Projecting the unit fund in an IFRS 17 VFA model

Projecting the unit fund is the starting point of an IFRS 17 variable fee approach (VFA) model for a direct-participation contract. Before any of the standard's variable-fee mechanics can apply, the model needs a month-by-month projection of the policyholder's own fund: the balance that charges are taken from, that death and surrender payouts are measured against, and that the entity's own fee income is calculated off. Getting the roll-forward right, and knowing which cash flows the survival probabilities attach to, is the foundation the rest of a VFA measurement sits on.

IFRS 17 writes the liability of a direct-participation contract as the net of three pieces (paragraph B104): the obligation to pay the policyholder an amount equal to the fair value of the underlying items, the entity's own share of that fair value, and the fulfilment cash flows that do not vary with it. Which of the fund's monthly movements lands in which of those three pieces, and therefore whether a financial movement reaches the contractual service margin (CSM) or bypasses it, is set out in paragraphs B110 to B115. The sections below build up to that split in the order a model actually projects it: the product shape, the roll-forward, where the survival probabilities attach, the death strain, the B104 split itself, and finally what the CSM absorbs.

1 — The product shape

A VFA-eligible unit-linked contract with direct participation features works like this: the policyholder's premiums buy units in an identified pool of assets, the insurer earns its income by taking charges out of that fund, and on death the policy pays the higher of the fund value and the sum assured. This is an example product: unit-linked designs come in many permutations, and this shape is used throughout the post to illustrate the VFA calculations. That product shape produces three families of cash flow. Money flows into the fund as allocated premium. Money flows out of the fund as charges, which is the insurer's own fee income. And the shareholder pays some amounts directly, outside the fund: the death-benefit top-up above the fund value, and the entity's own expenses. Every step that follows traces one of those three families through to the liability.

2 — The monthly fund roll-forward

The fund is projected month by month, per model point, on the assumption the policy stays in force. The order of the roll-forward matters, because charges are taken at different points in the month:

Allocated premium means the contractual premium net of any bid/ask spread or product-specific exclusion; not all of the premium reaching the fund is a given, and it depends on the contract wording. Which charges sit at the start of the month and which sit at the end follows the product's own charging rules: expense-type charges such as flat admin fees and premium loading are often deducted at the start of the month, while mortality-type and fund-based charges typically apply after the investment return has been credited. Some implementations deduct risk charges at the start of the month instead. The roll-forward should follow the product's actual charging order, not a fixed convention, and the build-up presentation is worth checking against the product's own charging rules before it is trusted.

3 — Where the survival probabilities attach

The fund itself is projected in force. It is never multiplied by a survival probability, and it is never decremented for a claim. The survival probabilities attach only to the cash flows that leave the fund or are paid against it:

• charge income: charges(t) × tpx(t), where tpx(t) is the probability the policy is still in force at time t
• death strain: max(sum assured − fund(t), 0) × qx(t), where qx(t) is the probability of death in the period
• surrender payout: fund(t) × the withdrawal probability for the period
• maturity payout: fund(t) × tpx at the term date

Multiplying the fund balance itself by tpx as well would double-count the decrement. The death, surrender, and maturity payouts already carry their own probability, and the charge stream is only earned for as long as the policy survives to pay it. The fund path stays a single, unweighted trajectory per model point; the probabilities live on the cash flows that peel off it, not on the balance itself.

4 — The death strain

On death, the policy pays the higher of the fund value and the sum assured. The unit balances fund the first part of that payout; the shareholder is on the hook only for the excess:

As the projected fund grows, that gap narrows. Once the fund exceeds the sum assured, the strain is zero and the insurer carries no further death-benefit exposure on that policy. What remains economically significant from that point is the loss of future charge income if the policy exits: the exposure shifts from a mortality risk to a persistency one, which on mature unit-linked business is usually the larger driver of value.

5 — From the fund to the liability: the B104 split

The fund projection above, and the death strain built off it, feed directly into the three-way split paragraph B104 prescribes for a direct-participation contract.

The fund at value. The unit fund measures the obligation to pay the policyholder (B104(a)) directly; it needs no separate cash-flow projection or discounting of its own. That shortcut rests on the replicating-asset logic in paragraph B46: where an asset's cash flows exactly match a set of contractual cash flows in amount, timing and uncertainty, its fair value can stand in for projecting and discounting those cash flows. The policyholder's own units are the exact replicating asset for the unit-funded payouts (maturities, surrenders, the fund-value part of a death claim) by construction. It also holds together as a discounting check: because the fund is grown at the discount forward rates in the roll-forward above, discounting the projected unit payouts explicitly would simply return today's fund value. Holding the fund at face is the shortcut form of the same calculation. And because a premium allocated to units raises the fund and the policyholder's obligation by the same amount at that instant, future allocated premiums drop out of the liability legs entirely; they reach the measurement only through the fund path, which is what drives the future charges and the future strain.

The entity's share. The variable fee's first leg is specifically the amount of the entity's own share of the fair value of the underlying items (B104(b)(i)). In plain terms: what the entity will earn from the underlying items. A cash flow belongs there only if the entity itself retains it as compensation. The charges taken from the fund each month, such as risk charges, admin fees, and fund-management fees, are the entity's own retained fee income: projected off the fund path, weighted by tpx as in Section 3, and discounted. This leg reduces the liability because it is income to the entity. The "entity's share" wording also draws a boundary: a fund-linked deduction the entity passes straight through to a third party, an external manager's own fee taken from the fund, say, is never the entity's own retained share, however much it varies with the fund. It stays inside the fund roll-forward itself, reducing the future fund value the way any unit-linked deduction does, rather than being carved out as a second income leg.

The non-unit cash flows. The variable fee's second leg is the fulfilment cash flows that do not vary with returns on the underlying items (B104(b)(ii)). In plain terms: what the entity has to pay that does not depend on the underlying items — the death-strain top-up from Section 4, rider strains, renewal expenses, acquisition costs. These are projected, survival-weighted, and discounted the same way as any general-model cash flow, and they increase the liability. The actuarial tradition has a name for the netted version of the entity-share and non-unit legs together: the non-unit reserve, or, on UK unit-linked business specifically, the sterling reserve. Neither term appears in the standard itself. They are long-standing unit-linked vocabulary for the same idea IFRS 17 now writes as the variable fee in B104(b).

The CSM closes the split at inception. At initial recognition the contractual service margin is the number that stops a profitable group producing a day-one gain: the entity's share, less the non-unit outgo, less the risk adjustment. With a fund of 1,000, expected charges worth 300, non-unit outgo of 180 and a risk adjustment of 8, the CSM is 300 − 180 − 8 = 112 — and the liability for remaining coverage at inception nets back to exactly the fund value of 1,000. The whole expected profit on the contract sits inside the CSM at the start, waiting to be earned over coverage.

6 — What the CSM absorbs

Once the liability is split, paragraphs B110 to B115 route each piece's movement to a destination, and this is where VFA measurement diverges structurally from the general model.

The practical effect: under VFA, the CSM absorbs the financial movements on the entity's own fee and the non-underlying-item financial-risk effects, and re-releases them over coverage units as the entity provides insurance and investment-related service. The investment return on the policyholder's fund reaches the P&L directly as insurance finance income or expense; it never touches the CSM, because paragraph B111 treats it as belonging to the policyholder from the outset. That investment-return leg is exactly the piece the next post in this series works with: which part of it an accounting policy election lets an entity park in other comprehensive income instead of profit or loss.

One contrast with the general model is worth stating explicitly, because it trips up readers trained on GMM mechanics: the VFA CSM does not accrete interest at a locked-in rate. Locked-in accretion belongs to contracts without direct participation features (B72(b)); under the VFA, the financial re-measurement of the CSM arrives through the B112 and B113(b) adjustments above, at current values. A VFA model has no locked-in curve for the CSM to carry.

7 — How the liability moves period to period

The split above is a point-in-time picture. In production, each leg gets its own roll-forward, and the reviewable view is a movement table with one column per leg and one row per movement family:

Reading it column by column, each column is its own small waterfall from opening to closing balance, and the CSM column tells the profit-emergence story on its own: it opens, absorbs the future-service movements from the other columns, and releases its coverage-unit slice. The unit fund sits beside this table, rolled forward separately as in Section 2 — the movement table covers the legs that are projected and discounted, and the fund covers the piece that is held at value.

From fund projection to VFA measurement

If you are building or reviewing a VFA model for a direct-participation contract, whether the unit fund roll-forward, the paragraph B104 split, or the CSM routing above, work on the methodology, the engine configuration, or the review framework sits within the scope of the IFRS 17 work on this site.