Source: SARB D5/2013.

Capital formulation

The amount of capital in each year equals the amount of capital from the previous year plus the net income in the current year, and is net of the dividend payout and the loan loss provisions.

1 The 10 asset classes include: total corporate, public sector, sovereign and central bank, local government, banks, securities, residential mortgages, retail revolving, retail SME, and retail others.2 Data on PDs for each asset class are available from January 2010. The monthly supervisory data are converted into quarterly frequency using an equally weighted average, which yields 16 data points from Q1 2010 to Q4 2013. Ideally, the model estimation should use data covering at least one business cycle, but this is not feasible partly because the definition of credit risk measures varied over time following the changes in regulations and accounting rules following BASEL best standard. The same data issue applies for other variables in the supervisory database.3 In particular, in the TD ST, it is assumed that all banks grow the balance sheet at the same rate as the nominal GDP.4 The realized RWAs include RWAs for credit risk, market risk, and operational risk. They are based on the total assets rather than on the loan book.5 The Pillar 2B charge and D-SIB charge are bank-specific and confidential information.

8. To assess market risks in the scenario-based analysis, a comprehensive set of instantaneous market shocks8 were added to the second year of the severely adverse scenario, following the US Comprehensive Capital Analysis and Review (CCAR) methodology. Trading positions, including those on interest rate, exchange rate, equity, commodity and credit-risk sensitive instruments, were evaluated to assess banks’ potential trading losses under the stressed market condition. The same valuation process was applied to financial assets classified as “available for sale” and “designated at fair value through profit and loss.” The shocks involve large and sudden changes in levels, spreads, and volatilities, which are calibrated to be consistent with those observed in 2008.

9. The scenario analyses for the banking sector are complemented by several sensitivity tests.9 Among them, failures by large financial and non-financial borrowers are introduced to study concentration risks and counterparty risks that could be significant in the highly concentrated and interconnected financial system. The sensitivity of the credit quality of the mortgage portfolio and unsecured lending to large interest rate hikes is also assessed. Finally, sensitivity to market risks (e.g., Prime-JIBAR (Johannesburg Interbank Agreed Rate) basis risk and widened credit spreads) on both the trading book and banking book is examined.

Insurance stress test

10. The main focus in the insurance STs was on market risks, while counterparty risks and specific insurance underwriting risks were also tested. A BU solvency ST was performed for both life and non-life insurers, based on a solvency ST that the Financial Services Board (FSB) conducts on a semiannual basis as part of its regular prudential supervision. South Africa will implement a new solvency regime (Solvency Assessment and Management, SAM) by 2016, and a quantitative impact study was ongoing in parallel to the ST exercise; therefore, a comprehensive stress testing approach based on the forthcoming regime was not seen as practical at this stage. The FSB ST includes single-factor shocks, as well as a combined economic scenario that addresses interest rate, equity, property, exchange rate, and (corporate) credit risks (see Box 2).

11. Overall, the stress level in the solvency test was higher for insurers than under the macro-financial scenario used for the banking ST. This results from the inherent conservativeness of the FSB’s own ST specification and the addition of further shocks to the FSAP ST (for more details see Appendix IV):

  • An increase in sovereign spreads, reaching 262 basis points for South African government bonds;

  • The default of the largest banking counterparty where the LGD on all contractual obligations was assumed to be 45 percent of the gross exposure, except for equity exposures for which a price decline of 100 percent was used instead of the standard 50 percent shock applied for other equity exposures;

  • As a separate single-factor sensitivity analysis: a catastrophic event of a magnitude that is expected to occur once in 100 years; following the event, the default of the largest reinsurer is assumed (with an LGD of 45 percent).

Insurance Stress Testing by the FSB

Following the global financial crisis, stress testing requirements were developed for the insurance industry. Since 2010, the major insurance companies in the South African market are required to submit stress testing results in respect of market risk on a semi-annual basis as part of the prescribed statutory returns. In addition, all insurers must submit stress testing results on both market and underwriting risks on an annual basis.

Market risks are covered in the form of single-factor shocks, and companies report the sensitivity to each of the shocks assuming them to occur instantaneously. In addition, a combined economic scenario is tested in which various asset classes are stressed simultaneously as specified below:

article image
Source: FSB.

Notes: 1/ The economic scenario stress performed by the FSB includes a decline in interest rates. Compared to an increase in interest rates, this is the more severe stress for life insurers in South Africa as the FSB sensitivity analysis as of end-2013 shows. While the increase by 50 percent results in the average CAR coverage ratio going up by 29 percentage points, a decline in interest rates by 35 percent results in a 37 percentage point drop in the ratio. To be more consistent with the macrofinancial scenario used for the banking ST, the FSAP insurance ST used an upward shock for interest rates (for details see Appendix IV). 2/ The table depicts only the specification of the FSB stress test for life insurers; the specification for non-life insurers differs slightly in some aspects.

Under the forthcoming SAM regime, the current STs are planned to be phased out. The calculation of the standard formula for the solvency requirement which is built on sensitivities to a broad set of shocks will then inform about the risk profile of an insurer. Stress testing will, however, continue to feature in two ways: Firstly, the own risk and solvency assessment (ORSA) will require insurers to conduct a number of STs, including reverse STs. These will provide insights into concentrations and stability risks. Secondly, it is proposed that a supervisory committee be established under the new Twin Peaks regime to periodically review industry-level statistics and consider what, if any, specific macro-prudential STs may be required. These will be conducted as-and-when deemed necessary, and will consider the emerging risk universe as potential stressors.

12. The statutory values of assets and liabilities, as calculated according to the South African insurance solvency regime, formed the basis of the ST calculations. Companies reported on a solo basis, and the impact of the stress scenarios on the capital adequacy requirement (CAR) was calculated based on the statutory valuation10 and the solvency regime in place at the reference date. To assess the medium-term ability of insurance companies to recover from the stress scenario, which was modeled to occur instantaneously, five-year projections were also requested. These had to include asset returns below historic averages and the very strict assumption that no new business is underwritten, which results in gradually declining premium income.

B. Results for the Banking Sector

Scenario Analyses

13. The STs confirm that banks’ capital levels are adequate to withstand credit losses in the adverse and severe scenarios. Resilience is partly due to the high capital buffers in the banking system. As of 2013Q4, the average common equity tier one (CET1) ratio of the six banks was about 11.8 percent of RWAs,11 well above the regulatory minimum of 6.74 percent.12

14. Banks’ capital positions remain resilient in the BU STs (Figures 3 and 5), since the net income will not be much affected in the stress scenarios, offsetting credit losses. Higher interest income from higher interest rates offsets the declines in non-interest revenue (down 7 percent from the baseline owing to lower business volume). As a result, gross operating income is only 4 percent lower than in the baseline scenario, serving as a formidable line of defense against credit losses, which are on aggregate 40 percent higher than in the baseline. Compared to the starting point of end-2013, the system-wide CET1 ratio only declines by one percentage point by year three, even though the average annual credit loss in the severely adverse scenario increases by 80 percent.

Figure 3.
Figure 3.

Bank—Scenario-based Stress Test Results

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: SARB supervisory data, bank reported results and IMF staff calculations.Note: the starting point capital ratio is different because the BU STs apply to both local operations and foreign subsidiaries, while the TD ST applies to the local banks only.
Figure 4.
Figure 4.

Bank—Top Down Stress Test System-wide Capital Shortfall

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: SARB supervisory data and IMF staff calculations.
Figure 5.
Figure 5.

TD ST Capital Ratio Distribution

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: SARB supervisory data and IMF staff calculations.

15. In the TD ST, the shocks have a larger negative impact on capital ratios, but the capital shortfall remains manageable (Figures 3, 4, and 5). The system-wide CET1 ratio declines from 11.2 percent to 7.0 percent in year three in the severely adverse scenario. Extending the test to year five reduces the CET1 ratio by another 100 basis points. Three of the four banks would have a capital shortfall with respect to their minimum CET1 requirements starting from 2016, while all banks would fall below their minimum capital requirements in 2017 and 2018. However, the maximum capital shortfall, at 1.6 percent of 2018 nominal GDP, is manageable, reflecting a high initial capitalization level. The decline is driven by a substantial deterioration in the credit quality of the loan portfolios, reflected in higher probabilities of default and loss given default across all asset classes. In particular, the PD for the performing floating-rate mortgage portfolio rises to a peak of 12 percent from 3.4 percent, partly the result of a sudden and large jump in interest payments. For the retail revolving credit portfolio, the performing PD rises almost four times to 19 percent from 4.4 percent.13 Credit losses exceed the net income in 2017 and 2018, resulting in a total loss equivalent to 4.4 percent of post-shock CET1 capital. Finally, higher PDs drive RWAs to grow by 16.8 percent per year on average, far exceeding the 2.5 percent nominal average growth rate of the entire portfolio.

16. The TD ST uses more stringent assumptions than the BU STs, which generate higher net income; lower PDs, LGD, and credit losses; and smaller RWAs.

  • In the BU tests, banks benefit from the higher interest rates in the severely adverse scenario, due to the positive, net rate-sensitive asset positions.14 This effect is not considered in the TD test, where net income grows with nominal GDP.

  • Moreover, while the TD exercises mostly rely on a statistical approach15 to project credit quality, banks use a combination of modeling and expert judgment and report smaller PDs.16 In addition, the TD ST uses more stringent assumptions on LGDs. Overall, the credit quality of loans deteriorates much more significantly in the TD ST than in the BU STs.

  • Another difference comes from the modeling of RWAs. In the BU STs, some banks use TTC PDs, which are lower than PIT PDs during periods of stress, to project RWAs for certain portfolios. This generates lower RWAs than those in the TD tests, which use PIT PDs to formulate capital requirements for all portfolios.17

17. Banks’ trading and fair valued banking books18 appear to be resilient to global market shocks, which are part of the macro-financial scenarios, and any losses are within group risk limits (Table 2). The largest losses come from interest rate directional movements and widening credit spreads. Higher interest rates and larger credit spreads reduce the mark-to-market value of sovereign and corporate bond portfolios in the trading book, which are partly held to meet liquidity needs. Other large losses are due to credit value adjustment (CVA), the market value of counterparty default risk, and incremental default risk (IDR), a capital charge to trading book exposures to credit-risk related and often illiquid products whose risk is not reflected in Value at Risk.

Table 2.

Results of Global Market Shocks

article image
Sources: SARB supervisory data and IMF staff calculations.

BU Sensitivity Analyses

19. Sensitivity tests indicate that the default of large exposures could impact banks substantially. Total exposure to the five largest non-financial borrowers was around 35 percent of total CET1 capital. A default by all the five borrowers would lower the CET1 ratio by 1 percent. The aggregate exposure to the five largest financial borrowers was larger still, amounting to 50 percent of CET1 capital. The CET1 ratio would decline by 1.9 percent due to a jump in default losses owing to the failure of all the five borrowers (Table 3).19 In particular, one bank would lose almost all of its equity if the shocks were realized. The banks’ staffs, however, noted that the shock is extreme since these were exposures to top tier entities and highly regulated banks in South Africa and abroad, and they were partly held to meet liquidity needs.

Table 3.

Sensitivity Analyses—Large Exposure Impacts

article image
Sources: SARB supervisory data and IMF staff calculations.

20. BU STs indicate that losses from unsecured credit exposures due to higher interest rates have a negligible impact on capital adequacy. Interest rate increases of 100 bps and 500 bps correspond to expected losses of 9.0 and 9.9 percent of total capital, respectively.20 Considering the level of provisions, these losses translate into reductions of 1.8 and 2.2 percent of total capital. Banks noted that the impact is limited partly because the unsecured lending book combined both floating-rate and fixed-rate loans, reducing the interest pass-through effects of higher rates; and for the large banks, the customer base comprises high-income professionals and/or government employees whose income and creditworthiness are less likely to be affected by macro downturns. Some banks noted that the legacy portfolio of nonperforming loans of unsecured lending to the lower income sector has already been cleaned up, so no further considerable deterioration of its credit quality is expected.

21. Similarly, the credit quality of the floating-rate mortgage portfolio is not much affected by higher rates, as reported in the BU STs. Increasing interest rates by 100 bps and 500 bps respectively would result correspond to expected losses as a percentage of total capital of 6.2 percent and 6.9 percent, reflected in capital level declines of 0.8 percent and 1.5 percent. The limited impact is due to the good quality of customers.

22. The above limited impact of interest rate driven credit risks on household debt is partly due to the caveat inherent in the ST approach (i.e., sensitivity analyses). Sensitivity analyses of interest rate shocks only account for changes in interest rates without including increases in unemployment or declines in house and asset prices. Banks noted that, while higher interest rates increase debt repayments, they will not have a large impact on credit quality unless house prices decline significantly. Moreover, sensitivity analyses usually assess the impact over a 12-month horizon and hence lagged effects (i.e., those that materialize beyond 12-months) are not considered.

Table 4.

Sensitivity Analyses—Interest Rate Hike Impacts on Household Debt

article image
Sources: SARB supervisory data and IMF staff calculations.

23. Banks would not suffer material losses on the trading book and the fair valued banking book under the single factor shocks in the BU STs. Positions on the trading and fair valued banking books, including interest rate, exchange rate, equity, commodity, and credit-risk sensitive exposures, were tested. The most adverse outcome is due to the failure of the three largest counterparty exposures, which would reduce the system CET1 ratio by 28 bps, with the highest impact on one bank being 42 bps. The limited impact is due to hedging practices as well as a daily monitoring process to ensure that exposures and losses are within risk appetite. A number of banks follow long gamma strategies, which benefit from increased currency or equity price fluctuations.21

Table 5.

Sensitivity Analyses—Market Risks on the Trading Book

article image
Table 6.

Sensitivity Analyses—Market Risks on the Trading Book

article image

24. The impact of interest rates change on the banking book is limited in the BU STs. Overall, banks remain asset sensitive, accumulating ZAR 235 billion of net, cumulative interest sensitive assets over a 6 to 12 month horizon. Sensitivity analysis indicates that a downward parallel move of the interest rate by 200 bps causes the CET1 ratio to fall by only 32 bps, while banks would benefit from an upward parallel move of rates by 36 bps if the interest rate increases by 200 bps. Due to the structural nature that assets are linked to the prime rate and liabilities are based on JIBAR, banks are subject to basis risks. The CET1 ratio would decline, on average, by 18 bps if the spread narrows by a relatively large amount of 50 bps. Given the current economic outlook, the downside risk on banking book earnings seems to be limited.

Table 7.

Sensitivity Analyses—Interest Rate Risks on the Banking Book

article image

C. Results for the Insurance Sector

25. The insurance ST showed a substantial impact on the participating life insurers, though on aggregate the sector is able to withstand the shock. The median CAR coverage ratio declines from 317 percent at end-2013 to 125 percent immediately after the shock, which is assumed to occur at the beginning of 2014 (Figure 6). While one company’s CAR coverage drops below 100 percent, the capital shortfall amounts to only ZAR 4 billion, which corresponds to 4 percent of the ST sample’s available capital before stress, or 0.1 percent of GDP. By end-2014, the median CAR coverage reverts back to 177 percent, and by end-2018, at the end of the projection horizon, the ratio is expected to stand at 300 percent, though still below pre-stress levels. The recovery is mainly driven by operating profits: the assumption that no new business is underwritten results in a reduction in variable costs (acquisition and set-up costs for new contracts22) and an increase in profitability.

Figure 6.
Figure 6.

Insurance—CAR Coverage Ratios

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: Participating companies and IMF staff calculations.

26. Non-life insurers show an even higher degree of resilience in the stressed scenario, although they start from a lower pre-stress CAR coverage ratio than the life insurers. This is because their asset allocation is more conservative, which makes them less sensitive to the market risk shocks applied. The median CAR coverage ratio drops from 175 percent at end-2013 to 138 percent immediately after the shock, with one company reporting a CAR coverage ratio below 100 percent. The capital shortfall is, however, limited and amounts to 2 percent of participating non-life companies’ available capital before stress. By end-2014, the median CAR coverage ratio is expected to increase to 151 percent, surpassing pre-stress levels already in 2015. By end-2018, the ratio rises to 218 percent.

27. The cumulative impact on net income is limited (Figure 7): In 2014, when all stresses are assumed to materialize, the aggregated profit of the five life insurers declines by 58 percent to ZAR 9 billion, after ZAR 21 billion in 2013. However, only one of the five life companies records a loss in 2014, which results in a very skewed distribution of return on equity (RoE) where the median increases slightly from 16.5 percent in 2013 to 17.6 percent in 2014. In the years 2015 to 2018, annual net income of the companies in the sample is expected to fluctuate around ZAR 16 billion, with all five companies reporting positive earnings from 2015 onwards. Non-life insurers would also experience a reduction in their profitability: Aggregated net income declines by 37 percent in 2014 and the median RoE drops from 22.3 percent to 16.8 percent. The range of results across companies, though, is much wider than for life companies.

Figure 7.
Figure 7.

Insurance—Return on Equity

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: Participating companies and IMF staff calculations.

28. The largest contribution to the deterioration in the life insurers’ solvency position comes from the equity shock, which reduces available capital on average by 28 percent. A second major contributor would be the default of the largest banking counterparty, which results in a reduction in available capital of about 23 percent (Figure 8). However, it should be noted that any second-round effects of a bank default (like contagion effects in the South African banking sector resulting in further defaults), as well as operational disruptions, have not been modeled in this ST, and so the results are likely to be significantly underestimating the overall impact. Finally, the interest rate shock also has a substantial impact by changing the value of both assets and liabilities, which both decline in the scenario of rising interest rates (Table 7).

Figure 8.
Figure 8.

Insurance—Reduction in Available Capital

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: Participating companies and IMF staff calculations.

29. Also for the non-life insurers, the equity shock is the main contributor to the decline in solvency ratios; available capital would drop by 19 percent on average. The banking default would cause a reduction in available capital of 13 percent. Overall, the scenario affects the non-life sector only via the reduced value in assets and subsequently via lower available capital. The value of liabilities is unchanged as these are not discounted according to statutory rules and are therefore insensitive to interest rate changes. Also, required capital remains constant for non-life insurers because it is not sensitive to the shocks applied in the ST.23

30. The additional sensitivity analysis with regard to catastrophe risks revealed a limited effect as the large non-life companies have wide-ranging reinsurance coverage with foreign reinsurers in place. On average, solvency ratios decline by only 14 percentage points when assuming a catastrophic loss that is expected to materialize once every 100 years. While South Africa has hardly been harmed at all in the past by large-scale weather-related catastrophes, the main risk, which was used to determine the 1-in-100 year loss, is a strong earthquake in the Johannesburg region; while the probability of such an event is low,24 the potential claims would be substantial due to high population density and economic activity.

Table 8.

Insurance—Effect on available and required capital

In million ZAR

article image
Sources: Participating companies and IMF staff calculations.

31. Life insurers benefit from the risk-sharing features of the main types of life insurance contracts. Investment losses in the stress scenario can partially be passed on to policyholders by reducing bonus allocations and smoothening the bonus allocation over time. This practice has been actively used by the large life insurers in the past; it follows a rather automatic mechanism with limited management discretion, and the mechanism is disclosed to policyholders. Furthermore, South African life insurers have been holding high levels of capital in recent years (also in anticipation of the new solvency regime to be implemented in 2016), which at the time of the ST provides them with a solid buffer. However, a prolonged period of lower or even negative investment returns could diminish this buffer.

D. Aggregating Banking and Insurance Stresses

32. On a conglomerate level, financial institutions could weather the combined losses from their banking and insurance operations. The banking operations would be able to withstand severe shocks; the recapitalization need, at 1.6 percent of nominal GDP at the maximum, is manageable, and the capital shortfall for the insurance companies, at 0.1 percent of nominal GDP, is even smaller. Recapitalizing the few insurance companies would be within the capacity of the corresponding conglomerates, owing to the small size of the capital required as compared to the level of aggregate capital in the group.

Liquidity Risk

33. Banks conduct BU STs using the Basel III framework for liquidity risk measurement and the regulatory standards proposed in BCBS (2013) and BCBS (2014). The tests cover the results for the Liquidity Coverage Ratio (LCR) and the Net Stable Funding Ratio (NSFR).25

  • The LCR measures whether banks have adequate levels of unencumbered, high-quality liquid assets that can be converted into cash to meet their liquidity needs for a 30 calendar day time horizon. The LCR is defined as the ratio between the stock of high-quality assets to the total net cash outflow over the next 30 calendar days. In particular, the liquidity risk stemming from funding withdrawals by related parties was assessed in the LCR by imposing a 100 percent runoff rate.

  • The NSFR complements the LCR by testing whether banks hold a minimum acceptable amount of stable funding based on the liquidity characteristics of an institution’s assets and liabilities over one year. The ratio is calculated as that between the available amount of stable funding and the required amount of stable funding.

34. The majority of banks are yet to meet the Basel LCR requirement. The LCR of five of the six large banks tested was below 100 percent, and for some banks it was below the 60 percent minimum requirement that will become effective from 2015,26 without using the proposed Committed Liquidity Facility (CLF) of the SARB. Conversations with banks and the authorities indicated that the banks have the capacity to meet the 60 percent minimum requirement, mainly through selling tri-party repo assets to acquire more high quality liquid assets (HQLA).

35. This shortfall results from a dependence on wholesale, short-term funding, and the limited availability of HQLA. In South Africa, a large chunk (60 percent) of bank deposits is wholesale funding from NBFIs and the corporate sector. Moreover, the average funding maturity has shortened in recent years.27 Compared with more stable funding sources, wholesale and short-term funding attract much higher run-off rates and thus larger liquidity needs in the LCR test. Regarding the supply of HQLA, South Africa has a limited pool of level 1 assets, and virtually no level 2 assets that satisfy the criteria specified in the BCBS (2013)28 29 (Figure 9).

Figure 9.
Figure 9.

Bank—Liquidity Stress Test

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Note, asset 1 to asset 6 on the LHS panel are: coins and bank notes; qualifying marketable securities from sovereigns, central banks, public sector senterprises (PSEs), and multilateral development banks; qualifying central bank reserves; domestic sovereign or central bank debt for nonzero risk-weighted entities; qualifying marketable securities from sovereigns, central banks, PSEs, and multilateral development banks (with 20percent risk weighting); and qualifying corporate debt securities rated AA- or higher.Sources: SARB supervisory data, bank reported results and IMF staff calculations.

36. The SARB’s CLF is necessary in a system with insufficient HQLA. Given the structural issue of a small retail deposit base and the limited supply of HQLA, SARB introduced the CLF to help banks meet the LCR requirement and, ultimately, liquidity needs under extreme market conditions. Without the CLF arrangement, Figure 9 shows that banks would be experiencing a shortfall in liquidity in less than 20 days, if the level of cash outflow assumed in the stress test were to be realized. The CLF, which provides up to 40 percent of HQLA, would help banks to meet the liquidity shortfall once they have met the initial 60 percent requirement on their own. Banks indicate that they will increase the securitization of home loans to boost the stock of collateral eligible for the CLF.

37. Regarding the NSFR, all banks except one would have less than 100 percent coverage. Banks indicated that they will have difficulty in fully complying with the minimum NSFR, effective from 2018, given the structural small retail deposit base and the difficulty in obtaining long-term funding from the capital markets, partly because it is expensive and funding sourced through the off-shore markets could also introduce FX risks. Banks noted that they may need to reduce the amount of long-term assets to meet the NSFR requirement. Banks could also consider raising the interest rate offered to attract more retail deposits and obtain more funding with maturity beyond one year. Our analysis indicates that the corresponding increase in interest rate payments is affordable given the high profitability of the banking sector.30

Contagion Risk

38. The contagion risk from the recent failure of African Bank was limited. The SARB acted decisively in resolving African Bank and in soothing market jitters when some money market funds with exposure to African Bank “broke the buck”31 in 2014. Capitec, the other boutique lender, saw only a slight decline in its share price as it was perceived to have a more conservative credit risk policy. Market participants also quickly realized that the Moody’s downgrade was not related to the soundness of the large banks but to a reduction in the “too big to fail premium” that had been implicit in the banks’ credit ratings.32 The vulnerability of financial institutions to contagion risk has remained low throughout the African Bank episode (Figure 10)33.

Figure 10.
Figure 10.

Financial System Spillover Coefficients (In percent)

Citation: IMF Staff Country Reports 2015, 054; 10.5089/9781498367608.002.A001

Sources: Moody’s KMV, IMF, Financial Stability Measurement, and IMF Staff calculations.

Recommendations

39. In terms of next steps, the SARB should continue to develop a TD macro stress testing framework. A TD ST, to be conducted at least annually, should complement the existing BUs exercises conducted by banks. The scenario should be derived from the macroeconomic baseline projection taking the latest risk assessments into account. More resources should be allocated to the validation of banks’ models and assumptions, as a substantial gap in the TD and BU results was found in this assessment.

40. For the insurance sector, a macroprudential stress testing framework should be developed. This ST should be:

  • - Severe (but plausible): Scenarios should entail a high degree of conservativeness and adequately address the structural specificities of the South African insurance sector, especially the concentrated exposures to domestic banks.

  • - Comprehensive: The test should also include risk factors that will not be covered by the standard formula of the forthcoming solvency regime, e.g., sovereign risk.

  • - Forward-looking: The test should capture long-term dynamics by using a multi-year projection horizon, and assess the feasibility and effectiveness of various risk-mitigating strategies.

  • - Easy to communicate: Each scenario should be based on a clear narrative.

41. The FSB, and in the future the prudential regulator, should establish a stringent monitoring framework for assessing the concentration risk of the insurance sector towards the largest domestic banks. It should request from insurance companies contingency plans that not only cover financial planning but also measures to address operational risks.

Appendix I. Risk Assessment Matrix (RAM)

article image
article image

Appendix II. Macro-financial Scenarios

Appendix Table 1.

Main Variables

article image
Appendix Table 2.

Global market shocks – interest rate directional shock

article image
Appendix Table 3.

Global market shocks – interest rate volatilities shock

article image
Appendix Table 4.

Global market shocks – cross-currency basis shocks

article image
Appendix Table 5.

Global market shocks – exchange rate directional shocks

article image
Appendix Table 6.

Global market shocks – exchange rate volatilities shocks

article image
Appendix Table 7.

Global market shocks – commodity shocks

article image
Appendix Table 8.

Global market shocks – equity shocks

article image
Appendix Table 9.

Global market shocks – sovereign credit spread shocks

article image
Appendix Table 10.

Global market shocks – corporate credit spread shocks IG HY Hot-rated

article image

Appendix III. Stress Test Matrix (STeM) for the Banking Sector

article image
article image
article image

Appendix IV. Stress Test Matrix (STeM) for the Insurance Sector

article image
1

Including underwritten pension funds and insurance policies of domestic pension funds.

2

Please refer to the South Africa FSAP Financial System Stability Assessment (FSSA) report for a full assessment, which considers mitigation factors.

3

The TD ST has a smaller coverage since the supervisory data on the probability of default (PD), loss given default (LGD) and exposure at default are available only for the four large internal ratings-based (IRB) banks and are based on domestic operations.

4

Other major, relevant global shocks potentially affecting South Africa include a short-term, sharp economic slowdown in China and a sustained decline in commodity prices (Appendix I).

5

See Appendix Table 1 for details.

6

The standard deviation is based on the historical sample of 1963 to 2013.

7

For a summary of the methodologies used by the BU and TD STs on the banking sector, see Box 1 and Appendix III.

8

See Appendix Table 2 to 9 for details.

9

See Appendix III for the complete set of tests.

10

Assets are generally stated at fair value if not directed otherwise by the FSB. Technical provisions in life business are discounted based on a government bond rate plus a margin. Non-life technical provisions are not discounted.

11

The capital ratio is for bank solo, including foreign and local banking operations. The number is based on data submitted in the BU STs.

12

This is the RWA weighted average of minimum CET1 requirements, which include the Basel III minimum CET1 requirements, Pillar 2A and Pillar 2B charges.

13

The magnitude of the increase in the performing PD is large. For instance, the proportion of defaulted loans would be 57 percent of the gross loan portfolio by the end of the five-year horizon, assuming no write-offs are taken, given a portfolio with default loans of 17 percent at the starting point.

14

The net interest income is only 0.6 percent lower than in the baseline, even though the amount of loans and advances is 3 percent smaller. The loans and advances are mostly prime-linked and therefore rate-sensitive. The non-rate-sensitive funding includes shareholders’ funds and non-repricing transactional deposits, which together account for a quarter of banks’ funding.

15

The TD approach is subject to the caveat that the time series is relatively short.

16

For instance, although the TD ST leads to a spike in PDs for the mortgage book, banks noted that, since interest rates rise gradually, allowing for adjustment in household expenditure, the impact of higher interest rates on household repayment rates would be limited in the short term. Compared to 2013 levels, the annual average credit loss in the BU STs is 78 percent higher and the annual average credit exposure is 10 percent higher. In the TD ST, the annual credit loss is 128 percent higher than the 2013 level, significantly higher than in the BU ST, even though the annual average credit exposure, which is 11 percent higher than the 2013 level, is not much different from the amount of credit exposure in the BU STs.

17

Moreover, since two out of the six banks use the standardized approach other than the IRB approach, the RWAs in the BU STs will always be less risk-sensitive than in the TD ST on aggregate. Other reasons for the differences include 1) due to management action, the balance sheet growth is slower in the BU STs than in the TD ST, yielding smaller credit losses and smaller RWAs. 2) Due to time constraints, some banks could not adjust their existing stress testing frameworks, which adopt a three-year horizon instead of a five-year horizon. As a result, they use a shorter horizon than in the TD ST. Conversations with banks confirmed that ST results could have been worse if the horizon was extended, due to the lagged effects. For instance, it would take 18 months for the macroeconomic downturn to affect the credit quality of some portfolios. 3) The TD ST assumes a maximum dividend payout ratio (Box 1) while the BU STs may restrict dividends.

18

Assets allocated to both the “available for sale” and those “designated at fair value through profit and loss” portfolios are subject to the application of the same valuation rules applied to the trading book on similar assets.

19

Banks assumed different LGDs, reflecting the variations in loan quality and the level of collateral. The average LGDs assumed are 27 and 54 percent for non-financial and financial borrowers, respectively.

20

The 9.0 and 9.9 percent include the expected losses under the current business condition and the added losses caused by the higher interest rates.

21

The success of these strategies relies on the solvency and creditworthiness of the trading counterparties. This raises concerns about counterparty risk, proper diversification of counterparties, and whether counterparties can manage their short-gamma positions adequately. Conversations with banks indicated that the counterparties are diversified and the markets are relatively liquid.

22

South African valuation rules allow for a zeroization of negative rand reserves (NRR). The NRR arises after the inception of an insurance contract when the present value of future premiums and fee income exceeds the value of benefits and expenses. The zeroization translates the build-up of the negative reserve into an accounting loss in the first year after the inception of the contract. If no new business is underwritten as assumed in this ST, this drain on profits is eliminated and net income is higher.

23

Technical provisions in non-life insurance are not discounted under the South African valuation framework which means that companies (and ultimately policyholders) have an additional buffer to withstand shocks.

24

The largest seismic event in this region in the recent past was in March 2005. It reached a magnitude of 5.3 and caused only relatively low damage above ground, though more severe damage was recorded in adjacent mines.

25

The results are based on SA banking operations as of March 31st, 2014.

26

The LCR requirements will be effective from 2015 with an initial requirement of 60 percent and an additional 10 percent each year thereafter until it reaches 100 percent.

27

Short-term deposits (6 months in maturity) rose to 63.3 percent of total deposits from 60.3 percent in 2008.

28

SARB guidance note 5/2012.

29

For the definition of level 1 and level 2 assets, see BCBS (2013).

30

Assuming that banks offer an interest rate of 8.7 percent to attract longer term funding, the additional funding cost would amount to ZAR 10 bn per year, equivalent to ⅙ of banks’ after tax profit as of end-2013. Assuming a proportionate reduction in return on equity (ROE), this would imply an ROE of 14 percent. The 8.7 percent rate could be attractive for investors. It is 3 percent higher than the historical average deposit rate of 5.7 percent, based on the historical data from 2010Q1 to 2013Q4. The 3 percent spread is greater than the average spread between the government bond yield and deposit rate of 2.7 percent, based on the same historical sample.

31

While African Bank’s debt accounted for only 1.3 percent of the assets held by the 43 money market funds, at least 10 of these funds “broke the buck,” i.e., the losses caused the unit price to fall.

32

Standard and Poor’s ratings do not incorporate an implicit subsidy of government support.

33

The vulnerability of a financial institution to spillover risks is measured by the conditional probability of its default given the default of another financial institution.

34

We prefer not to disclose the distribution by size since it may reveal bank specific information given the small number of banks included in the ST.

South Africa: Financial Sector Assessment Program-Stress Testing the Financial System-Technical Note
Author: International Monetary Fund. Monetary and Capital Markets Department