Guyton-Klinger guardrails vs fixed

Ayo is 55, has £700,000 invested and wants £28,000 a year out of it. That is a 4% start, the textbook number, and the withdrawal strategies tool says the plan lasts to 95 in 61.0% of a thousand simulated markets.

Change the rule to Guyton-Klinger guardrails and the same pot, the same spending and the same thousand markets come back at 99.9%.

Thirty-nine points for changing nothing about the money. That is the moment most people stop reading, and it is the wrong moment to stop, because the panel is already showing what it cost: in the leanest year of the unluckiest tenth of those paths, guardrails pays Ayo £5,765. The rule does not make his plan safe. It makes his income the thing that moves, and whether that is a good trade depends entirely on which parts of his spending he can actually stop.

Every run below is a thousand paths at 15% volatility from a seed that never moves, so your screen will match these numbers once you drag annual spending to £28,000. The panel opens at £32,000, which is a different plan: it leaves Ayo short of his own target and still saving for a few more years. The currency selector swaps the pound signs for dollars, euros or rupees without touching anything underneath.

What the rule does on 31 December

Once a year, guardrails compares this year's withdrawal against the portfolio it is coming out of. If that ratio has drifted more than 20% above where it started, spending gets cut by 10%. If it has fallen more than 20% below, spending rises by 10%. In between, nothing happens at all, which is the rule's whole reputation: a fixed plan's steady income, corrected only at the extremes. Hold on to that claim. The history does not support it.

The 10% comes off current spending, not off the original £28,000, so cuts compound. Two bad ones leave him on £22,680, not £22,400, and the gap widens every time the rule fires.

And nothing in the tool stops it firing. Jonathan Guyton and William Klinger's 2006 paper gave the cut an expiry date, switching the capital preservation rule off for the last fifteen years of the plan, on the reasoning that late cuts destroy purchasing power without buying much safety. The calculator implements the 20% trigger and the 10% cut and leaves the expiry out, along with the inflation and portfolio-management rules the authors ran alongside them. So read it as the guardrail mechanism rather than as the published system.

That omission is exactly why the worst-year number looks the way it does. £28,000 reduced by 10% fifteen times over is £5,765, to the pound. The p10 figure in the panel is not a rounding of some deeper calculation; it is fifteen cuts with not one raise between them, in a path that still counts as a success because the money never actually ran out.

The cohorts that broke the fixed plan

Simulated markets are only half the evidence. Run Ayo's plan through the historical backtest instead, which starts him in each of the 124 years from 1900 onwards and wraps the sequence round so every cohort gets a full forty years, and fixed withdrawals survive 93.5% of them. Eight cohorts fail: 1906 and 1907, then 1929 and 1930, then 1966, 1968 and 1969, and finally 2000. That page always runs the fixed rule, so to put the same 124 sequences through the other two, use the planner on the home page, which carries the strategy buttons once you switch it to historical mode.

Guardrails fails none of the 124.

What it does instead is intervene far more often than its reputation suggests. Across the 4,960 cohort-years those runs contain, spending is cut in 12.7% of years and raised in 37.1%, leaving almost exactly half untouched, and the median cohort gets 20 quiet years out of 40. Nor is it mainly a cutting rule: historically it handed out close to three raises for every cut. Anyone picturing a fixed income with rare emergency trims has the wrong picture.

The 1966 run shows both halves of that at once, and what the rescue costs. A retiree starting there meets real returns of -13, 21, 6, -14, -2, 11, 15, -22 and -35 per cent in their first nine years. The fixed plan answers every one of those by selling a fraction more of a pot that keeps shrinking, and empties at 85. Guardrails cuts instead, seven times, at 60, 63, 64, 65, 67, 68 and 71.

Which leaves Ayo drawing £25,200 at 60, £18,371 at 65 and £14,880 at 70. His leanest year is £13,392, less than half what he retired on, and 26 of his 40 retirement years come in under the £28,000 he planned for. He does not run out of money. He spends his sixties and seventies poorer than he intended, and that is the actual product being sold when a success gauge jumps to 100%.

Then the 1980s arrive, the prosperity rule starts handing back raises, and the same path pays £26,098 at 85 and £42,031 at 90, finishing with £1.2 million unspent. Guardrails saved the plan by moving Ayo's consumption from the decades he could use it to the decade he probably cannot. That is not an argument against the rule, since the fixed retiree was broke at 85 and had no decades left at all, but it is the shape of the trade, and the sequence-of-returns risk post covers why the early years are the ones doing the damage.

Notice the two engines disagree sharply: 61.0% simulated against 93.5% historical, for the identical plan. History is kinder because there are only 124 overlapping sequences in it and they mean-revert, while the simulator draws every month independently, at an annual 15% volatility, and will happily deal a decade no market has produced. Neither is the truth. The gap between them is roughly the size of the modelling assumption, which is worth holding in mind before treating any single percentage as a fact about your retirement. The Trinity authors made the same point from the other end in Portfolio Success Rates: Where to Draw the Line, where the survival odds for a given withdrawal move with the payout period and the data window, in their case US stocks and corporate bonds from 1926 to 2009. Both engines here also run on approximate US large-cap real returns, so a globally diversified portfolio is a rougher fit again.

Your starting rate stops mattering

Here is the part that changes how the tool should be used. Run the same pot under guardrails at three different starting rates and watch what the rule does to the income you actually receive.

Move from the top row to the bottom and Ayo declares 67% more income, then ends up with 16% more. The rule claws back nearly the whole difference, because a rate high enough to be worth declaring is a rate the guardrail spends the next forty years correcting. At a 3% start the correction runs the other way: the prosperity rule keeps handing him raises, and his median income lands above the number he asked for.

Guyton and Klinger concluded that initial rates of 5.2% to 5.6% were sustainable at a 99% confidence standard with at least 65% in equities. The success column here lands in the same territory without quite matching it, 98.4% at a 5% start, on a different set of assumptions about returns. What the middle column adds is that sustainable and received are different questions. If you are using guardrails, your starting rate is mostly a statement about how much variability you have signed up for, and the safe withdrawal rate tool answers a question that fixed plans ask and this one largely does not.

Two numbers that look alike and are not

Both rules report a low-income figure and they mean opposite things.

Fixed withdrawals at 4% show a worst year of £6,667. That is not a cut. Fixed real never cuts. It is the final stub of a year in the paths where the pot emptied, which is to say it is the sound of the plan ending, with nothing behind it. In the historical run those endings land at 75, at 84, at 85, at 89 and at 90.

Guardrails shows £5,765, and Ayo is still solvent, still drawing, and will draw again next year. Lower number, survivable situation. So the comparison that matters is not which figure is smaller but whether you would rather face an income that falls a long way or an account that reaches zero while you are alive to watch it, which is the trade the rich, broke or dead view prices directly.

There is an honest third option people skip. Fixed withdrawals at 3.25%, £22,750 a year, also survive all 124 historical cohorts. That income is known in advance, every year, for life. Guardrails at 4% delivers a median £26,284, about 15% more, in exchange for never knowing January's number in December. Some people should take the certain £22,750, and telling them otherwise because a success gauge reads 99.9% is bad advice dressed as arithmetic.

Whether you can actually take the cut

A guardrail cut is only real if there is something to cut. Ayo's £28,000 might be £11,000 of fixed costs and £17,000 of choices, in which case the rule can fire eight times, down to £12,053, before it reaches anything he cannot stop buying. The ninth cut is the one that takes food and heating. If most of it is a mortgage, or care costs, or supporting someone, then agreeing in advance to cut 10% is a promise he will break in the year it matters, and a broken guardrail plan is a fixed plan that started too high.

The practical version of that question in the UK is which pot the money comes from. Money in a stocks and shares ISA is available now and taxed at nothing on the way out, which makes it the flexible part of the plan. A SIPP or workplace pension is locked until at least 57 for anyone under 50 today, and the State Pension arrives later still on a date that depends on your birth year. The names are local and the arithmetic is not: a 401(k), an RRSP, an Australian super balance or a French contribution record all lock money away until an age somebody else chose, and every national scheme pays out for precisely as long as you are alive to collect. Whatever yours is called, find the age it starts and the amount it pays, because that last one matters more than it looks here. Guaranteed income indexed for life is the floor the guardrail rule does not have. Once it starts, the portfolio is covering a smaller number, the withdrawal ratio stops drifting upward so easily, and the compounding cuts lose most of their power. If your date sits well before it, the pension bridge is the piece to model first.

So set your own floor, since the calculator will not. Decide the annual number below which you would change the plan rather than take another 10%. Write it down now, in your own currency, while nothing is going wrong. Then run your plan under both rules and compare the worst-year figures against it. If guardrails can take you below that line and you would not follow it there, you are not running guardrails. The walkthrough of the three rules covers what each one does; where your own floor belongs is the part the tool cannot work out for you.