Building a Portfolio: What the Research Says About Stock/Bond Mixes, Bonds, Global Diversification and Property

This article summarises published research on building and running a portfolio. It is general information, not financial advice: it doesn't know your situation, your country's tax rules or what is available to you, and it doesn't recommend any product or fund. Past returns don't predict future ones.

VettedVids rates books and videos by what people report happened after using them. Research is different: we don't score it. We summarise what each study found, the market and period it covers and its limits, and where findings disagree we show both rather than settle it. Several findings come from fund managers or index providers (Vanguard, Morningstar, MSCI, AQR and commodity-index sponsors), and we say where. Much of the research uses US data. Every claim below links to its source, and the same findings are in the Research tab of our Build and run an investing portfolio page.

Choosing your stock/bond mix

• How far each mix has fallen in its worst year. From 1928 to 2025, the worst calendar year was -43.3% for 100% stocks, -39.2% for 90/10, -35.1% for 80/20, -26.9% for 60/40, -22.8% for 50/50 and -12.9% for 100% bonds. Average yearly returns over the same span ran from 11.8% (100% stocks) and 11.1% (90/10) to 9.1% (60/40) and 5.2% (100% bonds) (Vanguard 2026, Investment portfolios: Asset allocation models). This is published by a fund manager (Vanguard), using index data from Morningstar to 31 December 2025. Calendar-year losses understate peak-to-trough falls inside a year. Stocks are all-US until 1969, then 60% US/40% international; bonds are all-US until 1990, then 70% US/30% hedged international. These are index returns, with no fees or taxes.
• Shares over 30 years can still lose to inflation. Using stock-market data from 39 developed countries between 1841 and 2019, the authors estimate a 12% chance that a diversified stock investor with a 30-year horizon ends up behind inflation (Anarkulova, Cederburg & O'Doherty 2022, Journal of Financial Economics). It is a bootstrap simulation built from the historical record of many countries, including markets that did badly, so it is less US-centric than most long-run studies. It covers stocks only and does not test stock/bond mixes.
• Target-date funds changed how savers were invested. Target-date fund assets in US corporate retirement plans grew from $5 billion in 2000 to $734 billion in 2018. When plans added these funds, members' equity shares and bond exposures rose, holdings of cash and employer stock fell, and stock-specific risk dropped. The authors estimate low-cost target-date funds could raise retirement wealth by as much as 50% over 30 years (Mitchell & Utkus 2022, Journal of Pension Economics and Finance; Mitchell & Utkus 2020, NBER Working Paper 26684). This covers US workplace plans only, using Vanguard recordkeeping data. One author (Utkus) was head of investor research at Vanguard, a large target-date provider, as the paper discloses. The 50% figure is a model estimate, not an observed outcome.
• Future pay can act like a bond holding. In a calibrated life-cycle model, steady labour income substitutes for safe assets, so the model's optimal share in stocks starts high and roughly declines over life. Ignoring labour income altogether carried large welfare costs in the model (Cocco, Gomes & Maenhout 2005, Review of Financial Studies). This is a theoretical model, not observed investor outcomes. In the model, ignoring how risky income is mattered far less than ignoring income altogether, except when a disastrous income shock was allowed.
• Employer shares tie pay and portfolio to one company. About a third of the assets in large US retirement plans were in the employer's own stock. Staff at firms with the worst 10-year stock performance put 10.37% of their discretionary contributions into company stock, against 39.70% at firms with the best performance, yet those allocations did not predict future returns (Benartzi 2001, Journal of Finance). The data are from US retirement plans before 2001. Later evidence found that adopting target-date funds cut company-stock holdings (Mitchell & Utkus 2022).

Bonds and cash

• Bond prices fall when rates rise, more so for longer bonds. In the SEC's example, a 10-year Treasury with a 3% coupon bought at $1,000 would be worth about $925 a year later if market rates rose to 4%, and about $1,082 if they fell to 2% (SEC Office of Investor Education and Advocacy, Investor Bulletin: Fixed Income Investments). FINRA's rule of thumb is that a bond with a duration of 10 loses roughly 10% for each 1-percentage-point rise in rates (FINRA, Brush Up on Bonds: Interest Rate Changes and Duration). These are regulator worked examples and a rule of thumb, not a study. Duration is an approximation that works best for small rate changes, and a bond held to maturity still repays its face value, assuming the issuer doesn't default.
• Why stocks and bonds fell together in 2022. The correlation between US stock and government bond returns turned positive in mid-2021 and stayed positive, the first prolonged positive spell since the 1980s and early 1990s. The BIS links this to high, volatile inflation: inflation surprises push bond prices down and also hurt stocks through expected rate rises (Lombardi & Sushko 2023, BIS Quarterly Review, Box A). An earlier model-based study likewise found US Treasuries moved with stocks in 1979–2001 and against them in 2001–2011 (Campbell, Pflueger & Viceira 2020, Journal of Political Economy). Both use US data, and the BIS piece is a short analytical box, not a peer-reviewed paper. The stock–bond relationship has changed sign before and may change again; neither source forecasts it.
• Inflation-linked bonds: real protection, but prices still swing. In the US and UK, long-term real interest rates on inflation-linked government bonds fell sharply from the 1990s until 2008, then spiked during the 2008 financial crisis. The authors conclude that low yields and high short-term price volatility don't undo the basic case for these bonds as a safe asset for long-term investors (Campbell, Shiller & Viceira 2009, NBER Working Paper 15014). The study covers US (TIPS) and UK (index-linked gilt) markets up to 2009. The protection applies to holding to maturity; anyone selling early takes the market price, which can fall.

Diversifying globally

• Home markets are a small share of the world, but investors hold mostly home shares. At 30 September 2026, the US made up 64.21% of the MSCI ACWI global stock index, Japan 5.2%, Taiwan 3.46%, the UK 3% and Canada 2.9% (MSCI ACWI Index factsheet, 30 September 2026). The weights come from an index provider (MSCI), cover large and mid caps in 23 developed and 24 emerging markets (about 85% of investable stocks), and shift as prices move. Yet at the end of 1989, Japanese investors held more than 98% of their equities at home, US investors 94% and UK investors 82% (French & Poterba 1991, NBER Working Paper 3609). Home bias has fallen since then but not disappeared.
• Global diversification fails in crashes but works over long periods. Markets do become more correlated in short crashes, so international diversification gives little protection in a sudden panic. Over longer horizons, though, each country's own economic growth matters more than short-lived panics, and spreading across countries protected investors well against prolonged bad outcomes in any one market (Asness, Israelov & Liew 2011, Financial Analysts Journal). All three authors work at AQR Capital Management, a fund manager. The abstract states the conclusion qualitatively; the paper has the country data and periods.
• Currency hedging: near-full for bonds, less clear for shares. From 1975 to 2005, the US dollar (especially against the Canadian dollar), the euro and the Swiss franc tended to rise when world stock markets fell, so leaving some exposure to those currencies reduced risk for global equity investors. For global bond investors, the lowest-risk approach was close to a full currency hedge (Campbell, Serfaty-de Medeiros & Viceira 2010, Journal of Finance; NBER Working Paper 13088, 2007). The study is about minimising risk, not returns. Which currencies cushion a fall depends on the investor's home currency, and the paper found little evidence for changing hedges in response to interest-rate differences.

Building the portfolio

• What fund investors pay. The asset-weighted average expense ratio across US mutual funds and ETFs fell to 0.32% in 2025, from 0.80% in 2006. In 2025 it was 0.09% for broad index funds, 0.16% for strategic-beta funds and 0.57% for actively managed funds, and ETF fees averaged about half of mutual fund fees (Morningstar 2026, US Fund Fee Study). This covers US funds only and is published by a commercial research firm (Morningstar) that also sells fund data and indexes. Expense ratios exclude trading costs, adviser fees and platform fees, and fees in other countries differ.
• Investors in all-in-one funds kept more of their returns. Over the 10 years to the end of 2025, investors in US allocation funds earned 92% of those funds' aggregate total returns, and US stock fund investors 97%. Municipal-bond, alternative and taxable-bond fund investors captured only 49%, 62% and 65%, because of the timing of their purchases and sales (Ptak et al. 2026, Morningstar, Mind the Gap). This covers US funds and is also from Morningstar. The report notes that results can hinge on where funds are used, for example a retirement-plan menu versus a brokerage account.
• Asset location: which account holds what. A model of taxable and tax-deferred investing found a strong preference for holding taxable bonds in the tax-deferred account and equities in the taxable account (Dammon, Spatt & Zhang 2004, Journal of Finance). Yet in 1998 US survey data, many households with both kinds of account did the reverse; for nearly three quarters of them, moving less than $10,000 would have reached a tax-efficient arrangement (Bergstresser & Poterba 2004, Journal of Public Economics). Both rest on US tax rules of the late 1990s and early 2000s. Later changes (lower US dividend tax rates, Roth accounts) and other countries' accounts, such as ISAs, TFSAs and super, change the arithmetic.

We haven't summarised research on contribution limits or the tax treatment of each country's accounts; those are set by each government, and the page links to the official sources for six countries.

Running it over time

• Rebalancing by threshold or by calendar: small differences. In simulations of a global 60/40 portfolio, rebalancing only when the mix drifts 2 percentage points from target (back to 1.75 points) was expected to return 11–18 basis points a year more than monthly or quarterly rebalancing, mainly through lower trading costs. For target-date funds the expected benefit was 15–22 basis points a year versus monthly and 5–8 basis points versus quarterly during the saving years (Zhang et al. 2024, Vanguard Research, The rebalancing edge). This is published by a fund manager (Vanguard) that uses this method in its own target-date funds. The results are simulated and fund-level, with daily monitoring and institutional trading costs; individual investors face different costs and, outside tax-sheltered accounts, capital-gains tax.
• Tax-loss harvesting: about 1% a year in a US backtest. Applied to the 500 largest US stocks from 1926 to 2018, with long- and short-term capital-gains tax rates of 15% and 35%, a tax-loss-harvesting strategy produced a 'tax alpha' of 1.08% a year before transaction costs. Respecting the US wash-sale rule cut it to 0.82% a year (Chaudhuri, Burnham & Lo 2020, Financial Analysts Journal). This is a historical simulation under US tax rules, in taxable accounts only. Harvested losses usually lower the cost base, so part of the saving can be paid back when the holding is eventually sold, and results depend on tax rates, market path and costs. Co-author Andrew Lo founded AlphaSimplex Group, a fund manager.

Property and other assets

• Housing and shares since 1870: similar returns, less volatility in national data. Across 16 advanced economies from 1870 to 2015, housing and equities both returned about 7% a year after inflation. Housing returns were much less volatile (standard deviation 10% vs 22%), so compounded returns were higher for housing (6.6% vs 4.7%). Round-trip transaction costs averaged about 7.7% of a property's value in 13 of the 16 countries, cutting housing returns by less than 1 percentage point a year for a 10-year holding (Jordà, Knoll, Kuvshinov, Schularick & Taylor 2019, Quarterly Journal of Economics; figures from NBER Working Paper 24112, revised May 2019). These are national housing indices with estimated rents, not individual properties: they don't capture single-property concentration, leverage, vacancies or the effort of being a landlord.
• Property-level records show lower returns. Using property-level accounts of four large Oxford and Cambridge colleges from 1901 to 1983, real total returns after costs were about 2.3% a year for residential and 4.5% for agricultural property. Gross income yields started around 5%, and long-run real income growth was close to zero for all property types (Chambers, Spaenjers & Steiner 2021, Review of Financial Studies). This is one kind of owner (large institutional landlords in England) over 1901–1983 and may not represent other countries or recent decades. It contrasts with national-index estimates such as Jordà et al. above.
• Over long horizons, listed property (REITs) behaves more like property than like shares. Using sector-level data for the US, UK and Australia, long-run REIT performance was much more closely tied to the direct property market than to the general stock market, and the authors conclude REITs and direct real estate are fairly good substitutes in a long-horizon portfolio (Hoesli & Oikarinen 2012, Journal of International Money and Finance). The result concerns long-run co-movement. The paper models short-term movements separately and does not claim REITs avoid stock-market falls.
• Gold: a short-lived crash hedge, an unreliable inflation hedge. In US, UK and German data, gold was on average a hedge against stocks and a safe haven in extreme stock-market falls, but the safe-haven effect was short-lived (Baur & Lucey 2010, Financial Review). A separate study found gold hedges inflation only over centuries; over practical horizons it was unreliable, and periods with an above-average real gold price were followed by below-average real returns (Erb & Harvey 2013, Financial Analysts Journal). The two papers ask different questions (crash behaviour vs inflation protection) and support different roles for gold. Erb and Harvey's valuation point was made in 2012–2013, when the real gold price was historically high.
• Commodity futures: equity-like returns, or close to zero? (contested) An equally weighted index of commodity futures from 1959 to 2004 matched equities' return and Sharpe ratio while moving against stocks and bonds and with inflation (Gorton & Rouwenhorst 2006, Financial Analysts Journal); a 10-year follow-up found this largely held out of sample (Bhardwaj, Gorton & Rouwenhorst 2015, NBER Working Paper 21243). A competing study found the average individual commodity future earned an excess return of about zero, with 'equity-like' returns coming mainly from a rebalanced portfolio (Erb & Harvey 2006, Financial Analysts Journal). The 2006 study was funded by AIG Financial Products, then a commodity-index sponsor, and a co-author of the 2015 follow-up works at SummerHaven Investment Management, a commodity fund manager. The two camps differ on how much return to expect. Both study futures indices, not physical commodities or commodity-company shares, and index construction drives the results. Correlations with stocks rose temporarily in the 2008 crisis.

Stock-analysis basics

• Most stocks worldwide lag cash; a few create all the wealth. Among more than 64,000 stocks worldwide from January 1990 to December 2020, 55.2% of US stocks and 57.4% of non-US stocks had lower compound returns than one-month US Treasury bills. The top-performing 2.4% of firms accounted for all of the $75.7 trillion in net global stock-market wealth created (Bessembinder, Chen, Choi & Wei 2023, Financial Analysts Journal). An earlier US study covering 1926–2016 found the same pattern (Bessembinder 2018, Journal of Financial Economics). Both describe individual stocks held for their full listed life, not diversified portfolios; the point for portfolios is how easily a few holdings miss the rare big winners.
• How many stocks make a diversified portfolio? A well-diversified portfolio of randomly chosen stocks needed at least 30 stocks for an investor who borrows and 40 for one who lends, contradicting the common belief that about 10 stocks capture most of the benefit (Statman 1987, Journal of Financial and Quantitative Analysis). This is an older US study using a specific definition of 'well diversified'; the number depends on the period and the measure used. A broad index fund holds hundreds or thousands of stocks.
• Individual stock traders as a group lose to institutions. Using the complete trading records of all investors in Taiwan, the aggregate portfolio of individual investors suffered a performance penalty of 3.8 percentage points a year, equal to 2.2% of Taiwan's GDP. Institutions gained 1.5 percentage points a year, and almost all individual losses came from aggressive orders (Barber, Lee, Liu & Odean 2009, Review of Financial Studies). This is one market with very active retail trading, and it measures trading losses in aggregate; some individuals do well.
• Starting valuations and 10-year returns: useful for expectations, poor for timing (contested). Price-to-earnings and dividend ratios did poorly at forecasting earnings or dividend growth but were useful for forecasting long-run price changes, in US data from 1871 to 2000 and 12 countries since 1970 (Campbell & Shiller 2001, NBER Working Paper 8221). For the S&P 500 from 1926 to 2012, rolling decades starting in the cheapest tenth of Shiller P/E (5.2–9.6) averaged 10.3% a year after inflation; those starting in the most expensive tenth (25.1–46.1) averaged 0.5%, with a best case of 6.3% (Asness 2012, AQR Capital Management white paper, An Old Friend: The Stock Market's Shiller P/E). Asness, a fund manager at AQR, argues the Shiller P/E is meaningful for setting expectations but of 'very limited use for market timing'. A broad test found return predictors, valuation ratios included, performed poorly out of sample and would not have helped investors time the market (Welch & Goyal 2008, Review of Financial Studies). Critics also point to small samples: overlapping 10-year windows contain far fewer independent periods than the table suggests.

How we put this together

We collected research that answers questions people face when they build and run a portfolio: how much to hold in shares and bonds, what bonds do in a crash, how far to diversify abroad, what funds cost, how to rebalance, how property and other assets compare, and what the evidence says about picking individual stocks. The sources are peer-reviewed journals, working papers, regulator explainers and fund-industry and index-provider research, and we note which is which. Each finding is listed with the market and period it covers, the type of evidence and its limits in the Research tab of the Build and run an investing portfolio page, alongside rated books, official guides and tax-advantaged account pages for six countries, free tools and channels. Last checked 18 October 2026.