Apocalypse now or die another day? Stress testing APAC insurers’ corporate bond portfolios
Regulatory capital and solvency ratio implications of a moderate to severe recession
- Asia-Pacific (APAC) insurers are increasingly concerned about the credit risk of their investments. To quantify the potential balance sheet impact of defaults and downgrades in insurers’ corporate bond portfolios, we carry out several credit rating migration stress tests, with severity ranging from a typical recession to the worst year of the 1930s Great Depression.
- For each scenario, we estimate the typical level and tail value of mark-to-market losses on bond portfolios, as well as corresponding changes in regulatory capital requirements and in solvency ratios under several risk-based capital regimes.
- For most insurers’ portfolios, the main threat is not the risk of outright defaults, but rather the risk of rising capital requirements and shrinking capital resources due to downgrades of lower-quality investment-grade bonds.
- On average, APAC insurers appear adequately capitalized to withstand a very severe rating downgrade scenario. However, risk concentration in under-diversified portfolios can lead to significantly higher losses and more severe erosion of capital relative to well-diversified portfolios.
- The focus of this paper is on the capital and solvency impact of credit rating downgrades. A more thorough risk assessment of an insurer’s fixed income portfolio would require combining rating migration stresses with plausible spread widening scenarios.
Insurers, regulators, and stock analysts are increasingly concerned with the credit quality of insurers’ fixed income holdings and the potential balance sheet impact of credit downgrades in the event that a severe and/or protracted global recession is triggered by the Covid-19 pandemic.1
To help APAC insurers address these concerns, we carry out a number of stress tests aimed at quantifying the expected average impact and potential tail risk of corporate credit rating migrations on insurers’ regulatory capital requirements, capital resources, and solvency ratios. We consider a range of severe, yet conceivable, adverse scenarios based on historical experience, with severity ranging from a typical recession to the worst year of the 1930s Great Depression. The impact of these scenarios is modelled under the risk-based capital (RBC) regimes currently in force in Australia and Singapore, the forthcoming RBC regimes of Hong Kong and South Korea, and under the European Union Solvency II regime.
Our analysis suggests that insurers holding well-diversified portfolios of predominantly investment grade fixed income should be able to withstand a severe credit downgrade scenario, such as that of the 2008-2009 financial crisis. However, the tail risk of multiple simultaneous downgrades and defaults in excess of the expected levels could jeopardize the solvency of insurers holding more concentrated corporate debt portfolios.
How can insurers reduce their vulnerability to credit downgrades and defaults? A more global approach to fixed income portfolio construction could improve diversification and mitigate the tail risk of multiple simultaneous downgrades or defaults. Thorough fundamental credit analysis and active management could help steer insurers’ portfolios away from adverse credit events more adroitly. Thoughtful fixed income mandate design could help avoid forced sales of ‘fallen angels’ driven purely by guidelines compliance considerations. Closer integration between capital modelling, asset-liability management and portfolio management functions could help insurers anticipate potential changes in capital and plan for the unexpected. While key competencies should inarguably be developed and maintained in-house, APAC insurers could also access a wide range of advanced capabilities by partnering with experienced insurance asset managers.
Our main results are reported in the next paragraph. The rest of the paper covers the details of our modelling procedure: stress scenario specifications, composition of benchmark corporate bond portfolios, and calculation of key performance metrics before and after the credit stress. The simulation procedure for generating correlated rating migrations is summarized in the Appendix.
Our analysis suggests that insurers with well-diversified portfolios are adequately capitalized to withstand an extreme downgrade scenario in their fixed income portfolios, and could potentially use the current bond market dislocations to lock in attractive yields on BBB-rated and high yield debt. However, significant tail risks could jeopardize the capital position of smaller insurers that are more likely to hold less-diversified credit portfolios.
The balance sheet impact of corporate rating downgrades would depend on the accounting treatment of downgraded bonds: some insurers could soften the blow by not marking these bonds to market. However, even when insurers’ portfolios are fully marked to market and the risk of imminent downgrades is not reflected in pre-stress bond prices, a typical recession – or even the 2008-2009 Global Financial Crisis scenario – would not lead to significant downgrade and default losses on an average insurer’s fixed income portfolio (Exhibit 1). Under a Great Depression scenario (either ‘average’ or ‘peak’), widespread downgrades from BBB to high yield would lead to more sizeable, yet not ruinous, credit losses.
For the typical insurer, risk-based capital requirements for the fixed income portfolio would increase (in relative terms) by up to 5% above the pre-stress value under most risk-based capital regimes in the Typical Recession or the Global Financial Crisis scenarios; more significant increases in risk-based capital requirements could be expected under the Great Depression scenarios (Exhibit 2).
The impact of credit stresses on insurers’ solvency ratios is summarized in Exhibit 3. Most life insurers are highly leveraged – even a modest mark-to-market loss on investments can signify a sharp decline in capital resources. Coupled with an increase in risk-based capital requirements driven by a wave of rating downgrades, this can lead to a significant erosion in insurers’ solvency ratios. However, our analysis suggests that for the typical (median) insurer with a well-diversified fixed income portfolio, even a severe scenario, such as the average experience of the Great Depression, would lead to a relative decline in the solvency ratio of up to 5% (for example, from a 200% pre-stress solvency ratio to a 190% stressed solvency ratio). Under the scenario of the worst year of the Great Depression, with a 45% downgrade rate for BBB-rated bonds, the median insurer would take a hit to its solvency ratio of up to 20% in relative terms (for example, from 200% pre-stress to 160% post-stress).
While the ‘typical’ insurer is well-provisioned with capital to endure a severe credit storm, averages and medians do not tell the full story for the entire market. Insurers with under-diversified portfolios should anticipate greater credit losses, sharper increases in risk-based capital requirements, and much more severe deterioration in solvency ratios. The actual stress experience for these insurers could be closer to the estimated tail values (90th percentiles) plotted in Exhibit 1 to 3, suggesting a material risk of capital inadequacy under the more severe scenarios and the more demanding capital regimes.
The difference between the median and tail outcomes reflects the risk of credit concentration: even under a moderate default correlation, under-diversified portfolios could experience losses far in excess of the average level suggested by the corresponding scenario transition matrix. Smaller insurers are more likely to end up with more concentrated fixed income holdings and should thus pay particular attention to credit concentration risk.
Our results rest on the assumption that on average, credit spreads remain unchanged in the stressed environment. The experiences of the Global Financial Crisis and other stress events suggest that a global creditworthiness shock would almost certainly lead to a repricing of credit risk and a sharp widening of credit spreads, exacerbating mark-to-market losses and further eroding insurers’ capital resources. A more thorough risk assessment of an insurer’s fixed income portfolio would require combining credit rating stresses with plausible credit spread widening scenarios.
Despite significant improvements in the Covid-19 situation and a strong, coordinated response by multiple monetary and fiscal authorities at the time of writing, there remains significant uncertainty about the intensity and duration of the pandemic-induced recession and its implications for insurers’ balance sheets.
Reflecting this uncertainty, we consider several credit stress scenarios, ranging from a typical modern recession to the most extreme phase of the 1930s Great Depression (see Exhibit 4a). Each scenario is captured in a credit rating migration matrix based on the historical rating action frequencies observed over the corresponding reference period. Exhibit 4b summarizes the corresponding default and downgrade probabilities.
Relatively little is known publicly about the rating, industry, or currency composition of APAC insurers’ fixed income portfolios. Unlike their US peers, most APAC insurers do not file detailed investment disclosures that are made available by regulators.2 Unlike EIOPA (the European Union insurance regulator), national regulators in APAC typically do not publish detailed summaries of insurers’ investment holdings.3
With these limitations in mind, we stress two broad, diversified benchmark portfolios that we believe are relevant for APAC insurers (see Exhibit 5, all bond data is as of 31 May 2020):
- USD Asian Corporate Credit (Asian Credit) portfolio, which is comprised of corporate bonds that are included in the ICE BofA Asian Dollar (ADOL) index.
- USD Global Corporate Credit (Global Credit) portfolio, which is a 95%/5% mix of the ICE BofA US Corporate Index (C0A0) and the Global High Yield Index (HW00). The choice of weights reflects the preference for quality in insurers’ portfolios.
Risk-based capital requirements and solvency ratios
Each stress scenario is modelled as an instantaneous event: bond maturities remain unchanged, there is no amortization and no coupon income. To quantify the potential range of unexpected losses driven by default correlations, we simulate 20,000 random outcomes under each stress scenario and generate a distribution of capital requirements, mark-to-market portfolio values, and solvency ratios. An outline of the simulation procedure is provided in the Appendix.
Risk-based capital regimes
We estimate pre- and post-stress regulatory capital requirement for the fixed-income portfolio and for the overall balance sheet, as well as the stress impact on a hypothetical insurer’s solvency ratio under several APAC regulatory capital regimes that are currently in force or are expected to be implemented in the near future (see Exhibit 6).
Corporate bond portfolio capital requirement
To estimate the bond portfolio regulatory capital requirements, we calculate the capital requirement for credit spread risk (or, under K-ICS, credit risk) for each bond based on its remaining time to maturity, duration and credit rating before or after the simulated credit event. The portfolio capital requirement is the weighted average of capital requirements on individual bonds, weighted by the pre- or post-stress market values of the constituent bonds and expressed as a percentage of the overall bond portfolio value.
Overall balance sheet capital requirement
To estimate the overall investment risk capital requirement for the insurer’s balance sheet, we aggregate the spread/credit risk capital requirement with the equity risk capital requirement, accounting for diversification.
To focus on the impact of fixed income downgrades, we assume that mismatch risks, such as foreign exchange and interest rates, are fully hedged and thus carry zero capital requirements. To facilitate cross-country comparisons, we assume a simplified insurance balance sheet with an asset allocation of 60% corporate bonds, 10% listed equities and 30% home-country government bonds and cash (both treated as risk-free assets for regulatory capital purposes).4 Liabilities amount to 80% of market value of assets, leaving the insurers with 20% capital. The corporate bond allocation is the only element of the balance sheet directly impacted by the stress scenarios; changes in the value of the corporate bond portfolio lead to changes in available capital.
Risk-based capital regimes allow for diversification between the market risk of investments, underwriting risks of insurance liabilities and counterparty default risk. For most life insurers, the overall risk-based capital requirement is dominated by market risk; for simplicity, we assume that non-investment capital requirements are offset by diversification between market and underwriting risk, and are negligible.
Capital resources and the solvency ratio
The solvency ratio is the ratio of available capital resources (qualifying assets net of liabilities) to the overall balance sheet capital requirement. Our calculation of stressed capital resources assumes that defaulted bonds are liquidated, instantly recovering 30% of their face value5 – this allows insurers to avoid punitive capital requirements on defaulted exposures and is in line with the practice of insurance fixed income portfolio management.
For non-defaulted bonds, we consider two alternative treatments:
- No mark-to-market on non-defaulted bonds
Non-default rating actions (downgrades or upgrades) do not affect bond credit spreads and market prices.6 This could be the case for insurers that value all bond investments on historical cost basis regardless of their credit rating or for bonds for which the likelihood of a downgrade is already reflected in the market price.
- Mark-to-market on non-defaulted bonds
All bonds that experience a non-default rating action are re-priced based on credit spreads corresponding to their new credit ratings. The price impact of rating actions is a duration-convexity approximation based on the difference between the median option-adjusted spread (OAS) corresponding to the pre-stress and post-stress bond rating (see Exhibit 5).
The two approaches are at the extremes of a range of potential impacts of credit downgrades on insurers’ capital resources. For some bonds, the seeming inevitability of a downgrade or default is already reflected in spreads and prices; for others, rating actions could catch the markets off-guard, causing a sharp price reaction. In some jurisdictions, insurers are required to mark all their investments to market; in others, most bond investments are reported on historical cost basis, and revaluations are only triggered by permanent impairment to obligors’ creditworthiness. In several markets, regulators have temporarily allowed leniency in applying impairments to help mitigate the severe – but hopefully, temporary – impact of a global pandemic.
Simulation procedure: correlated migrations and defaults
For each bond in each portfolio under each stress scenario, we simulate a random distribution of credit ratings using a CreditMetricsTM-style framework. This allows quantifying the risk of correlated defaults and estimating a range of potential losses under each scenarios and their probabilities.7 Rating actions are based on one-year rating transition matrices, but are assumed to happen overnight; we do not model the effects of portfolio aging.
Each bond is mapped to an ultimate parent; a change in the bond credit rating is determined by the change in the ultimate parent’s creditworthiness. Under this approach, bonds issued by the same issuer and carrying different credit ratings pre-stress (for example, due to debt seniority) could have different simulated post-stress credit ratings, but directions of simulated rating changes would be nearly perfectly correlated. Bonds issued by the same issuer and carrying the same pre-stress rating will carry the same post-stress rating in each simulated outcome.
Ultimate parent creditworthiness is a linear combination of an independent firm-specific factor and a common global creditworthiness factor. The correlation between an issuer’s creditworthiness and the common factor is based on a three-year rolling correlation of the issuer’s stock return with that of MSCI ACWI index, as reported by Bloomberg. For unlisted issuers, we use the sample median stock index correlation. Median stock index correlation in our issuer universe is around 60%, corresponding to an average pairwise creditworthiness correlation of 36%. This value is broadly in line with the range of values typically used in third-party credit risk modelling applications.
To derive an accurate measure of tail risk, we generate 20,000 random outcomes for the joint distribution of portfolio credit ratings under each stress scenario.
1 Asian Investor, “Covid-19 could spark junk bond sales by Asian insurers,” (16 June 2020); Bank of England Prudential Regulation Authority, “Insurance Stress Test 2019 and Covid-19 stress testing: feedback for general and life insurers” (17 June 2020)
2 For a detailed analysis of US life insurers’ latest filings, see Wuhan Lin, Mark Snyder (7 April 2020) COVID-19 and U.S. life insurers – J.P. Morgan Asset Management Insurance Insights.
3 For a recent cross-country analysis of European insurance industry statistics, see Valérie Stephan, Bethsabée Illouz (8 June 2020) The impact of market volatility on European insurers – J.P. Morgan Asset Management Insurance Insights.
4 Average asset allocations vary widely from one country to another, with typical listed equity allocations ranging from under 10% in Japan to over 20% in Singapore. Investment asset allocations also vary significantly between companies in the same market. Insurers would be well-advised to carry out stress tests based on their specific portfolios and to avoid relying excessively on industry averages.
5 Given the low expected default rate in insurers’ fixed income portfolios, the recovery rate assumption is not material for our analysis; this assumption is also in line with the global supervisory practice – for example see EIOPA, “Technical documentation of the methodology to derive EIOPA’s risk-free interest rate term structures (EIOPA-BoS-19/408)” (12 September 2019)
6 The exception is bond default, which triggers a forced sale and a fractional recovery of the principal.
7 J.P. Morgan “Introduction to CreditMetrics™” (2 April 1997)
For more information
If you work at an insurance company and are interested in a further discussion or a detailed analysis of your fixed income portfolio, please contact your J.P. Morgan Asset Management insurance client advisor or email Insurance_Strategy_and_Analytics@jpmorgan.com