Learn / Advanced ALM

Actuarial Liability Modeling

One model point stands in for 10,000 identical deferred-income contracts. Mortality, lapse, and the discount rate change the projected cash flows and the present value. This is not seriatim valuation — valuing every contract one by one — not an experience study, and not a production assumption set.

01

Why a model point

A full policy extract is millions of rows. A model point is a bucket: same age, same deferral, same payment, same assumed decrements. The lab is one bucket. Policy-level modeling keeps every contract; the factory stage “Actuarial Model” is where that extract would be valued.

Analogy: instead of pricing every commuter’s ticket, the railroad prices one typical rider and multiplies. The typical rider is the model point. If that rider is a bad average, the total is wrong even if the arithmetic is exact.

Seriatim means one contract at a time. A seriatim run keeps each policy’s age, payment, remaining deferral, and guarantee, projects that policy’s cash flows, then adds the book. The word is Latin for “in series”: Ada, then Ben, then Chen. This lab does the other thing. One typical rider is projected once and scaled to 10,000. That is fast, and it is exact only when every contract really is that typical rider.

Teaching example — three deferred-income contracts, each promised $10,000 a year. The dollar marks below are classroom labels, not an output of the lab.

  • Ada, age 55, still in deferral. Teaching mark $80,000.
  • Ben, age 65, the typical rider this lab uses. Teaching mark $100,000.
  • Chen, age 78, already in payment. Teaching mark $140,000.

A model-point run treats all three as Ben: 3 × $100,000 = $300,000. A seriatim run keeps each contract: $80,000 + $100,000 + $140,000 = $322,000. The extra $22,000 is mix — older income already on, younger income still waiting — not a rate move and not a change in lapse. A production extract with millions of rows is the same idea at scale.

WhatModel pointSeriatim
UnitOne typical contract, then multiplyEach contract, then add
This page10,000 copies of one riderNot run here
When it misleadsThe average rider is a bad stand-inSlow; needs a clean in-force extract

Lapse is customers leaving. Mortality is the other decrement. Longevity is the opposite of a high death rate: people stay on the income longer. Withdrawal here is the lapse that pays a cash surrender value of two annual payments.

02

Projected cash flows

Teaching lab

One model point

10,000 identical contracts. Income of $10,000 a year starts after a 5-year deferral. Lapses receive two years of payment as a cash surrender value. Mortality and lapse decrements are annual. Not a seriatim run — every contract valued on its own — and not an experience study.

Liability PV
$560,165,844
Undiscounted cash
$870,273,181
In force at year 20
3437
Projected model-point cash flowsIncome plus surrenders by year for the classroom cohort.Projected cash flowsTeaching sketch135791113151719
Gold is surrender value in the deferral years. Blue is income after year 5. Raising lapse fattens the gold bars and thins later blue bars. Synthetic cohort.

Lapse is customers leaving. Mortality is the other decrement. If more people leave early, you pay surrenders now and fewer incomes later. If they live longer, the income tail fattens and the present value rises.

03

Behind the scenes

  1. Start with 10,000 contracts and a $10,000 annual payment.
  2. Each year, apply mortality q and lapse w to the remaining in-force.
  3. Lapses receive two years of payment as a teaching cash surrender value.
  4. After year 5, remainers receive income. That is the annuity guarantee.
  5. Discount the combined cash flow at the slider rate. Sum is the liability PV.

04

Best estimate, behavior, and experience

Best-estimate assumptions are the central q, w, and discount the model uses for economic value. Policyholder behavior — who lapses when rates rise, who annuitizes — sits inside w. An experience study compares actual deaths and lapses with expected (A/E) and proposes an assumption update. This lab does not ingest experience; it only lets you move the best-estimate sliders and watch cash flows and PV.

Longevity risk is a lower q: the income tail fattens and PV rises. A mass lapse fattens the gold surrender bars and thins later blue income. Dynamic lapse that depends on the rate path belongs with the ESG module, not this single-rate projection.

05

A real-world ALM use

A valuation actuary asks what a +2 percentage point lapse pickup does to the deferred annuity liability before anyone re-runs seriatim. Slide lapse from 4% to 6% and read PV and the cash-flow bars. The Phase 1 stress lab uses a linear lapse coefficient on a closed book; this page builds the cash flows that coefficient is approximating.

06

Limitations

One age, one payment, constant q and w, no mortality improvement, no dynamic lapse, no expenses, no statutory reserve method, and no seriatim in-force. A/E and assumption governance are named, not computed. Do not treat the PV as a filing number.

Related pages: Advanced ALM hub · ALM fundamentals · Economic scenario generators.

Trades | Learn. Practice. Analyze. Improve.