SOA MIM 2026: Four Years of Post-Pandemic Data, MP-2021 Still Unchanged

The SOA's 2026 Mortality Improvement Model extends historical data through 2023, adding four years of post-pandemic mortality experience to the underlying datasets, but leaves the MP-2021 projection scale unchanged. The SOA Retirement Plan Experience Committee concluded that "there is insufficient post-pandemic data to support the development of a new MP scale" (SOA, 2026). At a moment when pension risk transfer transactions are pricing at 99.7% of ABO (Milliman, June 2026) and VM-20 stochastic reserves embed that same improvement scale, the held assumption deserves explicit actuarial scrutiny.

What the 2026 MIM Update Changed, and What It Did Not

The 2026 update is the most substantive revision to the Mortality Improvement Model since the pandemic interrupted its calibration cycle. Three changes are material. First, the model now incorporates mortality data through December 2023, adding four full years of experience that spans COVID-19's acute peak in 2020 and 2021, the partial normalization in 2022, and the continued recovery in 2023. Second, the National Center for Health Statistics dataset covering 2011 through 2019 has been revised: the updated figures show mortality improvement estimates that are more closely aligned with Social Security Administration experience, narrowing a discrepancy between the two population data sources that had complicated model interpretation. Third, the retrospective workbook has been renamed the Mortality Trends Explorer, with enhanced graphing tools and downloadable Excel support (SOA, 2026).

None of these changes touch the MP-2021 projection scale. That scale, which governs prospective improvement rates used in pension funding, annuity pricing, and VM-20 principle-based reserve calculations, is derived from SSA data and has not been updated since its October 2021 release. The MIM-2026 prospective modeling framework also deliberately excludes data from 2020 onward when estimating improvement trends. Practitioners who want to study the 2020-through-2023 period can do so using the Mortality Trends Explorer, but those observations flow into the retrospective workbook, not into the prospective scale. The boundary is explicit: the post-pandemic mortality pattern is observable in MIM 2026, but it does not yet revise what actuaries project forward.

The Post-Pandemic Record That Did Not Produce a New Scale

To understand why RPEC held MP-2021, it helps to read the four-year data extension alongside the CDC's mortality record for the same period. U.S. life expectancy fell to 76.1 years in 2021, the lowest level since 1996, then recovered to 77.5 in 2022 and 78.4 in 2023 (CDC/NCHS, 2023). That is two consecutive years of improvement after two consecutive years of decline. COVID-19 deaths fell 73.2%, from 186,552 in 2022 to 49,932 in 2023 (CDC, 2023). Excess deaths, which peaked at approximately 1,098,808 in 2021, dropped to roughly 705,331 by 2023, a 36% reduction in a single year, though still materially above pre-pandemic baseline levels (PMC, 2025).

The partial normalization creates the analytical problem RPEC is navigating. The 2022 and 2023 improvement pattern looks, at the surface level, like a rebound toward the pre-pandemic secular trend. But several confounding dynamics prevent a clean read. First, post-COVID mortality improvements are partly artifact: they measure improvement against a 2020 and 2021 baseline that was artificially high, not against a counterfactual of what mortality would have been absent the pandemic. Second, excess deaths in 2023, while down sharply from the peak, remained elevated above the high-income country comparison group; the United States would have avoided more than 1.5 million deaths in 2022 and 2023 combined had it matched peer-country mortality rates (PMC, 2025). Third, seniors aged 65 and older still showed some elevated mortality as of mid-2025, a residual that the SSA data refresh factored into the revised RPEC baseline but that does not yet resolve cleanly into either a higher-improvement or lower-improvement post-pandemic trend (SOA RPEC, 2025).

RPEC's conclusion, stated in both the 2025 update and carried forward into MIM-2026, is that these confounds are not yet resolved. The post-pandemic signal is visible but not separable into a clean directional judgment. That is a defensible position. It is also, for actuaries building reserves or pricing annuities today, a load-bearing assumption rather than a neutral placeholder.

VM-20 Reserves: Where MP-2021 Lives in the Calculation

Under VM-20 principle-based reserving, life insurers calculate both a Stochastic Reserve and a Deterministic Reserve, taking the greater of the two as the binding floor. Mortality improvement assumptions enter through the prescribed mortality tables, which use the 2015 Valuation Basic Table projected with annual improvement factors derived from the SOA's individual life mortality improvement scale, updated each year for year-end AG-38 and VM-20 purposes. The 2025 edition of that scale, adopted by the NAIC Life Actuarial Task Force, reflects the same underlying MP-2021 framework (SOA, 2025).

The valuation standard does not require actuaries to project improvement beyond the valuation date for purposes of computing the minimum reserve itself. The NAIC PBR guidance notes that the "absence of mortality improvement beyond the valuation date" functions as an implicit margin in the Deterministic Reserve, providing conservatism without requiring the actuary to make an explicit forward improvement assumption. What changes when MP-2021 eventually revises is the prescribed scale underlying that implicit margin: if future improvement rates embedded in the scale shift higher, the implicit margin compresses; if they shift lower, it widens. The direction of that shift matters, and the post-pandemic data as of mid-2026 does not yet tell a clean story about which way the scale will move.

For companies with large in-force blocks of term and whole life issued prior to the pandemic, the improvement assumption is a second-order driver of reserve adequacy relative to lapse and premium persistency. For companies with significant permanent life or universal life with secondary guarantees, where the liability tail extends 30 to 50 years, a sustained upward or downward shift in long-term improvement rates has a present-value impact that is not trivial. A 10-basis-point change in the long-cohort improvement rate, compounded over a 40-year projection horizon, can move a present-value estimate by 1 to 3 percentage points. At that scale, the difference between MP-2021 and a hypothetical revised scale matters to reserve adequacy decisions being made today.

Pension Risk Transfer: Margin so Thin That Improvement Is the Swing Factor

Nowhere in actuarial practice is the mortality improvement assumption more immediately load-bearing than in pension risk transfer pricing, and the current pricing environment makes this especially acute. Milliman's Pension Buyout Index showed competitive PRT cost declining from 100.9% of ABO in March 2026 to 100.1% in April, then to 99.7% in May 2026 (Milliman, June 2026). The duration-7 annuity purchase rate for May was 4.94%, and the duration-15 rate was 5.02%, among the highest levels observed since June 2025 (October Three, May 2026).

At 99.7% of ABO, the pricing margin is essentially at the accounting liability itself. The ABO assumes no future salary growth and discounts at a corporate bond rate; it does not incorporate the longevity loading that carriers build into their pricing. At that price point, the carriers' longevity margin, the spread between their pricing mortality assumption and the ABO mortality assumption, is being competed almost entirely away. That compression matters for the mortality improvement assumption in a specific, mechanical way.

Each additional year of unanticipated life expectancy at age 65, roughly equivalent to a 1% increase in mortality improvement rates, increases pension liabilities by 4% to 5% (October Three, 2026). At a transaction priced at 99.7% of ABO, a carrier that has embedded MP-2021 improvement rates into its pricing and then experiences 25 basis points of additional longevity beyond the scale would face a liability outrun on that block. Using portfolio-specific mortality assumptions rather than generic tables can itself produce a present-value impact of 3% to 5% in either direction, comparable to gaining or losing 30 to 40 basis points of annual asset returns (October Three, 2026). These are not theoretical sensitivities. They are the actual swing factors at current PRT pricing levels.

The 2025 PRT market closed at $49 billion across 700 transactions, a 6% decline from 2024, with 22 carriers actively offering group annuity contracts (Mercer, 2025). The decline was partly attributable to fiduciary litigation uncertainty that caused some jumbo transactions to pause. The market's structural trajectory has not changed: improving funded status, record annuity sales, and growing insurer capacity continue to push sponsor interest toward buyout. For every transaction that clears at or below 100% of ABO, the mortality improvement scale is effectively the pricing residual. There is no other margin to absorb longevity error.

The Threshold Question: When Does RPEC Pull the Trigger?

RPEC has not published a formal threshold for what data would be sufficient to support a new MP scale. The committee's 2025 update stated that "the worst effects of the pandemic on mortality have subsided" but that insufficient post-pandemic data prevented a new scale. The MIM-2026 overview carries the same language. Reading the two statements together, the committee appears to require both the subsidence of acute COVID effects, which has occurred, and a period of stable post-pandemic mortality data long enough to separate the recovery rebound from a genuine shift in the secular improvement trend.

The next data that would inform this judgment is 2024 and 2025 mortality experience. CDC preliminary estimates for 2024 will likely be available in late 2026, and SSA will incorporate that experience into its intermediate-cost projections in the 2026 Trustees Report. If 2024 shows continued normalization without elevated excess deaths, and if the SSA data begins to show a stable post-pandemic improvement trend at or near the pre-pandemic trajectory, RPEC will have the building blocks for a new scale in a 2027 or 2028 MIM cycle.

The directional uncertainty cuts both ways. One reading of the post-pandemic data argues for lower long-term improvement rates: the United States has persistently higher mortality than peer countries, excess deaths remain elevated even after acute COVID subsided, and the socioeconomic gradients in pandemic mortality may have permanently shifted the distribution of who survives to older ages. The opposite reading argues for higher near-term improvement rates: the 2022 and 2023 recovery was rapid, COVID-19 deaths are now below pre-pandemic flu mortality levels, and the aging of a relatively healthy boomer cohort into their 70s supports continued secular improvement through the 2030s. MP-2021, which was calibrated to the pre-pandemic trend with modest long-term convergence, may prove either too optimistic or too conservative depending on which of these dynamics dominates the 2024 and 2025 data.

Reviewing principle-based reserve filings and PRT pricing models that embed mortality improvement assumptions, the decision to hold MP-2021 through a fourth consecutive year while adding four years of data to the underlying model is regulatory conservatism of a particular kind: it avoids committing to a post-pandemic direction before the signal is clear, at the cost of leaving every VM-20 valuation and PRT pricing model anchored to a 2021-vintage assumption about a phenomenon that is now four years older. That is a reasonable choice given genuine uncertainty. It is not, however, a free lunch. The actuaries using MP-2021 in reserves and pricing today are implicitly making the same directional judgment RPEC is deferring: that the pre-pandemic improvement trend is a reasonable proxy for the post-pandemic one, at least until the 2027 or 2028 cycle provides a cleaner answer.