The retirement models that show 95% survival probabilities are not deceiving you. But they are making a quiet assumption about your behaviour that most people never examine: that when the model says “cut your spending by 20%”, you will actually cut it by 20%. Not notionally. Not in a spreadsheet. In real life, in a bad market year, when your portfolio has already dropped and you are wondering whether this is a temporary correction or the beginning of something worse.
This assumption is doing enormous work. Dynamic withdrawal strategies — the rules that adjust your spending based on portfolio performance — are mathematically elegant and historically well-supported. But their survival probabilities are conditional on compliance. If you cannot cut to where the model tells you to cut, the model’s probability is not your probability. It is the probability for a more flexible version of you that does not exist.
The fix is not a more sophisticated withdrawal rule. It is a more honest number: your withdrawal floor.
1. Three numbers that define your actual plan
Before building any withdrawal model, you need three figures:
- W — your planned withdrawal (what you intend to spend in a normal year)
- F — your floor withdrawal (the absolute minimum you could sustain — essential housing, food, utilities, healthcare, with no discretionary spending)
- P — your portfolio value at retirement
From these, a single ratio reveals how much flexibility your plan actually contains:
φ = (W − F) / W
This is your flexibility ratio. It tells you the maximum percentage by which you could cut your withdrawal before hitting a level you genuinely cannot sustain.
| Example | W | F | φ | Interpretation |
|---|---|---|---|---|
| Comfortable buffer | $80,000 | $55,000 | 0.31 | Can absorb a 31% cut before hitting bone |
| Moderate buffer | $80,000 | $65,000 | 0.19 | 19% maximum cut available |
| Thin buffer | $80,000 | $75,000 | 0.06 | Almost no room to manoeuvre |
A φ near zero is not simply uncomfortable — it is structurally significant. Dynamic withdrawal strategies are built on the premise that you will cut when prompted. If φ is small, that premise fails regardless of how sophisticated the rule is.
2. What dynamic withdrawal rules actually require from you
The most widely cited dynamic strategy is the Guyton-Klinger guardrails framework (Jonathan Guyton and William Klinger, 2006). In simplified form:
- If your portfolio falls below a lower guardrail, reduce your withdrawal by a fixed percentage (typically 10%)
- If your portfolio rises above an upper guardrail, increase your withdrawal by a fixed percentage
The elegance of this system is that it extends portfolio longevity by preserving capital during downturns. Simulations show it comfortably outperforms rigid fixed-rate strategies in most historical sequences. But those simulations assume the retiree actually follows through on the cut.
If W = $80,000 and the lower guardrail triggers a 10% reduction, you are now expected to live on $72,000. If your floor F is $75,000, you cannot comply — you withdraw $75,000 regardless, because $72,000 does not cover your fixed costs. The model expected $72,000; you took $75,000. Your portfolio is now on a slightly worse trajectory than the model predicted. In a good year, this difference is negligible. Repeated across a bad sequence of years, it compounds into something material.
The sequence of returns risk post covers why early-sequence downturns are disproportionately damaging. The floor constraint makes that damage worse, precisely when it matters most.
3. Adding the floor constraint to the model
The mathematical correction is straightforward to state and significant in its implications. Let W(t) be the withdrawal at time t, governed by whatever dynamic rule you prefer. Now add a single constraint:
W(t) = max[ rule(P(t), W(t−1)), F ]
This says: apply the dynamic rule, but never withdraw less than your floor. In a well-resourced plan with a large φ, this constraint rarely binds. In a plan with a small φ, it binds frequently — exactly when the portfolio can least afford it.
To see the true impact, the right tool is a Monte Carlo simulation with the floor constraint built in:
- Run N scenarios (10,000 is standard) using historical or bootstrapped return sequences
- At each annual timestep, apply the dynamic rule, then clamp the result to F
- Count portfolio depletion events
The result is a floor-constrained ruin probability: P(ruin | F). This will always be at least as high as the standard unconstrained figure — and for plans with a small φ, it can be substantially higher. The gap between these two numbers is what most published retirement calculators do not show you.
4. The flexibility-adjusted safe rate
The 4% rule, derived from William Bengen’s 1994 research and confirmed by subsequent Trinity Study work, has become shorthand for “a safe starting withdrawal rate over a 30-year horizon.” But it is not the only relevant rate in your plan.
Define your flexibility-adjusted initial withdrawal rate:
r_adj = F / P
This is the withdrawal rate that governs your worst-case behaviour — the rate you are implicitly locked into when the dynamic rule pushes you to the floor. It exists whether or not you have modelled it.
| Scenario | P | W | F | Planned rate | r_adj |
|---|---|---|---|---|---|
| Healthy buffer | $2,000,000 | $80,000 | $55,000 | 4.0% | 2.75% |
| Tight buffer | $2,000,000 | $80,000 | $75,000 | 4.0% | 3.75% |
| No buffer | $2,000,000 | $80,000 | $80,000 | 4.0% | 4.0% |
In the third row, the “dynamic” strategy is not dynamic at all — it is a fixed withdrawal at 4%. The Guyton-Klinger guardrails cannot function because there is no space between the floor and the plan. A common planning threshold is that r_adj should sit below 3.5% to give dynamic strategies meaningful room to operate.
If r_adj is already near your planned rate, you should either increase P before retiring, reduce F by renegotiating fixed costs, or accept that your true survival probability is lower than the model suggests.
5. Stress-testing the floor: sequence risk and the critical portfolio level
Average bad returns are not the real threat — early bad returns are. To stress-test your floor, model a scenario where years 1–5 draw from historically severe sequences: the 1966–1970 stagflation period, the 2000–2002 dot-com crash, or the 2008–2009 global financial crisis. In each of these, equity markets fell 30–50% in nominal terms while inflation continued to erode purchasing power.
In this stress test, track how quickly P(t) declines when W(t) is clamped to F at every step. The key output is P* — the critical portfolio level.
P* is the portfolio value below which, even with average returns going forward, the floor withdrawal exceeds the portfolio’s sustainable rate. Below P*, recovery becomes mathematically implausible regardless of what markets do next. Once your portfolio crosses below P*, you are in a ruin trajectory that only extraordinary returns or a genuine reduction in F can reverse.
Knowing P* in advance converts an abstract risk into a concrete early-warning threshold. If your portfolio approaches P* in year three of retirement, you are in structurally dangerous territory — not just temporarily uncomfortable.
The retirement smile research from David Blanchett (2014) shows that real spending tends to decline naturally through the middle phase of retirement. This is encouraging, but it is not guaranteed — and healthcare costs can produce a sharp late-retirement spike that partially offsets the decline. Do not assume that your floor will fall over time without explicitly modelling how it might change.
6. What an honest model should tell you
A complete withdrawal analysis should output at minimum three figures, not one:
| Metric | What it tells you |
|---|---|
| φ — flexibility ratio | How much you can cut before hitting your genuine floor |
| P(ruin | F) — true ruin probability | Your actual failure probability given that the floor constraint binds |
| P* — critical portfolio level | The portfolio value below which recovery is implausible given F |
Most people enter retirement knowing their planned withdrawal rate and a single modelled ruin probability. The ruin probability they have seen is the unconstrained one — the number that assumes compliance with cuts that may not be feasible for them specifically.
The floor-constrained ruin probability is harder to compute and less flattering. It is also more honest, and therefore more useful.
What this means for your FIRE plan
The floor calculation is not an exercise you do once and file away. It deserves deliberate, granular attention before you retire — and periodic revisiting after.
Some questions worth sitting with:
- What is actually in your floor? List every fixed cost: rent or mortgage, essential food, utilities, insurance, healthcare premiums, minimum debt service. Remove all discretionary spending. That number is F.
- How honest is that number? Many people underestimate F because they include optimistic assumptions about what they could cut. If you have never lived at F, you do not know whether it is actually survivable.
- What is your φ? If it is below 0.15, your plan has limited dynamic flexibility. Below 0.10, question whether the strategy you are modelling is the one you will actually execute.
- What would cause F to rise? A fixed mortgage converting to market rent, a medical diagnosis, caring responsibilities, a change in geography — all of these can move F upward after retirement. Stress-test against an F that is 20–30% higher than today’s estimate.
- Where is your P*? That number should be somewhere on your dashboard before you retire, not discovered in year four.
Use our Retirement Calculator to model your plan under different floor assumptions — run the same scenario at your planned withdrawal W, then again at your floor F, and compare the survival trajectories. The gap between those two curves is the honest range of your dynamic strategy’s outcomes. Pair it with our Life Expectancy Quiz to calibrate how long your plan needs to hold — because the longer the horizon, the more consequential a small φ becomes.
This article is for general educational purposes and does not constitute financial advice. Withdrawal rate research is based on historical market data and does not guarantee future outcomes. Individual circumstances — tax treatment, income sources, healthcare costs, and living expenses — vary significantly and are not captured by generalised models. The mathematical frameworks presented here are illustrative; a full floor-constrained Monte Carlo analysis should be conducted with tools calibrated to your specific situation. For advice tailored to your circumstances, consult a qualified financial adviser.