Key takeaways
- In 2022 the MSCI World Index returned -18.14% and the Bloomberg Global Aggregate Bond Index -16.25%. A 40/60 blend of the two, levered 1.5 times, lost 25.51% before any borrowing cost.
- An unlevered 60/40 built from the same two indices lost 17.38% that year. Both portfolios carried 60% equity exposure, and 8.13 percentage points separated them.
- Over 1926 to 2010, the levered risk parity portfolio in Asness, Frazzini and Pedersen earned 7.99% excess against 4.65% for a 60/40, at 15.08% volatility against 11.68%.
- The same paper's robustness table shows that edge over the market portfolio falling from 4.15% a year to 1.81% when financed at LIBOR instead of Treasury bills.
- UK conventional gilts fell 23.83% in 2022 while the FTSE 100 Index gained 4.67%, a gap of 28.50 points in favour of the asset usually called the risky one.
Two ways to hold 60% in equities, and only one of them borrows
You've read that a bond-heavy portfolio has better risk-adjusted returns than a stock-heavy one, and that borrowing to hold more of it gets you the same expected return with better diversification. So you want to know what a leveraged portfolio actually does to your outcomes, next to the plain 60/40 you already own.
Here's the arithmetic on one side. Hold 40% equities and 60% bonds, then borrow 50p for every pound of your own money. Your 150p now buys 60p of equities and 90p of bonds. Your equity exposure matches the 60/40 exactly. Your bond exposure is 90% of your capital rather than 40%.
Run both through 2022 and they part company badly. The 60/40 lost 17.38%. The levered 40/60 lost 25.51%, and that's before you pay a penny of interest. The equity exposure was identical. What differed was the 50 extra points of bond exposure, in the year bonds behaved like equities.
That is the whole piece in three paragraphs. The rest explains why the strategy has good evidence behind it anyway, and where that evidence stops covering you.
The case for borrowing into bonds is 85 years old and it is not stupid
The serious version of this argument is Leverage Aversion and Risk Parity, by Clifford Asness, Andrea Frazzini and Lasse Pedersen, published in the Financial Analysts Journal in January 2012. Their starting observation is uncomfortable for anyone holding a 60/40.
Across US stocks and Treasuries from 1926 to 2010, stocks returned an average 10.8% a year at 18.9% volatility. Bonds returned 5.2% at 3.4% volatility. Run those two points through the standard mean-variance machinery and the highest-Sharpe combination isn't 60/40 at all. In their words, "the ex post tangency portfolio invests 88 percent in bonds and 12 percent in stocks."
Bonds delivered more return per unit of risk. A 60/40 loads most of its actual risk onto the smaller, wilder allocation. If you want the bond-heavy mix and the return of a stock-heavy one, borrowing is the only route. That's the case, and it isn't a marketing claim. Their Table 2 shows a levered risk parity portfolio earning 7.99% in excess of Treasury bills over 1926 to 2010, against 4.65% for the 60/40, with Sharpe ratios of 0.53 and 0.40. The alpha carries a t-statistic of 4.30.
The "same risk" claim doesn't survive the table it comes from
Look again at those two rows and the framing slips. The levered risk parity portfolio ran at 15.08% volatility. The 60/40 ran at 11.68%. It wasn't matched to the 60/40 at all. The paper says so plainly: the weights "are multiplied by a constant to match the ex post realized volatility of the value-weighted benchmark," which is the whole market, and the market was a riskier thing than a 60/40.
So the comparison isn't equal risk with more return. It's 3.40 more points of annual volatility, and more return. That's a different, weaker claim, and it's the one the numbers support.
The leverage involved is larger than most readers picture. The unlevered risk parity portfolio ran at 4.25% volatility; the levered one at 15.08%. That's a multiple of roughly 3.5 times your capital, sustained for 85 years, rebalanced monthly. The 1.5 times in the worked example above is a mild version of the same idea.
Financing cost ate more than half the edge in the paper's own robustness check
Appendix B of the same paper is the most honest page in it, and it's the page nobody quotes. The authors rerun the strategy charging five different borrowing rates. Over 1926 to 2010, risk parity beat the market portfolio by 4.15% a year when financed at Treasury bills. At one-month LIBOR, averaging 62.3 basis points above bills, that margin drops to 1.81%, and the t-statistic falls from 2.95 to 1.29. More than half the edge, 2.34 percentage points of it, goes to the lender.
Now price your own borrowing. Interactive Brokers quotes 5.244% on the first tier of a sterling margin loan, described on its rates page as benchmark plus 1.5%. Borrowing 50p per pound at that rate costs 2.62 percentage points of your total capital every year, in a good year and a bad one alike. That is roughly the entire edge the paper found once it was honestly financed, handed over before the strategy does anything.
The paper's best answer is that futures avoid the problem, because futures returns, in its phrase, "are already excess returns" and no financing assumption is needed. That's true, and it's a real rebuttal. It also means the version that works cleanly is a futures programme, not a margin loan on a fund.
2022 is what a leveraged portfolio's failure mode looks like
The chart below plots calendar 2022. The two building blocks fell almost together: the MSCI World Index, measured net of withholding tax, returned -18.14%, and the Bloomberg Global Aggregate Bond Index returned -16.25%. Both figures are the benchmark rows on iShares fund factsheets dated 31 July 2026.
Work through the blends. The unlevered 60/40 lost 17.38%. The unlevered 40/60 lost 17.01%, so the conservative mix did its job, just barely. Apply 1.5 times leverage to that same conservative mix and the loss becomes 25.51%, before financing.
The mechanism is not that bonds are secretly risky. It's that leverage multiplies whatever correlation turns out to be. A 40/60 protects you when bonds and equities move apart. Multiply it by 1.5 and you've also multiplied the year they don't. The correlation instability that makes diversification unreliable is precisely the input a leveraged portfolio is most sensitive to.
The UK version was worse, and the gilt market needed the Bank of England
For a sterling investor the 2022 numbers are starker still. The FTSE Actuaries UK Conventional Gilts All Stocks Index fell 23.83%. The FTSE 100 Index rose 4.67%. The supposedly safe leg lost nearly a quarter of its value while the risky leg made money, a spread of 28.50 points the wrong way round.
What happened next is the clearest live demonstration of the failure mode there has ever been, and it happened to professionals. In a November 2022 speech given at ISDA and AIMA, the Bank of England's Sarah Breeden described liability-driven investment funds, which hold leveraged gilts on behalf of pension schemes. Long-dated gilt yields moved violently in late September that year, by her account 130 basis points on the 30-year nominal yield in just a few days. What that did to the funds is worth quoting exactly. The rise in yields, she said,
caused a significant fall in the net asset value of these leveraged LDI funds, meaning their leverage increased significantly. And that created a need urgently to delever to prevent insolvency and to meet increasing margin calls.
Read the middle clause again. Losses don't reduce a leveraged position, they concentrate it. Your borrowing stays fixed in pounds while your equity shrinks, so the ratio between them gets worse exactly when you can least afford it. Breeden's diagnosis of the whole episode came down to three words: "poorly managed leverage." The Bank's response, in her account, was a "13 day and £19.3 billion intervention" in the gilt market.
Margin rules protect the broker, and they trigger later than you'd guess
People assume a margin call arrives early enough to matter. The arithmetic says otherwise. Regulation T sets initial margin for a long equity position at "50 percent of the current market value of the security," under 12 CFR 220.12, which caps you at two times your capital. FINRA Rule 4210 then sets ongoing maintenance at "25 percent of the current market value of all margin securities."
Put £100 in, borrow £50, hold £150 of assets. For the maintenance test to fail, your assets have to drop 55.6%. At that point your equity is £16.67, so you've lost 83% of your money before the broker asks for anything. Your capital is gone entirely at a 66.7% fall.
Set that against the deepest historical case. The UBS Global Investment Returns Yearbook 2026, compiled from the Dimson, Marsh and Staunton database, reports that since 1900 "equities and bonds have on several occasions lost more than 70% in real terms. yet a 60:40 equity:bond blend has never declined more than 50%." A 50% fall on a portfolio levered 1.5 times takes 75% of your capital. Your equity to assets ratio at that moment is 33.3%, comfortably above the 25% maintenance floor. Three quarters of your money is gone and the margin system registers nothing wrong, because it was never designed to notice. This is the loss recovery arithmetic problem in its harshest form.
The strongest case for the other side
Two objections deserve a fair hearing, and the second one is good.
The first is that 2022 was a single year chosen after the fact, and that's fair. One calendar year proves nothing about a strategy tested over 85. The 1926 to 2010 evidence is real, statistically significant on its own terms, and reproduced across ten other countries in the paper's Table 3.
The second is more interesting, and it cuts against the paper. In that same Table 3, the United Kingdom over 1986 to 2010 shows a 60/40 earning 2.62% excess and risk parity 3.14%, a gap of 0.52 percentage points with a t-statistic of 0.58. That is nowhere near significant. Sharpe ratios were 0.26 against 0.31. The result that looks overwhelming in the pooled US data is, in the home market of a sterling investor, close to noise. The mechanism behind it, that safer assets are priced too expensively because investors avoid borrowing, is the same one that drives the low volatility anomaly in equities, and it has the same problem: it depends on cheap, reliable leverage that most people can't get.
What this evidence cannot tell you
Start with the bond volatility figure. Bonds ran at 3.4% annual volatility over 1926 to 2010, and that number does a lot of work in making a levered bond position look sensible. Much of that sample sits inside an era of pegged and heavily managed interest rates. Gilts fell 23.83% in 2022. A strategy whose case rests on bonds being five and a half times steadier than equities is a strategy resting on a regime, not a law.
The performance in the paper is also gross of the thing that kills leveraged investors. The authors say it themselves: their simulated results do "not reflect any adjustment of the returns for the costs of leverage, such as financing spreads and costs associated with deleveraging." Forced deleveraging is not a cost you can average into an annual figure. It is a specific event on a specific afternoon, at prices nobody would have chosen.
The long sample is US only, monthly rebalanced, and assumes you can always meet a margin call. The 2022 arithmetic above is one calendar year, uses index returns rather than a real portfolio, ignores fund charges, spreads and tax, and applies no financing cost at all, which flatters the leveraged case rather than hurting it. Backtests aren't forecasts. None of this covers what happens if your broker raises its margin requirement mid-drawdown, which is the moment it usually happens.
What would change the conclusion
If your financing genuinely cost the risk-free rate, most of this reverses. The paper's own table shows the edge surviving intact at bill rates and dying at LIBOR. Futures get closer to that than a margin loan does, which is why institutions run this strategy with futures and households mostly can't.
If bonds returned to 3.4% volatility and a reliably negative correlation with equities, the 2022 result becomes the outlier it's often described as. That's a claim about the inflation regime, not about portfolio construction, and it isn't settled.
If your holding period is long enough to sit through a 75% drawdown, the argument moves from arithmetic to temperament, and the arithmetic can't referee that. What it can tell you is the size of what you'd be sitting through. The US real returns in the 2026 Yearbook are 6.6% a year for equities and 1.6% for bonds since 1900, a gap of 5.0 points, and leverage does not create that gap. It rents exposure to it.
The number worth watching isn't your allocation. It's your gross exposure, the total assets you control divided by the money that's actually yours. A 60/40 holds that at 1.00 by construction. Anything above it is a second decision, taken separately, and it's the one that decides how a bad year ends. LedgerTouch reports gross exposure alongside weights; so does a spreadsheet with one more column in it.