Leverage, margin and borrowing against your portfolio (advanced)

This is the most counterintuitive chapter in the protections part. It is about borrowing, in a book devoted to not running out of money. So let me state the position up front: for the vast majority of withdrawal plans, leverage has no place at all. It amplifies the exact risk, sequence risk, that everything else in the book works to reduce.

But "no leverage" deserves to be a conclusion, not a taboo. Three uses hold up in the withdrawal phase, with real and documented logic behind them. The first is borrowing for liquidity: a line secured by the portfolio acts as a bridge, so you never have to sell at the bottom, an alternative to the cash buffer. The second is capital efficiency, return stacking (Return stacking, overlays and portable alpha: stacking the premia). You borrow not to hold more stocks, but to add defensive assets without diluting the engine. That is the idea behind 90/60 funds. The third is ERN's countercyclical leverage: borrow in the holes instead of selling (Parts 49 and 52 of his series).

This article takes all three seriously. It starts with why "naive" leverage is toxic precisely in retirement. Then it walks the instruments, from the mortgage you kept to the securities-backed line, from leveraged ETFs (no) to capital-efficient funds (maybe). It ends with the safety rules, non-negotiable, for the rare plans that qualify.

The instruments, from legitimate to toxic§

Seven instruments, gentlest first. One test really decides, whether the thing can produce a margin call, so read the second column before the others.

The mortgage you keep is the gentlest leverage there is, with a rate fixed for the life of the loan and a payoff date you already know (Real estate in a FIRE plan (your home, and rentals)). Its two conditions fit on one line: a payment that sits inside a floor with room to spare, and a balance the portfolio covers several times over.

A securities-backed line and a policy loan are the same move, borrowing against what you own instead of selling it. Borrowing against a cash-value policy is the quieter of the two: the money comes from the insurer, against your own contract, with nothing surrendered and no taxable sale, as long as the policy stays in force. A line secured by the portfolio is more flexible and can be paid down whenever you like, but the lender sets the advance rate and keeps the right to change it. This is the instrument behind the legitimate uses below, and behind the abuses too, when the "temporary" line quietly turns permanent.

The box spread, wholesale funding. There is a way to borrow at the market rate from inside a brokerage account, by selling a "box" of options. Four European options with the same expiry, on two strikes, assemble so that the package is worth exactly the gap between those strikes at expiry, whatever the index does. Selling it therefore means taking in cash today against a fixed, dated debt, a synthetic zero-coupon bond cleared by the clearing house. The implied rate is read off the discount between the two amounts.

That rate is the whole argument. Measured on SPX options from 1996 to 2023, it sits on average about 35 basis points above Treasuries of the same maturity, with a spike toward 130 points in October 2008 (van Binsbergen, Diamond and Grotteria, 2022; New York Fed, 2023). It is a wholesale rate, far below what a retail broker charges on margin. But financing is judged against your alternative, not against zero. In the summer of 2026, a four-month SPX box came out around 4.5% a year, while one large broker's margin schedule ran from 4.4% to over 6% depending on balance and account type. Against the best tier of a large account, the advantage disappears entirely.

Then come the frictions, and they are serious. The structure only makes sense on European, cash-settled options such as SPX or the Euro Stoxx 50: on American options, the buyer exercises the in-the-money legs early and the box comes apart into a naked directional position. That is exactly what cost a Robinhood user about $60,000 in 2019, on a $5,000 account, with a box built on an ETF with American options. You then need permission to sell spreads at your broker, and an execution of all four legs as a single order, never leg by leg. The round trip costs little on a liquid book and a great deal more if you need to unwind in a hurry. Above all, this is a hard debt with a fixed term, repaid at expiry or bought back at the prevailing rate, where a securities-backed line is repaid at par on any day you choose. The collateral, meanwhile, is marked to market every evening, and the real risk is that mismatch between a frozen debt and assets that move. Cboe itself presents it as a tool for short durations and clearly defined needs, not as long-term financing.

None of this shifts the rule of the chapter. It is an instrument for a knowledgeable trader, it lives on the wrong side of a margin-callable account, and the core of a withdrawal plan has no business there. The tax treatment of such a structure is a subject of its own, one for the tax part of this book and for current professional advice.

Daily-leveraged ETFs (2x, 3x) are the textbook case of volatility drag, where a 7% arithmetic average leaves 2.5% compounded (Arithmetic mean, geometric mean, and volatility drag). Trading instruments, not building blocks for a plan.

And it is not leverage alone that costs, it is its rhythm. The same 2x exposure reset every day, every month or once a year does not return the same ending capital, and the gap is not a rounding error. Over the choppy decade of 2000-2009, a daily 2x on the S&P 500 ends at ×0.38, a monthly one at ×0.48 and a one-year term loan at ×0.31, while the bare index ends at ×0.91. Both ends lose, and not for the same reason. The daily reset pays beta slippage at full strength, the realized variance, every year (Arithmetic mean, geometric mean, and volatility drag). The term loan pays none of it, but it lets the leverage drift upward as the market falls. At the November 2008 trough, the equity of a 2x left alone since January had fallen to 3% of the stake for an effective leverage of 33, meaning a certain margin call long before the decade was out. Over the trending decade of 2010-2019, the three land within a few points of each other. So the reset frequency is not an execution detail. It is the dial that trades slippage against drift, and what it costs depends entirely on the regime you go through.

What a daily leveraged fund does inside a portfolio. Three things, two of them against the holder. Financing first, since the borrowed leg pays the short rate, invisible when that rate was zero and worth several points a year as soon as it comes back. It is the arithmetic of the cash posted as margin behind a futures programme, running the other way (Managed futures and trend following: the diversification that works in a crisis). Risk next. In a crisis, a 2x equity fund carries a correlation near 1 with the rest of the equity sleeve and twice its beta, so its share of variance runs far above its share of capital. In the simulator, a portfolio's risk-budget view puts those two side by side, line by line, and it is the only number that says whether a small line is really small. Rebalancing last, and that one works for you. Held at a fixed weight and rebalanced, such a fund is sold after gains and bought back after losses, which recovers at portfolio level a share of what the reset loses at product level. Hence the only arguable form, a small dose, rebalanced. Left to drift, the line concentrates and then behaves exactly as its standalone curve warned it would.

"Capital efficient" funds (90/60 and cousins). Here is the interesting category. Such a fund holds 90% in stocks and 60% in bonds through futures, 150% of exposure in all. Leverage of 1.5 is moderate and permanent, and it lives inside the fund rather than in your account. The idea is called return stacking (Return stacking, overlays and portable alpha: stacking the premia). Leverage buys you not more of the same risk but room to stack diversifiers without shrinking the engine (Defensive assets: the map and the roles). An example: on $100, $67 in a 90/60 plus $33 in gold or trend gives you roughly the exposure of a plain 60/40 with a real regime sleeve on top. Diversification per dollar invested goes up. A handful of these funds are listed in the US, and the shelf keeps growing (Building it with US-listed ETFs: the retiree's shopping list). Three things still need watching. The implicit financing cost first: futures pay the short rate, and when that rate is high the leverage gets expensive. The embedded duration next: the bond leg took 2022 on the chin like everyone else (Bonds in retirement: types, duration, the exact job). And traceability, with the same checks you would run on any fund.

That leaves the one flat prohibition of this chapter, raw brokerage margin and options, leverage with a margin call attached to the heart of the plan. Reg T lets you borrow half a position's value to open it, but it is the maintenance level that sells you out, and the broker picks the moment. A withdrawal plan holds nothing that can be liquidated for you on a panicky Tuesday.

The three defensible uses, with their numbers§

Use 1: the liquidity bridge (the credit line as a buffer). The need is simple: fund 6 to 24 months of spending during a downturn without selling beaten-down assets. That is exactly what the cash buffer does (The cash buffer: size, cost, and what it is really for). The borrowed version does away with the permanent buffer: the capital stays invested and the opportunity cost disappears. Once the drawdown threshold is breached, the same one that governs refills (Drawing on a buffer and refilling it: the rules that work), you draw on the line to live. You repay when calm returns, with sales you can make with a clear head. The arithmetic makes the case. Borrowing 18 months of spending, about 6% of the portfolio, at 4% for two years costs half a point of portfolio in total. The equivalent buffer costs 0.3 to 0.4% a year, every year. On cost alone the bridge wins, then, as long as the downturns do not come back to back. What it loses is the psychological safety of cash (The cash buffer: size, cost, and what it is really for): the real value of a buffer is that it stops panic, and a debt growing through a crash does the opposite. Verdict: technically better, behaviorally demanding. It fits quantitative types with a covered floor, and a policy loan is the door in.

Use 2: capital efficiency (a dose of 90/60). This need comes from somewhere else: regime protections, gold, trend and linkers, take up room. A regime sleeve of 30 to 40% shrinks the engine by the same amount (All-weather portfolios: Browne, All Weather, Golden Butterfly, Dragon). The stacked version parks the stock-and-bond core in a capital-efficient block and spends the freed-up room on diversifiers. At equal engine exposure, the portfolio gains its defenses. Leverage funds diversification here, the one purchase that improves the mix without inflating variance (Arithmetic mean, geometric mean, and volatility drag). This is the intellectually cleanest of the three. Sized properly, with the plan's overall leverage capped around 1.35 and no margin call anywhere, it is closer to allocation engineering than to speculation. The thing to watch: do not stack the illusion, because adding more stocks instead of diversifiers takes you straight back to naive leverage.

Use 3: ERN's countercyclical leverage (Parts 49 and 52). The idea: rather than sell during drawdowns, borrow against the portfolio for a while and repay on the way back up. In the Part 52 version, the line only opens in deep holes: past a 20 to 30% drawdown, withdrawals are funded by borrowing instead of selling, and the balance is repaid as the market climbs back toward its highs. ERN's simulations credit the disciplined version with a real but modest gain on the worst vintages, on the order of 0.3 points of withdrawal rate. What they mostly show is the obvious: everything rides on mechanical execution. This is rebalancing with the volume turned up, run through an instrument someone else can switch off. Advance rates shrink in a crash, and the lender cuts the line at the exact moment you need it. Verdict: for oversized plans hunting the last bit of optimization, never for tight plans hunting salvation.

Modeling leverage, and an example§

The modeling. Putting numbers on a levered plan takes two separate tools, and the first mistake is asking both the same question. Start one floor down, at the portfolio. A portfolio visualizer replays your holdings one by one over history, and the full-featured ones let the weights add up to more than 100%, with a borrowing cost charged on the financed part. That, and only that, is where you get an honest comparison between a plain core and a stacked core, financing included. Then move up to the plan. A decumulation engine no longer sees holdings; it runs withdrawals against the aggregate portfolio (Under the hood: how the simulator computes this book describes one in detail). Leverage enters that engine upstream, through the returns of the portfolio once stacked, either the history of the capital-efficient fund itself or a mean and a volatility you set by hand. The credit-line bridge is approximated there by a cash buffer with the same flows, but with its return marked down to the cost of the line. Both floors, in that order, and never the second without the first.

The essentials§

• Leverage changes sign in retirement. It amplifies volatility drag and sequence risk, and it brings forced liquidation into the picture. "No leverage" is the right default answer, as a conclusion and not as a taboo.
• Three defensible uses, and only three. The bridge: a secured line or a policy loan, drawn on a written drawdown trigger, a buffer without the opportunity cost, for quantitative types. Capital efficiency: a 90/60 that funds diversifiers, overall leverage capped around 1.35, no margin call anywhere. ERN's countercyclical version: borrow in the hole instead of selling, to fine-tune plans that are already safe.
• The bans: daily-leveraged ETFs, where the drag grinds you down, brokerage margin on the core with its forced liquidation, and any leverage in the service of a tight plan. Leverage helps the rich and finishes off the poor.
• Leverage has a size and a rhythm. The daily reset pays beta slippage at full strength; the term loan pays none of it but lets the leverage drift upward through a fall. Over 2000-2009, the three versions of one 2x end between ×0.31 and ×0.48, all far below the bare index at ×0.91.
• The box spread on European options borrows at the wholesale rate, on the order of 0.3 to 0.6 point above Treasuries at the same maturity. It is a hard debt with a fixed term, backed by collateral marked every evening, and it stays an instrument for a knowledgeable trader.
• Five rules if you use it: loan-to-value under 20 to 25% at the largest planned draw, no forced liquidation possible, a known spread retested as rates move, a dated exit, the spouse test.
• Leverage gets measured on two floors. At the portfolio, with weights above 100% and an explicit borrowing cost, to compare a stacked core with a plain one. At the plan, on the aggregate portfolio, where the bridge shows up as a buffer with a negative return. Measure first, as always. That is what separates an instrument from a bet.

Going further§

• Early Retirement Now, Parts 49 and 52 ("Leverage in Retirement", "Timing Leverage"): the full countercyclical case (ERN's Safe Withdrawal Rate series: a reader's guide).
• Ayres and Nalebuff, Lifecycle Investing (2010): the theory of leverage by age. And why it reverses in the withdrawal phase.
• returnstacked.com and the capital efficiency literature (Newfound, ReSolve): stacking explained by the people who sell it, worth reading with the usual skepticism.
• The Cboe and OCC notes on box spreads (mechanics, margin, settlement) and van Binsbergen, Diamond and Grotteria (2022), "Risk-Free Interest Rates": the implied rate, measured across twenty years of SPX options.
• In this book: Real estate in a FIRE plan (your home, and rentals) (the mortgage you keep), The cash buffer: size, cost, and what it is really for and Drawing on a buffer and refilling it: the rules that work (what the bridge replaces), Arithmetic mean, geometric mean, and volatility drag (the drag that governs everything).