Bonds in retirement: types, duration, the exact job

Bonds are the most misunderstood asset a retiree owns. People think they are safe, and 2022 destroyed that belief: broad investment grade indexes lost double digits, long government bonds 30% and more, worse than stocks. People think they are complicated, when the mechanics come down to two ideas. And people ask them to do a job, to pay, that is not their job in a withdrawal plan. This chapter puts all of it back in order. First the mechanics, the minimum that is enough: price and rates, duration as the unit of both risk and horizon, yield to maturity as the only honest expectation. Then the three services a retiree wants from the bond sleeve, and the regimes in which each one works or breaks. Then the decisive history of the stock-bond correlation, the most important and least stable parameter in the whole allocation. Then the four concrete decisions: quality, duration, the linked share, currency. And finally the stable value fund, a bond substitute with unusual properties in both directions, and the full lesson of 2022.

By the end, your bond sleeve will have a written spec instead of being "the rest" of the portfolio.

The mechanics, no more than needed§

Price and rates. A bond is a stream of fixed cash flows. When market rates rise, those old fixed flows are worth less, so the price falls. It falls by exactly what it takes to bring the buyer's yield back to today's rate. Duration measures that sensitivity. It runs around 2 for a short-term fund, around 7 for the standard aggregate, 15 to 20 for long government funds. It is also, and this is no accident, the horizon over which your return is locked in. Hold a bond fund for roughly its duration in years and you earn about the YTM you bought at. Rate rises along the way hurt the price, but reinvesting the coupons at the new rates pays them back. Hence one basic consistency rule: the sleeve's duration must stay at or below the horizon it serves. Money earmarked for the next three years of withdrawals and parked in duration 15 is a bet on rates, not a reserve.

Funds against individual bonds, a false debate. "I hold my bond to maturity, so I never lose" is the most stubborn illusion in the asset class. Holding to maturity locks in the nominal yield you were promised. If rates have risen, you take the same loss the fund took, as an opportunity cost. You collect 2% while the market pays 4, instead of a visible drop in price. Final wealth is identical; only the mental accounting changes. The constant-duration fund and the bond held to maturity are two legitimate tools for two jobs. The fund serves permanent sleeves, the shock absorber you rebalance. The individual bond, or a dated-maturity fund, serves dated liabilities. That is the ladder principle, covered in Bond ladders (and the ladder of linkers).

A seven-year bond position, a two-point rate shock in the twelfth month, and the same position under two accountings: the fund marks to market and prints an 11% fall at once, the bond held to maturity stays at amortized cost and prints nothing, but pays 2% while the market pays 4. The band between the two curves is the same loss under its two names. At maturity the two wealths are the same number, to the cent.

The three services, and what each one needs to work§

The retiree's bond sleeve delivers three services (The stock and bond allocation in retirement), and each has its regime conditions (Market regimes (growth × inflation, sticky bears) and why they matter).

Service 1: cushioning stock crashes. It works when the crash is a growth shock, a recession or a panic. Central banks cut rates and bonds rise. That is what happened in 2008 (+8% for the aggregate, +25% for the long end, while stocks fell 40%), and again in 2000-2002 and in 2020. It breaks when the shock is inflation. Rates rise and bonds fall alongside stocks, as in 1973-74 and in 2022. The service is conditional on the regime, and its power grows with duration. Hence a dose of long bonds held as explicit deflation insurance (Defensive assets: the map and the roles).

Service 2: funding withdrawals through the holes. This is the logistics service. While stocks recover, which takes anywhere from one to seven years (Market regimes (growth × inflation, sticky bears) and why they matter), the withdrawals come out of the bond sleeve, which is flat or up. This is funding the withdrawal by selling the overweight (The fixed inflation-adjusted withdrawal (Bengen): the benchmark rule), and the real mechanism behind the bucket intuition (The bucket strategy: the promise, and the case against it). It demands quality (see below) and a short to intermediate duration, because the sleeve has to be stable at the moment you spend it.

Service 3: capping total volatility. This is the statistical service. It brings the portfolio's drawdown back inside what your rule and your nerves can hold (The stock and bond allocation in retirement). It works by plain dilution, even when the correlation is unhelpful, because a sleeve at 4% volatility always dilutes one at 16%. But it works twice as well when the correlation is negative. Which brings us to the parameter at the center of everything.

The stock-bond correlation, a story in three acts. Before 2000 it was positive on average. The world lived under the threat of inflation, and bad inflation news hit both assets at once. From 2000 to 2021 it turned negative. The world was disinflationary, the bad news was growth shocks, and growth shocks are good for bonds. That was the golden age of the 60/40, when the bond was perfect insurance and got paid for it. Since 2022 it has turned positive in episodes, following the inflation regime. Research has identified what drives it: the inflation regime itself. When inflation is low and stable, the correlation is negative; when it is high or uncertain, it is positive. The practical consequence is large. The insurance quality of your bond sleeve is not a constant you can copy out of the 2000-2020 backtests. It depends on the world you land in. Hence the backups in the table of defenses, gold, linkers (inflation-linked bonds) and trend, for the worlds where the bond sleeve is not there (Defensive assets: the map and the roles).

The rolling correlation between stocks and bonds is not a constant. It changes sign with the inflation regime. Positive through the inflationary years of the 1970s and 1980s, it turns clearly negative from 2000 to 2021, the golden age of the 60/40, when bonds cushioned the crashes. Then it goes positive again in 2022, when stocks and bonds fall together. So the insurance quality of the bond sleeve is never a given. It depends on the world.

⚠ 2022, the year that taught the lesson

Read 2022 again with the grid in hand. Inflation at 8 to 10%, policy rates off the floor and up several points in eighteen months. Duration 7 lost in the teens, duration 15 and up lost 30 to 40%. Stocks fell 15 to 20% at the same time, the signature of a positive correlation. Long linkers fell too, because real rates were rising (Inflation-linked bonds: the only contract written in real terms). The only winners were money market funds (a rising rate is their friend), gold (flat, then at record highs) and managed futures (+20 to 30%, having sold the bond downtrend, Managed futures and trend following: the diversification that works in a crisis). A "prudent" 60/40 retiree lived through the worst real year of the half century. Not through bad luck, but because the entire defensive sleeve was betting on a single regime. The doctrine of this whole part of the book, diversify your defenses by regime, is the lesson of 2022. Putting it in place cost about a point of expected return in 2021; skipping it cost fifteen points of portfolio.

The four decisions§

1. Quality: government and investment grade, full stop. High yield (junk) carries default risk that shows up in recessions, with a correlation to stocks around 0.7 in a crisis. It delivers none of the three services. It is an equity position in disguise, and one taxed as ordinary income at that (Defensive assets: the map and the roles). Investment grade credit (well-rated corporate borrowers) is acceptable in moderate size, since it adds 0.5 to 1 point of yield for a limited increase in crisis correlation. But the core stays Treasuries and quasi-sovereigns. No indulgence for complexity. Emerging market debt, subordinated bank debt and CoCos stay outside a retiree's spec.

2. Duration: by service, not by a view on rates. Each service sets its own horizon, so its own duration, and the grid at the end of the chapter fixes them block by block. Most of the work happens in intermediate, the compromise between cushioning power and rate risk, with the broad aggregate or a 5 to 10 year government fund. The long dose stays optional and deliberate (All-weather portfolios: Browne, All Weather, Golden Butterfly, Dragon). The same rule decides the cautious sleeve of a glide path, which exists to be spent over the first ten years of retirement (Glide paths: the bond tent, rising equity, and the fragile window). What you do not do is steer duration off rate forecasts. The rate market is the most efficient in the world, and your view has no edge in it (Simulator traps: independence, US bias, survivorship shows that timing destroys value everywhere).

3. The linked share: 25 to 50% of the sleeve. Linkers cover the service the nominals never deliver, inflation by surprise. The chapter devoted to them takes apart the vehicle and the traps (Inflation-linked bonds: the only contract written in real terms). The split rule is simple. The more of your floor the portfolio funds, and so the longer the uncovered phase, the higher the linked share goes.

4. Currency: home, or hedged. The bond service is a stability service. An unhedged foreign bond stacks the currency's volatility (8 to 10%) on top of its own. It no longer delivers service 2 or service 3. Simple rule: hold the bond sleeve in your own currency, or hedged into it. In a hedged share class, the rate differential you pay for the hedge is offset by the higher yield on the foreign debt, and only 0.2 to 0.3% a year of friction is left. Take the portfolio's currency exposure on the equity side, where it cushions (International diversification (and home bias)).

The stable value fund, in its right place§

One vehicle deserves a paragraph of its own, because nothing else on the menu behaves like it: the stable value fund offered inside many workplace plans. It is a managed bond portfolio wrapped by an insurer, who guarantees the principal and smooths the yield out of its own reserves. Its virtues for a retiree are real. Its reported volatility is zero, which makes it the natural home of the cash buffer (The cash buffer: size, cost, and what it is really for) and of the short sleeve: service 2 with no price risk at all, in a tax-deferred account. Its limits are the mirror image. The credited rate follows market rates with years of delay, the time it takes the underlying bond book to turn over, so it lagged by a point or two through the whole climb of 2022 to 2024. It has no crisis convexity, since it cannot jump 20% in a deflation: no service 1 in the insured version. And the guarantee rests on the insurer behind the wrapper and on contract terms that can restrict or delay transfers in extreme conditions, which argues against overweighting it. Its right place is clear. Stable value beats money market funds and short duration for the same job, up to roughly half the bond sleeve. It replaces neither the long duration that insures against deflation nor the linkers. Where no such fund is on offer, money market funds and T-bills fill the same slot.

The spec, in one grid§

Here is the opening promise, written down. One line per building block, with the service it delivers, the horizon over which it delivers it, the regime that puts it out of action, and the share it takes. Shares are counted over the bond sleeve alone; gold and the other defenses belong to the table of roles (Defensive assets: the map and the roles).

Block	Service	Duration	What breaks it	Share
Stable value, money market	2, withdrawals in the holes	1 to 3	Falling behind rising rates	Up to 50%
Intermediate Treasuries	3, and 1 in disinflation	5 to 8	An inflation shock	The remainder
Long Treasuries	1, deflation insurance	15 to 20	An inflation shock, amplified	10 to 20%
Short TIPS	Inflation by surprise	3 to 5	Rising real rates	25 to 50%
Investment grade credit	3, plus some yield	The core's	A severe recession	Moderate size
High yield, emerging, CoCos	None	Not applicable	Any recession	0%

The duration column is what makes the grid executable. It ties each block to the horizon of its use, and that consistency, not a view on rates, is what decides what you buy.

🧮 A bond sleeve, written down

The plan: $1.6M, 65% stocks, 35% defensive, a 14-year uncovered phase. The defensive sleeve fills the grid service by service, in points of portfolio. Buffer and near-term withdrawals, 8% in stable value and T-bills (service 2, no volatility). Cushioning core, 12% in government bonds of 5 to 8 years (service 3, and service 1 in a disinflationary regime). Deflation insurance, 4% in government bonds of 15 years and up (service 1 in its powerful form, a deliberate dose). Inflation line, 8% in short linkers plus 3% in gold (Inflation-linked bonds: the only contract written in real terms, Gold in a retirement portfolio). Checked in simulation against the same allocation held as "35% aggregate", the central case comes out the same, failure is 1.2 points lower under sequence stress and 1.8 points lower under the broad sample, and the 1973 vintage gets through. The sleeve has a spec, and every holding a job. That is all "bonds" ever meant.

The essentials§

Two ideas are enough: price ≈ −duration × the change in rates, and today's YTM = the expected return out to the duration horizon. Bonds are the only asset with their expectation printed on the label. "Holding to maturity" protects nothing, it just re-books the same loss.
Three services (cushion the crashes, fund the withdrawals in the holes, cap the volatility), all conditional on the regime. The stock-bond correlation follows the inflation regime (negative from 2000 to 2021, positive before it and since 2022). Do not copy the backtests of the golden age.
Four decisions, one grid line each: quality (government or investment grade, never high yield), duration by service, the linked share, and currency. That last one is not negotiable. The bond sleeve is held in your own currency or hedged into it, and the currency exposure is taken on the equity side.
Stable value beats money market and short duration for the same job (guarantee, smoothing, tax deferral), but it replaces neither crisis convexity nor the linkers. Up to half the sleeve, no further.
2022 is the lesson to frame. A one-regime defensive sleeve costs fifteen points the year its regime breaks. Defense gets diversified (Defensive assets: the map and the roles), and a simulator puts it to the test on the inflationary vintages.

Going further§

Ilmanen, Expected Returns, the chapters on rates and the term premium: the reference on what bonds really pay.
The research on the stock-bond correlation and the inflation regime (AQR, "A Changing Stock-Bond Correlation", and the BIS work): the central parameter, documented.
Early Retirement Now, Part 21 (bonds and the mortgage) and the allocation analyses (ERN's Safe Withdrawal Rate series: a reader's guide).
In this book: Inflation-linked bonds: the only contract written in real terms and Bond ladders (and the ladder of linkers) (the two direct companions), Defensive assets: the map and the roles (the table of roles), The stock and bond allocation in retirement (the ratio).