Withdrawal Strategy Lab

Five rules, one set of market paths — 60/40, single-regime engine

$1,000,000 starting balance, retirement ages 65–95, 2.5% inflation, monthly rebalancing, JPMorgan LTCMA 2026 capital-market assumptions. All income figures are in today's dollars. Worst-decile (p10) / median (p50) shown for income at ages 80 and 94.

^* Fraction-of-balance rules recompute spending from the current balance every year, so they cannot hit zero before the horizon — their risk appears as declining income (see the p10 columns), not depletion. † Share of paths with at least one year of real income more than 25% below year-1 income.

Read the table as a three-way trade: lifetime income (VPW highest, by spending everything), income stability (the 4% rule is perfectly flat — until it isn't), and terminal balance (fixed-percentage leaves the most behind, because it spends the least when markets fall). The RMD-style rule is the quiet standout for late-life income: it starts lowest ($29,070) but its income rises with age on most paths, and it has the lowest deep-cut rate of any dynamic rule.

The same five rules when bad months cluster — 60/40, regime-switching engine

Same paths-per-scenario, same expected return, same rules — but returns drawn from a two-state regime model fitted to monthly data 1871–2026 (the live regime monitor documents the fit), where stressed months arrive in episodes instead of spreading evenly. This is sequence-of-returns risk made explicit.

^*/† as above.

Two results stand out. First, the 4% rule's depletion rate roughly doubles (14.1% → 29.8%) with no change in expected return — clustering alone does that. Second, and less obvious: guardrails lose most of their protection (1.1% → 19.6% depletion). Guardrails adjust spending once a year in 10% steps; a clustered bear episode can take the portfolio down faster than those steps can follow. Dynamic rules keep their no-depletion property, but their worst-decile income at 94 falls by roughly half versus the single-regime engine (fixed-percentage $17,779 → $8,707; VPW $25,445 → $11,066; RMD-style $31,555 → $15,266). On a more conservative 40/60 allocation the same pattern holds (4% rule: 11.3% → 28.0%; guardrails: 0.3% → 15.8%) — shifting to bonds helps far less under the regime engine than the independent-draw engine suggests.

Explore: income paths by age

Median real income by age — all five rules

Median (p50) annual real income at each age for the selected allocation and engine. Identical market paths across rules.

Income range for one rule

10th–90th percentile band (shaded) and median (line) of annual real income by age.

All metrics for the selected scenario

^* Fraction-of-balance rules cannot deplete before the horizon; risk appears as income decline. † Paths with ≥1 year of real income more than 25% below year-1.

How each rule works

4% rule (fixed real) — static

Withdraw 4% of the starting balance in year one ($40,000 here), then adjust that dollar amount for inflation every year regardless of what markets do. Perfectly smooth income while the money lasts; all of the risk is concentrated in the possibility that it doesn't.

Guardrails (Guyton-Klinger-style) — dynamic

Start higher — 5% ($50,000) — and adjust along the way: if the current withdrawal rate drifts 20% above its initial level (portfolio shrinking), cut spending 10%; if it drifts 20% below (portfolio growing), raise it 10%. The inflation adjustment is skipped after a losing year while the rate is elevated, and cuts stop after age 80 in this implementation. Higher starting income and far less depletion than the 4% rule — paid for with repeated spending cuts on poor paths.

Fixed percentage of balance — dynamic

Withdraw 4% of whatever the balance is each year. Mathematically incapable of depleting, and leaves the largest median estate of any rule here — but income tracks the portfolio down without any floor.

VPW-style (PMT annuitization) — dynamic

Each year, withdraw the payment that would exactly exhaust the current balance by age 95, computed with the PMT formula at the allocation's expected real return (3.42% for 60/40, 3.01% for 40/60 under these assumptions). The most lifetime income of any rule, a rising-then-volatile path, and a balance designed to reach roughly zero at the horizon.

RMD-style (balance ÷ life expectancy) — dynamic

Withdraw the current balance divided by remaining life expectancy from the IRS Uniform Lifetime table (extended below age 73 by adding one year of factor per year of age). Starts lowest, rises with age on most paths, and shows the smallest deep-cut rate of any dynamic rule — late-life income is its strength.

Methodology

• Setup: $1,000,000 at 65, withdrawals to 95 (30 years), 2.5% inflation, monthly rebalancing, no taxes or Social Security (the full planner models both). Equity sleeve split 70/30 US/international; 4% cash in both allocations.
• Market assumptions: JPMorgan LTCMA 2026 long-term capital-market assumptions — the same default the QuantCalc planner uses — with the full asset correlation matrix.
• Common random numbers: within each allocation × engine scenario, all five rules see the identical 10,000 market paths, so differences between rules are exact, not sampling noise.
• Two return engines: a single-regime lognormal engine (independent monthly draws), and a regime-switching engine using the two-state model fitted to 1871–2026 monthly real S&P returns documented on the regime monitor — with the unconditional expected return held equal to the baseline by construction (drift re-centering plus variance compensation), so the differences shown are clustering, not pessimism.
• Validation: the pipeline replicates the production engine's return conventions, and the 4%-rule scenario was run through the actual QuantCalc engine for comparison — results agree within ~1.1 percentage points on both engines (details in the dataset's cross_validation block).
• Limitations: rules are tested in pure form with annual adjustments; real retirees blend rules, hold cash buffers, and have lumpy spending. A fixed 30-year horizon understates longevity risk past 95 and overstates it for shorter retirements. All figures are pre-tax, in today's dollars.

Full simulation methodology for the QuantCalc engine is on the methodology page. Related reading: Dynamic vs. Static Withdrawal Strategies, what regime-aware Monte Carlo does to a plan, and the 4% rule's known problems.

FAQ

What is the difference between static and dynamic withdrawal strategies?

A static strategy sets spending once and only adjusts for inflation — the 4% rule is the classic example. A dynamic strategy recalculates spending from the current balance, so spending falls after poor markets and rises after good ones. The trade is fundamental: static rules concentrate risk in running out entirely; dynamic rules essentially never deplete but pass market risk through to annual income.

How often does the 4% rule run out of money?

Here — $1M, 60/40, ages 65–95, 10,000 paths on JPMorgan LTCMA 2026 assumptions — the 4% rule depleted on 14.1% of paths under a single-regime engine and 29.8% under the regime-switching engine. The doubling comes entirely from return clustering, not from any change in average returns.

Can a percentage-of-balance strategy run out of money?

Not before the horizon — a percentage of the current balance always leaves something. But the risk moves into income: 4%-of-balance started at $40,000 real and on the worst tenth of paths was down to $17,779 by 94 (single-regime) or $8,707 (regime engine). Depletion risk becomes income risk.

Do Guyton-Klinger guardrails work?

They cut depletion dramatically — 14.1% → 1.1% on identical paths — at the cost of repeated spending cuts (46.6% of paths saw at least one cut deeper than 25%). But under the regime engine they retained 19.6% depletion: clustered bear episodes can outrun annual 10% adjustments.

Which strategy produces the most retirement income?

VPW-style annuitization: $1,706,258 median lifetime real income vs $1,200,000 for the 4% rule on the same 60/40 paths (+42%) — by design spending the portfolio to near zero at the horizon. Lifetime income, income stability and terminal balance are the three quantities every rule trades against each other.

How does a regime-switching model change the picture?

With expected return held identical, clustering roughly doubled 4%-rule depletion, raised guardrails depletion from 1.1% to 19.6%, and cut the dynamic rules' worst-decile income at 94 roughly in half. Independent-draw simulations understate sequence-of-returns risk; the regime monitor documents the underlying model.