Bonds: that which cannot be made simpler

Bonds have a dual nature. They are in part the stately road to riches. Let interest compound without interruption and the exponential function does the rest. Their conservative credentials are well known.

And they are part wild beast. A British government bond issued in May 2020 has since lost more than three quarters of its price, without missing a payment.

Take the base case. You buy a government bond of whatever maturity and hold it until redemption. You know precisely what you get, to the last penny. If the rate is attractive, you can park cash and forget. You know the cash flow coming in over time, but you don’t know what it will buy. Since the income is fixed at purchase, there’s no further adjustment for inflation or other risks.

Should you have a change of heart and want out early, you can. The price you get is not fixed. Even if you do not sell, the risks remain. You hold a fixed income contract while the market offers better rates. Your investment ‘sensitivity’ is to interest rates. If they rise, the price of your bond falls. And if they fall, it rises. That much is certain.

Above all, bonds get the job done. Which is why looking at them properly, from first principles, sounds dull. Bonds are all maths, goes the saying. But who wants to spend their investment days hunched over a calculator when there are enticing equity stories to enjoy?

But this is not the whole story. Beyond the base case is where things get interesting.

Three jobs

A government bond is reliable on three counts.

• It pays income.
• It returns capital on a known date.
• It rises when growth disappoints, offsetting equity downturns.

Though a portfolio manager would tick each of these as desirable, the three primary tasks of a bond holding do not always align. A long gilt bought as a diversifier is a poor instrument if the return of capital needs to be sooner than the bond’s redemption date. Then you are dealing with a secondary market with market-set prices, which are not necessarily what the contract says it is worth.

The third job, as a hedge for growth risk, comes with a condition. Recessions come in different flavours, inflationary and deflationary. Bonds do not hedge inflation risk; they expose you to it. If general prices keep rising while growth falls, bond prices typically fall along with equities. No more hedge.

The bonds-as-diversifier argument is a description of an environment, not a mechanism. In 2022 both fell together. A cautious allocation in a balanced portfolio surprised many by becoming a speculation. Bond investors like predictability, not surprises.

Corporate bonds and other debt increase the complexity. But the base case of a long government bond is the foundation.

The first question about any bond investment is which of the three jobs it is for.

The price of the promise

Once the bond is purchased, you cannot change the terms of the promise. The only thing that can move is the price of the promise.

And it moves with mathematical precision. The direction of movement is determined by market yields. The mechanism setting yields is price discovery, not government instruction.

Price moves inversely to yields. A bond bought when yields were 1%, held while yields go to 3%:

Maturity	Price at 1% yield	Price at 3% yield	Capital loss
2 years	100.00	96.15	−3.9%
5 years	100.00	90.78	−9.2%
10 years	100.00	82.83	−17.2%
20 years	100.00	70.08	−29.9%
30 years	100.00	60.62	−39.4%
50 years	100.00	48.38	−51.6%

1% coupon, bought at par — at 100, which is what the bond repays at the end. Coupons paid twice a year. The same two points of yield across the maturities on offer.

These are the same bond contract, and each kept its promise. The only difference between them is how long you agreed to wait before getting repaid. That length of time determines how sensitive you are to changes in yields. The market has a name for that sensitivity: duration.

2022 is the example of a wild year for bonds. Gilts lost 23.8%, their worst year in records going back to 1899. The thirty-year gilt began the year yielding 1.12% and ended it at 3.95%. On the arithmetic above, a seller at 3.95% lost about half their capital. A seller at the September low in prices, when the yield closed at 4.99% on the 27th, lost 60%. Nobody defaulted; every payment was made on time. The capital loss could be made up by waiting till maturity, some thirty years hence. The investor who needed the money, or got jittery, in 2023 paid for the maturity mismatch.

The starting level matters as much as the move. The same calculation above, starting at 5% and moving to 7%, costs a thirty-year holder 25% rather than 39%. The coupon is the cushion. Low starting yields offer nothing to ameliorate the capital loss. The negative-rate years stand out as truly wild.

A wild beast

Held from the negative-rate years to today. No defaults. No missed payments.

Bond	Price at issue	Price now	Capital loss
UK 0½% gilt 2061	96.87	23.02	−76.2%
German 0% Bund 2050	103.47	41.14	−60.2%
Austria 0.85% 2120	98.01	30.30	−69.1%

Prices at issue against the market now. The gilt was sold in May 2020 at a yield of 0.59% and yields 5.45% today; the German zero-coupon was sold at −0.11% and yields 3.77%; the Austrian century bond was sold at 0.88% and yields 3.21%. The gilt’s 5.45% sits below the 5.75% on the benchmark thirty-year gilt, which is not an error: capital gains on gilts are free of UK tax, so a bond trading at 23 delivers most of its return as an untaxed gain and gets bid up by holders who pay tax on income.

Every one of these will pay exactly what it promised, on the day it promised. The British government has not wavered. The price paid was wrong. Which was reasonably knowable looking at the coupon alone.

This runs both ways. A thirty-year bought at par when yields are 3% is worth 151.73 when they reach 1% — a 52% return from a government bond. Forecast the direction of rates correctly and bond investing suddenly looks more interesting.

Better still, the two directions are not proportional. For a two-point move either way in yields, that thirty-year loses 30.9% going up and gains 51.7% coming down. The instrument pays more for being right than it charges for being wrong.

Forecasting deflation and inflation is not a simple task. But for any coming deflation, the longest thing you can find pays best. Then the dullest asset in the market goes stellar. For an inflationary period, short maturity or cash is where to wait for better times ahead.

The buyers of bonds at sub-zero were not unlucky. They were late to a long trend that had worked. But eventually trends result in excessive pricing. A low yield on a long date is not a cautious position. It is the most leveraged bet on interest rates available to a private investor without borrowing.

But long rates are above 5% now, half a point higher than a year ago. Real yields are back again.

A bond is not a product. It is a set of terms, and you are agreeing to them.

Nobody reads a product. Everybody reads a contract, or knows they should.

The terms are four: what it pays, when it repays, who is promising, and where you stand if they cannot.

Is there a date?

A bond is a dated promise of cash. Coupon, par, yield and the rest are the bookkeeping entries.

This is the fundamental difference between owning a bond directly and owning a bond fund. A gilt has a date when it returns your cash, as does a corporate bond. A bond fund does not.

A fund holds bonds that mature and buys more with the proceeds. There is no day on which it pays par. A single bond’s sensitivity to yields falls every year towards nothing at redemption; a fund’s does not, because it keeps replacing what ages. That is why a fund can still be down years after the bonds inside it would have matured and paid par. A bond’s return is fixed on the day it is bought, a fund’s on the day it is sold. Liquidity is what the fund gives instead of a date.

Perpetuals do not have that date. Nor do most structured notes.

This is not an argument against any of them. It is an argument against reviewing one as if it were a bond. The date is the feature that makes the asset checkable in advance. Give it up if you have a reason to — but know you have given it up.

The six lines that matter

Every bond arrives with a page of terms. Six lines on it decide almost everything.

The line	What it actually tells you
Issuer	Who owes you.
Coupon and frequency	Your income, and when it is paid. Fixed at purchase.
Maturity	The date, determining your interest-rate risk.
Ranking	Secured, senior, subordinated. Where you stand in event of failure.
Call provisions	An option granted to the issuer to hand your money back early. Exercised when it suits them, not the holder.
Denomination	The minimum piece, often £100,000, which decides whether the offer was ever for you.

The coupon line is the first of the four terms, and the one most often misread: a fixed number of pounds, not a share of anything, and not indexed unless the bond says so.

Who is promising

Governments borrow because spending exceeds taxes. It is calendar-driven and largely indifferent to price. Companies borrow because debt is cheaper than equity and the window is open. It is opportunistic and sensitive to price.

Insider shareholder activity is informative as a real indication of management beliefs. So too with treasury activity. When companies substantially increase their issuance of debt, or the overall maturity, they are making a forecast. A company increasing its issuance of long-term debt considers the terms it is getting favourable. Refinance when the market lets you, goes the saying, not when you want to.

Where you rank, and what you are paying

Two bonds from the same company at the same maturity can pay very different yields. The difference is its position in the capital structure. A subordinated bond paying two points more than the senior is not a better deal — it is a different deal.

The bond market is not an exchange. There is no central order book, and often no commission line on the contract note either: the dealer takes the other side and the margin sits inside the price quoted. Some brokers charge a flat fee instead and quote a tighter price. Some do both. The cost is real on every route, it differs by broker and by size, and the only way to know what it is on your trade is to ask before you place it.

Two questions bring it into view: the gap between the price you are being quoted to buy and the price the same dealer would pay to take it back, and what that second number would be a month from now. The difference is the cost of changing one’s mind, and it widens exactly when the holder most wants out.

Then set it against the coupon. A spread that sounds small as a price sounds different expressed as months of income, and that is the conversion worth doing before agreeing to anything.

A dealer who cannot answer either question in a sentence has told you something.

What has to be true

Bond documentation is written to be filed, not read. What really matters comes down to one question: what has to happen for the bond investment to pay off? Every fixed income proposition reduces to a simple primary risk. These aggregate from interest rate risk in government bonds to rate risk and credit risk in corporate bonds.

The condition, bond by bond.

A long gilt. That rates do not go materially higher than where it was bought.

A high-yield bond. That the company can refinance when it matures.

A bond fund. That yields are no higher on the day it is sold, because no date arrives to make the holder whole.

An index-linked gilt. That inflation runs higher than the market has already priced.

A familiar company’s ten-year bond. That the name is still investment grade when it matures.

Four of those five come down to one risk: what rates do. The high-yield bond is the exception, and it is the doorway to a second risk sitting on top of everything in this guide — whether the borrower pays at all, and what the market charges for the doubt. That is credit, and it has a guide of its own coming. Until then what transfers is the discipline above: it is a contract, it has a date, and something has to be true for it to pay.

Nobody is offering you a rate. The offer is the market’s forecast, priced. Errors in that forecast are where the opportunity lies, and anything beyond taking the income is a position on it.

Take the extreme case. The German thirty-year in Part One was sold at a yield of −0.11%. That was the market’s forecast of deflation, priced and dated, and the only question for a buyer was whether to agree with it. It has since lost 60% of its price.

Today’s yields carry a forecast in the other direction. Part Four puts a number on it.

The curve

Plot the yield on government bonds against how long each has to run and you get a line. That line is the yield curve.

22 Sep 2026 22 Sep 2025 22 Sep 2021

Each shape has a headline meaning and at least one other explanation that fits equally well.

Shape  ·  the usual reading	Equally consistent with
Upward sloping Growth ahead	Investors demanding more to hold duration. No growth content at all.
Flat Late cycle	No view — or two large offsetting views.
Inverted — short yields above long Recession coming	Policy is tight now and expected to loosen later. A statement about policy, not GDP.
Steep at the long end Growth returning	Supply, term premium — the extra yield demanded simply for lending longer — or doubt about fiscal credibility.

Four shapes, and the right-hand column is the one nobody quotes.

The discipline that makes it usable is one question: which end moved?

A curve that steepens because the front end fell means something entirely different from one that steepens because the long end rose. The first is a central bank. The second is the bond market forming an opinion about the government.

The inversion signal is a correlation with a story attached, not a mechanism. The 2022–24 inversion did not arrive on the usual schedule.

What you are actually paid for

The curve already contains the market’s expected path for rates. Every yield on it is a consensus forecast with a price attached.

Which means this: two things pay. The difference between what happens and what was already priced, and the speed at which the market closes the gap.

“Rates will fall” is not an investment view: the curve may already assume they fall further than you do. “Rates will fall further than the curve implies” is a view.

Time pays a second way. On an upward-sloping curve a ten-year becomes a nine-year, and nine-year paper yields less, so the price rises for no reason other than the calendar. That is real money on top of the coupon, for a holder who sells before maturity.

It is also checkable, and worth checking before anyone sells it to you as a reason. On the curve of 22 September the ten-year yields 4.96% and a nine-year, interpolated, 4.94%. Hold for a year and sell and the price has risen to 100.17: a total return of 5.13% against a 4.96% coupon. Seventeen basis points. On a curve this flat that is what the calendar is worth; on a steep curve it is a good deal more.

Four conditions, and only one is easy

Level and direction give four conditions, and duration is your friend or your enemy in each.

	Rates falling	Rates rising
Rates high	Long duration. A high coupon and a capital gain on top.	Short, or cash.
Rates low	Long but with caveat. This is a trade. Capital gains are real but a low coupon is a handicap.	Short, or out. 2021–22. No income, all of the loss.

The boxes pay only where the market has not already priced the move. And this is prior to any consideration of credit risk.

The number that is not the number

A yield of 5% is not a return. It is a return before inflation, and the promise is fixed in pounds.

There is a way to read the market’s inflation forecast directly. Governments issue two kinds of bond: ordinary ones and index-linked ones. The gap between the two yields is the rate of inflation the market is assuming for the life of the bond. It has a name: the breakeven.

The shape that keeps repeating

What goes wrong is rarely the forecast. It is what people do once they believe it. Bond disasters do not look alike on the surface. Underneath they are the same three things: certainty, leverage, and a funding requirement.

In September 2022 British pension funds using leveraged hedges became forced sellers into a falling market. A hedge that requires collateral is a leveraged speculation in waiting. That is the 60% in Part One: not the loss of a holder, but the loss of a seller who had no choice about the date.

In 1994 the Federal Reserve took its policy rate from 3% to 6% in thirteen months. Orange County was among the casualties, not because it was wrong about direction but because it had borrowed to express certainty and had to fund the position daily.

The hallmarks of hubris. Anyone who is certain, has borrowed, and has to fund the position daily is one adverse move from being a seller at the worst moment — and their selling is what makes it the worst moment.

The private investor’s version is milder and more common: certainty about rates, expressed at long maturity, with money that turns out to be needed.

The rate forecast is legible. Read it before you agree with it.

At the close of 22 September 2026. The last two numbers are the whole of Part Four. Parts One to Three do not date; this one does, and is reviewed quarterly.

Management

Every security has management. For a company you ask whether they do what they say and whether the plan is funded. The long bond is the rare security whose management announces in public the price it would like the market to pay.

“We are going to make a market in these,” the Secretary said on 19 August, adding that yields do not reflect the underlying fundamentals. He would like the market to take another look and reconsider. This is bravado talk, but the intervention in the market is real. And every intervention gets tested.

That intervention was rejected inside a day, and the long bond has not obliged since. It was 5.21% on 28 August, 5.27% on 1 September and 5.29% on 22 September.

Whether that makes the bond cheap is an argument, and arguments belong to the other strand — Making a market in bonds takes it up. The use of it here is narrower. Management has stated in public the price it would like, and the market is ruling on it daily, in figures anyone can read. No equity holder is ever given that.

On 9 September the Treasury made its buyback increase permanent for the rest of the quarter, and changed the wording. Buybacks at the longer end are now “at least” four billion dollars an operation. The next day the thirty-year made its high for the year at 5.37%. The thirty-year gilt reached 5.93% on the same day, its highest since 1998.

Management has spent a month telling the market what it would like the price to be. The market has moved the other way.

Evaluate the management as you might a company board. Is the plan funded? A central bank’s bid has no ceiling, because it can print. The Treasury’s bid has a ceiling, because every dollar it spends buying a long bond it must first borrow. It is not retiring debt. It is swapping duration — buying the long, funding it short. If the shorter debt matures into a rising rate environment the terms of the overall debt get incrementally worse. This is the same trade that has gone wrong for every leveraged holder in Part Three.

The Secretary is eyeing the Treasury General Account, something close to a trillion in cash. A bigger balance to move around. Interventions tend to grow in size. And the tell is the language. The initial fixed number became “at least”. That is the escalation authorised and not yet used.

What’s in a price

What you get as a holder is 5.29% a year for thirty years and your money back at the end. The market prices that promise at 3.04% a year after inflation — the highest real yield the thirty-year has offered since the Treasury began publishing the series in 2010, and the whole of the bull case.

The gap between those two numbers is 2.25%. That is the breakevenThe inflation rate at which an ordinary bond and an index-linked one of the same maturity pay the same. Nominal yield minus real yield. Above it the index-linked wins; below it the ordinary one does.: the inflation rate the price assumes, not for a year but for thirty.

Core PCE is running at 3.34%.

The price contains an assumption that inflation falls more than a point below where it currently sits and stays there for three decades. Buy it and you are not taking a view on rates. You are underwriting that forecast.

If the forecast is right, you earn 3.04% real. If inflation simply runs where it runs today you earn 1.89%, and the hundred dollars returned in 2056 buys what thirty-seven buys now.

The case for

Access to capital carries a price again. You get paid to lend. A 3.04% real yield is not a crisis number. It is far above anything available between 2010 and 2022.

The move is global. British, German, Japanese and American long yields have risen together. These countries are not having simultaneous sovereign crises — they are competing in one capital market that has more borrowers in it than it used to.

And higher yields make their own buyers. The hundred largest corporate pension schemes in America were 112.1% funded at the end of July, a surplus of $139bn, which turns the problem from earning a return into locking one in. At five and a quarter on a government bond you no longer need equity risk to hit those targets.

The policy response to a fiscal problem is not bearish for bonds either. Buybacks, quantitative easing, pressure on banks and insurers to hold government paper, yield curve control — every instrument on that list is designed to support the price of the thing you own. The bear who is right about the deficit can still be wrong about the bond.

The case against — the cushion

The bear case does not need a fiscal crisis. It needs inflation to stay where it already is.

At 5.29% the thirty-year has a durationHow far a bond’s price moves when yields move: roughly this many per cent of price for every one point of yield. It rises with maturity and falls as redemption approaches. of just under fifteen — fifteen per cent of its price for every point of yield, in either direction. Which means a year’s entire income is wiped out by a rise of thirty-five basis points. A third of one per cent and your income is gone.

Maturity	Yield	Duration	Rise that wipes a year’s income
2 years	4.71%	1.89	250 bp
5 years	4.83%	4.40	110 bp
10 years	4.96%	7.81	64 bp
20 years	5.33%	12.21	44 bp
30 years	5.29%	14.96	35 bp

US Treasury par curve at the close of 22 September 2026. Modified duration computed at par, coupons twice a year. The thirty-year pays 58 basis points more than the two-year; on 9 July it paid 89.

That makes the case for a shorter duration. The two-year pays 4.71% and absorbs two and a half points of rise before you are down on the year. A seventh of the protection for an eighth more income.

Behind the arithmetic, the supply keeps coming. Debt is 122.6% of GDP with no political route out; interest expense rises even if yields fall, because cheap old debt matures into new debt about a hundred basis points more expensive; and the government is no longer the only large borrower in the queue, with the corporate market now funding a data-centre and power buildout on a scale that did not exist five years ago. None of that says what the bond is worth. It says why there is always more of it to buy.

What it’s worth

Held one year, and held five, if the thirty-year yield ends at:

Yield in one year	Move	Return over one year	Five years, annualised
3.79%	−150 bp	+31.5%	+8.84%
4.54%	−75 bp	+17.3%	+6.93%
5.29%	unchanged	+5.3%	+5.17%
6.04%	+75 bp	−4.9%	+3.55%
6.79%	+150 bp	−13.6%	+2.06%
8.29%	+300 bp	−27.5%	−0.56%

5.29% coupon, bought at par at the close of 22 September 2026. One-year column is price plus one coupon. Five-year column assumes coupons reinvested at 4%.

There are two different assets in that table.

Over one year the long bond is a leveraged position on interest rates that happens to pay a coupon. Seventy-five basis points the wrong way and the year ends down five per cent: the income gone, and as much again on top.

Over five years it is close to unloseable in nominal terms. Yields have to rise close to three full points before the holder is down at all.

Putting aside any rate fortune-telling, the question is what is the money for. If it has to be available, the long bond is the wrong instrument at any yield. If it does not, the arithmetic is on your side and the noise is somebody else’s problem. And for a sterling holder there is a second question the yield does not answer: a US thirty-year is also a thirty-year position in the dollar, and over that distance the currency can move further than the coupon.

Rate cycles run for decades. The price paid at the start decides most of what a holder ends up with — and that price is always somebody’s forecast.

Here the forecast is 2.25% inflation a year for thirty years, against 3.34% today. Whoever buys the long bond underwrites that number, whatever they think about rates.

Which is what to carry into the next bond that is offered: not whether the yield looks attractive, but what the price is asking you to believe.