What a Lost Decade for Stocks Does to a 60/40 Portfolio

Key takeaways

• Of 1,750 rolling ten-year windows in Robert Shiller's US data since 1881, 193 of them, or 11.0%, returned zero or less from equities after inflation.
• In 151 of those 193 windows, 78% of them, bonds lost money in real terms as well, at a median of minus 1.54% a year.
• With equities flat and gilts at the 1.937% ten-year real spot rate of 29 October 2026, £250,000 in a 60/40 ends the decade at £270,059 in today's money.
• Repeating June 1972 to June 1982 leaves the same £250,000 at £164,897, and at £81,381 once £10,000 a year is being drawn from it.
• With the bond leg at 1.937% real, equities can lose 1.29% a year after inflation for ten years before a 60/40 ends the decade flat.

What a lost decade for stocks did to £250,000

You want to know what happens to your money if shares go nowhere for ten years. The short answer is that a lost decade for stocks, on its own, isn't the disaster it sounds like. A decade in which stocks and bonds both lose money is, and in the record those two mostly arrived together.

The table runs £250,000 in a 60/40 portfolio, rebalanced once a year, through six ten-year scenarios. Everything is in today's money, so inflation has already been taken out. The last column repeats the exercise while £10,000 a year comes out of the pot.

The equity and bond columns are the real total returns Shiller's series records for those exact windows. The chart above plots the fourth column. Notice the gap between row two and row six: the same pot, the same ten years, and £105,162 of difference between equities that stood still and equities that fell while bonds fell with them.

Flat is the mild version of a lost decade, and it has almost never happened

A portfolio stress test usually starts with a round number, and zero is the roundest. It's also the wrong shape. Shiller's monthly series carries a real total return index for US equities and another for ten-year government bonds, running from 1871 to October 2026. Line them up into every overlapping ten-year window and you get 1,750 of them, ending between January 1881 and October 2026.

Equities came in at zero or less, after inflation, in 193 of those windows. That's 11.0% of the record, so roughly one ten-year stretch in nine. But almost none of them were flat. They were falls. A lost decade for stocks has nearly always meant a falling decade rather than a still one. The median of the 193 is minus 2.11% a year, and the worst of them, the ten years to March 2009, took 5.92% a year off equities in real terms. Over the full decade that's a loss of 45.7%.

So "equities return zero for ten years" is not the pessimistic case. It's the optimistic version of the pessimistic case. Asking what a flat decade does to you answers a question that history has rarely posed.

The 40 did not rescue the 60, in 78% of the bad decades

Here's the part that changes how the stress test should be built. In 151 of those 193 windows, bonds also returned zero or less after inflation. That's 78% of them. The median real bond return across the 193 was minus 1.54% a year.

The reason isn't mysterious. The thing that kills a decade of equity returns is usually inflation, and inflation is exactly what destroys a nominal bond. The 1970s are the clean example. Between June 1972 and June 1982, US equities lost 3.76% a year in real terms and ten-year government bonds lost 4.55% a year. The diversifier was the worse holding.

Run those two numbers through a 60/40 and the portfolio ends the ten years down 32.4%. Across the whole record, a 60/40 returned zero or less in real terms in 182 of the 1,750 windows, or 10.4%. Within the 193 equity-negative windows, it was negative 85% of the time. The 40 moved the arithmetic a little. It did not change the answer.

The 2000s are the decade everyone pictures, and the least representative one

When someone pictures a lost decade for stocks, they're usually picturing the 2000s, and that episode is genuinely the friendly one. From March 1999 to March 2009, US equities lost 5.92% a year in real terms, the deepest ten-year equity loss in the whole series. Ten-year government bonds made 4.06% a year in real terms over the same window. A 60/40 came out of it down 15.9%, which for the worst equity decade on record is remarkably mild.

That is the decade that taught a generation of investors that bonds catch you. It is also one of the 42 windows out of 193 where they did. Building a lost decade scenario on 1999 to 2009 is building it on the kindest bad decade in the data, and the table shows what that costs you: £205,774 against £164,897 for the 1970s version of the same question.

The 193 windows fall into five clusters of end dates: 1918-1922, 1939-1940, 1946-1947, 1974-1983 and 2008-2011. Bonds made a positive real return in only two of them. The decades ending in 1939-1940 ran at a median real bond return of 6.24% a year, and those ending in 2008-2011 at 3.97%. Both were deflationary. In the other three clusters, 151 windows in total, bonds lost real money every time.

Today the bond leg of the stress test is priced rather than guessed

There's one way the present is genuinely different from 1972, and it's worth more than any forecast. On 29 October 2026 the Bank of England's government liability curve put the ten-year real spot rate at 1.937%, against a nominal ten-year rate of 5.276% and an implied inflation rate of 3.339%. A ten-year index-linked gilt bought at that price and held to maturity delivers roughly 1.937% a year above inflation, whatever equities do.

That turns the stress test from two unknowns into one. If the bond leg of a 60/40 is pinned at 1.937% real, equities can fall 1.29% a year in real terms for ten years before the whole portfolio ends the decade flat. That's the breakeven. Compare it with the record: in the 193 bad windows, equities were below minus 1.29% a year far more often than not.

It's also a narrower claim than it looks. Index linked gilts deliver that return only if they're held to maturity and only against the inflation measure they track. A bond fund has no maturity date, and the UK's inflation rate was 3.5% in the year to September 2026 against a Bank of England target of 2%. The ten-year implied inflation rate of 3.339% is what separates the nominal and real curves, and it sits above that target rather than on it.

Withdrawals turn a bad decade into a different problem altogether

The last column of the table is where the arithmetic stops being academic. Take £10,000 a year out of £250,000, in real terms, and the median decade still leaves £296,907. The 1972 to 1982 repeat leaves £81,381, which is 32.6% of what went in. That's not a smaller version of the same outcome. It's a pot with far less left to recover from.

This is sequence risk, the plain fact that two decades with the same average return can leave you in very different places depending on when the bad years land. The sequence of returns matters because a withdrawal in a down year sells more units to raise the same cash, and those units aren't there for the recovery. The safe withdrawal rate literature exists because of this one mechanism.

The table understates it. Holding a return constant at minus 3.76% a year strips out the lumpiness that does the damage. Running the actual monthly path of June 1972 to June 1982 instead, the same £250,000 ends at £168,576 without withdrawals and £82,607 with them. Along that path the portfolio's deepest fall from its own running peak was 36.7% in real terms. The 1999 to 2009 path bottomed 25.6% below its peak, and the 1910 to 1920 path 43.0% below. A reader who only saw the end value would never know how far down it went in between.

Japan is the case where the decade didn't end

Every number above comes from a market that recovered. Japan is the counter-example, and it's the reason a ten-year horizon can be the wrong unit entirely. TOPIX closed 1989 at 2,881.37. It closed 1999 at 1,722.20, down 40.2%. It closed 2009 at 907.59, down 68.5% from where it started twenty years earlier. The index did not close a year above its 1989 level until 2025, when it finished at 3,408.97.

That's a price index, so dividends are excluded and the total-return picture is better than those figures suggest. The direction isn't in doubt though. Thirty-six years is not a lost decade. It's a lost working life, and no ten-year stress test has a row for it.

Britain's 1970s were worse than America's, before dividends

UK readers have a home example that rarely gets cited. The Bank of England's long-run dataset carries a spliced UK share price index back to 1700 and a consumer price index alongside it. Between 1972 and 1982 UK share prices rose 71.4% while consumer prices rose 243.9%. In real terms, share prices fell 50.2%.

That figure excludes dividends, and UK dividend yields in the 1970s were high enough to change the total-return answer materially, so treat minus 50.2% as the price leg rather than the whole story. Even so, it sets the scale. On the same price-only basis, US shares lost 48.3% in real terms over those same ten years, so Britain's inflationary decade was marginally the worse of the two.

The strongest objection to all of this

The honest criticism of a table like the one above is that it is built out of history, and history had different starting prices. A decade that begins at a low valuation is not the same bet as one that begins at a high one, and the same ten-year return can follow from either. Critics of historical stress testing are right that the windows aren't independent: 1,750 overlapping monthly windows contain far fewer than 1,750 independent decades, so "11.0% of windows" is not the same as "an 11.0% chance".

The second objection is about regime. Three of the five clusters end before 1948, inside a world of gold standards and world wars that no UK investor lives in now. Strip them out and the sample shrinks to two episodes, the 1970s and the 2000s, which isn't enough to estimate anything from. What survives the cut is one inflationary decade and one deflationary one, which is exactly the ambiguity the table is trying to price.

And the recent evidence points hard the other way. US equities returned 11.74% a year after inflation over the ten years to October 2026. MSCI's World index returned 13.37% a year gross in dollars over the ten years to 30 September 2026, against 9.02% a year since the end of 1987. Someone who started investing in 2016 has never seen a decade from the bottom half of the table.

What this evidence cannot tell you

The equity and bond series are US data. Shiller's series is long enough to contain five of these episodes, which is why it's used here, but a US sample measures one market that won two world wars and ended the century as the world's reserve currency. Applying it to a global portfolio is an assumption, not a finding. The UK leg above is a price index without dividends, and the Japanese leg is too.

The bond series is a single ten-year government bond series, not the mix of gilts, corporates and short duration that sits inside most real-world 40%. Backtests are not forecasts. Nothing in 1,750 historical windows constrains what the next ten years do, and the overlapping-window problem above means the frequencies are softer than they look.

The scenario rows also hold each year's return constant, which is not how a decade behaves. Where the actual month-by-month path has been run alongside it, the constant-rate version came out lower without withdrawals and lower with them, so the table is on the conservative side of the real path for those three windows. That is three comparisons, not a rule.

What would change the conclusion

If inflation stayed near target, the central finding weakens considerably. The reason bonds failed in 78% of the bad equity decades is that most of those decades were inflationary. A world where CPI sits at 2% is a world where the 40 behaves the way the brochure says, and where the 1999 to 2009 row is the right template rather than the exception.

If real yields went back to zero, the breakeven moves against you immediately. At a 1.937% real yield the bond leg absorbs 1.29% a year of equity losses. At nothing, it absorbs nothing, and the 60/40 simply inherits whatever equities do. It is the one input in the whole table that gets repriced every trading day.

If your horizon isn't ten years, none of this transfers. The Japanese case is the warning: a stress test with a ten-year box has no way to represent a market that takes thirty-six years to get back. Looking at stock market returns by decade alongside the question of how long bear markets last gets closer to the shape of the risk than a single decade ever does.

The number to keep an eye on isn't an equity forecast, which nobody has. It's the real yield on the bond half, because that one is quoted every day and it sets how much equity pain the rest of the portfolio can absorb before the decade goes backwards.