How much should be in safe assets? The two answers you will meet first treat that as a question about you — your age, your months of expenses, your tolerance for a bad Tuesday. It is not. It is a question about what the money has been promised to. And before you can size the safe part you have to say which part of the book — my word throughout for the whole portfolio — it is, and neither rule of thumb tells you. In the Incomestead Stack it is two Layers counted as one quantity: Bedrock plus Cash — the part you will not be forced to sell, and will not trade around. We call that sum The Receding Floor.
Across four reference shapes a book can be built to, that floor reads 52 / 29 / 13 / 5. Deep when the money must not break; thin when it has been left alone to compound. Those four are illustrations of four shapes, not a prescription for your book.
TL;DR
- The safe part of a book is a sum, not a vibe: Bedrock plus Cash, read as one share of the whole book. The name for that sum is The Receding Floor — the part you will not be forced to sell, not the part that cannot fall.
- Its right size is not a fact about you — not your age, not your months of expenses. It follows from what the money has been promised to, which is why the floor is supposed to recede.
- Across the four reference shapes it reads 52 / 29 / 13 / 5. Illustrations of four shapes — and there is no governed "correct floor" for your book, which is the honest answer rather than a missing one.
- Where the points go has a name of its own: The Counterweight. End to end the floor falls 47 points and growth rises 51 points; the other four Layers net four points between them.
- A floor that shrank because the plan changed is not drift and is not fixed by a trade. Amend the written target first; then a rebalance has something to aim at.
I write as a practitioner, not a licensed adviser. I run a leveraged, multi-strategy book myself, and for years my own answer here was a feeling — a sense that I had "enough" boring stuff under the interesting stuff. What I actually had was a number I had never written down, never named, and therefore could never be wrong about. This post names it.
The two answers everyone gives, and what they are actually about
There are two answers in general circulation and they are both older than any of us. The first is an age rule: subtract your age from a hundred and hold that much in equities, the rest in something steadier. The second is a liquidity rule: keep three to six months of expenses in cash. You will meet both on the first page of the search that brought you here.
Neither is stupid. For somebody with nothing written down, either beats the alternative — a number chosen by whatever happened last. But look at what each takes as its input. The age rule reads your birthday. The liquidity rule reads your spending. Both produce a figure about the investor, and then hand it to you as though it were a figure about the portfolio. Advisers who think past the rules of thumb already size the safe part by dated obligations: goals-based wealth management, liability-driven investing and the bucket approach all work that way, and none is fringe. What I have rarely seen any of them publish is which holdings count — the half this post is about.
An age rule reads your birthday. A months-of-expenses rule reads your spending. Both are reaching for the obligation, and neither one asks.
There is a second problem, and it is the one that actually stops you governing anything. Suppose you accept "twenty-five percent in safe assets." Twenty-five percent of which holdings? Is a dividend fund safe? A short-dated bond ladder you bought for income rather than for ballast? A number with no stated membership cannot be checked, and a number that cannot be checked cannot be breached, amended, or handed to anybody else. It is a feeling wearing a percent sign.
So the question splits in two, and the order matters. First: which part of this book is the part you will not be forced to sell? Then, and only then: how big should that part be, given what this money has been promised to? This post answers the first question once, and refuses on purpose to answer the second one for you.
What is the floor? Bedrock plus Cash, counted as one thing
The Incomestead Stack divides a book by the job each strategy does rather than by asset class. There are eight Layers — Bedrock, Pillars, Long-Horizon Growth, Short-Horizon Growth, Income Engine, Precious Metals, Crypto and Cash — and the pillar post sets out what each Layer is for and how the set is governed. Two of those eight share a property nothing else in the book has.
Bedrock is the broad, boring, low-cost core — the part most of the book rests on, the part you do not have a clever opinion about. In instruments, Bedrock is equity: the engine classifies it EQ_CORE, inside the Equities bucket, so when equities fall Bedrock falls with them — and broad equity falls a long way. On the peak-to-trough figures used elsewhere in this library, the S&P 500 lost 56.80% in 2007–2009 and 33.90% in 2020 (Yardeni Research, Stock Market Historical Tables: Bull & Bear Markets, 21 January 2024 edition). Cash is dry powder and buffer, and it is the part of the floor that does not move with the market at all. So the floor is not the part of a book that cannot fall — on a Preservation shape, thirty-two of its fifty-two points can. It is the part you have decided you will not be forced to sell and will not trade around. The other six — Pillars, the two growth Layers, the Income Engine, Precious Metals and the sliver of Crypto — are doing work a bad market can interrupt, delay or reverse.
Counted severally, they are two Layers among eight. Counted together, they are one governed quantity, and that is the move this post is making: The Receding Floor is Bedrock plus Cash, read as a single share of the whole book. The reason to sum them is simple. "How much should be in safe assets" is a question about a total. A total you have not named is a total you cannot govern — it has no target, no band, and no line in any document.
A total you have not named is a total you cannot govern. It has no target, no band and no line in any document — which is why it is the first thing to go.
One caution, because the word is already taken. The retirement-income literature uses floor in a stricter sense than I do: near-certain income — an annuity, a ladder of inflation-linked bonds held to maturity — and it excludes anything carrying market risk by design. Mine is looser, and you should know that before you carry the word anywhere else: the Incomestead floor is the part I will not be forced to sell, not the part that cannot fall.
Two tests in the library look adjacent to this one and are not. One asks whether every job in a book is being done by somebody — the difference between a portfolio and a pile of good stocks is a coverage-and-concentration question with a fixed pass mark. The other asks whether your biggest strategy has quietly become the whole book. Both read a ceiling. The floor test reads the opposite end of the same distribution, and it has no fixed pass mark at all.
The idea that a portfolio is really several portfolios, each with its own job, is not ours and is not new. Shefrin and Statman's behavioral portfolio theory describes investors who segregate a book into separate mental accounts and size each one against what it is for: in their words, "the low aspiration layer is designed to avoid poverty while the high aspiration layer is designed for a shot at riches" (Shefrin & Statman, Journal of Financial and Quantitative Analysis 35(2), 2000, pp. 127–151).
The same paper then argues against the method. Shefrin and Statman show that in their model a portfolio built this way is "neither mean-variance efficient nor BPT-SA efficient", because an investor who segregates a book into separate accounts, in their words, acts "as if they overlook covariances" between them — which is exactly what sizing each Layer against its own job does. I take that cost knowingly, and you should know I am taking it: I am buying a number I can write down, band and audit next year, at the price of an efficiency I have no honest way to measure on my own book. It is a positive theory in any case — a description of how investors build portfolios, not a recommendation of how big either layer should be — so it hands you no number, and neither will I.
Why the floor is supposed to shrink
If the floor's size follows from what the money is for, then it has to move when that answer moves. In the Investment Charter — the written policy that governs the book — "what is this money for?" has four published answers, and the Charter pillar defines all four and the rules behind them. Named with their glosses, because the gloss is what stops the noun reading as a risk profile:
What is this money for?
Preservation — it must not break.
Provision — it pays the bills.
Accumulation — it compounds, untouched.
Acceleration — it's the swing: money you have decided in advance you can lose in full.
Each of the four points at a different shape of the same eight Layers, and each shape is built to seven named construction rules rather than to taste. They are named, never numbered — a number is a position in a list; a name is a claim you can check. Two of those names are the subject of this post, and The Receding Floor is one of them.
| Layer | Preservation | Provision | Accumulation | Acceleration |
|---|---|---|---|---|
| Bedrock | 32 | 20 | 10 | 4 |
| Cash | 20 | 9 | 3 | 1 |
| The Receding Floor — Bedrock + Cash | 32 + 20 = 52 | 20 + 9 = 29 | 10 + 3 = 13 | 4 + 1 = 5 |
| Long-Horizon Growth | 5 | 10 | 27 | 32 |
| Short-Horizon Growth | 0 | 3 | 12 | 24 |
| The Counterweight — the two growth Layers | 5 + 0 = 5 | 10 + 3 = 13 | 27 + 12 = 39 | 32 + 24 = 56 |
| Pillars | 20 | 24 | 22 | 15 |
| Income Engine | 15 | 27 | 18 | 15 |
| Precious Metals | 8 | 6 | 5 | 4 |
| Crypto | 0 | 1 | 3 | 5 |
| The other four Layers | 43 | 58 | 48 | 39 |
| The Declared Whole | 100 | 100 | 100 | 100 |
Points of the whole book, which is declared at 100 in every column. The Receding Floor reads 52 / 29 / 13 / 5; The Counterweight reads 5 / 13 / 39 / 56; the other four Layers carry 43 / 58 / 48 / 39 between them. Provenance, in plain English: these four shapes are my own judgement rather than an optimiser's output. I set the weights by hand against seven named construction rules published alongside them, and adopted the set on 2026-08-17; they live at governance-score.vP.json → illustrative, set anchors/1. Four illustrations of four reference shapes — not recommendations, not forecasts, and no expected return is attached to any of them.
52 / 29 / 13 / 5. The part you have decided you will not be forced to sell is a little over half of a Preservation shape and one-twentieth of an Acceleration one. That is not four risk appetites. It is four answers to a question about obligations: as the money moves from being spent to being left alone, the share of it that has to survive a bad market without being sold falls, because less of it is promised to anything with a date on it.
The floor is supposed to recede. Watching it shrink is a finding about your plan, not a failure of nerve.
Three things have to be said plainly here, and the post is dishonest without all three.
These are illustrations, not recommendations, and they carry no expected return. Not a hidden one, not an implied one. I do not have a reliable forward return for any of these shapes, I have never seen one I would run a book against, and the moment a shape carries a return it stops being a construction and starts being a forecast.
There is no governed "correct floor" for your book, and that absence is correct. There are numbers I publish as binding and keep fixed; how deep your floor should be is not one of them, because there is no right answer to fix, and inventing one here would be exactly the personalized prescription this is not. What the four readings give you is a scale: somewhere to stand while you choose your own number, and four worked points to argue with.
And the trade runs in both directions. A deep floor costs you compounding — that is not a side effect, it is the price, and a Preservation shape pays it every year on purpose, and the cash inside that floor carries the mirror cost, losing purchasing power in any year inflation runs above what it earns. A shallow floor is only honest if nothing has been promised: if there is a mortgage, a school fee, or a date you told somebody you would stop working, a thin floor is not conviction, it is an unfunded obligation. Neither direction is the safe default. The default is writing down which one you have chosen.
One more thing, and it is the less flattering half. The recession is not even. The floor gives up 23 of those 47 points on the very first step, from Preservation to Provision — its largest single step, and just under half of the whole move. Most of what a floor can lose, it loses the moment the money stops being untouchable and starts merely paying the bills.
Where the points go: The Counterweight
You cannot say the floor gives up forty-seven points without a reader asking where they went. The answer has a name of its own: The Counterweight — the two growth Layers, Long-Horizon and Short-Horizon Growth, summed the same way. Read end to end across the four shapes, the floor falls 47 points and the counterweight rises 51 points. One movement seen from both ends.
Those two numbers are deliberately not equal, and the difference is the interesting part. The floor and the counterweight are not exact inverses. The remaining four Layers — Pillars, the Income Engine, Precious Metals and Crypto — net four points down between them across the whole dial, and not one of them moves more than five points end to end. Read "end to end" literally: the Income Engine finishes exactly where it started, but it climbs 12 points in the middle, where the job of the money is income. Every one of those moves is in the table above, so none of this is a set you have to take on trust.
The whole is declared at a hundred. Every point you add to the floor is a point taken from the thing that compounds — and that is the trade, stated in the only units that make it visible.
This is why the floor is not a dial you can turn on its own. "I would like a deeper floor" is never a standalone wish; it is a decision to hold less of the thing that compounds, and in practice the points come almost entirely out of growth.
One definition this post owes you, because I run a borrowed book and some of you will too. The whole book here means the gross market value of everything you hold, before any loan is netted off it. A margin loan is a dated obligation on the lender's timetable rather than yours, and measured gross the floor does not net it out: a borrowed book's floor is a share of assets the lender has a prior claim on, and that loan is what can force the very sale the floor exists to prevent.
A floor that shrank because you were right is a different problem
There are two ways a floor ends up thinner than it should be, and they look identical on a screen.
In the first, the book moved. Growth had a good year, the floor shrank as a percentage even though you never sold a share of it, and the gap between your written target and your real weights widened on its own. That is allocation drift — a measurable gap with a tolerance band and a persistence test, and it has its own post. The fix is a trade.
In the second, the book did not move at all. The plan did. That is this one, and until this post the library had no page for it.
Take an illustrative book built, deliberately and some years ago, to the Accumulation shape. Its floor target reads 13 points of the whole book. Nothing has gone wrong: the owner wrote it down, the weights have held, every Layer is inside its band. Then his job changes. The money that was compounding untouched is now the money a date has been attached to — it has been promised to something. The honest answer to "what is this money for?" is no longer Accumulation. It is Provision, where the floor target reads 29 points.
The gap is 16 points of the whole book. On a $1,000,000 book that is $160,000, and I give the dollar figure only because points are hard to feel — it is illustrative and it is not a target.
Nothing drifted. The book is exactly where it was put. The target moved — and a rebalance cannot aim at a number nobody has written down yet.
Now notice what this is not. It is not a drift finding. Both of those readings are target shapes, not holdings; no position moved, no weight slipped, and a drift report run the week before would have come back clean on every Layer. And it is not fixed by a trade, or at least not first — because until the written target is amended there is nothing for a rebalance to aim at. Rebalancing a book to a target its owner no longer believes in is a very tidy way of getting further from where you need to be.
The sequence is: amend the Charter, rebalance against the amended target, then measure drift against it from the next weekly reading on. The order is what makes the trade governable rather than improvised.
The sixteen points do not come from one place. Between those two shapes growth gives up 26 points; the Income Engine, Pillars and Precious Metals take 12 points of that back between them, the Income Engine taking most of it as the money starts having to pay for things; and Crypto gives up 2 points. Sixteen of growth's twenty-six end up in the floor. A change of purpose is not a tweak to one Layer — it is a different shape.
Writing the floor down so next year can read it
A number you have decided and not written down is not a policy; it is a mood with a good week behind it. The Investment Charter is where a floor target belongs, and the distinction it turns on is one we use constantly: an objective is a destination, a constraint is a guardrail. Your floor target is an objective. The band around it is a constraint. In my experience most self-directed investors have written down neither, and very few have both.
So here is the artifact, and it is one line plus three fields. Write it this week, on your own book, in whatever document you already keep.
The floor line — write it this week
- Purpose, one sentence. What this money is for — and whether anything has been promised to it with a date attached. If something has, say what and when.
- Which holdings count. Name them. In the Stack the floor is Bedrock plus Cash; if you use your own groupings, write down which of yours you will not be forced to sell and will not trade around.
- The target, in points of the whole book, and today's date. Your number, chosen by you against the purpose in line one. Not a number from this page.
- The band, and what you will do at the edge. How far it may sit from target before it is a breach, which reading you will act on, and what the action is.
Amendments get a new date and a new version; they never overwrite the old line. The point of writing it down is that next year can read what this year decided.
No tool can choose the number for you. A tool can hold the line you write and give it a date; only you can say what the money has been promised to, and that is the question which sets the number. What you get back is the thing neither rule of thumb can give you: next year, after a loud quarter, you can read what you decided in a quiet one.
Do this next
Write the line this week, while it is still a decision and not a memory. The Investment Charter asks six questions — what your money is for, what it must reach and by when, how much you will borrow, and who reads it if you can't.
What you get back is a dated charter as a PDF and a machine-readable charter.yaml, built in your browser. It does not read your account and it does not choose your floor number — that part is yours, and it is the part that sets the number.
Frequently asked questions
How much should be in safe assets?
There is no universal figure, and anyone handing you one is answering a question about you rather than about your money. The honest method has two steps. First, decide which of your holdings you will not be forced to sell and will not trade around — in the Incomestead Stack that is Bedrock, the broad low-cost core (which is equity, and falls when equities fall), plus Cash. Second, size that sum against what the money has been promised to. Across four reference shapes of the Stack that sum reads 52 / 29 / 13 / 5 points of the whole book, deepest where the money must not break and thinnest where it is being left alone to compound. Those four are illustrations, not a recommendation for your book.
What is The Receding Floor?
The Receding Floor is the Incomestead name for two Layers of the Stack counted as one governed quantity: Bedrock plus Cash — the part of a book you have decided you will not be forced to sell. It is not the part that cannot fall: Bedrock is equity, and equity falls. It is called receding because it is meant to shrink as the money's job moves from being spent to being left alone, not because it has gone wrong. It is one of seven named construction rules the four reference shapes are checkable against; they are named rather than numbered, because a number is only a position in a list while a name is a claim you can check.
Is cash part of my allocation, or separate from it?
Part of it. The commonest way a floor gets mis-measured is that the cash doing the reassuring is left out of the measurement: you feel covered because of it, then exclude it from the denominator, and the governed number stops describing the book you own. Note which way that error runs — it makes a book read more aggressive than it is, not less. $600,000 of equities beside $400,000 of bonds reads 60 / 40; count the $200,000 of cash held alongside and the same book is 50 / 33 / 17. Cash is a Layer of the Stack with a job of its own, and in the floor it is summed with Bedrock. The bright line is the date: money with a dated claim already on it — a tax bill falling due, a deposit you have committed to — is a liability rather than a holding, so subtract it before you measure. Money with no claim on it, however safe it feels, is in.
Should my safe assets shrink as I get older?
Not because of your age, no — that is the age rule wearing a different hat. What moves the floor is what the money has been promised to. A sixty-year-old whose book has nothing attached to it and a thirty-year-old funding a date he has told somebody about are in different shapes, and age ranked them the wrong way round. Age correlates with obligations, which is why the rule of thumb survives, but the correlation is the only thing it is tracking. Write down the obligation, and the obligation sizes the floor.
Does a bigger floor mean lower returns?
It means less of the book is in the Layers whose job is to compound, and that is the trade — stated as structure, not as a forecast. I attach no expected return to any of these shapes, so I will not tell you what a deeper floor costs you in percentage points; I do not have a reliable figure and I have never seen one I would run a book against. What I will say is that the cost is real and it is paid every year, on purpose, by anyone running a Preservation shape — and that the opposite error is just as real. A thin floor is only honest if nothing has been promised.
This is education, not personalized investment advice. I write as a practitioner who runs his own leveraged, multi-strategy book, not as a licensed adviser, and nothing here is a recommendation of any particular allocation for you. The four shapes are illustrations published as teaching data; they are not forecasts and no expected return is attached to any of them. Your own weights depend on your circumstances and are your decision.
Incomestead recommends. You decide.