--- name: retirement_decumulation_planning description: Build a defensible retirement spend-down plan: set a starting withdrawal rate from history and valuations, defend against sequence-of-returns risk with dynamic spending rules, sequence taxable/tax-deferred/Roth withdrawals for taxes, and compute RMDs. Triggers when someone asks how to spend down retirement savings, what withdrawal rate is safe, how to handle RMDs, or how to withdraw tax-efficiently in retirement. --- # Retirement Decumulation Planning ## Purpose Turn "how do I not run out of money" into a concrete, research-backed spend-down plan: an honest starting withdrawal rate, dynamic rules that adapt when markets misbehave, tax-efficient account sequencing, and an RMD schedule that never misses a deadline. Educational only. This is not financial or tax advice, and it never presents backtested history as a guarantee of the future. ## Workflow 1. **Frame the horizon and the floor.** Ask: retirement age, planning horizon (30/35/40 years), guaranteed income (Social Security, pension, claiming ages), account mix (taxable / tax-deferred / Roth), current market context. The 4% rule was studied for 30 years; longer horizons need lower starting rates or dynamic rules. Guaranteed income is a floor that shrinks the portfolio-covered spending gap; price the gap, not total spending. 2. **Set the starting withdrawal rate with history AND valuations.** - Base case: 4% of starting portfolio in year one, inflation-adjusted each year after (Bengen 1994; Trinity study, ~95% success on balanced portfolios over 30 years; 1966 is the historical worst-case start year, not 1929). - Valuation adjustment (Kitces/Pfau): starting valuations predict the first decade, and the first decade decides most of the outcome. Rich starting valuations (CAPE well above history) argue for 3 to 3.5%; cheap valuations historically supported 5%+. - State the failure modes plainly: high inflation, a decade-long equity slump, too-conservative allocation, or fees can break the fixed rule even at 4%. 3. **Install a dynamic spending rule, not a fixed one.** Fixed inflation-adjusted withdrawals ignore reality. Pick one and commit to its cuts: - **Guardrails (Guyton-Klinger style):** start near 5% of the starting portfolio; guardrails at +/-20% of the initial rate. If the current withdrawal rate breaches the upper guardrail, cut spending 10%. If it falls below the lower guardrail, raise 10%. Skip the inflation bump after a negative-return year. Higher starting income, paid for with real variability. - **Kitces ratcheting:** start at 4 to 5%; when the portfolio exceeds 150% of its starting balance, take a 10% raise (plus inflation), at most one raise every 3 years. Raises only, never cuts; pairs well with a lower starting rate. - **RMD-style percent-of-balance:** withdraw a horizon-and-age-based percentage of the current balance each year. Spending flexes with markets, and the portfolio cannot mechanically hit zero. Compare policies with `bin/decumulate.py simulate` (below). A policy the retiree will not actually follow is worse than a conservative fixed rate. 4. **Blunt sequence-of-returns risk.** The order of returns matters more than the average. Practical defenses: - Hold 1 to 2 years of spending in cash or short-term bonds so bear markets do not force sales at the bottom; refill from equities after recovery. - Consider a bond tent: heavier bond allocation in the first ~10 to 15 years, gliding back toward equities later (Kitces/Pfau rising-equity glidepath). - Time matters most early: Pfau estimates the first decade drives the bulk of the final outcome. A bad first five years is the plan-killer; protect it. 5. **Sequence withdrawals for taxes.** Default order: RMDs first (mandatory), then taxable accounts (capital gains at lower rates, return of basis untaxed), then tax-deferred, Roth last (longest compounding, best estate asset). Then improve on the default: - **Roth conversion window:** in low-income years before RMDs begin, convert tax-deferred dollars up to the top of the current bracket. This shrinks future RMDs and smooths lifetime brackets. - **Bracket-filling, not just order:** blended or proportional withdrawals across account types can beat strict sequencing; the T. Rowe Price research found large gains from dynamic approaches. - Watch the cliffs: RMDs taxed as ordinary income, Social Security taxation thresholds, IRMAA Medicare surcharges, and the fact that RMDs cannot be converted or rolled over. 6. **Schedule RMDs explicitly.** Use `bin/decumulate.py rmd` for a projected schedule, then verify against IRS Publication 590-B. Rules baked into the tool: - Start age: 73 if born 1951-1959, 75 if born 1960 or later (SECURE 2.0). First RMD can wait until April 1 of the next year, but that doubles up two RMDs in one tax year. - RMD = prior Dec-31 balance / IRS Uniform Lifetime Table factor (e.g. 26.5 at age 73). Penalty for a miss is 25% of the shortfall, 10% if corrected within 2 years. - Roth IRAs have no lifetime RMDs; Roth 401(k)s exempt since 2024. IRAs may aggregate (one withdrawal covers all IRAs); each 401(k) needs its own. Still-working exception covers the employer plan only, not IRAs. - QCDs from age 70 1/2 can satisfy RMDs without adding taxable income. 7. **Write the withdrawal policy statement.** Document: income goal, account inventory, starting rate, chosen dynamic rule with exact triggers, account sequencing order, Roth conversion plan by year, RMD calendar, and the pre-agreed response to a 20% portfolio drop. Review it yearly. ## Output Contract A usable decumulation plan contains: - Starting withdrawal rate with the valuation reasoning behind it, stated as a range with conditions, never a single guaranteed number. - One dynamic spending rule, named, with exact triggers the retiree agreed to follow. - Sequence-of-returns defenses: cash buffer size, allocation glidepath, first-decade focus. - Year-by-year account sourcing (which accounts fund spending, conversion targets) for the pre-RMD window at minimum. - RMD schedule with ages, divisors, deadlines, and aggregation notes. - A dated withdrawal policy statement the retiree can actually read. ## Operating Rules 1. Verify every factual claim before it ships. Never invent numbers, dates, quotes, or events. Tax figures reflect law as of the research date; flag anything that may have changed and point at IRS Publication 590-B. 2. History is a guide, not a promise. Bengen and Trinity results describe US markets over specific periods; say that. 3. A spending rule only works if the retiree follows it. Prefer a plan they will keep over a theoretically optimal one they will abandon. 4. Never present a backtest as advice. Every plan carries the plain-language disclaimer: educational, not financial or tax advice. 5. The divisors in `bin/decumulate.py` come from IRS Publication 590-B Table III (2022 update). If the IRS revises the table, the tool must be updated from the primary source, not from memory.