Milevsky Ch. 3: Probability of Ruin — The Only Retirement Number That Actually Matters

"Ruin is not an event. It is a probability that you are choosing, whether or not you know you are choosing it." — the actuarial framing of retirement withdrawal

Investment Background

Every retirement plan produces a single headline number. Monte Carlo software prints it, advisors quote it, and clients remember it: your plan has a 92% probability of success.

Milevsky's contribution — and his warning — is that this number is simultaneously the most important output in retirement planning and the most systematically misunderstood. It is important because it is the only figure that combines the three risks that actually determine outcomes: how long you live, what returns you earn, and how much you withdraw. It is misunderstood because of what it does not say.

Start with the word "success." A 92% success rate does not mean you will have 92% of what you need. It means that in 8% of simulated lifetimes, the portfolio hit zero while you were still alive. The word "failure" in this framework describes a specific and catastrophic event: an 89-year-old widow discovering that the account is empty and the only remaining income is Social Security.

Now the harder part. The probability tells you nothing about the size or the timing of the failure, and those are the two things that determine whether the failure is survivable. A plan that fails at 94, with the shortfall covered by downsizing a paid-off house, is a rounding error. A plan that fails at 79, with fifteen years of uncovered essential spending ahead and no house to sell, is a household disaster. Both show up in the software as one point of failure probability.

Neither is what a retiree actually wants to know. What they want to know is: what do I change?

The Wall Street Translation

What the Number Actually Responds To

Probability of ruin is driven by four inputs, and they are wildly unequal in leverage:

Input	Typical effect on ruin probability	Under your control?
Withdrawal rate	Enormous — the dominant lever	Yes, fully
Spending flexibility (ability to cut)	Very large	Yes, mostly
Longevity	Large	No
Asset allocation	Moderate	Yes
Sequence of returns	Large	No
Fees	Moderate but permanent	Yes, fully

The ordering is the lesson. Retirees and their advisors spend the overwhelming majority of their attention on the one input in the middle of that list — asset allocation — and comparatively little on the two at the top, which they fully control.

The rough sensitivities are worth memorizing. Over a 30-year horizon with a diversified portfolio, moving the initial withdrawal rate from 4% to 5% typically takes success from the low nineties into the seventies. Moving from 4% to 3.5% pushes it toward the high nineties. Meanwhile, moving equity allocation from 50% to 70% typically changes success by a handful of percentage points — real, but a fraction of the withdrawal-rate effect. And a 1% annual fee, compounded across thirty years, costs roughly the same as raising your withdrawal rate by close to a full percentage point.

Why 100% Is the Wrong Target

Retirees hear 92% and ask how to get to 100%. This is the most expensive question in retirement planning, and the answer is that you cannot — and should not try.

You cannot, because 100% requires assuming a horizon longer than any human lifespan and a return sequence worse than any recorded history. Software that reports 100% has usually just been given a conservative enough set of assumptions to make the answer meaningless.

You should not, because the last few percentage points are ruinously expensive in a different currency. Getting from 90% to 99% typically means cutting spending by roughly 20–25% for your entire retirement. You are buying insurance against a scenario with a one-in-ten chance by permanently reducing the quality of the ninety-in-a-hundred scenarios. Milevsky's framing is blunt: the cost of a 99% plan is a materially worse life, paid with certainty, to avoid a possibility.

The Reframe: From Probability to Magnitude

The useful question is not "what is my probability of ruin" but "in the failing scenarios, how bad is it, when does it hit, and what would I do?"

This converts an abstract percentage into a plan. Run the failing paths and ask three things. What year does the portfolio deplete? What is my income in that year from sources that cannot fail — Social Security, pension, any lifetime income? What assets exist outside the plan — home equity, a second property, whole life cash value?

A household with $58,000 of inflation-adjusted Social Security, a paid-off $500,000 home, and a plan that fails at 91 in the bottom decile is in a completely different position from a household with $26,000 of Social Security, renting, and failing at 78. Their probability-of-ruin numbers may be identical. Their actual risk is not remotely comparable.

Note what this reframe does to the portfolio question. It does not argue for abandoning the diversified index core, and it does not argue for market timing to avoid the bad paths. The core stays broad, low-cost and indexed, because the failing paths are mostly caused by withdrawal rate, fees and inflexibility — three things you control directly — not by holding the wrong fund.

Execution Rules

1. Target a success band of 85–90%, not 95% or 100%. Below 80% the plan is genuinely fragile. Above 90% you are almost certainly underspending. The band from 85 to 90 is where the trade between shortfall risk and a diminished life is balanced, and where dynamic adjustment can carry the remainder.

2. Buy flexibility instead of buying probability. Adopt a guardrail rule and write it down before you need it: if the portfolio falls more than 20% below the plan's projected path, cut discretionary spending by 10% until it recovers. Historical analysis consistently shows that modest, rules-based spending adjustments raise success rates by roughly as much as a 0.5–1.0 percentage point cut in the initial withdrawal rate — while letting you spend more in the good states.

3. Stress the failing decile explicitly, in dollars. Ask for the median depletion year, the annual shortfall in that year, and the income remaining. Convert the shortfall into a concrete backstop: home equity, a reverse mortgage line established early as a standby, downsizing, or family support. A plan without a named backstop is not finished.

4. Attack fees before attacking allocation. Every basis point of cost is a permanent, guaranteed deduction from the success probability, with no offsetting upside. Moving from a 1.1% all-in advisory-plus-fund cost to 0.25% is worth more to your ruin probability than most allocation debates, and it is the only lever with a certain payoff.

5. Recalculate annually with the actual balance and the actual conditional horizon. Probability of ruin at 65 assuming 30 years is not the same question as probability of ruin at 74 with the real balance and a 20-year conditional horizon. A plan run once and filed is not a plan; it is a memento.

Retirement Application

The most common practical failure is treating the success percentage as a verdict rather than a dial. Consider a couple, both 66, with $900,000 and combined Social Security of $52,000, who want to spend $88,000 a year. That requires $36,000 from the portfolio — a 4.0% initial rate. Software reports something in the high eighties.

The wrong reaction is to cut spending to $80,000 to reach 96%. That is a permanent $8,000-a-year reduction — over thirty years, roughly a quarter of a million dollars of foregone living — bought to avoid an outcome that was already unlikely and that they had other resources to absorb.

The right reaction is a set of three commitments. First, adopt a guardrail: if the portfolio drops below $720,000, discretionary spending falls by $6,000 until it recovers. Second, delay the higher earner's Social Security to 70, converting portfolio dependence into inflation-linked lifetime income and directly shrinking the tail. Third, establish home equity as a named, deliberate backstop rather than an unspoken hope.

Those three commitments improve the real risk profile more than the spending cut would, and they cost nothing in the ninety scenarios out of a hundred where nothing goes wrong. The portfolio itself does not change: it stays a low-cost global index core, because that core is not what was creating the risk.

Risk Management

• Success rates are model outputs, not facts. Monte Carlo results are extremely sensitive to the assumed return and inflation inputs. A plan showing 94% on 7% nominal equity returns may show 81% on 5.5%. Always ask what returns the software assumed, and re-run at lower ones.
• Sequence risk is concentrated in the first decade. The failing paths are overwhelmingly those with poor returns in the first ten years of withdrawals. A cash and short-bond buffer covering two to three years of withdrawals directly attacks the mechanism, by removing the need to sell equities into a decline.
• Inflation is the quiet killer of the number. A 3% assumption versus a 4% realized outcome over thirty years is the difference between a comfortable plan and a failing one. Inflation-linked income — Social Security, TIPS, I-bonds — is a more direct hedge than any equity allocation change.
• The number ignores long-term care entirely. Most Monte Carlo plans model level spending. A two-year nursing home stay at $120,000 a year is not in the simulation. Model it as a separate, explicitly reserved contingency rather than hoping the success rate absorbs it.

What This Chapter Cannot Do

This chapter cannot give you your personal probability of ruin, because that requires your balance, your spending, your Social Security, your health and your portfolio, and because the number is only as good as the return assumptions behind it.

It also cannot make the number mean more than it does. Probability of ruin measures one specific hazard — the portfolio reaching zero while you are alive — and is silent on long-term care, on family obligations, on divorce, on a business failure, and on the retiree who has plenty of money but no idea what to do with their days.

And it makes no case against indexing. Nothing in a ruin calculation suggests that active management, tactical allocation or complex products improve the odds. The levers that move the number are withdrawal rate, flexibility, fees and lifetime income. The core stays exactly where it was.

---

Key Takeaway: A 92% success rate is not a grade; it is a dial you are setting. Aim for 85–90%, buy flexibility with a written guardrail rule instead of buying the last few percentage points with a permanently smaller life, cut fees before debating allocation, and insist on knowing not just the probability of failure but its year, its size, and your named backstop.