How do you forecast bond returns?

Mostly by looking at the yield. A bond fund's expected return equals its yield, plus a small amount from bonds ageing along the yield curve, plus or minus the effect of interest rates moving, minus the losses expected from borrowers who fail to pay. The yield is by far the largest part, and for government bonds it is nearly the whole answer.

Why is the current yield such a good predictor of bond returns?

Because a bond fund's holdings mature and are replaced at whatever rates prevail. Leibowitz and Homer showed that a fund keeping a steady duration earns roughly its starting yield when held for about twice that duration, whatever happens to rates in between. Their empirical work found the starting yield came within a point or two of the realised average return at that horizon. Research Affiliates' own review of a decade of forecasts found their bond predictions had an R-squared of 75 percent and were off by only 42 basis points on average.

Do rising interest rates always mean losses on bonds?

No, and this is the most common misunderstanding. Rising rates cause an immediate fall in the price of bonds you already hold, but every pound reinvested afterwards earns the new higher rate. The loss happens once; the extra income arrives every year. Roughly speaking, if you hold for longer than twice the fund's duration, a rate rise leaves you better off than if rates had stayed put.

Duration is a measure of how sensitive a bond fund's price is to interest rates. A fund with a duration of 5 will lose roughly 5 percent of its value if interest rates rise by 1 percentage point, and gain roughly 5 percent if they fall by 1 point. It is measured in years because it also indicates roughly how long you need to hold before the higher income makes up for a price fall.

Why does the TIPS row look so much worse than the others?

It is not worse, it is measured differently. TIPS returns are shown after inflation and every other row is shown before inflation. A TIPS forecast of 2.8 percent after inflation and a Treasury forecast of 4.6 percent before inflation are close to the same thing if inflation runs at about 1.8 percent. The row is marked 'real' for that reason.

How we forecast bond returns

8 bond types · 3, 5 and 10 year horizons · written for investors rather than bond traders

Bonds have a reputation for being complicated. They are not, particularly. Forecasting what a bond fund will return is far easier than forecasting shares, because most of the answer is printed on the tin. This page explains what the Fixed Income tab is doing and why, without assuming you know what convexity is.

The short version

When you buy a bond fund, you are lending money at a known interest rate. That rate is the yield, and it is most of your answer. Four things then adjust it:

expected return
  = the yield it pays today
  + a small gain as bonds age (roll-down)
  +/- the effect of interest rates moving
  - what you lose when borrowers fail to pay

For government bonds the last item is nothing and the middle two are small, so the yield is nearly the whole story. For high yield bonds the losses are large and matter a great deal. That is the entire model.

Why the yield tells you most of it

A bond fund does not hold the same bonds forever. They mature, and the money is lent out again at whatever rates apply then. That churn is what makes bond returns predictable in a way share returns never are.

Leibowitz and Homer showed that a fund keeping a steady duration earns approximately its starting yield if you hold it for about twice that duration, and that this holds whatever rates do in between. Their work on actual bond data found the starting yield landed within a point or two of the realised average return at that horizon.

This is not a theory that only works on paper. Research Affiliates reviewed ten years of their own published forecasts and found the bond predictions had an R-squared of 75 percent, with a one-for-one relationship to what actually happened and an average error of 42 basis points. Their equity forecasts over the same decade were considerably less accurate. Bonds are the part of this that genuinely works.

Duration, explained properly

Duration is the number that tells you how much a bond fund moves when interest rates change. The rule of thumb is simple:

A fund with a duration of 5 falls roughly 5 percent if interest rates rise by 1 percentage point, and rises roughly 5 percent if they fall by 1 point. Duration 10, roughly 10 percent. Duration 2, roughly 2 percent.

It is quoted in years, which confuses people, because it is not how long the bonds last. It is closer to how long you must hold before higher interest income makes up for a price fall. A short-duration fund barely moves when rates change; a long-duration fund is genuinely volatile, and in 2022 long bond funds fell further than most stock markets.

The durations used here are typical for each type of fund: about 4 years for short Treasuries, 6 for a broad US bond index, 7 for corporate bonds and around 4 for high yield, which is shorter than people expect because those bonds are issued for shorter terms.

Why rising rates can be good news

Almost everyone believes rising interest rates are bad for bonds. Over a few months that is true. Over the period most people actually hold for, it is often the opposite, and the tab reflects that.

When rates rise, two things happen. The bonds you already own fall in price, immediately and visibly. Then every pound the fund reinvests earns the new, higher rate, quietly, for as long as you hold. The first effect happens once. The second happens every year.

effect of a rate change = change × (0.5 - duration / years held)

Which is negative if you sell soon, exactly zero at about twice the duration, and positive after that. A fund with duration 4 held for 10 years is better off if rates rise. The same fund held for 3 years is worse off. Same fund, same rate rise, opposite answer, and the only thing that changed is how long you left it alone.

This is a correction to what this page used to do. An earlier version counted the price fall but never the higher income that follows it, so it showed rising rates hurting at every horizon. That was wrong, and it made the bear case look worse than it should have for anyone holding for the long term.

Credit losses, and why high yield is not a free lunch

A government that prints its own currency will pay you back. A company might not. The extra yield you are offered for lending to companies is partly compensation for that risk, and part of it will be consumed by borrowers who default.

Type of bond	Expected annual loss	Why
Government	None	Assumed to pay in their own currency.
Investment grade corporate	0.3%	Defaults are rare and recoveries are decent.
High yield	2.5%	Around 4% default in a typical year, recovering roughly 40%.
Emerging market sovereign	1.0%	Defaults are uncommon but not rare, and restructurings are messy.
Emerging market local currency	1.5%	Default risk plus the currency.

The high yield figure is the one that changes how the table reads. Those funds are showing a yield above 7 percent, which sounds wonderful until 2.5 points a year are removed for the borrowers who do not pay. That is why high yield ends up forecasting close to government bonds despite advertising nearly three points more income.

These loss rates are long-run averages and are held constant across the three scenarios. In a real downturn defaults rise sharply at exactly the moment everything else is falling, so treat the high yield bear case as optimistic.

The one row measured differently

TIPS are US government bonds whose value rises with inflation. Because their whole point is to be measured after inflation, that is how their yield is quoted, and the tab shows their forecast the same way with a “real” tag on the row.

So a TIPS forecast of 2.8 percent and a Treasury forecast of 4.6 percent are not the gap they appear to be. If inflation runs at around 1.8 percent the two end up in much the same place, which is roughly what the market is pricing. The difference between them is not expected return, it is who carries the inflation risk. Buy a Treasury and you are betting inflation stays below what the market expects. Buy TIPS and you are handing that risk to the government.

Where every number comes from

Input	Source	Notes
US Treasury yield curve, 3 months to 30 years	FRED, US Treasury constant maturity	Refreshed each time the page loads.
Investment grade and high yield spreads	ICE BofA indices via FRED	Option-adjusted spread over Treasuries.
Emerging market spread	ICE BofA public sector emerging markets via FRED	Public sector issuers, so sovereign and state-owned rather than ordinary companies.
Expected inflation	10-year breakeven rate via FRED	The difference between Treasury and TIPS yields.
Euro area yields	European Central Bank	German government curve.
UK gilt yields	Bank of England
Japanese yields	FRED, OECD series	Needed because Japan is around a sixth of a world government bond index.
Durations and credit loss rates	Portfolio Lab assumptions	Typical values for each fund type, listed above.

The world government bond row blends these roughly in the proportions a global index holds them: the United States about 42 percent, the euro area a quarter, Japan a sixth and the UK only about a twentieth. An earlier version left Japan out entirely and gave gilts six times their real weight, which overstated the yield by several tenths of a point.

What this cannot tell you

The world government row is unhedged. Most people who hold foreign government bonds hold a currency-hedged version, and hedging changes the return by roughly the difference between short-term interest rates in the two countries. For a dollar investor holding Japanese bonds that is currently a large adjustment upward. This model does not make it.

Credit losses are long-run averages applied in every scenario, so the bear case understates what a genuine credit crisis does to corporate and emerging market bonds.

The rate scenarios are assumptions, not forecasts. Rates rising or falling by one and a half points is a reasonable range to think about and nothing more. Nobody, including us, knows what rates will do.

Everything here is before fund charges and tax, and before any difference between the index a fund tracks and the fund itself.

Related reading. The equity forecaster methodology covers the share side, which is a great deal less reliable and says so. The main methodology page covers the optimiser, Monte Carlo engine and backtester.

Portfolio Lab is a research tool. Nothing on this page or in the forecaster is investment advice or a recommendation to buy or sell any security. Forecasts are estimates and will be wrong to some degree.