Technical Notes#

Status: Draft, 2026-08-03 (all cited sources accessed 2026-08-03).

Scope note. These notes specify a reference liability cash-flow projection model for the standardized composite product defined in product-spec.md (same directory). This is not any single insurer’s product. [S#]/[R#] tags refer to the source list in sources.md, whose numbering is carried verbatim from _research/income-protection.md; [REG-R#] tags refer to the cross-product reference library references/regulatory-and-actuarial-references.md (its own R-numbering; research provenance in _research/regulatory-actuarial.md). std marks standardizations introduced for the reference implementation; unverified marks claims not confirmed against a retrieved document. Parameter values are identical to those in product-spec.md.

Model scope and conventions#

Purpose. Project gross best-estimate liability cash flows (premiums, benefit outgo, expenses) for a single-policy model point of full-term guaranteed-premium own-occupation IP. Reserves are not computed (see Valuation and reserve pointers).
Model structure. Three-state multiple-state model — healthy/active (H), sick and in claim payment (S), dead (D) — the structure of the CMI graduations: CMIR 12 introduced the healthy–sick–dead multiple-state model for UK PHI, applied to the IPM 1991-98 graduations and carried through to the IP11 Series R4 R1. (The CMIR 12 report title/date is recorded from search summaries only unverified.) Lapse is an additional exit from H. Experience is parameterized exactly as the CMI publishes it: claim inception rates by sex, deferred period and occupation class, and claim termination rates split by recovery and death, duration-dependent R1 R2.
Projection frequency. Monthly grid, matching the monthly-in-arrears benefit [S1] [S3] [S10]. Annual assumption rates are converted to monthly per the formulas below std.
Timing conventions std. Premiums received at the beginning of the policy month (BOM) from lives in H; state transitions occur at end of month (EOM); benefit for month t is paid at EOM to lives in claim payment throughout month t — in S at BOM and still in S at EOM (monthly in arrears [S1]; a claim incepting at EOM t receives its first payment at EOM t+1). The contractual daily pro-rating of partial claim months [S1] [S3] [S10] is replaced by whole-month payment std. Escalation applies at BOM of each anniversary month.
Age basis. Age nearest birthday at entry, advancing with policy year std. No public statement of the IP11 age definition was retrieved (the briefing note records graduated age ranges 17–65 M / 17–60 F, extended to 70 R1); the choice is a pure convention and must be revisited by CMI Authorised Users.
Claim duration. Measured in months since claim (payment) inception, i.e. since the end of the deferred period std convention; IP11 termination rates are two-dimensional in age and sickness duration, with run-in periods of increasing recovery rates at early durations for DP4/13/26 R1. Whether the CMI duration clock runs from sickness onset or payment start was not extracted from public documents — subscribers must align the convention with the tables they license.
Currency and units. GBP; benefit and premium in £/month; rates are probabilities per period unless labelled “per mille”.
Model points. Single-policy model points projected on an expected (probability-weighted) basis; survivorship/state probabilities multiply per-policy cash flows. In-force portfolios need both active cells and claims-in-payment cells (with claim duration as a model-point attribute).

Model point attributes#

Attribute	Type	Example (base cell)
`entry_age`	int	35 std
`sex`	enum {M, F}	M std — IP11 rates are sex-split R1
`occ_class`	enum {1, 2, 3, 4}	1 std — CMI occupation classes OC1–OC4 R1
`benefit_monthly`	currency (£/month at issue)	2,000 std
`earnings_annual`	currency (underwriting record)	40,000 std
`deferred_weeks`	enum {4, 8, 13, 26, 52}	26 std
`expiry_age`	int (50–70)	65 std
`escalation`	enum {none, RPI}	RPI (capped 10%, premium ×1.5) [S1] [S2]
`premium_monthly`	currency (£/month at issue)	35 std — rates not public
`premium_basis`	enum {guaranteed} (reviewable/age-costed out of scope)	guaranteed
`status`	enum {active, in_claim}	active
`claim_duration_months`	int (in-claim cells only)	0

Smoker status is not an attribute: the IP11 rate structure is sex / deferred period / occupation class R1; any smoker differentiation sits in insurer pricing, which is not public.

State variables#

Variable	Description	Updated
`l_H(t)`	Probability in state H (active premium-payer, incl. any sickness spell still inside the deferred period — see Deferred-period mechanics) at EOM t	monthly
`l_S(t, z)`	Probability in claim payment at EOM t with claim duration z months (z = 1, 2, …)	monthly, two-dimensional
`l_S(t)`	Total in-claim probability = Σ_z l_S(t, z)	derived
`B(y)`	Escalated monthly benefit in policy year y	at anniversaries
`P(y)`	Escalated monthly premium in policy year y	at anniversaries
`AP(y)`	Amount payable per month of full incapacity (spec formula; base: = B(y))	at anniversaries
`n(t)`	New claim inceptions during month t	monthly
`rec(t)`, `dth_S(t)`, `dth_H(t)`, `lps(t)`	Exits: recoveries, deaths in claim, deaths in H, lapses	monthly

There is no account value, surrender value or unit fund: the contract has no cash-in value at any time [S4] [S5] [S7], so the only state is the insured population itself.

Assumption inputs#

Three classes are distinguished explicitly.

(a) Contractual / guaranteed elements (cited; from the spec)#

Input	Value	Basis
Deferred period d	26 weeks (base cell)	menu [S6]; pick std
Benefit formula parameters	65% / 50% bands, £60,000 breakpoint, £20,000/month cap, £1,500 guarantee, 90% tolerance	[S1] [S2] [S5] [S7]; picks std (spec footnotes 6–8)
Escalation mechanics	j(y) = min(max(RPI_y, 0), 0.10); B ×(1+j); P ×(1+1.5j); continues in claim	[S1] [S2]; multiplier pick std
Premium guarantee	Guaranteed level apart from escalation	[S1] [S3] [S5] [S7]
Waiver of premium	No premiums from lives in S (payable through the deferred period)	[S5] [S7] [S10] [S11]; convention std
Linked claims	Same-cause recurrence within 52 weeks: no new deferred period	[S1] [S3] [S5] [S7] [S10] [S11]; window pick std
Proportionate benefit	(A − B)/A × C on partial return to work	[S7]; common structure [S1] [S3] [S5] [S10] [S11]
Expiry	All cover and claim payments cease at the policy end date (age 65 base cell)	[S1] [S3] [S5] [S7] [S10]; age pick std
Grace	60 days, cancellation without value	[S1] [S3]; pick std

(b) Insurer-discretionary current elements#

For the guaranteed-premium full-term composite these are deliberately thin: there are no bonuses, no reviewable charges, and no market value reductions — premiums are guaranteed [S1] [S3] [S5] [S7] and there is no surrender value [S4] [S5] [S7]. Recorded for the variations only:

Reviewable premiums (variation): fixed for 5 years, then reviewed with no contractual cap [S1] [S4] [S6] [S10]. Review formulas are discretionary and undisclosed (research file gap) — any reviewable-premium model needs a std review rule; none is specified here.
Holloway surplus participation (out-of-scope variation): surplus and bonus allocations plus a discretionary terminal bonus, on With-Profits Actuary advice [S11] [S12] — requires a capital-account state not present in this model.
Escalation index snapshot: future RPI is an economic input, not insurer discretion; the reference snapshot is RPI = 3.0%/yr flat std, so j = 0.03, premium growth 4.5%/yr while premiums are payable.

(c) Behavioral / experience assumptions (modeler’s view)#

The authoritative UK experience basis is the CMI IP11 Series (individual IP, 2007–2016 data): claim inception rates by sex, deferred period (DP1/4/13/26/52) and occupation class (OC1–OC4); termination rates split recovery vs death, two-dimensional in age and sickness duration; table naming IP11 {M/F} DP{d} OC{n} {Inc/Rec/Dth} R1. The rate values are restricted to CMI Authorised Users (the working papers and the IP Rate Table Tool are subscriber-only R2 R3 R5; the CMI access model is per REG-R22), so the reference basis below is a std proxy shaped like the IP11 structure — the values are NOT IP11 values and carry no CMI authority. Known data issue: IP11 inception rates are understated due to exposure errors; the CMI published indicative adjustments alongside WP136 (terminations unaffected) R1 R2 R3 — users with table access must apply them.

Input	Reference basis	Basis tags
Claim inception rates ι_a(a)	std proxy table below (structure: M, DP26, OC1 R1)	values std
Recovery rates ρ_a(z)	std proxy table below, by claim duration year (IP11 is duration- AND age-dependent R1; age suppressed std)	values std
Mortality in claim q_S_a(z)	std proxy, flat 3%/yr all durations (IP11: duration-dependent to 5 years, age-only beyond R1)	value std
Active-life mortality q_H_a(a)	ONS UK national life tables qx (sex-specific), ×100% factor	table REG-R32; factor std (1)
Mortality/morbidity improvement	None in base std; CMI_20xx with a std long-term rate is the projection convention for mortality	REG-R30
Lapse w_a(y)	std table below; no public UK IP lapse study was retrieved	std
Maintenance expense	£60/policy/yr, inflating 3.0%/yr	std
Claim management expense	£300/yr per claim in payment, inflating 3.0%/yr	std
Offset/guarantee effect	AP(y) = B(y) — amount-payable ratio 1.0 (offsets and guarantee assumed not to bite)	std (2)
Claim severity factor k	1.0 (proportionate/rehabilitation claims not projected separately)	std (2)
Discount	PRA risk-free term structure for valuation R7 REG-R1; flat 3.0%/yr in the worked example only	rate std

ONS national life tables are the only freely redistributable UK mortality source (Open Government Licence); population mortality is heavier than insured experience REG-R32. CMI assured-lives table names are public (e.g. AM92/AF92 REG-R24) but current insured tables are Authorised-User-restricted REG-R22. Active-life mortality is a minor decrement in IP; the ×100% factor is a placeholder to be replaced with portfolio experience.
Offsets, the minimum benefit guarantee, and proportionate benefits change the amount paid relative to the chosen benefit (spec, Contractual mechanics). The base model pays the full escalated benefit; portfolio calibrations should set AP/B < 1 or k < 1 from claims experience.

std proxy claim inception rates (annual, per mille of lives in H; male, OC1, DP26; linear interpolation between pivot ages; pure placeholders):

Age a	30	35	40	45	50	55	60	64
ι_a(a) ‰	1.0	1.3	1.8	2.6	4.0	6.5	10.0	14.0

std proxy claim termination rates (annual, by claim duration year since payment inception; pure placeholders):

Claim duration year	1	2	3	4	5+
Recovery ρ_a	0.40	0.25	0.15	0.10	0.05
Death in claim q_S_a	0.03	0.03	0.03	0.03	0.03

The declining-with-duration recovery shape mirrors the qualitative structure the CMI publishes (duration-dependent termination rates R1); the IP11 “run-in” feature (recovery rates increasing over the first weeks of claim for DP4/13/26 R1) is not reproduced at this granularity — a monthly refinement point for table licensees.

std lapse table (annual rates from H; lives in S do not lapse — premiums are waived and the benefit is valuable std):

Policy year	1	2	3–5	6+
w_a(y)	10%	8%	6%	4%

Cash flow components and recursions#

Notation (defined once, used throughout)#

Symbol	Meaning
t	policy month, t = 1..T; T = 12 × (expiry_age − entry_age) = 360 (base cell); y = ceil(t/12); attained age a = entry_age + y − 1
B(y), P(y), AP(y)	escalated benefit, premium, amount payable (£/month); B(1) = 2,000, P(1) = 35 std
j	escalation rate = min(max(RPI, 0), 0.10); snapshot 0.03 std
ι_m(a)	monthly claim (payment) inception rate = 1 − (1 − ι_a(a))^(1/12) std
q_H_m(a), w_m(y)	monthly active mortality and lapse, same annual-to-monthly conversion std
ρ_m(z), q_S_m(z)	monthly recovery and death-in-claim rates at claim duration z months (from the annual rate of the duration year containing z) std
s_S(z)	monthly in-claim survival = (1 − ρ_m(z)) × (1 − q_S_m(z)) (independent decrements std)
e_m(y), ec_m(y)	monthly maintenance and claim-management expense = 60/12 and 300/12, × 1.03^(y−1) std
v(t)	discount factor to time t (valuation: PRA risk-free curve R7 REG-R1; worked example: 1.03^(−t/12) std)
l_H(t), l_S(t, z)	state probabilities (state-variable table); l_H(0) = 1, l_S(0, ·) = 0 for an at-issue cell

Dimensional check: ι, q, w, ρ are monthly probabilities (dimensionless); B, P, AP, e are £/month; every cash flow below is £ per month per policy issued.

Monthly processing order std #

At month t (skip all steps from t > T; at t = T all cover and any claim in payment terminate without value [S1] [S3] [S5] [S7] [S10]):

Anniversary (BOM, months t = 13, 25, …): B(y) = B(y−1) × (1 + j); P(y) = P(y−1) × (1 + 1.5 × j) [S1] [S2]. In-claim benefit escalates identically [S1] [S2] — the same B(y) applies to lives in S.
Premium income (BOM): PREM(t) = P(y) × l_H(t−1). Lives in S pay nothing (waiver [S5] [S7] [S10] [S11]); lives in H still inside a deferred period pay normally (see Deferred-period mechanics).
Transitions (EOM), from H — order death, then lapse, then inception among survivors std:
- dth_H(t) = l_H(t−1) × q_H_m(a)
- lps(t) = l_H(t−1) × (1 − q_H_m) × w_m(y)
- n(t) = l_H(t−1) × (1 − q_H_m) × (1 − w_m) × ι_m(a) (new claims at duration z = 1)
Transitions (EOM), from S — order recovery, then death std, per duration cohort z:
- rec(t, z) = l_S(t−1, z) × ρ_m(z)
- dth_S(t, z) = l_S(t−1, z) × (1 − ρ_m(z)) × q_S_m(z)
- l_S(t, z+1) = l_S(t−1, z) × s_S(z)
- l_S(t, 1) = n(t)
State update: l_H(t) = l_H(t−1) × (1 − q_H_m) × (1 − w_m) × (1 − ι_m) + Σ_z rec(t, z) (recovered lives return to H and are again exposed to inception std; see the linked-claims limitation below).
Benefit outgo (EOM): BEN(t) = k × AP(y) × [l_S(t) − n(t)] — i.e. paid to the surviving cohorts z ≥ 2 only (equivalently Σ_z l_S(t−1, z) × s_S(z)): benefit is monthly in arrears [S1], so new inceptions n(t) = l_S(t, 1), seeded at EOM t, receive their first payment at EOM t+1. (Including n(t) in BEN(t) would pay a full month’s benefit at the instant of payment inception and break the inception-annuity equivalence in Active-lives valuation.) Whole-month convention std.
Expenses (EOM): EXP(t) = e_m(y) × [l_H(t−1) + l_S(t−1)] + ec_m(y) × l_S(t−1).
Discount cash flows at v(t) and accumulate.

Net cash flow (insurer perspective): CF(t) = PREM(t) − BEN(t) − EXP(t). Death and lapse generate no payment (no death benefit, no surrender value [S4] [S5] [S7]; the out-of-scope £5,000–£10,000 death benefits in two sampled contracts [S3] [S5] would add a dth × DB term).

Deferred-period mechanics#

The contract pays after d weeks (26, base cell) of continuous incapacity, with premiums payable through the deferred period and waived from benefit start (spec). The model embeds the deferred period in the inception basis: ι is a claim payment inception rate specific to DP26 — exactly the quantity the CMI publishes per deferred period R1 — so sickness spells that recover inside the deferred period never leave H, and lives sick within the deferred period remain in H (still premium-paying, matching the contractual waiver-from-payment-start convention std, spec footnote 16). Consequences:

No separate “sick, not yet in payment” state is needed; the d-week lag between onset and payment is absorbed into ι’s calibration. A timing refinement (shifting inception cash flow impact by d weeks) is second-order at DP26 std.
Dual deferred periods and sick-pay-linked NHS/teacher deferreds [S1] [S3] [S5] [S7] [S10] would need spell-level modeling and are out of scope.
Linked claims limitation std: contractually, a same-cause recurrence within 52 weeks of payments stopping restarts payment without a new deferred period (spec). The base model returns recovered lives to H with the standard DP26 inception basis, which understates short-horizon re-inception. Refinement: a post-recovery flag with a loaded ι for 12 months; not specified further here.

Claims-in-payment valuation (disabled-life annuity)#

A claim in payment is valued as a disabled-life annuity: expected present value of the escalating benefit until recovery, death or expiry — the “claim annuity values” the CMI Rate Table Tool produces for subscribers R5. For a claim at duration z0 months, attained age a0, with T_rem months to expiry:

a_dis(a0, z0) = Σ_{m=1}^{T_rem} [ Π_{i=1}^{m} s_S(z0 + i − 1) ] × k × AP(y(m)) / AP(y(0)) × v(m)

so that claims-in-payment BEL outgo per £1/month of benefit in payment is a_dis, and the cell’s benefit liability is AP × a_dis + the claim-expense annuity (same survival, ec_m in place of AP). Escalation enters through AP(y(m)) (step-ups at policy anniversaries [S1] [S2]); the annuity truncates at expiry — payments stop at the policy end date [S1] [S3] [S5] [S7] [S10].

Active-lives valuation#

Active-life BEL cash flows are steps 1–8 run from the valuation date: premium income from l_H, benefit outgo from claims yet to incept (each n(t) seeds a new duration cohort), and expenses. Equivalently, the benefit side can be written as Σ_t v(t) × n(t) × [k × AP × a_dis(a(t), 0-month equivalent)] — the inception-annuity decomposition of the same multi-state projection.

Alternative: inception-annuity method [brief]. The historical alternative prices each year’s claim cost as (inception rate) × (disabled-life annuity at claim start) without tracking in-claim cohorts through time — adequate for premium rating, but it cannot roll claims-in-payment forward or produce per-period BEL cash flows, which is why the multi-state formulation is the reference structure. The research file records the pre-CMIR 12 history (Manchester Unity sickness-rate basis; inception-annuity vs multi-state reserving) as unverified textbook knowledge — no public IFoA source was retrieved; the CMIR 12 → IPM 1991-98 → IP11 multi-state lineage itself is verified at landing-page level R4.

Policyholder behavior modeling#

All dynamic formulas are std reference constructions; no public UK IP policyholder-behavior study was retrieved.

Base lapse std. w_a(y) per the table above; monthly w_m = 1 − (1 − w_a)^(1/12). Applied to H only; lives in S never lapse (premiums waived, benefit in payment) std.
Premium-shock lapse std. With escalation on, premiums rise 1.5 × j each year; the model multiplies lapse by M_esc(y) = 1 + 2 × max(0, 1.5 × j(y) − 0.05) in anniversary years (lapse response to premium increases above 5%; e.g. j at the 10% cap gives 1.5 × 0.10 = 15% premium growth and M_esc = 1.2). Contract anchor: sampled insurers let policyholders decline escalation increases, with the option lapsing after consecutive refusals — two consecutive cancelled increases end the option in one contract [S5]; declining three consecutive increases removes it in another [S11]; declines are modeled as lapse of the escalation margin only at portfolio level — the base single-cell model keeps escalation always-on and uses M_esc as the aggregate proxy.
Economic-cycle morbidity link [std note]. Claim inceptions are widely believed to rise (and recoveries to slow) in recessions — job insecurity raises claim propensity on an own-occupation definition. No sampled document or public CMI output quantifies this; it is recorded here as a scenario overlay ι × M_cycle, ρ / M_cycle with M_cycle = 1 in base std, not as a calibrated assumption.
GIO take-up, alterations, career breaks: held at zero (spec: out of scope).

Worked example#

Claims-in-payment recursion for the base cell’s level-cover variant std (the escalation step falls outside the 3-month window shown): a claim in payment from duration z = 0, benefit AP = B = £2,000/month, duration-year-1 proxy terminations (ρ_a = 0.40, q_S_a = 0.03 std), discount 3.0%/yr flat std.

Monthly factors (derived): ρ_m = 1 − 0.60^(1/12) = 0.041675; q_S_m = 1 − 0.97^(1/12) = 0.002535; in-claim survival s_S = 0.958325 × 0.997465 = 0.955895; v = 1.03^(−1/12) = 0.997540.

Month m	l_S start	Recoveries ρ_m × l_S	Deaths (1−ρ_m) q_S_m × l_S	l_S(m) = l_S × s_S	Benefit 2,000 × l_S(m)	v^m	PV
1	1.000000	0.041675	0.002429	0.955895	1,911.79	0.997540	1,907.09
2	0.955895	0.039837	0.002322	0.913735	1,827.47	0.995086	1,818.49
3	0.913735	0.038080	0.002220	0.873434	1,746.87	0.992638	1,734.01

Three-month PV of benefit outgo: £5,459.59 per claim in payment. Trace, month 1: survival s_S = (1 − 0.041675) × (1 − 0.002535) = 0.955895; expected benefit paid at EOM = 2,000 × 0.955895 = £1,911.79 (in-arrears convention: exits during the month receive nothing under the whole-month simplification std; contractually they would receive a daily pro-rated amount [S1] [S3] [S10]); PV = 1,911.79 × 0.997540 = £1,907.09. Claim expense follows the same survival column at ec_m = 300/12 = £25.00 per month std. On the active-lives side, the same conventions give month-1 premium income P × l_H(0) = £35.00 and expected new inceptions n(1) ≈ ι_m(35) = 1 − (1 − 0.0013)^(1/12) = 0.000108 — each seeding this in-claim recursion at z = 1.

Valuation and reserve pointers#

This library projects gross best-estimate liability cash flows; valuation layers consume them and are NOT reproduced here:

Solvency UK. Technical provisions = best estimate + risk margin; the best estimate is the probability-weighted average of future cash flows discounted at the relevant risk-free term structure R7 REG-R1 — i.e. exactly the active-life and claims-in-payment projections above, both premium and claim sides. Risk margin: cost-of-capital method at 4% with life λ-tapering 0.9 REG-R4 — cited-not-specified. The claims-in-payment element of IP is an MA “eligible element” where organised and managed separately R8 REG-R2: the disabled-life annuity cash flows are the portion a UK insurer may discount at risk-free + MA under an MA permission.
IFRS 17. UK-adopted IFRS 17 (effective 2023) applies to IFRS-reporting UK life insurers REG-R38; the fulfilment-cash-flow engine consumes the same projections with its own discounting, risk adjustment and CSM layers (cited-not-specified).
Professional standards. TAS 100 (all technical actuarial work) and TAS 200 (insurance-specific) govern UK actuarial use of such models R10 (fetched_ok=false in the product research pass; verified via REG-R33 REG-R34).

Key sensitivities and model risks#

Dominant assumptions, in rough order:

Claim inception and termination (recovery) rates. They set both sides of the liability: inceptions drive new-claim frequency, recoveries drive claim length — a small recovery-rate change compounds across the whole disabled-life annuity. Both proxy tables here are std placeholders; the real IP11 basis is restricted R1 R2 R5 REG-R22, and IP11 inceptions carry a known understatement requiring the WP136 indicative adjustments R1 R3. Sensitivity-test ι and ρ first, and independently.
Morbidity trend. The base holds morbidity level std; cause-mix shifts (notably mental-health claims, which interact with own-occupation assessment) move both ι and ρ. No public quantification was retrieved — treat as a scenario axis, not a calibrated input [std note].
Economic sensitivity of claims. Recession-linked inceptions and slowed recoveries (the M_cycle overlay) are the classic IP experience risk on own-occupation business — a std scenario note, deliberately not calibrated here.
Escalation/inflation. RPI-linked benefit escalating in claim [S1] [S2] makes the disabled-life annuity inflation-sensitive precisely when it is longest; the 10% cap is an embedded inflation option. The premium side compensates only ×1.5 on actives, and not at all on claims in payment (waiver).
Lapse. Second-order for claim cost but first-order for premium income and deferred-acquisition economics; the table is std with no public anchor.

Known modeling pitfalls:

Duration dimension. Collapsing l_S(t, z) to a single bucket with duration-independent termination rates materially misstates claim runoff — the duration gradient (0.40 → 0.05 in the proxy) is the defining feature of IP terminations R1.
Premium waiver double-count. Projecting premium income from l_S lives overstates premiums; premiums come from l_H only (waiver [S5] [S7] [S10] [S11]).
Expiry truncation. The disabled-life annuity must truncate at the policy end date [S1] [S3] [S5] [S7] [S10]; an untruncated annuity materially overstates liabilities for claims incepting near expiry.
Amount payable vs chosen benefit. Offsets, the £1,500 guarantee and proportionate benefits mean amounts paid can differ from B; modeling AP = B std overstates outgo where the maximum-benefit formula bites (and understates nothing — AP ≤ B always).
Linked claims. Returning recovered lives to the standard inception basis ignores the waived deferred period on 52-week recurrences (see limitation note) — understates outgo for short-recovery portfolios.
Run-in periods. IP11 recovery rates increase over the first weeks of claim for DP4/13/26 R1; annual duration-year granularity std smooths this away — significant for short deferred periods, less so at DP26.
Basis-structure mismatch. The proxy termination rates drop the age dimension std; IP11 is two-dimensional (age × duration), with claimant mortality duration-dependent to 5 years and age-only beyond R1. Table licensees should restore both dimensions.
Escalation timing. B escalates on policy anniversaries in this model std; some contracts escalate in-claim amounts on claim anniversaries with index-lag rules (e.g. RPI five months prior, post-claim catch-up [S10]) — align with the contract being modeled.