Assumptions & methodology

Architecture

Year-by-year simulation from current age to retirement age. Each year: select career phase → apply wage growth → calculate employer SG → calculate concessional contributions (with carry-forward) → apply 15% contributions tax → apply Div 293 → apply NCCs and bring-forward → apply investment return (interpolated if glide path) → deduct fees → apply Div 296 → grow non-super account.

Contribution timing convention

Contributions are applied before the investment return each year (start-of-year convention). This means contributions earn a full year's return in the year they are made. The mid-year convention — where contributions arrive evenly throughout the year and earn approximately half a year's return — would be marginally more accurate for monthly payroll and quarterly SG cycles, but the difference is small relative to the opening balance and is a common simplification in annual projection models.

Career phases

Multiple phases can be defined, each with a start age, salary, SG rate, and contribution amounts. The engine selects the phase with the highest start age ≤ current age each year. Wage growth is applied from the phase's start age.

Concessional cap

$32,500/yr (2026-27). Employer SG plus personal concessional. If contributions exceed the cap, the excess is counted as NCCs. Maximise concessional option automatically tops up to the cap minus employer SG each year. ⚠ Model limitation: the cap is held at the 2026-27 value for the first year and then projected forward using a fixed long-run AWOTE rate of 3.5% p.a. in $2,500 increments — consistent with RBA and Treasury long-run estimates. This is independent of the user's personal wage growth setting. In practice the cap rises by government decision and may not match this projection in any given year.

Carry-forward

Unused concessional cap from the prior 5 financial years, available when TSB < $500,000. User enters per-year amounts from ATO MyGov. The engine depletes oldest years first (FIFO). Entries older than 5 years at simulation start are discarded.

Contributions tax

15% flat rate on all concessional contributions. Applied before balance update.

Division 293

Additional 15% tax on concessional contributions if income (salary + concessional) exceeds $250,000. User selects whether deducted from super or paid from other income. The $250,000 threshold is legislatively frozen — there is no indexation mechanism — and has been unchanged since 2012. The model treats it as permanently static, which is correct under current law.

Non-concessional cap (NCC)

$130,000/yr (2026-27). Regular NCC and lump-sum contributions share this cap. Bring-forward contributions override this cap in the trigger year. The NCC cap, bring-forward cap ($390k), and all TSB-relative bring-forward thresholds are derived from the projected concessional cap (NCC = 4× CC) and Transfer Balance Cap, and rise with them over the projection using the same fixed AWOTE and CPI rates.

Bring-forward rule

Allows up to 3 years' NCC cap in a single trigger year. Allowable limit gated by TSB at trigger age: full 3-year ($390k) if TSB < $1,840,000; 2-year ($260k) if TSB < $1,970,000; standard 1-year ($130k) if TSB < $2,100,000; $0 if TSB ≥ Transfer Balance Cap ($2,100,000). NCC cap locked to $0 for 1 year (2-year bring-forward) or 2 years (3-year bring-forward) after the trigger.

Downsizer contribution

One-off, exempt from NCC cap. Available from age 55. Capped at $300,000 per person. Applied to super in the trigger year before the investment return step.

Government co-contribution

ATO matches 50c per $1 of NCC up to $500/year. Full $500 if income ≤ $49,293; phases out linearly to zero at $64,293 (2026-27 thresholds, indexed annually). Applied automatically when eligible.

Division 296

Enacted, effective 1 July 2026: tiered additional tax on notional super earnings. Extra 15% on earnings attributable to TSB above $3M (combined 30%); extra 25% on earnings attributable to TSB above $10M (combined 40%). Thresholds CPI-indexed. Marked with disclaimer in the UI.

Investment return

Applied to the post-contribution, pre-fee balance. Flat rate when glide path off; linearly interpolated from start to end rate when on (lerp). Same interpolation applied to MC volatility.

Fund fees

Three components deducted after investment return: (1) investment fee (% of post-return balance); (2) flat annual admin fee ($); (3) insurance premium ($). All default to $0.

Non-super account

Grows at the user's specified post-tax return rate. Same lognormal z-value as super in MC runs (correlated returns). Annual contribution added each year before return.

Distribution

Lognormal draws using Box-Muller transform. Mean = (expectedReturn/100 − σ²/2), σ = volatility/100 per year.

Glide path in MC

Per-year μ and σ interpolated between start and end values using the same lerp as the deterministic engine.

Correlation

Super and non-super use the same z-value per year — positively correlated but not identical returns.

Output

P10/P25/P50/P75/P90 per year at retirement for super, non-super, and combined. 1,000 simulations by default.

Architecture

Binary search on three separate dimensions, each holding all other inputs constant (including partner balance, which is always fixed at user input).

Extra NCC per year

Searches in $100 steps from $0 to $130,000 — the minimum additional NCC needed to reach the target balance.

Earliest retirement age

Searches from currentAge+1 to 80 — the earliest age at which the target balance is reached.

SG boost equivalent

Searches in 0.1% steps from 0% to 15% — the additional SG rate that would close the gap.

Bridge analysis

Compares non-super balance at preservation age against (annualIncome × bridgeYears). Signals whether non-super covers spending from early retirement to age 60.

Preservation age

Fixed at 60 (current legislated age).

Year-by-year simulation

The model runs one year at a time from retirement. Each year: apply investment returns → receive income → calculate spending → fund spending from accounts → apply Age Pension means test → record results.

Annual time steps

The simulation uses annual (not monthly) time steps. Intra-year timing is not modelled. This is standard for long-horizon retirement planning models.

The three accounts

Cash account (overflow buffer, earns cash return rate, accessed first); Sequencing buffer (deliberate defensive reserve, earns cash return rate, accessed second); Main super (primary growth portfolio, minimum drawdown rules, accessed last).

Withdrawal waterfall

Each year, spending is funded in strict order: (1) Cash account, (2) Sequencing buffer, (3) Super and/or non-super investments. Super is inaccessible before preservation age (60).

Non-super investments

When enabled, tracked alongside super. Returns scaled proportionally to super in historical and MC modes. Subject to Age Pension deeming and assets test each year.

Downsizer contribution

At the configured age, contribution added to super, remaining proceeds to non-super account, homeowner status updates — all before the Age Pension means test for that year.

Income floor first

All fixed income (defined benefit pension, Age Pension, other income) applied before drawing from the portfolio. Only the shortfall is funded from accounts.

Return application

Returns applied to main super balance at start of each year before withdrawals. Buffer and cash earn the separate cash return rate (default 3%).

Minimum drawdown

ATO requires minimum annual super withdrawals (4% at 60–64, rising to 14% at 95+). Excess above spending flows into the cash account.

CPI indexation

Spending, pension income, and Age Pension rates all indexed to CPI each year. ⚠ Model simplification: the legislated indexation for Age Pension payment rates is the higher of CPI or MTAWE (Male Total Average Weekly Earnings), not CPI alone. The model uses CPI only, which may understate Age Pension income in periods where wages outpace inflation. Asset test thresholds and income test free areas are CPI-indexed only, so those are correctly modelled.

Guardrails interaction

Guardrails adjust base spending — not minimum drawdowns, debt repayments, or aged care costs. The pension income floor prevents cuts below fixed income. A user-set spending floor, if any, is applied after both the guardrail adjustment and the spending pattern.

Simulation horizon

Single mode: Modelled Death Age minus Retirement Age plus 1 year. Couple mode: Couple Horizon Years setting. Parametric MC with stochastic mortality: each run samples its own death age. Historical MC: always fixed 35 years. Formal stress test D1: fixed 45 years.

Fees

Not explicitly modelled. Reduce your expected return assumption by your estimated fee percentage (e.g. 0.5% fee on 7% return → set 6.5%). Typical Australian industry super fees: 0.5–1.0% p.a.

Maximum rates

Single: $31,224/yr ($1,200.90/fn). Couple combined: $47,071/yr ($1,810.42/fn).

Asset test — homeowner

Full pension below $333,000 (single) / $499,000 (couple). Cuts out at $733,500 (single) / $1,102,500 (couple). Reduces $3/fn per $1,000 above lower threshold.

Asset test — non-homeowner

Full pension below $600,000 (single) / $766,000 (couple). Cuts out at $1,000,500 (single) / $1,369,500 (couple).

Income test free area

$5,876/yr (single), $10,296/yr (couple combined). Pension reduces 50c per dollar above threshold.

Deeming rates

1.25% on first $66,800 (singles; $110,600 couples), 3.25% above. Applied to all financial assets including super.

Work Bonus

$7,800/yr per eligible partner (employment/self-employment income only, age 67+). Unused amounts accumulate up to $11,800/person.

Eligibility age

67 (current legislated age).

Means test logic

Both asset test and income test applied each year — the result that produces the lower pension is used.

DB pension deductible amount

An optional input that reduces the assessable DB pension income in the income test before applying the taper. Modelled as a fixed nominal deduction (does not CPI-index) consistent with how Services Australia applies it. Defaults to $0 — without it the income test is applied to the full pension, producing a conservative (lower) Age Pension estimate. Enter the amount Services Australia calculates from your contribution history.

Minimum drawdown rates

4% (ages 60–64), 5% (65–74), 6% (75–79), 7% (80–84), 9% (85–89), 11% (90–94), 14% (95+). Applied annually regardless of spending needs.

Transfer Balance Cap

$2.1M (from 1 July 2026). Super above this cap cannot be held in pension phase tax-free. ⚠ Model limitation: the TBC is held at $2.1M for the entire simulation. In practice the TBC rises in $100k CPI increments — over a 20-year projection it may reach approximately $2.7–2.9M. This means the model is slightly conservative for users near the current TBC with long horizons, as it overstates the portion of earnings subject to accumulation-phase tax.

Tax treatment

Super income stream is tax-free after age 60 in pension phase. Accumulation phase earnings taxed at 15%.

Pre-60 retirement

⚠️ Super cannot be accessed until you meet the conditions of release, usually not before the preservation age (60). The calculator enforces the preservation age — pre-60 spending must be funded from other sources.

Concessional contributions cap

$32,500/yr (2026-27). Fully modelled in the Projection Calculator — carry-forward, maximise flag, Div 293.

Non-concessional contributions cap

$130,000/yr (2026-27). Bring-forward rule allows up to $390,000 over 3 years. Fully modelled in the Projection Calculator.

Low-Rate Cap

$260,000 (2026–27). Applies to taxable component withdrawals before age 60. Not modelled — enter after-tax amounts if retiring before 60.

Division 293

Modelled in the Projection Calculator (accumulation phase). Not modelled in the drawdown simulation.

SAPTO / LITO

Effectively make income up to ~$32,279 (single) or ~$28,974 each (couple) tax-free for eligible pensioners. Modelled in super drawdown calculations.

Default CPI

2.5% per annum (user-adjustable 0–10%).

Parametric MC — distribution

Lognormal return distribution. Default: 6.5% mean, 10% standard deviation (balanced growth). Presets: Conservative (5%/8%), Balanced (6.5%/10%), Growth (8%/14%), High Growth (9%/18%).

AU Equities — XAOA

1980–2024 (45 years). Source: ASX/LSEG.

AU Balanced 60/40

1990–2024 (35 years). Composite of AU equities and bonds.

US S&P 500

1928–2025 (98 years). Source: Shiller/Yale CAPE data, Ibbotson SBBI.

Historical MC sampling

Block bootstrap (5-year blocks), shuffled years, or 35-year complete blocks.

Source

Australian Bureau of Statistics Life Tables 2020–2022 (ABS 3302.0.55.001).

Default death age

Interpolated from ABS period life tables by sex and current age. Example: a 60-year-old male defaults to age 84; a 55-year-old female defaults to age 87.

Stochastic mortality (MC)

Each parametric Monte Carlo run samples a death age from a normal distribution centred on ABS life expectancy, SD = 10 years. Historical MC always uses fixed 35-year windows.

Probabilistic entry

Each simulation path draws ONE uniform random value and compares it each year against a cumulative entry-probability curve (inverse-transform sampling of the entry age). The curve is calibrated to Australian residential aged care data — not to mortality tables: it yields a 41.9% lifetime probability of entering permanent residential care from age 65, a median entry age of 85, and 54.8% of entries at 85 or older. Published benchmarks: 39–46% lifetime (Cooper-Stanbury 2025 puts females at 46%; the AIHW death-linkage study found 43% of decedents aged 65+ had used permanent care), median admission age 85 and ~54% of admissions at 85+ (AIHW/GEN).

Who enters care

Aged care is modelled as a PER-PERSON health event, evaluated independently for each partner. It does not depend on the death scenario. In deterministic mode you set an entry age for each person and tick who to model — either partner, both, or neither. In probabilistic mode both partners are assessed independently against the incidence curve, with separate draws, so their care histories differ.

Stay duration

In deterministic mode the stay is exactly the length you set. In probabilistic mode with death in care enabled, that figure is the AVERAGE stay: each year of a stay carries a mortality hazard of 1/duration, so stays are right-skewed — most are shorter than the figure you set, a minority run considerably longer, and the mean matches your input. A fixed length for every resident put every death at the same age and shifted the median simulation run by more than a decade.

Entry is one-way

A person who leaves residential care does not re-enter, and pays only one Refundable Accommodation Deposit. Permanent residential care is not a state people cycle in and out of.

Death in aged care

Optional. In deterministic mode the resident dies at the end of the modelled stay — a definite outcome for a scenario you specified. In probabilistic mode each year of the stay carries the mortality hazard above, so deaths spread across the stay; with a 1/duration hazard almost everyone who enters care dies there, which is what the setting is intended to express. A death in care applies the same consequences as any other death: super transfers to the survivor, the defined benefit pension drops to the reversionary rate, and household spending moves to the single level. NOT applied when the projection is set to have neither partner die — care still occurs and its costs still apply, but nobody dies.

Not differentiated

Entry probability does not vary by sex, living arrangement or health status. Women enter residential care more often and stay longer than the population average this curve represents.

Deterministic entry

Entry age set per person by the user. Used in non-MC scenarios. Note that with the PROBABILISTIC approach selected, a single (non-Monte-Carlo) projection uses the median draw of 0.5, so it shows care beginning at age 90 — the age at which cumulative incidence first exceeds 50%. The full spread of outcomes is visible only in Monte Carlo.

RAD (Refundable Accommodation Deposit)

Withdrawn from super at entry as a lump sum. Refunded on exit LESS the provider retention introduced by the Aged Care Act 2024 for entries from 1 November 2025: 2% per year calculated on the declining balance, capped at 5 years — so a five-year stay returns 90.4% of the deposit (9.6% retained), not the flat 10% often quoted. Residents who entered before 1 November 2025 are grandfathered as fully refundable; every entry this planner models is post-reform. The refund is returned in nominal dollars, so a long stay also erodes its real value. Default: $400,000.

Annual ongoing costs

Basic daily fee plus means-tested care fee, applied PER PERSON in care. Not refundable; indexed to CPI. Default: $65,000/yr. National median residential stay: approximately 3 years.

Income streams after a death

Each additional income stream belongs to one partner, and you set what share of it continues once that person dies — 100% for income that passes to the survivor in full, 0% for income that simply stops, or a reversionary fraction in between. The default is 100%, which is NOT right for every stream: a UK State Pension under the post-2016 new State Pension rules generally cannot be inherited by a spouse, and many overseas and private pensions end with the recipient. Check your own scheme rather than assuming the income continues. Only applies when a death is modelled.

Paying for the room

You choose how the accommodation is paid: a refundable deposit (RAD), a daily accommodation payment (DAP), or a combination. The daily payment is derived from the room price at the government's Maximum Permissible Interest Rate (MPIR), which is 8.43% from 1 July 2026: DAP per day = room price × MPIR ÷ 365. The rate is FIXED on the day a resident enters care and does not change for the rest of their stay, so the daily payment is a flat nominal amount that inflation erodes over a long stay. A deposit is refunded less the provider retention; a daily payment is never refunded.

Age Pension treatment of the deposit

A refundable deposit balance IS an asset (Social Security Act 1991 s 11(3AA)), but it is expressly excluded from the definition of 'financial investment' (s 9(1D)(c)-(d)) so it is NOT subject to deeming, and its value is disregarded when calculating assessable assets (s 1118(1)(v)). The model reproduces this without special-casing: assessable assets are derived from your actual portfolio balances, so paying a deposit removes that money from both the assets test and deeming, while choosing the daily payment leaves it invested where both apply. For a part-pensioner this can matter more than the interest rate does — run your scenario both ways.

Household spending while in care

Derived each year from who is in care. Both at home: your base spending. One partner in care and one at home: the person-at-home rate (default 70%) plus that person's care costs. BOTH in care, or a single person in care: the nobody-at-home rate (default 15%) plus care costs for each — nobody is living at home, but rates, insurance and maintenance continue. That second figure decides how much a second person entering care really costs: at 15% their care fees are largely offset by the drop in home spending, so raise it if you would keep and maintain the home. Both rates are settings, not fixed assumptions.

Constant Real

Annual spending grows by CPI. Purchasing power maintained throughout retirement.

J.P. Morgan Curve

Based on Blanchett (2014) and J.P. Morgan (2024) research. Real spending declines 1.0% p.a. (years 1–10), 1.5% p.a. (years 11–20), 0.5% p.a. (years 21+). Aged care modelled separately.

Forgo Inflation modifier

Skips CPI uplift in years following a negative portfolio return. T. Rowe Price method. Morningstar 2025: lifts safe withdrawal rate from 3.9% to 4.3% at 90% success.

Guardrails (Guyton-Klinger)

Adjusts spending when withdrawal rate drifts above or below initial rate. Default: ±20% trigger, 10% adjustment. Morningstar 2025: supports 5.2% starting withdrawal rate at 90% success.

Accumulation-phase super

Super held in accumulation phase rather than as an account-based pension is modelled as a separate pool. It earns the same GROSS return as your pension pool but NET OF 15% super earnings tax, since accumulation-phase earnings are taxed and retirement-phase earnings are not — a 7% gross return compounds at 5.95% in this pool. If your pension pool is exhausted while this pool still holds money, the model converts it to pension phase — but only up to the Transfer Balance Cap space you have left. Cap space is counted when a pension STARTS and is not restored by drawing one down, so a member who used their full cap at commencement keeps that money in accumulation, still taxed at 15%. In couple mode the accumulation pool and any contributions are attributed to partner 1, so partner 1's own cap and total super balance govern both — these are per-person limits, and a couple has two of them. Partner 2's unused cap space is NOT available to partner 1: modelling that would require per-partner ownership of the accumulation pool, which this version does not track. Personal caps index only by the share never used — use 100% and your cap is frozen permanently; use 80% and you receive 20% of each future increase — so partial users gain a little room over time and full users gain none. (Such a member could instead withdraw it tax-free after 60 and hold it outside super, which may work out better depending on their marginal rate; the model does not assume this.) It is counted in the Age Pension assets test, but has NO legislated minimum drawdown, and the model draws on it only AFTER the pension pool is exhausted. Default: 0.

One-off super contributions

After-tax (non-concessional) contributions made during retirement land in the accumulation pool, not the pension balance — you cannot contribute to a retirement-phase income stream, and its earnings are taxed at 15% where pension earnings are tax-free. If the pension pool is later exhausted, the model converts what it can back to pension phase, limited by remaining Transfer Balance Cap space (see the accumulation-phase super row above). Anything that cannot be converted keeps funding spending from accumulation, so the projection does not end early. Contributions are limited to the annual non-concessional cap ($130,000 in 2026-27, indexed forward), blocked entirely once total super balance reaches the Transfer Balance Cap, and not permitted from age 75. Amounts the rules disallow are reported, not silently dropped. Bring-forward is not offered here: split a larger amount across years. (The Pre-retirement calculator does model bring-forward.)

Spending floor (optional)

A minimum annual spending level in today's dollars, applied AFTER the spending pattern and any guardrail cut. It represents essential costs you will not go below. Order matters: the J.P. Morgan curve is a behavioural projection (retirees choose to spend less with age) while the floor is a hard constraint, so the floor overrides the curve — without this, a floor would be silently breached by about 21% by year 20. Aged care, health costs and one-off expenses sit on top and are never reduced to meet it. Setting a floor LOWERS modelled success rates: the plan draws down and can deplete rather than cutting below your essentials. That trade-off is the point of the setting. Default: 0 (disabled).

Simulation mode

All scenario lines use deterministic (constant-return) simulation regardless of which mode is active in the main calculator. This ensures all lines are comparable — each difference reflects only the parameter being varied, not random path variation.

Spending

Spending is fixed at the user's configured baseSpending for all non-spending scenarios. The solver does not re-run for each scenario. This means the scenario lines show the portfolio impact of a pure parameter change, holding financial behaviour constant.

Stochastic irregular expenses

Non-recurring expenses (one-offs, windfalls, irregular health costs) are excluded from all scenario lines. Including stochastic expenses would cause lines to diverge at random points unrelated to the parameter being tested, making the comparison meaningless. The same exclusion applies to the Sensitivity Analysis grid.

Sequencing buffer

The sequencing buffer allocation is held constant across all scenarios (computed once from the base scenario). Previously, re-running the solver for each return scenario caused the buffer allocation to change, shifting the starting balance of each line.

Current settings line

The 'Current settings' line in the Scenario Comparison uses the same deterministic engine as the other scenarios, not the main simulation. This means it may differ slightly from the main balance chart if stochastic features (irregular expenses, mortality) are active. This is intentional — the comparison is designed to be internally consistent, not to reproduce the main chart.

Architecture

The Manager calls the same runSimulation function as the Retirement Readiness Calculator. There is no second simulation engine. A thin adapter (manageAdapter.ts) maps the Manager's 4-bucket account model and store fields into a SimulationInput; the engine produces a SimulationYearResult[]; the adapter reads source-attribution fields back to produce per-account draw recommendations.

Engine extensions

Two optional fields were added to SimulationInput to support the Manager: accumulationSuperBalance (separate super pool with no minimum drawdown) and apOverrideYear0 (year-1 override of the Age Pension means-test result). Both default to no-ops; existing Readiness callers behave identically. Engine extensions are protected by per-year exact-value golden fixture tests across nine representative scenarios.

Per-account recommendation

The cashflow plan reads source-attribution fields from year 1: superDrawnForMinimum + superUsedForSpending + superUsedForRAD → ABP draw; accumulationSuperUsedForSpending + accumulationSuperUsedForRAD → super accumulation; bufferUsedForSpending + bufferUsedForRAD → user cash; nonSuperWithdrawn → shares. The engine's internal cashAccount (built up from min-drawdown surplus and excess income) is NOT counted as a user cash draw — that money originated from the ABP.

Account kinds

ABP (account-based pension) · Super accumulation · Cash · Shares non-super. Multiple accounts per kind are allowed; balances are summed by kind for engine input. Each Account is identified by a stable id so the actuals ledger can record per-account end-of-year balances.

ABP

Subject to legislated minimum drawdown (4–14% by age — see next row). Earnings tax-free inside the fund. Drawn first within the super pool when spending exceeds income.

Super accumulation

No minimum drawdown. Earnings still tax-free in the model (a v1 simplification — accumulation phase is taxed at 15% in reality, but the user's expectedReturn is treated as the net rate they receive). Drawn after the ABP within the super pool.

Cash

Drained first when spending exceeds income (after the engine's internal cashAccount float, which is recycled ABP min-drawdown). Earns the cashReturnRate. Stored as the engine's sequencingBuffer field; the engine's separate cashAccount (always initialised at 0) is reserved for internal recycling and is not addressable from the Manager UI.

Shares non-super

Drained last by default (controlled by nonSuperDrawdownOrder = 'non_super_first'). Earns expectedReturn − sharesTaxDragPct (see Tax treatment). No CGT or franking modelling.

Couple-mode attribution

In couple mode the engine ignores SimulationInput.mainSuperBalance and reads super from partner1.superBalance + partner2.superBalance. The adapter attributes ALL ABP balance to partner1, with partner2.superBalance = 0. Accumulation super is treated as a single primary-owned pool. Material per-partner splits are not modelled in v1.

Source

Schedule 7, Superannuation Industry (Supervision) Regulations 1994. Same rates the Readiness Calculator uses. Implemented in app/engine/superannuation.ts.

Rates by age

Under 65: 4% · 65–74: 5% · 75–79: 6% · 80–84: 7% · 85–89: 9% · 90–94: 11% · 95+: 14%. Applied to opening balance of the ABP at the start of each financial year.

Pre-60 behaviour

Minimum drawdown only applies in pension phase (age ≥ 60). Before 60, ABP balances are not legally accessible. The Manager flags pre-60 spending shortfalls as insufficientFunds.

Surplus handling

If the legislated minimum exceeds the user's spending need, the surplus accrues in the engine's internal cash float (reported as a 'Surplus → cash float' row in the cashflow plan). It is NOT lost. Subsequent years draw from this float first before touching other balances.

Accumulation pool exemption

Minimum drawdown is NOT applied to the accumulation super balance. This is the engineering rationale for distinguishing the two pools — without it, the engine would force-deplete super accumulation balances each year contrary to legislation.

ABP earnings

Treated as tax-free. Matches Australian pension-phase super tax law.

Super accumulation earnings

Treated as tax-free in v1. The user's expectedReturn is interpreted as the post-tax rate they will actually receive. ⚠ A more rigorous model would deduct a 15% tax on accumulation earnings within the super wrapper. The user can compensate by lowering expectedReturn for portfolios with significant accumulation balances.

Cash returns

cashReturnRate is treated as gross. Income tax on cash interest (typically marginal rate) is NOT modelled. Conservative for retirees in low brackets; understates tax for higher-income retirees.

Shares (tax drag knob)

A single sharesTaxDragPct field is subtracted from gross return on shares-non-super accounts. nonSuperReturnRate = expectedReturn − sharesTaxDragPct. nonSuperBaseReturnRate = expectedReturn (so MC paths scale shares proportionally with super returns).

What tax drag does NOT model

CGT events (timing of realised gains), the difference between assessable income and capital gains, franking credits, the CGT discount on assets held >12 months, the interaction between dividend income and the Age Pension income test, or the difference between marginal rates. Treat results as directional when shares are a large fraction of the portfolio.

Division 293 / Division 296

Not modelled in the Manager (these primarily affect contribution and high-balance phases, less relevant for the post-retirement use case).

Storage

ActualYear[] in the useManagerStore Zustand store, persisted to localStorage under the key 'managerStoreV2'. Sorted ascending by FY string ('YYYY-YY'). Each row carries: fy, endingAge, accountBalances (Record<accountId, number>), actualSpending (nullable), actualAgePension (nullable), notes.

Replay rule

When projecting, the most recent ActualYear's accountBalances replace the live state.accounts[].balance values for projection purposes. Implemented via the accountBalanceOverrides option on toSimulationInput. The live accounts panel is reference only when the ledger is non-empty (a banner indicates this).

Identification of latest

Sort actuals by FY string descending; first row is 'latest'. Lexicographic comparison of 'YYYY-YY' strings happens to match chronological order for FY values within the same century.

Overrides for unknown account ids

Silently ignored. If an account is deleted after a ledger row references it, the projection still works — the deleted account contributes zero to its kind's bucket sum. Stale account ids ARE preserved in the ledger row (visible in the expanded view as '[deleted account ...]').

Editing semantics

Inline edit replaces the row's balances/spending/AP/notes. The FY field is locked when editing — changing the FY is effectively a delete + re-add, achieved via the ✕ button. Multiple ledger rows with the same FY can exist if the legacy migration produced duplicates; they collapse on the first add/edit for that FY (addActual filters by fy match).

Algorithm

Binary search over annual spending levels. At each candidate level, run N Monte Carlo paths (default 500) by calling runSimulation with fresh stochastic return sequences. For each enabled goal, count paths where the goal's success criterion is met at the goal's targetAge. A spending level passes if all goals' achievedProbability ≥ goal.successThreshold.

Search bounds

Lower = essentialSpendingFloor. Upper = max(floor × 4, totalAssets / max(5, yearsToHorizon × 0.25)). Runs an infeasibility check at floor first (returns infeasible flag if even floor fails) and an unconstrained check at ceiling (returns unconstrained flag if even ceiling passes).

Termination

30 max iterations or $100 tolerance, whichever first. Result rounded to nearest $100. ~15,000 simulations per solver call, typically 1-3 seconds in browser.

Default objective (no goals)

When no goals are enabled, the solver targets 'portfolio survives horizon at 85% confidence'. Confidence threshold parameterised via SolverOptions.defaultSuccessThresholdPc.

Return distribution

Lognormal draws using the Box-Muller method (sampleLognormalReturn from monteCarlo.ts). Same parameterisation as the Readiness MC: μ_log = ln(1 + expectedReturn/100) − σ²/2, σ = volatility/100.

Binding goal identification

The goal with the smallest (achievedProbability − successThreshold) margin at the recommended level is flagged as the binding constraint. Marked with ◆ in the goal outcomes section.

What the solver does NOT do

Does not optimise across multiple years (each MC run uses the same fixed annual spending — it does not search for a varying spending path). Does not model dynamic spending strategies like guardrails. Does not consider variance reduction techniques (antithetic variates, etc.).

Source

AGE_PENSION_MAR_2026 in app/data/agePensionParams.ts. Same parameter block the Readiness Calculator uses.

Project mode (default)

Engine means-tests every year of the projection: assets test (using projected balances + nonSuperAssets) and income test (deeming financial assets at the published thresholds). Final pension is the lower of the two binding tests, capped at the maximum rate. Rate indexed annually by max(CPI, MTAWE). MTAWE approximated as CPI + 0.75% (long-run historical differential).

Pin year-0 mode

Replaces the year-1 AP figure with the user's pinned Centrelink amount (entered as fortnightly $ × 26 = annual). Year 2 onwards reverts to means-testing. Pinning forever would be incorrect — balances change as the user draws down. agePensionDetails on the year-1 row is set to null when overridden (the audit trail no longer represents an engine calculation).

Accumulation super in assets test

In v1 the assets test treats accumulationSuperBalance as fully assessable. The Centrelink rule that exempts a partner's accumulation super if they are under pension age is NOT applied. Conservative when both partners are 67+; understates AP when one partner is under 67.

Survivor pension

The store carries a survivorAgePensionFortnight field but it is not yet plumbed through to the engine. v1 estimates the survivor pension via the (max single / max couple-each) ratio applied to the couple's combined annual figure.

Deterministic projection

Horizon = max(1, 100 − currentAge). Mortality is NOT sampled. The projection runs to age 100 by default for headline display.

Solver MC paths

Each path uses the deterministic horizon — mortality is not sampled per-path in the solver. This is a v1 simplification matched to the legacy Manager's behaviour. The Readiness Calculator's MC sampler uses ABS period life tables; integrating that into the Manager's solver is a future enhancement.

Couple-mode death scenarios

The Manager models survivor scenarios via the deathScenario field (both-alive / partner1-dies / partner2-dies) and scenarioDeathAge. When a partner dies, income streams with survivorFraction revert at that age, and the Age Pension reassesses at the singles rate. The summary panel shows the key before/after differences at the death year.

Single accumulation pool

accumulationSuperBalance is treated as held by one person, not split per partner. In couple mode it is attributed to partner 1.

Accumulation in AP assets test

No under-67 exemption applied to the accumulation portion. Conservative when one partner is under 67.

Pre-60 super access

Uses the primary's age, not per-partner. Edge case for couples with a large age gap.

Tax drag is a single knob

No CGT, franking, or marginal-rate modelling. Treat results as directional for shares-heavy portfolios.

DB pension survivor reversion

Income streams with survivorFraction revert correctly when a survivor scenario is active (deathScenario = partner1-dies / partner2-dies). Reversion is NOT automatic outside of an explicit survivor scenario — it requires the user to activate the scenario in the Profile card.

Solver mortality

Solver does not sample stochastic mortality. Each MC path runs to the deterministic horizon. Goal probabilities therefore measure 'probability of meeting goal CONDITIONAL on living to that age'.

Goal types

Four types only (balance_at_age, funded_event, estate, ruin_avoidance). All score against totalBalance — no per-account goals or income-stream goals.

What-if comparison

Year-1 only. Multi-year comparison and goal-probability comparison for what-if scenarios are not yet exposed.

Same-FY ledger duplicates

Migrated rows from the legacy Manager may produce multiple ActualYear entries per FY. Cosmetic — collapses on first add/edit for that FY.