The CSM release ratio, the share of an insurer's opening Contractual Service Margin recognized as profit in a period, has become the primary lens analysts use to judge IFRS 17 earnings quality, displacing embedded value and value-of-new-business as the life insurer KPI. KPMG's April 2026 review of 55 global insurers' 2025 annual statements found it varying widely by product line.

The variation is the point. Under IFRS 4 the number was an internal actuarial planning figure. IFRS 17 made it public and comparable.

Key Takeaways

  • 55 insurers reviewed, with the comparison deliberately restricted to General Measurement Model and Variable Fee Approach business. Premium Allocation Approach carriers hold no comparable CSM stock, so this is a life and long-duration health metric.
  • $6,224 million released by AIA during 2025 against a CSM balance that still grew 15% to $64,945 million, roughly 9.6% of the closing balance, which is what new business outpacing runoff looks like.
  • Lapse is a release-timing driver, not a secondary assumption. Surrenders shrink the remaining service obligation and accelerate recognition; higher persistency stretches the profile and understates near-term earnings.
  • No smoothing reserve absorbs a miss. Adverse experience on future service adjusts the CSM directly, and once a contract group's CSM is exhausted, further deterioration hits current-period profit as a loss component.
  • A fast release ratio has two opposite meanings, orderly runoff of a legacy block or unlocks eating the stock faster than pricing assumed, and the ratio alone does not distinguish them.

How the Ratio Is Built

The CSM release ratio is CSM released during the period divided by the opening CSM balance, before the release is deducted. KPMG restricted its comparison to insurers applying the General Measurement Model or the Variable Fee Approach, since non-life carriers mostly apply the Premium Allocation Approach and carry no comparable CSM stock. Any cross-industry comparison ignoring measurement model mix produces a distorted read.

Even inside the GMM and VFA universe the ratio is period-length and product-type dependent. A ten-year term life block releases CSM on a different amortization schedule than a forty-year whole life block or a unit-linked savings product with an open-ended coverage horizon. Third-year data shows insurers refining how they disclose the components of that release, with finer breakdowns of new contracts, changes in estimates and the unwind of the risk adjustment.

Product lines are separating into distinct behavior. Protection business, term life and group risk with fixed, relatively short coverage periods, produces the most stable ratios, because coverage units decline mechanically with in-force counts rather than with markets. Savings and unit-linked contracts under VFA are far more sensitive, because the CSM absorbs the insurer's share of underlying fund performance, so a strong equity year moves the ratio independent of underwriting experience. Fixed annuities and traditional whole life sit between the two.

Product categoryMeasurement modelRelease-ratio behavior
Term life, group risk (protection)GMM, PAA in some marketsStable, predictable; tied mechanically to in-force decrement
Fixed annuities, traditional whole lifeGMMModerate sensitivity via discount-rate unwind
Savings, unit-linkedVariable Fee ApproachHigh sensitivity; CSM absorbs market-performance pass-through

Company disclosures show the range. AIA's CSM balance grew 15%.pdf) to $64,945 million at year-end 2025 after releasing $6,224 million into operating profit, roughly 9.6% of the closing balance. Allianz reported a EUR 1.3 billion release in the third quarter of 2025, in line with its own expectations. Both sit at the healthy end because new-business CSM addition is running ahead of in-force runoff. A legacy-heavy insurer running off a closed savings block shows a shrinking balance at a comparable ratio, which reads very differently.

Behavior Drives the Release, and Unlocking Makes It Visible

Actuaries used to treating lapse and persistency as secondary to mortality and morbidity need to reverse that hierarchy for CSM release modeling.

Lapses and surrenders shrink the remaining service obligation, which accelerates CSM recognition into the current period. Higher persistency does the opposite, stretching the release profile forward and understating near-term earnings relative to a higher-lapse block. On savings and unit-linked products, benefit elections at maturity or surrender pull release forward or push it back depending on how coverage units are defined, since IFRS 17 allocates the CSM across coverage units in proportion to the insurance contract services still owed.

That has a direct pricing consequence. A block priced on an optimistic persistency assumption shows a slower release ratio at inception, which reads to an analyst as earnings conservatism, until realized lapses come in above assumption and the ratio jumps as the coverage-unit base shrinks faster than modeled. That jump is not organic income growth. It is an assumption catch-up running through the same profit and loss line as organic release, and nothing in the headline figure separates the two.

Unlocking is what gives the ratio its force. Any deviation between actual and assumed experience on future service, whether mortality, lapse, morbidity or expense, adjusts the CSM directly rather than flowing through a separate experience-variance line the way legacy frameworks handled it. For a profitable contract group, an adverse variance shrinks the stock available for future release. Once the CSM is exhausted, further deterioration recognizes immediately as a loss component in current profit.

The actuary who set the original assumption carries that exposure directly. An over-optimistic lapse or morbidity assumption baked in at transition or at new-business recognition no longer produces a reserve estimate that develops slowly over years. It produces a disclosed unlocking event traceable to a specific assumption change, a specific reporting period, and a specific block.

The Ratio Alone Does Not Say Which Story It Is

A rapidly declining CSM balance supports two readings that point in opposite directions, and the ratio does not choose between them.

It can reflect deliberate, orderly runoff of a legacy block the insurer no longer writes, in which case a fast release is the expected mechanical unwind of business nearing the end of its coverage period. Or it can reflect adverse experience unlocks eating into the stock faster than the original pricing assumed, in which case the identical ratio is a signal about assumption discipline. Reading the ratio without the balance trend and the disclosed drivers of change cannot separate them, which is why the finer component breakdowns KPMG observed matter more than the headline percentage.

The comparison problem gets harder for groups reporting under both frameworks. US insurers under ASU 2018-12 separate market risk benefits, including variable annuity guarantees such as guaranteed minimum death and income benefits, from the host contract and fair-value them each period through net income, routing only the own-credit component to Other Comprehensive Income. IFRS 17 takes a different view of the same exposure: for products like fixed indexed annuities it treats underwriting and financial risk as too intertwined to separate, so there is no host-and-embedded-derivative split and market-driven volatility on VFA business flows through the CSM itself.

The CSM release ratio and the LDTI net income impact of MRB remeasurement are therefore not interchangeable views of the same economics. A market move producing a sharp MRB swing under one standard can produce a muted CSM movement under the other, depending entirely on which risks each framework carved out for fair-value treatment.

The profession has not settled the underlying practice either. The Actuaries Institute's February 2026 post-implementation survey, covering the Australia and New Zealand market, found continued practitioner demand for worked examples and technical guidance specifically on CSM, loss component and related mechanics three years after transition. Reinsurers surveyed separately said the general IFRS 17 guidance was not sufficiently tailored to reinsurance contracts, which leaves reinsurance CSM calculations running on interpretation without market-specific precedent behind them.

Further Reading

Sources