Executive summary
- The risk adjustment covers non-financial risk only: insurance risk, lapse risk and expense risk, not market or credit risk.
- Common methods are quantile approaches such as value at risk and tail value at risk, cost-of-capital approaches, and margins applied to individual assumptions.
- Whatever the method, the entity must disclose the confidence level to which the adjustment corresponds, which forces comparability at the point of disclosure.
- Diversification benefit may be recognised only to the extent it reflects the entity's own view of compensation required, and the chosen entity level must be applied consistently.
- The risk adjustment unwinds over time; its release is part of insurance service result, not investment result.
What the risk adjustment does and does not cover
The risk adjustment answers a specific question: what compensation does the entity require for bearing the uncertainty in non-financial cash flows, over and above the expected value? It is entity-specific by construction. Two insurers facing identical risk may legitimately hold different risk adjustments because their risk aversion, diversification and capital position differ.
- In scope: insurance risk, lapse and persistency risk, expense risk, and the uncertainty in claims timing.
- Out of scope: market risk, credit risk on assets, and operational risk not arising from the fulfilment of insurance contracts.
- Excluded from the CSM: the risk adjustment is a separate building block and is presented separately in the reconciliations.
The three method families
Quantile methods estimate a distribution of the fulfilment cash flows and take the difference between a selected percentile and the mean. Cost-of-capital methods project the capital required to run off the obligations and charge a rate for holding it. Explicit margin methods load individual assumptions and aggregate the effect.
RA_VaR = Q_alpha(L) - E[L], RA_CoC = sum_t c * SCR_t / (1 + r_t)^t
- Q_alpha(L)
- — Selected quantile of the non-financial loss distribution
- E[L]
- — Probability-weighted mean of the same distribution
- c
- — Cost-of-capital rate applied to required capital
- SCR_t
- — Capital required for non-financial risk in year t of run-off
| Method | Strengths | Practical difficulties |
|---|---|---|
| Value at risk / quantile | Directly interpretable, aligns with the disclosure requirement | Needs a credible full distribution, sensitive in thin-data lines |
| Tail value at risk | Captures severity beyond the threshold, more stable for skewed risks | Higher data demands, less intuitive to non-technical audiences |
| Cost of capital | Consistent with Solvency II risk margin, economically grounded | Requires a capital projection and a rate that must be justified |
| Explicit assumption margins | Simple, auditable, workable for small portfolios | Aggregation is crude, confidence level equivalent must still be derived |
Diversification and the level of aggregation
IFRS 17 permits recognition of diversification benefits only to the extent that they reflect the compensation the entity requires. In practice the entity chooses a level — legal entity, group, or portfolio — calculates the risk adjustment there, and allocates it down to groups of contracts on a documented, consistent basis. The allocation basis is itself a judgement that affects which groups appear onerous.
The confidence level disclosure
Entities using a technique other than a confidence level must disclose the confidence level to which their result corresponds. This translation is not always straightforward: a cost-of-capital result must be mapped back onto a distribution, which requires the distribution the method was designed to avoid estimating. Most practitioners solve this with an approximate calibration exercise, documented and reviewed periodically.
Practical example
A medium-sized non-life insurer holds discounted best-estimate liabilities of 400 for incurred claims. A bootstrap of reserve variability suggests a coefficient of variation of 12 per cent and mild right skew. A 75th percentile risk adjustment implies roughly 32; a 90th percentile implies roughly 62. Both are defensible, and the disclosed confidence level is what makes the difference visible to a reader.
Under a cost-of-capital calibration with a 6 per cent charge and a run-off pattern averaging four years of capital at 130, the result is close to 31 — broadly the 75th percentile. Presenting both calculations side by side is the strongest evidence that the number is a considered position rather than an accounting plug.
Limitations and caveats
- Entity-specific compensation is a conceptually clean idea that resists external verification; auditors test process and consistency more readily than level.
- Confidence level equivalents derived from cost-of-capital methods are approximations and can drift as the portfolio changes.
- Diversification recognised at group level may not be available at the legal entity where the obligations actually sit.
- Thin-data lines produce unstable quantile estimates, and stability is often bought by judgement rather than data.
Conclusion
The risk adjustment is where an insurer states, in a single number, how much uncertainty it thinks it is carrying. Consistency across periods matters more than the method chosen.
Firms that align the risk adjustment with their internal capital view get a coherent story across accounting, capital and pricing; firms that treat it as an accounting exercise get three inconsistent narratives.
References
- IFRS 17 Insurance Contracts — International Accounting Standards Board, 2020
- Risk Adjustments for Insurance Contracts under IFRS 17 — International Actuarial Association, 2018
Author bio
Jonas Mohamed Osman Abdelghafour, known as Yonas Osman is an actuary, FRM and financial risk professional specialising in banking, insurance, model risk, capital modelling and quantitative risk management.