Every VM-20 historical improvement scale has to bridge the years between the last insured-data point and the valuation date, and for 2026 that bridge spans 2024 and 2025. The SOA Life MI Subgroup filled it with the Social Security intermediate projection again, although its own slide calls 2026 "the first year where we have actual data for these catch up years" (SOA Life MI Subgroup, August 2026). That scale now carries 11.5 years of drift from the 2015 VBT's July 1, 2015 date.
Dated August 27, 2026, the revised recommendation replaced a July 27 deck that carried a "New Proposed Method" estimating the two catch-up years from general-population deaths directly. The Life Actuarial (A) Task Force exposed the 2026 rates at its August 10 meeting in Columbus for 21 days ending September 1, then re-exposed the revised deck and spreadsheet for 14 days ending September 15.
Key Takeaways
- -1.28% a year is the smoothed 2026 male HMI at attained ages 32 to 39 in the exposed spreadsheet, and -0.84% the female rate at 27 to 35; the 2025 troughs sat near -1.25% over male ages 28 to 36 and -1.0% over female ages 22 to 36.
- 15.7% above the 2015 VBT rate is where a male age 35 lands once -1.28% compounds over the 11.5 years to year-end 2026 that VM-20 Section 9.C.3.g requires companies to cover.
- 10.8% is the 2025 fall in the general-population death rate at ages 25 to 34 (NCHS, July 2026), and 7.3% the SOA's annualized improvement at ages 20 to 49 in the year to September 2025, the data the July 27 method would have carried into the catch-up.
- 2027 is when the subgroup says a direct calculation "will be explored in greater detail", leaving the 2026 catch-up on the SSA 2023 to 2043 intermediate average, which runs near 1% a year on the deck's own chart.
- September is the month by which VM-20 says each year's rates "should be adopted by LATF and published on the SOA website"; comments close September 15, and a missed timeline lets each company elect the 2025 rates.
What Changed Between the July 27 Deck and the August 27 Revision
Slide 4 draws the standard method in three parts. A historical component takes the geometric average of ten years of industry data, 2013 to 2023. A catch-up carries that average from the historical end point to 2026, filled with the geometric average of the Social Security intermediate projection for 2023 to 2043, which the deck says "limits volatility in short term HMI calculations". On the July 27 chart that 20-year average sits close to 1% a year at nearly every age.
Beside it, the July 27 deck put a catch-up built from measured deaths. The new method would "use new SOA Quarterly Mortality Monitoring Report tool data to develop an adjustment based on general population results through from 2023 to the end of 2025", assuming insured lives improved the way the population did. Under it, unsmoothed 2026 HMI for men in their mid-20s ran near +2.5% a year on the deck's chart, against about -1.3% under the prior method.
Five weeks later the recommendation reads: "The 2026 scale recommendation for HMI is to continue to apply the standard methodology applies in 2025 and earlier rather than applying a direct calculation method using general population data for catch up years" (SOA Life MI Subgroup, August 27, 2026). A direct calculation moves to 2027, "to better understand potential volatility of the method."
Even so, the scale moved. In the exposed spreadsheet, improvement under age 20 turns negative for both sexes, -1.30% for girls at ages 6 to 8 and -1.38% for boys at 10 to 16, where the 2025 scale carried +0.3% through age 19. The female working-age trough narrows to ages 27 to 35 at -0.84%, and the male trough sits at -1.28% over 32 to 39. Plateaus of 1.18% for women at 60 to 79 and 0.87% for men at 52 to 82 run two to three years older than in 2025. Both sexes now grade to zero at 92, against 84 and 88 a year ago.
How the Negative Band Compounds Into the Year-End 2026 Reserve Basis
One word changed in 2021 turned the scale from an option into a requirement. APF 2020-10, adopted by LATF on June 10, 2021 for the 2022 Valuation Manual, rewrote VM-20 Section 9.C.3.g. Historical improvement from the industry table's date, "July 1, 2015, for the 2015 VBT", to the valuation date now "shall be incorporated" using the SOA factors adopted by LATF. Its guidance note adds that "mortality improvement may be positive or negative (i.e., deterioration)". At year-end 2026 that is 11.5 years of drift, and a negative rate raises the table.
| Attained age | 2026 smoothed HMI, male | Reserve-basis rate vs 2015 VBT, male | 2026 smoothed HMI, female | Reserve-basis rate vs 2015 VBT, female |
|---|---|---|---|---|
| 30 | -0.89% | +10.7% | -0.84% | +10.1% |
| 35 | -1.28% | +15.7% | -0.84% | +10.1% |
| 40 | -1.11% | +13.5% | -0.07% | +0.8% |
| 45 | -0.29% | +3.3% | +0.70% | -7.8% |
| 65 | +0.87% | -9.6% | +1.18% | -12.8% |
Rates are the smoothed 2026 HMI values in the exposed spreadsheet, rounded. Each multiple is (1 minus the rate) raised to the 11.5 years from July 1, 2015 to December 31, 2026, applied at a constant attained age, before the prescribed margin and before any company-experience blending under Section 9.C.6.
A term block issued at ages 30 to 45 carries the load; the permanent block at 65 gets a 10% to 13% credit. Male age 40 takes the largest step from last year: the 2025 scale was already grading toward its plateau there, at roughly -0.5% on the June 2025 chart, and the 2026 scale holds it at -1.11%. A woman of 45 moves the other way, from the 2025 scale's 1.2% plateau to +0.70%.
Future improvement runs the same way for five projection years. Section 9.C.7.f requires "twenty years of future mortality improvement" at rates "no greater than the loaded factors determined by the SOA". FMI starts at the HMI value, grades to the long-term rate by 2036, holds it to 2041 and grades to zero in 2046. Ages 26 to 40 keep the early projection-year adjustment. A male 35 runs -0.64% unloaded in 2027, zero through 2031, then grades to a 0.66% long-term rate; loaded, he starts at -0.80%, sits at -0.15% for 2028 to 2031 and reaches 0.50%. At 65 the loaded male rate moves from 0.65% to 0.68%.
NCHS provisional data put the 2025 age-adjusted death rate at 689.2 per 100,000, down 4.6% from 722.1, with ages 25 to 34 down 10.8% to 111.1 and 35 to 44 down 4.4% (NCHS, July 2026). The SOA's Quarterly Mortality Monitoring Report shows annualized improvement of 7.3% at ages 20 to 49 in the year to September 2025, with male deaths at 20 to 29 at 86.0% of their 2019 level (SOA Research Institute, January 2026). This site tracked the record low in August; the gap between a 1% assumed catch-up and a measured 7% is the "potential volatility" the subgroup wants another year to study.
The September Publication Clause and the Open VBT Question
APF 2020-10 also wrote a calendar into VM-20. Rates for a year-end valuation "should be adopted by LATF and published on the SOA website by September of YYYY. If this timeline is not met, then at the company's option they may use the mortality improvement rates for the prior year (year YYYY-1)." Comments on the revised scale close on September 15, after a first exposure that had already run to September 1 before the August 27 revision reset it. Adoption falls in the month the clause names, with publication still to follow.
An election of the 2025 rates cuts both ways by age. Under the 2025 scale's -1.0% band a woman of 30 lands 12.1% above the table over the same 11.5 years, against 10.1% under the 2026 scale; a man of 40 lands about 6% above it against 13.5%. Two companies holding the same term block, both compliant, could book year-end 2026 reserves on different mortality bases, with the difference concentrated where the working-age band moved.
The other end of the same assumption was the next item on the August agenda. LATF's meeting summary records that it "exposed the Society of Actuaries (SOA) request for directional feedback on the development of a new Valuation Basic Table (VBT) for a 14-day public comment period ending Aug. 25" (NAIC, August 2026). The 2015 VBT rests on 2009 to 2013 experience, and VM-20's guidance note says it "has already been improved from the mid-point of the underlying data supporting the table to 2015". A later central year would absorb the negative working-age band into the base rates and shorten the drift the HMI scale has to cover.
Both decisions are the same assumption seen from two ends, and each is being made without the other. A direct calculation explored in 2027 would measure 2024 and 2025 deaths against a base year LATF has not chosen. The volatility it is meant to test already sits inside the 2025 scale, which any company may elect this December if publication slips past September.