VM-22 became effective January 1, 2026 for new non-variable annuity business, replacing the prescribed factors of the Commissioners Annuity Reserve Valuation Method with company-specific prudent-estimate assumptions modeled stochastically at a CTE70 tail standard (NAIC Valuation Manual, 2026). Insurers have until January 1, 2029 to apply it to all business, meaning old-basis and new-basis reserves will sit on the same balance sheet for three annual statement cycles while U.S. annuity sales run near a record $464.1 billion (LIMRA, 2026).

What CARVM Actually Did, and What Replaces It

CARVM's defining feature was uniformity: every company calculated the greatest present value of future guaranteed benefits using the same prescribed mortality tables and discount rates, regardless of how the company actually managed lapses, credited rates, or asset risk. That uniformity made reserves comparable across companies but disconnected from the economics of any individual block. VM-22 replaces that with prudent estimate assumptions the company sets itself for mortality, policyholder behavior, and expenses, run through the framework's Stochastic Exclusion Test and Single Scenario Test to determine whether the filed reserve is the deterministic reserve, the stochastic reserve, or, in narrow cases, a continuation of CARVM (Milliman, VM-22 readiness, 2026). Where the stochastic reserve governs, it is measured as the Cumulative Tail Expectation above the 70th percentile, or CTE70, across the company's full range of generated economic scenarios.

The mechanical shift that actuaries who lived through VM-20's rollout for life insurance will recognize immediately is the move from a table lookup to a full asset-liability projection. Each in-force cell now runs forward under hundreds of stochastic economic paths, with the reserve set at the average of the worst 30% of outcomes rather than a single prescribed factor. RGA's assessment of the transition frames the conservatism bluntly: field testing has produced "over-conservative reserves with redundancies, in some cases with resulting reserves higher than those required by the outgoing CARVM methodology" (RGA, 2026), an outcome that inverts the industry's working assumption that principle-based reserving is a capital release exercise. A structural driver of that redundancy is the cash surrender floor built into the stochastic calculation, which RGA describes as "essentially equivalent to requiring an insurance company to reserve for a 100% mass-lapse scenario" in every modeled path (RGA, 2026). For a fixed indexed annuity block with a guaranteed lifetime withdrawal benefit, that floor can dominate the reserve even when the company's own lapse studies show persistency running well above any plausible mass-lapse trigger.

The Comparability Problem Nobody's Statutory Filing Will Flag

Having modeled reserve transitions through VM-20 for life insurance, the VM-22 rollout for annuities follows a familiar script, except this time it lands on the industry's fastest-growing product segment rather than a mature, slow-growth one. VM-20's three-year phase-in for term and universal life ran against a book of business where new issue volume was flat to declining, so the parallel-regime distortion in reported reserve trends was modest in dollar terms. VM-22 phases in against a market where RILA sales just posted their 30th consecutive quarter of year-over-year growth (LIMRA, May 2026), which means the share of the in-force book sitting on the new stochastic basis grows every single quarter of the transition, not gradually across a flat issue pattern.

That has a specific consequence for anyone reading filed annual statements between now and 2029: year-over-year reserve trend for non-variable annuity lines will not be interpretable as a clean signal of underlying experience. A company writing $2 billion of new FIA and RILA premium annually under VM-22 while its 2024-and-prior in-force block runs off under legacy CARVM will show a blended reserve growth rate that mixes two fundamentally different valuation bases. An analyst, rating agency, or reinsurance counterparty comparing 2026 statutory reserves to 2025 statutory reserves for the same company is implicitly comparing a partially CARVM number to a partially stochastic number, and the blend ratio changes every quarter as new business accumulates on the VM-22 side. The NAIC's VM-22 Subgroup is aware of the distortion at the edges: a retrospective application exposure draft, which would extend VM-22 to business issued as far back as January 1, 2017, closed its 90-day comment period on June 22, 2026 (NAIC Valuation Manual (A) VM-22 Subgroup, 2026), precisely because leaving older blocks on CARVM indefinitely widens the comparability gap the longer the transition runs.

GOES and the Negative-Rate Modeling Gap

VM-22's stochastic engine runs on the Generator of Economic Scenarios, the replacement for the prior American Academy of Actuaries interest rate generator that VM-20 and VM-21 also now draw on. The technical requirement actuaries have not previously had to build models around is explicit: companies must ensure their asset-liability models "can handle negative interest rates" under the new generator (Milliman, VM-22 readiness, 2026). The prior generator's scenario set was bounded in a way that made near-zero and negative paths a tail curiosity rather than a modeled outcome with real reserve weight. GOES removes that bound, and because VM-22's CTE70 standard is explicitly a tail-weighted measure, scenarios that spend meaningful time at or below zero on the discount curve now feed directly into the average that sets the reserve, not just into a stress test run alongside it.

For a company whose ALM infrastructure was built to VM-20's original economic scenario generator, this is not a parameter update. Asset cash flow projections, hedge cost models, and crediting rate algorithms calibrated against a generator that never produced sustained negative rates can behave unpredictably when the new generator does, particularly for floating-rate and option-embedded assets whose cash flow logic was never tested against that regime. The Academy and Society of Actuaries jointly govern GOES going forward, replacing the split maintenance structure that existed under the prior generator, which at minimum consolidates where model updates originate but does not reduce the validation burden on the company side.

What the Sales Numbers Say About Where the Exposure Concentrates

VM-22's reserve mechanics do not land evenly across the annuity product shelf. The products most exposed to the stochastic tail calculation, those with embedded guarantees, index-linked crediting, and withdrawal benefit optionality, are also the fastest-growing segment of the market this transition is running against.

Product (Q1 2026) Sales YoY Change VM-22 Reserve Sensitivity
Total U.S. annuity sales $104.6B -2% 10th consecutive $100B+ quarter
RILA $21.2B +21% High: index-linked crediting, cap/floor optionality
Fixed indexed annuity (FIA) $26.6B -4% High: GLWB mass-lapse floor exposure
Fixed rate deferred (FRD) $34.0B -16% Moderate: primarily target spread and lapse assumption

Total U.S. annuity sales came in at $104.6 billion in the first quarter of 2026, 2% below the same period a year earlier but still the tenth consecutive quarter above the $100 billion threshold (LIMRA, May 2026). "Although first quarter sales were slightly below prior year's results, the threshold for annuity sales appears to be stabilized above $100 billion," said Bryan Hodgens, LIMRA's senior vice president and head of research (LIMRA, May 2026). RILA sales reached $21.2 billion, up 21% year over year and the 30th straight quarter of growth for the product category. Keith Golembiewski, LIMRA's assistant vice president and head of annuity research, attributed the momentum to distribution breadth: "This product has tremendous tailwinds. As new RILA products and riders continue to be introduced and more carriers and distribution adopt RILAs into their portfolios, the market capacity for these products broadens" (LIMRA, May 2026). FIA sales were $26.6 billion, down 4%, while fixed rate deferred annuities fell 16% to $34.0 billion as the credited-rate advantage over other guaranteed products narrowed.

RILA is the product where the VM-22 transition and the sales trend intersect most directly. Its cap-and-floor index crediting structure is inherently path-dependent, which is exactly the feature stochastic reserving is designed to capture and CARVM's formula factors could not. A carrier growing RILA new business 21% a year is adding stochastically-reserved exposure to its balance sheet at a faster rate than almost any other segment of the market, which means its blended reserve basis shifts toward VM-22 faster than a company with a flatter or more FRD-weighted new business mix, independent of any change in the underlying risk being written.

Pricing and Valuation Assumptions Now Have to Match, Explicitly

CARVM's uniform factors created an unusual convenience: pricing actuaries could set lapse and crediting rate assumptions for profitability testing without those assumptions having any mechanical connection to the statutory reserve, since the reserve was a prescribed table lookup regardless of what the pricing model assumed. VM-22 removes that separation. The prudent estimate lapse and target spread assumptions that drive the stochastic reserve are company-specific, which means the valuation actuary's assumption set and the pricing actuary's assumption set are now describing the same product with the same tools, and a divergence between them is no longer invisible to the reserve.

If pricing assumes higher lapse rates to support a richer credited rate or commission structure than the valuation team's prudent estimate assumption supports, that gap surfaces directly as reserve strain at issue, not as a footnote reconciling two unrelated bases. Companies that historically ran pricing and valuation as loosely coordinated functions, common where CARVM's mechanical reserve made tight coordination unnecessary, now have a structural reason to formalize an assumption governance process that ties the two together, with the target spread and lapse assumptions reviewed on a shared cadence rather than independently by each function.

Flow Reinsurance and Captive Cessions Built for a Different Reserve

A large share of the fixed and indexed annuity market runs some or all of its new business through flow reinsurance treaties or captive cessions, structures generally built around the predictability of a CARVM reserve credit that both sides of the treaty could calculate the same way from a published factor table. VM-22 disrupts that predictability at the mechanism level, not just the number level, because the ceded reserve credit is now itself a company-specific stochastic output rather than a shared formulaic reference point.

The direction of the effect is not uniform across products. Where VM-22 produces a lower reserve than CARVM, insurers may see reduced need for the reserve financing structures, letters of credit, and captive arrangements that existed specifically to fund CARVM's redundant reserve levels (RGA, 2026), which weakens the economic case for treaties priced around that redundancy. Where VM-22's cash surrender floor and tail-scenario weighting push the stochastic reserve above the old CARVM level, particularly likely for FIA blocks with guaranteed lifetime withdrawal benefits given the redundancy field testing has already surfaced, the ceding company's reserve strain increases and the case for reinsurance support strengthens, but the treaty has to be repriced against a reserve basis that moves with the ceding company's own assumption set rather than a table both parties could independently verify. RGA's own assessment concludes that reinsurance solutions under VM-22 have to be customized deal by deal because "the company-specific nature of VM-22 guidelines requires that any reinsurance solution be customized for each insurer" (RGA, 2026). A flow treaty negotiated in 2023 on the expectation of a predictable CARVM credit is not a treaty that renews cleanly onto a VM-22 basis; both sides need to revisit the economics with the new reserve mechanics in view before the cedant's block crosses fully onto the new basis by 2029.

The 2029 Deadline for Companies Without Stochastic Infrastructure

The three-year transition window is generous on paper but compresses quickly for any company that has not already built or bought the modeling capability VM-22 requires. Running the deterministic and stochastic reserve tests, the Stochastic Exclusion Test, and the full CTE70 scenario projection demands an actuarial software platform capable of asset-liability stochastic projection at a scale most mid-size and smaller non-variable annuity writers have not historically needed, since CARVM never required more than a factor lookup against in-force data.

For those companies, the practical choices narrow to three: license and build internal stochastic modeling capability, likely a multi-year, seven-figure infrastructure investment given the asset and liability data granularity VM-22 requires; outsource the valuation function to a consulting actuary or software vendor already running GOES-compatible platforms for other clients; or cede enough new business through reinsurance that the stochastic reserve calculation becomes the assuming reinsurer's problem rather than the ceding company's. Each path has a different timeline. Vendor and consulting capacity for VM-22 modeling is not infinite, and with mandatory application landing January 1, 2029 for every company regardless of size, the companies that wait until 2028 to start building or contracting for that capability will be competing with every other small and mid-size annuity writer facing the identical deadline, in a market for actuarial modeling talent and platform capacity that is already stretched by parallel VM-20, VM-21, and now VM-22 implementation work across the industry.

Why This Matters

VM-22 is the second major life and annuity line, after universal life and term under VM-20, to move from prescribed factors to company-specific stochastic reserving, and it is landing on a segment growing faster than any comparable point in VM-20's own rollout. The transition period's comparability problem is not a temporary reporting inconvenience; it is three years of statutory filings that mix two valuation bases in proportions that shift every quarter, complicating trend analysis for analysts, rating agencies, and reinsurance counterparties alike. The GOES negative-rate requirement is a genuine model-risk item, not a documentation formality, for any company whose ALM infrastructure assumed the prior generator's bounds. And the reinsurance and captive structures built around a predictable CARVM credit need renegotiation, not renewal, as ceded blocks cross onto the new basis. Actuaries working non-variable annuity valuation, pricing, or reinsurance treaty design should treat the next three annual statement cycles as an active reconciliation project between two reserve regimes, not a scheduled cutover to be revisited once in 2029.

Further Reading

Sources

  1. NAIC, Valuation Manual (A) VM-22 Subgroup, content.naic.org, 2026
  2. NAIC, PBR Data and Valuation Manual, Jan. 1, 2026 Edition, content.naic.org, 2026
  3. Milliman, “VM-22 Readiness: Key Areas for Consideration,” Milliman.com, 2026
  4. Milliman, “Current State of Principle-Based Reserving for Non-Variable Annuities (VM-22),” Milliman.com, 2026
  5. RGA, “VM-22 Is Here: Where Does Reinsurance Fit In?” RGAre.com, 2026
  6. LIMRA, “U.S. Annuity Sales Notch Tenth Consecutive $100 Billion+ Quarter,” LIMRA.com, May 2026
  7. LIMRA, “Final U.S. Retail Annuity Sales Set New Sales High, Totaling $464.1 Billion in 2025,” LIMRA.com, 2026