The Department of Labor’s Interpretive Bulletin 95-1 (DOL 95-1) outlines criteria that pension plan fiduciaries should consider when evaluating insurers for a potential pension risk transfer (PRT) transaction when annuities are purchased on behalf of plan participants. The DOL’s guidance is very clear that fiduciaries should not rely on a single metric and that they instead should “conduct an objective, thorough and analytical search for the purpose of identifying and selecting providers from which to purchase annuities. In conducting such a search, a fiduciary must evaluate a number of factors relating to a potential annuity provider’s claims paying ability and creditworthiness.” The guidance states that fiduciaries should select the “safest annuity available” and clarifies that more than one provider can meet that standard in a given transaction.
While this guidance was originally written to dissuade fiduciaries from relying solely on ratings provided by credit rating agencies, it also serves as a clear reminder that no single metric can provide comfort that an insurer is sufficiently strong to provide pension risk transfer annuities. NISA Investment Advisors (NISA) has published two articles outlining its “Economic Loss to Beneficiaries” (ELB) metric and touts this as something that fiduciaries should use to disqualify certain insurers that are active in the PRT market. While we are supportive of developing innovative ways of evaluating insurers as part of a holistic approach, and there are some situations where market-based metrics add value, this ELB metric is highly flawed as an evaluation tool. The name, and suggested use of the ELB, implies beneficiaries may actually lose benefits equivalent in value to this metric; however, the ELB likely overstates expected default losses faced by policyholders by many multiples, potentially by a factor of hundreds. We calculate expected losses over 10 years, before allowing for the additional protections PRT policyholders experience, in the 0.2-0.5% range compared to the ELB which is often in excess of 10%. The NISA approach is not an acceptable substitute for a robust approach to evaluating insurers and transaction-specific details by fiduciaries.
Based on our expertise in the insurer due diligence review market, we discuss why NISA’s ELB measure provides an incomplete view of determining the safest annuity available. We also provide five reasons why NISA’s approach to evaluating insurer creditworthiness in the PRT market is misleading.
Agilis is a leading provider of annuity purchase services both as a placement consultant and as an Independent Expert, which includes due diligence reviews and opinions related to the Department of Labor’s Interpretive Bulletin 95-1 (DOL 95-1). Since 2015, we have assisted in over 220 annuity purchases totaling over $24B in transferred pension liabilities. In 2022 we opined, as per DOL 95-1, on approximately 20% of all pension risk transfer transactions in the US market by value.
Why FABN spreads are an incomplete measure of the “safest annuity available”
NISA’s ELB is calculated using observed credit spreads on Funding Agreement Backed Notes (FABNs) issued by insurers. The ELB for a given insurance company is the credit spread of a FABN issued by that insurer, minus the credit spread on a reference FABN, multiplied by a factor of 10. The reference FABN is that with the lowest credit spread of PRT market insurers (New York Life at the time NISA’s articles were written). The ELB is meant to quantify the amount of relative credit risk, as implied by bond markets, that an annuity policyholder is taking on, expressed as a present value over ten years. ELB differs substantially between insurance companies; New York Life by definition has an ELB of 0% and Athene the highest at 12.8% at the end of May 2023. Other insurers lay in the middle of this range (e.g., Principal cited as having an ELB at 6.5%).
Taken at face value, NISA’s ELBs imply that most annuities purchased by pension plans for participants have been dramatically overpriced and that there is an extremely high level of risk of default and loss being borne by PRT policyholders at many insurers. NISA assert that the difference in ELB between insurers should translate into higher annuity payments to participants to compensate for higher risk (e.g., an increase as high as +12.8% for annuities placed with Athene compared to New York Life, at the end of May 2023).
This approach dramatically over-states the risk of loss to participants of annuities issued by different insurers, fails to account for robust insurance industry protections and inflates the reliability of FABN spreads as a true measure of default risk.
5 Reasons Why NISA’s Approach to Evaluating Insurer Creditworthiness is Misleading
1. Policyholder losses incurred due to insurer failure are expected to be very low, even before allowing for the presence of state guaranty systems
This view is supported by i) a low historic default rate, which we expect to continue for the insurers currently active in the PRT market, and ii) a high recovery rate should insurer failure occur.
When we refer to policyholder default losses, we are referring to the unlikely event that an insurance company fails. In this case, there would be a “rehabilitation” period—the outcome of which may lead to all policyholders being made whole, or alternatively, a loss being imposed on policyholders (and FABN holders who choose to hold on to their bonds through this entire process).
Expected default loss = [Probability of Default] x [Loss Given Default] = [Probability of Default] x [ 1 – Recovery Rate]
Losses imposed on policyholders have historically been much lower than corporate unsecured debt. In several large life insurer failures, the loss has been zero: policyholders ended up receiving 100% of the value of their policies. Conversely, the average recovery rate experienced by policyholders of Executive Life Insurance Company, a California-domiciled insurer that infamously failed in 1991, was 87.2%, with claims in excess of guaranty association coverage always greater than 77.7%. To be solvent, insurers start from a position where assets must exceed liabilities; therefore, even conservatively allowing for substantial loss of asset value at the point of failure we would still expect recoveries to be in the 75%-90% range, and they could well be higher. To quantify expected losses over 10 years, we multiply historical default rate experience for all corporate debt by a recovery rate assumption as follows:
- A best estimate of general account policyholder expected default loss could use Moody’s 1983-2023 default data. Using the historic default rates of A-rated corporate debt and a 90% recovery: 9% x (1-90%) = 0.19%. This is almost 70 times lower than the highest ELB cited by NISA. A 75% recovery would give a 0.48% expected loss.
- An extremely conservative estimate of general account policyholder default loss uses BBB-rated debt default experience since 1920 and a low recovery rate: 5.2% x (1-75% recovery) = 1.3% , which is still approximately 10 times lower than the highest ELB cited by NISA.
- Using life insurance company experience, rather than general corporate debt, may result in even lower loss numbers; however, we use the experience of the latter, as there have been so few defaults of well-rated insurance companies that it is difficult to draw statistically robust conclusions.
Use of credit ratings and historic experience
The above default loss estimates are based on long-term history allowing for starting credit ratings. Both could be inaccurate in the future, but we see no compelling reason why future default loss would be materially different from long-term historic averages. Use of ratings in this way is also the approach taken by regulators of banks and insurance companies when estimating losses on the bonds they own. While new risks will inevitably emerge, history does also include distressed periods and periods under weaker regulatory oversight than today.
To calculate the very conservative estimate above, we used the experience of BBB-rated corporate debt. That BBB rating is lower than all current PRT market providers. In summary, even if we prudently scale up the default losses of history, this would only explain a small component of current FABN credit spreads, and thus a small fraction of the ELB cited by NISA.
2. The credit risk priced into bonds issued by insurers is not equivalent to expected policyholder losses
The difference in credit spreads between FABNs drives NISA’s evaluation which is, by definition, a market price of relative credit risk between the insurers. If expected losses upon default are so low, as described in the previous section, why are credit spreads (hence the NISA ELB) so much larger? FABN holders may be demanding compensation for uncertainty and other forms of credit risk such as downgrade risk and price volatility, in addition to illiquidity risk. If FABNs get downgraded to sub-investment grade, for example, many bond holders will either be forced sellers over time or have to hold much more capital against these positions. However, PRT contracts are not liquid or redeemable, nor do the beneficiaries of PRT contracts receive a market price for their contract. While the marginal bond investor may rationally demand some higher yield to compensate for downgrade risk and price instability, when investing relative to other opportunities, it would be inaccurate to directly equate these risks and associated losses to those faced by a PRT policyholder.
3. NISA’s approach does not allow for the substantial additional protection provided by separate accounts and by the State Guaranty Association system
NISA acknowledge that their approach does not allow quantitatively for additional protections that are provided to pension participants through the use of separate accounts and the State Guaranty Associations. The only justification for suggesting that participants should demand increased benefits from different insurers, in line with the market price of bonds issued by the insurer, is if participants could invest their pension benefits in such bonds and receive an equivalent level of safety. But this is not the case – most FABN holders will effectively receive a very small level of protection indirectly through the State Guaranty Association system. However, PRT policyholders are often substantially if not fully protected from the losses, and it is common to see expected recovery rates for PRT transactions in excess of 99% allowing for the State Guaranty Association system.
Also, as described in our previous article, the use of separate accounts in PRT transactions materially reduces expected participant losses relative to policies in the General Account; typically, by around a factor of ten from what is already a very low number.
Two insurers with FABNs at higher relative credit spread levels are Athene and F&G. These companies, as of mid-November 2023, have credit spreads approximately 100 bps higher than the lowest spread insurers. Most of this difference has emerged since the end of 2021. If one accepted that FABN spreads could be used to infer relative creditworthiness reliably, this suggests that Athene and F&G have a slightly higher expected loss under General Account policies. However, it is important to note that both these insurers utilize separate accounts for PRT transactions, as do many other insurers.
Allowing for the loss mitigation provided by separate accounts, it is possible to come to the conclusion that Athene and F&G have an expected loss that is substantially similar in magnitude to the loss faced by PRT policyholders at the other providers, even if one believes that the expected loss faced by General Account policyholders at these companies is relatively higher.
4. Plan sponsor failure is generally much more likely than that of a well rated insurance company
If, as argued by NISA, participants should demand additional benefits to compensate them for a measure of increased risk they are taking on with some insurance providers, then that logic could be extended to participants who remain in a pension plan that is maintained by their employer. That would imply that significant benefit increases (perhaps >20%) are owed to plan participants based off the credit spreads of the plan sponsor’s debt, if that is observable, which are often higher than those of the PRT insurers. Either way, this approach ignores the guarantees provided by either the PBGC or State Guaranty Associations, which can’t be ignored when addressing benefit security.
5. Credit spreads on Funding Agreement Backed Notes have limited reliability as a metric
For the US insurers that write PRT transactions, the credit spread on their FABNs has typically been between +0.60% to +2.0% per annum since January 2022:
Source: Bloomberg (YAS_YLD_SPREAD). CUSIPs references described in Appendix 2. Insurers with incomplete spread histories excluded from chart.
There is a substantial noise and potential unreliability in the observed spreads:
- FABN Spreads are volatile through time, and like any other corporate bond spreads, are a function of supply and demand. Spreads tend to over-react in periods of stress and be correlated with wider market sentiment. For example, Athene and F&G’s spread has oscillated within an approximately 100bp range since June 2022. We see no evidence from a fundamental evaluation of these companies that underlying credit worthiness or claims paying ability has changed by that magnitude relative to other insurers.
- Spreads tend to be higher for longer duration issues, all else equal, which complicates comparing the spreads of the insurers. Looking at multiple issues from several insurers, we calculate an additional 16 bps of spread per year of duration on average in 2023. The FABN shown for Principal has a duration of 5.3 years, which is significantly longer in duration than the average of other insurers. Based on our calculations, allowing for this average duration impact on spreads would result in a 34 bps reduction to spreads for Principal and a 13 bps reduction for Pacific Life, for example, which are significant reductions from those presented by NISA in determination of their ELB. This highlights the instability of a measure based on a single bond, that will not be a like-for-like comparison with other bonds.
- Spreads on FABN do not always appear consistent with those of debt issued by the holding company or with Credit Default Swap levels (where the actual default risk should be higher than those for FABNs).
- Even if markets are efficient at reflecting publicly available information, investors may not have access to the same level of information as fiduciaries, rating agencies, and other experts evaluating life insurers.
The relative risk of PRT insurers and evolving credit risk
Although there is a low forward-looking expectation of insurer failure, particularly relative to a typical pension plan sponsor, any insurer has the possibility of failure. We acknowledge that if the market spread of certain insurers is materially higher than others, investment markets are implying a higher risk of downgrades, price volatility, and ultimately a higher potential loss level. We do not entirely dismiss the market’s perception of such higher credit risk as irrelevant, as it can flag concerns, even if it is not directly relevant to the losses PRT policyholders face directly. FABN credit spreads can be used as a data point among many others.
Conclusion and Application to DOL 95-1
The DOL 95-1 guidelines require fiduciaries to select the ‘safest annuity available’ and describe factors that fiduciaries must consider in that evaluation. The history of failure of financial firms shows us that reliance on single measures, such as the credit spreads of FABNs, can be extremely misleading and dangerous. Aside from the drawbacks of excluding otherwise sound insurance companies, reliance on market-based measures could result in insufficient due diligence on insurers with relatively lower spreads.
Markets can be very useful at signaling certain concerns, particularly if there are emerging issues or loss of confidence at an insurer, giving rise to a significant and rapid widening of spreads. However, there is no substitute to performing fundamental, bottom-up evaluation of insurance companies when selecting a PRT provider.
1. Introduction to Funding Agreement Backed Notes (FABNs)
Funding Agreement-Backed Notes (FABNs) are utilized by insurance companies as an additional source of financing. Total amounts issued are typically much smaller than traditional life insurance reserves for the insurers active in the Pension Risk Transfer (PRT) market. Rather than issue debt directly, the insurance company involved will issue an insurance policy to a Special Purpose Vehicle (SPV), and that SPV will issue debt to the market. That insurance policy ranks alongside other General Account policyholders in the event of insurer failure. Thus, the potential losses of the SPV (and ultimately of the FABN holders) should be broadly equivalent to the loss General Account-backed PRT policyholders face, in absence of the State Guaranty Association system. However, the ultimate loss a typical FABN holder faces upon an insurer failure will be much higher than an individual policyholder, who may be substantially protected by the State Guaranty Association system. The use of separate accounts can also significantly reduce losses experienced by PRT policyholders.
The price of FABNs infers a particular yield, and that yield can be expressed as a credit spread over and above the yield on US Treasuries. The observed price, hence yield and spread, are a function of supply and demand of buyers and sellers.
The larger the spread, the greater compensation investors are demanding for forms of future losses and/or for the uncertainty of losses. However, only a small portion of the observed spread is directly relevant to individual policyholders, i.e., compensation for actual default losses. Allowing for historical default rates implies there are other much larger components of the observed spreads that are compensating the holder of those notes for the risks they face compared to holding US Treasuries, namely downgrade risk, price volatility and illiquidity.
2. FABN spreads, with duration adjustments to spreads at 11/15/2023
Source: Bloomberg, Agilis calculations
 Pension Risk Transfers May Be Transferring Risk to Beneficiaries (Oct. 13, 2022), https://www.nisa.com/perspectives/pension-risk-transfers-prt-may-be-transferring-risk-to-beneficiaries/; NISA, Liquidity Refinements to Potential Economic Loss to Beneficiaries (ELB) in PRTs (Jun. 16, 2023)
 Funding Agreement Backed Notes described further in Appendix 1
 Includes liquidity adjustment as described in “Liquidity Refinements to Potential Economic Loss to Beneficiaries (ELB) in PRTs (NISA, Jun. 16, 2023)
 As at end 2006, the last date recovery estimates are available. US and Japan Life insurers Insolvencies Case Studies, Lessons Learned from Resolutions – The Geneva Association, 2015. https://www.genevaassociation.org/sites/default/files/research-topics-document-type/pdf_public//ga2015-insurance-resolution.pdf
 Moody’s 1983-2023 default data. Annual Default Study: Corporate Default Rate will rise in 2023 and peak in early 2024, 13 March 2023, Moody’s Investors Services, pg 48
 10-year cumulative default history from Moody’s experience, 1920-2023 for BBB rated debt. Annual Default Study: Corporate Default Rate will rise in 2023 and peak in early 2024, 13 March 2023, Moody’s Investors Services, pg 48
 Annuity Purchases and Guaranteed Separate Accounts (January, 2022)