compoundcoast
Live off it · signature

Safe withdrawal rate — three strategies, side by side, net of tax

Most SWR tools show one strategy, one start date, and pretend tax doesn't exist. This rolls fixed real (Bengen), Guyton-Klinger 20/10 guardrails, and VPW across every start year since 1928, then converts gross withdrawals to net-spendable with 2026 IRS brackets — where taxable accounts pay capital-gains tax on the gain portion only. Every output is a range.

S&P 500 total return · asOf 2026-01 3-mo T-bill · asOf 2026-01 CPI-U · asOf 2026-01 IRS Rev. Proc. 2025-32 · tax year 2026 Updated 2026-07-06

Assumptions

live · every start year since 1928
Starting balance the withdrawals draw from.
Year-1 withdrawal for fixed real and guardrails. VPW sets its own percentage from the formula.
Longer horizons leave fewer rolling windows to test against.
Remainder sits in 3-month T-bills (this dataset's cash series) — more conservative than Bengen's intermediate Treasuries.
Account type
Traditional: whole withdrawal is ordinary income. Roth: untaxed. Taxable: LTCG on the gain portion only, cost basis tracked year by year.
Filing status
Sets the 2026 standard deduction and bracket thresholds.
100 bps = 1.00%/yr, dragged off the portfolio return every year.
Run the three strategies across every rolling start year since 1928.
—
Initial net-spendable — Income floor p5 — ▲ —
Median annual income (net, real) Range across start years (p5–p95)
Set your inputs — this rolls the plan across every start year, not one lucky date. Markets vary, so the honest answer is a range, not a single number.

Three strategies, one table

all figures are ranges, real (today's) dollars
Fixed real (Bengen)GK 20/10 guardrailsVPW

Success = the portfolio covers every scheduled withdrawal through the horizon. VPW shows 100% by construction — it withdraws a percentage of whatever is left, so its risk lands in the “lowest real year” row instead. Compare the income-floor row, not just the success row.

Ending balance after the horizon

real $ · p5 · median · p95

Fixed real (Bengen)

GK 20/10 guardrails

VPW

Coral = 5th-percentile, aqua = median, mint = 95th-percentile ending balance in today's dollars. VPW ends near zero by design — it spends the portfolio down over the horizon rather than leaving a bequest.

Methodology & sources
Show the math ›
Run the backtest to see the worked numbers for your inputs.
Assumptions & sources ›
AssumptionValueSource · asOf
Equity return S&P 500 total return Damodaran (NYU Stern), annual · asOf 2026-01 · dividends reinvested
Fixed-income sleeve 3-month T-bill Damodaran (NYU Stern) · asOf 2026-01 · more conservative than the intermediate Treasuries in Bengen's study — disclosed, not hidden
Inflation CPI-U, Dec/Dec BLS · asOf 2026-01 · drives the fixed-real and GK inflation step and all real-dollar output
Federal tax, 2026 brackets + standard deduction IRS Rev. Proc. 2025-32 · asOf 2025-10 · ordinary tables for traditional, 0/15/20 LTCG for taxable-account gains; withdrawal assumed to be the only income; state tax and NIIT out of scope (v1)
Guardrail rules ±20% band, 10% adjust Guyton & Klinger, Journal of Financial Planning · asOf 2006-03 · capital-preservation + prosperity rules only; the paper's inflation-cap and portfolio-management rules are omitted for transparency
VPW percentage p = r / (1 − (1+r)−n) Bogleheads wiki (PMT formula) · r = geometric-mean real return of your mix over 1928–2025, net of your fee — derived from the dataset, not a forward-looking guess
Mechanics start-of-year withdrawal, annual rebalance coverage 1928–2025; real withdrawals taxed against 2026 tables (the standard “brackets keep pace with inflation” assumption)

Rolling windows are overlapping start years, so they are not independent samples — a p5 across the windows is an illustrative historical range, not a probability of the future. Sequence risk is the point: the same average return with a bad first decade (1966-style, 2000-style) is what breaks a plan.

Written by Author to be finalized before launch · Updated 2026-07-06

Related