Fixed indexed annuities (FIAs) with Guaranteed Lifetime Withdrawal Benefits (GLWBs) are among the most widely used retirement income products. These products combine guaranteed lifetime income with continued access to the remaining account value, providing policyholders with both income security and liquidity.
Current CARVM reserves for FIAs with GLWB are highly conservative because they are based on the most adverse guaranteed policyholder election path. For products with richer guarantees, day zero statutory reserves sometimes fall between 115% to 130% of the initial premium. VM-22 replaces the formulaic CARVM reserve methodology with a principle-based stochastic framework that produces reserves that are more closely aligned with the underlying economics of the business, while still maintaining an appropriate level of prudence for statutory valuation.
This article examines the key drivers of the VM-22 reserve for a block of FIAs with GLWB by comparing it with a hypothetical economic reserve calculated under a single deterministic forward-rate path. We then explore how asset portfolio construction—specifically as it relates to duration and key rate duration, credit spread profile, and hedging strategy—affects reserve efficiency under VM-22.
Modeling Framework under VM-22
At the core of VM-22 is an ALM projection model that projects assets and liabilities under a set of stochastic economic scenarios prescribed by the NAIC. The projections generally reflect the insurer's prudent estimate assumptions for both assets and liabilities, subject to prescribed guardrails. For this study, we focus exclusively on the stochastic reserve (SR) to highlight the mechanics of the stochastic methodology, while excluding other elements of VM-22 such as the Stochastic Exclusion Test and Deterministic Reserve.
Asset and liability cash flows were modeled from first principles using Agam's proprietary pALM platform, which is an integrated asset-liability modeling platform that projects cash flows, reinvestment, and hedging strategies consistently across a range of economic scenarios.
Asset cash flows, liability cash flows, hedge payoffs, and reinvestment were projected across 1,000 prescribed stochastic scenarios[1] from the Generator of Economic Scenarios. For each scenario, the model applied the direct iteration method to determine the minimum starting asset required to eliminate projected deficiencies throughout the projection horizon. The SR was then calculated as the CTE70, defined as the average of required starting assets across the worst 30% of scenarios.
Liability Description
For this study, we modeled a representative FIA product with a GLWB rider that reflects common product features and serves as a generalized example for this analysis. The specific results of the VM-22 analysis will vary depending on the characteristics and risk profile of the underlying liability. (See Table 1)
Table 1
Summary of Product Features and Modeling Assumptions
We modeled a representative block of FIA with $1 billion in total account value, spanning a range of issue ages and a mix of male and female policyholders.
Comparison of VM-22 Reserve to Economic Liability
To illustrate and understand the various components of the VM-22 reserve, we compare the resulting reserve with a best estimate liability (BEL) and quantify the incremental impact of key VM-22 requirements. For the purposes of this analysis, the BEL is defined as an economic liability measure based on best estimate liability cash flows under a single forward interest rate scenario, based on the intrinsic path of the forward rates implied from the prevailing yield curve at that date, discounted at the net earned rate of the backing asset portfolio. For the representative FIA with GLWB block modeled in this study, the calculated VM-22 stochastic reserve is approximately 15% higher than the BEL.
To better understand the attribution between an economic BEL and the VM-22 stochastic reserve, we also express those impacts in terms of changes in the implied discount rate solved for as the rate that equates the discounted cash flows to the reserve incorporating the corresponding incremental VM-22 requirements. The baseline implied discount rate is an indication of portfolio yield applicable to our reserves. The implied discount rate for the BEL is approximately 6.0%.
The resulting decomposition provides an illustrative perspective on the key drivers of the difference between an economic BEL and the VM-22 reserve for the representative FIA with GLWB modeled in this study. The magnitude and relative contribution of these components will vary depending on the characteristics of the underlying liability, asset portfolio, and other modeling assumptions. (See Table 2)
Table 2
Summary of Differences Between Economic Reserve and VM-22 Reserve
- Prudent Estimate Liability Assumptions
VM-22 requires provision for adverse deviation (PADs) on top of best estimate assumptions, including mortality, policyholder behavior (such as lapse rates and rider utilization timing), and expenses. Furthermore, VM-22 requires a 1.5% haircut on options payoffs related to hedging FIA index credits.[2] For this analysis, we modeled reasonable margins based on industry benchmarks and survey data, which are estimated to increase economic reserves by approximately 2% – 5%, an impact of 20-30 bps reduction on the implied discount rate. - Prescribed Reinvestment Strategy
VM-22 effectively requires a prescribed portfolio consisting of 5% cash, 15% AA-rated corporate bonds, and 80% A-rated corporate bonds.[3] Accordingly, any benefit of higher-yielding in-force assets persists only for the duration of the existing asset portfolio, with all subsequent reinvestment assumed to occur in the prescribed portfolio. Our modeling suggests that the prescribed reinvestment assumptions increase reserves by approximately 1% – 3%, equivalent to an aggregate portfolio spread compression of 10 – 15 bps driven by reinvestment. The magnitude of this impact can vary significantly depending upon the yield assumed under the company's actual reinvestment strategy. - Prescribed Defaults Cost (Spread Cap)
VM-22 adopts the prescribed default cost methodology first introduced in Section 9.F of VM-20. The methodology effectively limits the earned spread recognized on in-force assets through a prescribed default cost adjustment applied during the first three years of the projection. When a portfolio's net spread exceeds that of a comparable Baa (BBB)-equivalent portfolio, an additional default cost is applied to offset the excess spread. This adjustment then grades off over the first three projection years. We expect this methodology to have a significant impact on reserves during the initial three years of projection for portfolios with substantial allocations to shorter structured assets and associated higher yields. Our modeling suggests that the spread cap assumption increases the reserves by an estimated 2% – 4%, with a cost of 20 – 25 bps in implied discount rate. - Impact of worst 30% scenarios
The VM-22 stochastic reserve is determined using CTE70, defined as the average required starting assets across the worst 30% of stochastic scenarios, rather than the average projected outcome across all scenarios (i.e., CTE0).[4] For the block modeled in this study, the worst scenarios are generally characterized by lower equity returns coupled with declining interest rates, which increase projected policyholder benefits and the starting assets required to support them. Our model suggests that the CTE70 reserve is approximately 4% – 8% higher than the CTE0 reserve, estimated as ~60 bps to 80 bps of implied discount rate. - ALM Mismatch
Intrinsically based on principles of prudent Asset Liability Management (ALM), VM-22 naturally rewards well-matched portfolios. However, even well-matched portfolios cannot perfectly align asset and liability cash flows across all scenarios, and differences in timing of asset and liability cash flows can lead to losses during reinvestment and sales based on different scenarios relative to our single forward scenario leading to a margin within the reserves that can be attributed as an ALM mismatch. Additionally, our modeling for VM-22 here assumes current generalized approach of static reinvestment tenors and does not dynamically modify tenors based on remaining cash flow duration (to align with the common reinvestment practice applied at many insurers today and to estimate the ALM mismatch that many insurers will see in practice). Our modeling suggests that the impact of this ALM mismatch is approximately 1% – 3% of the reserve and ~10 bps to 20 bps of implied discount rate differential.
Collectively, these adjustments explain the key drivers of differences between the VM-22 reserve and the corresponding single scenario BEL as described above. The analysis also highlights the areas where changes to asset strategy and hedging may have the greatest impact on VM-22 reserves.
ALM and Portfolio Construction Considerations under VM-22
To assess the impact of asset portfolio construction on the VM-22 reserve, we calculated the CTE70 reserve using five alternative investment portfolios. The portfolios were designed to represent realistic investment strategies using readily available public corporate bonds and public structured assets. They differ in asset allocation, spread, and duration while being intentionally stylized to isolate and illustrate the impact of these characteristics on the VM-22 reserve. Table 3 below summarizes the key asset assumptions and portfolio allocations.
Table 3
Summary of Asset Assumptions and Portfolio Allocations
We calculate CTE70s with the six portfolios, and we express the results as a percentage of account value. The analysis, along with other useful metrics, is summarized in Table 4 below.
Table 4
Summary of Portfolios and CTE70s
1. Trade-Off Between Yield and ALM Mismatch
A higher portfolio yield generally lowers the reserve. A 10% allocation to higher yielding Public Structured Assets (Portfolio A) reduces the reserve by approximately 1% relative to the Base portfolio. However, increasing the allocation further to 25% (Portfolio B) has the opposite effect. While it increases overall portfolio yield, it also increases the asset-liability duration mismatch, largely offsetting the benefits from a higher portfolio yield.
The trade-off is more pronounced for Portfolio D, which includes a 50% allocation to floating-rate Public Structured Assets, which have effectively zero interest rate duration. This reduces the duration of portfolio D significantly and the increased duration mismatch fully offsets the benefit of the higher portfolio yield. In the adverse scenarios (typically ones where rates are falling) our shorter-duration floating rate portfolio is exposed to the dual effect of declining floating coupon and lower reinvestment yields as assets mature, thereby increasing the overall CTE 70 reserve relative to baseline.
Implication: Under VM-22, constructing the asset portfolio involves balancing the benefits of higher portfolio spreads against the potential adverse impact of asset-liability mismatches under the stochastic scenarios.
2. Techniques Aimed at Reducing ALM Mismatch Improve the Efficiency of VM-22 Reserve
In portfolio E, we test a static hedging strategy by adding a receive-fixed, pay-floating interest rate swap to the allocation in Portfolio D. This synthetically adds duration to the portfolio and eliminates mismatch while preserving the underlying asset mix and net spread. The portfolio can therefore retain the benefit of higher spread from its 50% allocation to public structured assets without incurring the penalty associated with ALM mismatch. In the declining interest rate scenarios that are likely to drive CTE70, the interest rate swap will result in positive payoff and reduce CTE70. In effect, the interest rate swap decouples duration management from spread generation, allowing the portfolio to capture the benefits of both.
Implication: Risk mitigating instruments can be used strategically along with asset selection to improve the outcomes, emphasizing that principles of duration management are not solely tools for effective risk management, but also key drivers of economic value and profitability.
3. It's Not Just the Spread—It's How Long You Keep It
Under VM-22, the impacts of higher yielding in-force assets are subject to two key guardrails:
- Prescribed Reinvestment Portfolio
Under VM 22, positive cash flows are required to be reinvested into a prescribed portfolio consisting of 5% Treasury securities, 15% public AA-rated corporate bonds, and 80% public A-rated corporate bonds.[5] Consequently, shorter-duration assets transition to the lower spread prescribed reinvestment portfolio sooner, reducing the overall earned yield within the projection.
- Prescribed Default Costs
Excess spreads on existing assets are capped during the first three projection years by applying a prescribed default cost factor. As a result, the benefit of shorter-duration assets with attractive spreads may be limited under VM-22, while longer-duration assets can continue to earn their contractual spreads after the temporary adjustment grades away, allowing more of the yield advantage to emerge over time.
Due to the above, assets with longer tenors are less impacted by VM-22 guardrails on asset yields. Portfolio C illustrates this by extending the maturity of the Public Structured Asset allocation from five to 10 years relative to Portfolio B. While both portfolios have similar current market yields at inception, Portfolio C performs better for longer over time. The longer maturities not only preserve the benefit of higher spreads beyond the initial spread-capping period, but also delay reinvestment into the prescribed reinvestment portfolio, while simultaneously improving alignment with the long-duration FIA liabilities.
Implication: Under VM-22, consider both asset spreads and tenor implications in portfolio construction. A well-matched portfolio with moderate, durable spreads may produce a more efficient reserve than one with higher initial spreads and shorter tenors due to the factors mentioned above.
Conclusion
VM-22 fundamentally changes fixed annuity reserving by linking statutory reserves to the interaction of liability characteristics, investment performance, stochastic market scenarios, and prescribed reinvestment assumptions. Our analysis demonstrates that the difference between statutory and economic reserves is driven by several identifiable factors, including asset-liability mismatch, limitations on recognized asset spreads, adverse stochastic scenarios, and prescribed liability assumptions.
The results also highlight the importance of evaluating product design, asset portfolio construction, and hedging strategies within an integrated ALM framework. For products with features such as GLWB riders, the interaction between liability characteristics and the underlying asset and hedging strategies can have a meaningful impact on VM-22 reserves. Insurers that incorporate these considerations into their broader product, investment, and risk management frameworks may be better positioned to manage statutory requirements while maintaining prudent risk management and competitive product economics.
This article is provided for informational and educational purposes only. Neither the Society of Actuaries nor the respective authors’ employers make any endorsement, representation or guarantee with regard to any content, and disclaim any liability in connection with the use or misuse of any information provided herein. This article should not be construed as professional or financial advice. Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries or the respective authors’ employers.
Gaurav Rastogi, FSA, FCIA, CFA, is partner and global chief actuary at Agam Capital Management. He can be reached at grastogi@agamcapital.com.
Jack Cannon, FSA, is an executive director at Agam Capital Management. He can be reached at jcannon@agamcapital.com.
Peter Yang, FSA, CFA, is an executive director at Agam Capital Management. He can be reached at pyang@agamcapital.com.
Abhishek Mittal, FIA, is a director at Agam Capital Management. He can be reached at amittal@agamcapital.com.
Endnotes
[1] We tested the model using 10,000 stochastic scenarios and found that, for the product and assumptions used in this study, the resulting reserve, overall scenario dispersion, and key scenario analytics were sufficiently similar to those produced using 1,000 scenarios.
[2] Per VM-22 Section 4.A.4.b.i.b:
An Index Credit Hedge Margin for these hedge instruments shall be reflected in both the “best efforts” and the “adjusted” runs, as applicable, by reducing index credit hedge payoffs by a margin multiple that shall be justified by sufficient and credible company experience and be no less than 1.5% multiplicatively of the portion of index credit that is hedged. This margin is intended to cover sources of potential error due the hedging itself and the ability for the company to accurately model it. In the absence of sufficient and credible company experience, a margin of 20% shall be assumed. There is no cap on the index credit hedge margin if company experience indicates actual error is greater than these minimums.
[3]Per VM-22 Section 4.D.3.b:
Notwithstanding the above requirements, the aggregate reserve shall be the higher of that produced by the modeled company investment strategy and that produced by substituting an alternative investment strategy in which the fixed income reinvestment assets have the same weighted average life (WAL) as the reinvestment assets in the modeled company investment strategy and are all public non-callable corporate bonds with gross asset spreads, asset default costs, and investment expenses by projection year that are consistent with a credit quality blend of:
- 5% Treasury
- 15% PBR credit rating 3 (Aa2/AA)
- 80% PBR credit rating 6 (A2/A)
[4] Per VM-22 Section 3.D.2:
The Stochastic Reserve amount for any group of contracts shall be determined as CTE70 of the scenario reserves following the requirements of Section 4.
[5] See endnote [3]