# Designing a portfolio: the method, not the model Most books and most forums hand you a finished portfolio. The 60/40, All Weather, the Golden Butterfly, a recipe ready to copy. Copying a recipe has one deep flaw. It answers somebody else's questions. The right allocation is the one that answers **your** risks and that **you** can hold through a bad decade. This chapter gives you the method, not the model. A portfolio is not a list of funds. It is a series of answers to a series of questions, in order. And the order runs against instinct. You do not start with the assets ("how much gold?"). You start with the risks ("what can break my plan?"), then you decide which ones to defend, and only then do you pick the cheapest building block for each risk you defend. The allocation falls out at the end, as a result. Seven questions, in that order. This chapter covers the design of the portfolio itself. The full plan, with spending, the withdrawal rule and taxes, comes next ([[building-your-plan]]). ::: cle The idea in one sentence Design by risks, not by assets. Name the risks that can break your plan, decide which ones you accept and which ones you defend, then pick the cheapest building block that answers each risk you defend. The allocation is the output of the method, never its starting point. And a block you could not hold for ten years without proof is not in your portfolio, whatever the spreadsheet says. ::: ## 1. What are your risks? Start by naming the enemies of the plan. First come the market risks, the ones the portfolio can fight. Sequence risk, the order of returns in the early years, is enemy number one ([[sequence-of-returns]]). Then the long bear market (a lost decade), persistent inflation ([[inflation-and-withdrawal-rates]]), deflation, and a crisis of confidence in the currency ([[market-regimes]] for the growth × inflation grid). Then come the risks the portfolio does **not** handle. Longevity ([[horizon-and-life-expectancy]]), long-term care, divorce, a personal accident. Name every one of them, because the classic mistake is to defend one market risk twice and forget longevity. The portfolio answers market risks. The rest of the plan answers the others, through annuities, reserves, flexibility and insurance. ## 2. Which ones do you accept? You cannot defend everything, and you should not try. Every defense is paid for in carry or in return, because gold, cash, linkers and options all cost their price ([[defensive-assets]]). So separate two kinds of risk. The risk you are **paid** to carry is the engine of the plan: equity risk, which pays the premium that funds everything else ([[risk-premia]]). Insuring it away would mean insuring your returns away. The risk you **defend** against is the one that breaks the plan and pays you nothing in exchange: sequence and inflation. The whole art is choosing which tail to insure and which one to accept. Two ideas help you decide ([[deciding-under-uncertainty]]). Capacity is the loss the plan can absorb without breaking. Tolerance is the loss you can live through without cracking. Size on the smaller of the two. Accepting a risk explicitly, in writing, is a decision. Accepting it by forgetting about it is how plans die. ## 3. What is your horizon? The horizon drives every other answer. Retiring at 45 means planning over fifty years, retiring at 65 over twenty-five or thirty. The longer the horizon, the more the plan **needs** the equity engine, because nothing but stocks beats inflation over fifty years. But the longer it is, the more it exposes you to the sequence of those first years. The bridge years, before the pensions start, are where the portfolio carries everything and where sequence does its damage ([[horizon-and-life-expectancy]]). The horizon also sets the survival quantile you plan on, a prudent one rather than the average. A short horizon tolerates less equity risk, because a lost decade weighs heavily over a short span. A long horizon calls for more stocks **and** more sequence defense at once. The horizon is an input to the design, not a detail. ## 4. What future income? The portfolio never works alone. A future pension, Social Security, side income, rent, an annuity: these are floors with a bond-like nature. They change how much market risk the portfolio has to carry ([[pensions-and-other-income]], [[annuities-and-safety-first]]). So think in terms of a full balance sheet, the "total portfolio" idea. A large future pension you can count on is already a big "bond" sitting on your balance sheet. The liquid sleeve can then carry more stocks and defend less. A plan with no floor other than the portfolio has to defend harder. List every future income stream, its start date, how certain it is, and whether it is indexed to inflation. The bridge years, before those floors kick in, are the ones the portfolio has to cross alone. They dominate the design. ## 5. Which assets answer each risk? Now, and only now, the map. One risk, one answer, the cheapest one that works. - Sequence: a cash buffer, spending flexibility, and a little government duration ([[cash-buffer]], [[glidepaths]]). - Persistent inflation: inflation-linked bonds ([[inflation-linked-bonds]]), gold ([[gold-in-retirement]]), trend ([[managed-futures]]), a value tilt ([[factors-in-retirement]]). - Deflation or a disinflationary crash: high-quality government duration ([[bonds-in-retirement]]). - A monetary crisis: gold, and reserve-currency exposure (an unhedged global equity fund already carries some, [[international-diversification]]). - The engine (long-term growth, longevity): global stocks ([[risk-premia]]). One risk from question 1 deliberately gets no line in this map, the long bear market. That is equity risk itself, the risk you are paid to carry (question 2). So you accept it in writing rather than insure it, on pain of insuring your own engine. Two rules govern the map. First, every block earns its place by answering a single question, "in which regime does this block win while the rest loses?" ([[why-diversification-works]]). No answer, no diversification, just fees. Second, beware of disguises, because plenty of products look defensive without being defensive ([[false-defensive-assets]]). Pick the cheapest block that answers a risk you **defend**, and never stack blocks that are correlated with each other. The allocation falls out here, as the sum of the answers, weighted by how hard you defend each risk. ::: figure risques-briques The map is the article's own: the risks named in question 1 on the left, the blocks chosen in question 5 on the right. The thickness of each link is an editorial reading of how hard the risk is defended, not a measurement. ::: ## 6. How do you test it? A design is a hypothesis. You test it on the tails, not on the average. The average always looks acceptable, while the plan dies on the worst paths. In a simulator ([[using-the-fire-simulator]], [[reading-a-fan-chart]]), read three things. The infamous vintages, 1966 and 2000, to see whether the design survives the start years that broke the naive plan ([[valuations-and-cape]]). The stress models and the broad sample, not just the central case ([[simulator-traps]]). And the 5th percentile plus the safe withdrawal rate at **your** tolerance for failure, not the median ([[failure-probability]]). The verdict to read is the tradeoff. A good defensive block lifts the worst paths while costing a little on the median. If a block does not move the worst paths, it defends nothing. If it damages the median, it is too big. Test the design, then size each block by the tradeoff you are willing to accept. ## 7. When do you change it? Rarely, by rule, and never on the news. Rebalancing mechanically back to target is not changing the design, it is maintaining it ([[the-annual-review]]). The design itself changes only when a parameter of the **plan** changes: a pension confirmed, a horizon shortened, a risk newly accepted or newly defended, a life event. That decision belongs to the annual review, in writing, never in a market trough and never on a headline ([[when-to-worry]], [[bear-markets-in-retirement]]). The written thesis behind each block, the one from question 5, is exactly what lets you **not** change under fire. When a block disappoints, you reread its thesis instead of selling it the day before it earns its keep ([[the-psychology-of-spending]]). A portfolio redesigned at every crisis has no design at all. ::: exemple A design, start to finish Léa and Marc, both 50, $1.3M, a 15-year bridge before two modest but secure pensions. 1. **What are your risks?** Sequence (the 15-year bridge, dominant), inflation (their nightmare regime), a long crash. Longevity and health are handled outside the portfolio, through flexibility and a reserve. 2. **Which ones do you accept?** They accept equity risk, their engine, and they defend sequence and inflation. 3. **What is your horizon?** 45 years, so the plan needs plenty of stocks, but the first 15 years are fragile. 4. **What future income?** The future pensions, modest as they are, are a bond already on the balance sheet, so the liquid sleeve can carry more stocks. 5. **Which assets answer each risk?** The map: world stocks (the engine), a little duration and a buffer (sequence), linkers, gold and a pinch of trend (inflation). Add it up and you get, say, 65% stocks, 15% bonds and linkers, 10% gold, 5% trend, 5% buffer. The allocation was not the starting point. It fell out at the end. 6. **How do you test it?** Tested on 2000 and on the blocks of the 1970s, the inflation tilt softens the worst vintages at almost no cost on the median. 7. **When do you change it?** The map moves only when the pensions are confirmed or the horizon shortens. ::: ::: astuce Test the design, do not guess it A design is judged on the tails, never on intuition. A simulator lets you put two versions side by side, yours and its variant, and read not the central case but the worst vintages and the 5th percentile ([[reading-a-fan-chart]]). The right question is not "which one has the better average?" but "which one lifts the worst path the most, and at what cost on the median?". That is the only tradeoff that matters to a retiree. ::: ## The essentials - Design by risks, not by assets. Name the risks that can break the plan, decide which ones you accept and which ones you defend, then pick the cheapest block for each risk you defend. The allocation is the output of the method, not its starting point. - The portfolio answers market risks: sequence, inflation, deflation, a monetary crisis. The other risks, longevity, health, divorce, are handled elsewhere, through flexibility, reserves and annuities. - You do not defend everything. Equity risk is the engine, so you keep it. You defend what breaks the plan without paying you anything, sequence and inflation. Capacity and tolerance set the dose. - Your future income streams are floors already on the balance sheet. The bigger and the safer they are, the more stocks the liquid sleeve can carry. The bridge years, before those floors start, dominate the design. - Every block has to answer "in which regime does it win while the rest loses?". Test the design on the tails (the 1966 and 2000 vintages, the stress models, the 5th percentile), and change it only when a parameter of the plan changes, in writing, never on the news. --- ## Going further - Ben Mathew, "Total Portfolio Allocation and Withdrawal" (TPAW): thinking of the portfolio and future income as a single balance sheet. - Dimson, Marsh and Staunton, *Triumph of the Optimists*: a century of returns country by country, the raw material of a realistic design. - In this book: [[defensive-assets]] and [[false-defensive-assets]] (the map of blocks and the decoys), [[why-diversification-works]] (why every uncorrelated block counts), [[market-regimes]] (the regimes you defend against), [[stock-bond-allocation]] (the central dial of the plateau), [[choosing-your-strategy]] (the withdrawal rule, the other half of the plan), [[building-your-plan]] (what comes next, the full plan).