How do you build a long-term portfolio when the US is no longer a given?

For the long term, the most robust portfolio is one that holds up even if your view on the US turns out to be wrong. As of 30 October 2026, the US makes up 72.8% of the MSCI World index, and ten US companies alone make up 26.6% of it (MSCI).

Yet the world's leading stock market changes: Japan was 40% of the global market at the start of 1989 and is about 6% today (Dimson, Marsh, Staunton). copio, a portfolio analysis app in private beta across Europe, sums up the method in one question: how many independent return engines does your portfolio really have?

This guide makes no forecast and is meant to stay valid as markets move. copio's dated view of the decade ahead is in Was your portfolio built for the last decade?, and the correlation data behind return engines is in What is diversification, really?.

Key takeaways

1. The most robust long-term portfolio is one that holds up even if your view on the US turns out to be wrong. As of 30 October 2026, an ETF tracking the MSCI World is 72.8% invested in the US and 29.8% in technology.
2. No stock market has kept its place as the world's leader by default. The dominant market was London in 1900, Tokyo in 1989 and New York in 2026, and Japan went from 40% of the global market at the start of 1989 to about 6% today.
3. Buying the leader at the top can mean a very long wait, and for a euro investor the dollar is a second bet. The S&P 500 returned -0.95% a year from 2000 to 2009 while emerging markets returned +10.1% a year, and in 2025 a euro investor made 3.4% on US stocks against 17.3% for a dollar investor.
4. Spreading across regions protects against backing the wrong leader, not against a crash, because all the major equity regions form a single return engine. Every major equity region is above copio's expected-correlation threshold of 0.6 with the others, while euro aggregate bonds (0.24 to 0.30 with equities) and gold (-0.18 to 0.05) behave as separate engines.

Is your World ETF really global?

No: as of 30 October 2026, an ETF tracking the MSCI World is 72.8% invested in the US (MSCI). Japan weighs 5.7%, the UK 3.5% and France 2.2% of the same index, the French share being a copio calculation from the MSCI France and MSCI World market values (MSCI France).

The chart shows the five largest countries MSCI names on the same factsheet; France, at 2.2%, is now counted in "Other". The index is weighted by market capitalisation: it buys what has risen most, so its make-up reflects the past decade rather than the next one.

In January 2012 the US made up 52.4% of the same index, the ten largest holdings 10.2% and technology 12.3% (MSCI factsheet, archived). By October 2026 those shares had reached 72.8%, 26.6% and 29.8%. In 2012 the top ten holdings included ExxonMobil, General Electric and Nestlé. In 2026 they are ten US companies, from Apple (5.7%) to Eli Lilly (1.3%). The MSCI ACWI, which adds emerging markets, is still 64.6% invested in the US (MSCI).

Concentration is just as high inside the US market. The ten largest stocks make up 40.6% of the S&P 500's market value, for 36.1% of its earnings (J.P. Morgan, Guide to the Markets). That share peaked at about 27% in 2000, at the height of the dot-com bubble (RBC; J.P. Morgan, Guide to the Markets).

The authors of the Global Investment Returns Yearbook note that by the end of 2025, US market concentration was at its highest for at least 100 years (UBS, 2026). France's market regulator, the AMF, also lists "the strong concentration of indices and performance in a limited number of large technology stocks" among market vulnerabilities (AMF, Cartographie 2026).

Young French investors are directly exposed. More than 1.1 million French people placed at least one ETF order in 2025, 83% more than in 2024 (AMF). In 2025, more than half of French retail ETF transactions came from investors under 35 (AMF, Cartographie 2026). In the first quarter of 2025, ETFs on US indices made up about 40% of retail ETF transactions, and MSCI World ETFs about 20% (AMF, Cartographie 2025). A portfolio split half S&P 500, half World is therefore about 86% invested in the US.

Will today's dominant market still dominate in 20 years?

History offers no such guarantee: the world's leading market has changed three times in 125 years.

The fall can be steep for the former leaders too. The UK went from 24.2% of the world market at the end of 1899 to 3.7% at the start of 2024, and France from 11.2% to 2.8% (Dimson, Marsh, Staunton).

In 1989, the safe bet was Japan

On 29 December 1989, the Nikkei 225 index closed at 38,915.87. It only closed above that record on 22 February 2024, at 39,098.68, 34 years later (Bloomberg). Even with dividends reinvested, investors had to wait until February 2021, 31 years (Nikkei Inc.). In the late 1980s, the Japanese market traded at up to 70.9 times earnings (French and Poterba, 1991). The arguments of the time sound familiar: a dominant technology economy, global champions, and a market that "deserved" its premium. The Asia article follows Japan's recovery and the reforms since.

In 1900, the safe bet was London

At the end of 1899, the UK made up 24.2% of the world market, ahead of the US (14.5%), Germany (12.6%) and France (11.2%) (Dimson, Marsh, Staunton). Russia made up 5.9%, and its investors lost everything after 1917. The US then delivered the highest common-currency real return of any Yearbook country (UBS Global Investment Returns Yearbook 2024), 6.6% a year from 1900 to 2025 (UBS, 2026). Nobody could have known that in 1900.

The Yearbook draws the lesson itself: the global record of returns reveals "the US to be the exception and not the rule" (UBS Global Investment Returns Yearbook 2024). A study of 39 developed countries since 1841 puts the probability that a stock investor loses money in real terms over 30 years at 12.1%, against 1.2% if you only look at US history (Anarkulova, Cederburg and O'Doherty, 2022). Judging equity risk on the US example alone is like judging the lottery on its winner alone.

What happens when you buy the leader at the top?

You can wait a very long time, as the "lost decade" of 2000 to 2009 showed. The S&P 500, with dividends reinvested, returned -0.95% a year over those ten years, or about -9% in total (Willenbrock, 2011). After inflation the loss reached 29%, and it took until May 2013 to regain the 2000 peak in real terms (Shiller data).

Over the same decade, other engines took over.

The Nasdaq Composite lost 77.9% between March 2000 and October 2002, then took fifteen years to regain its record, in April 2015 (FRED). Cisco was the world's most valuable company on 27 March 2000, at $569 billion (Cisco). Its share price did not close above its March 2000 record of $80.06 until 10 December 2025, at $80.25, more than 25 years later (Dow Jones Market Data). Cisco did not go bust: the business kept growing, but it had been bought at too high a price.

Are US stocks expensive today?

On the most widely followed long-term measure, yes. Shiller's CAPE, which compares price with average earnings over the past ten years, stood at 40.1 on 30 October 2026 (multpl, Shiller data). The only other period above 40 since 1871 is 1999-2000. CAPE says nothing about the next year, but from 1926 to 2011 it explained about 43% of the variation in ten-year real US returns (Vanguard, 2012).

The nuance matters. On expected earnings, the S&P 500 trades at 19.0 times, against 25.2 times in March 2000 (J.P. Morgan). On MSCI's measure, US equities traded at 18.7 times expected earnings on 30 October 2026, against 14.1 times for Europe, 14.6 times for Japan and 8.7 times for emerging markets (MSCI). A high valuation lowers the expected return without dating the correction.

Why does the dollar matter for a euro investor?

Because buying US stocks also means buying dollars, and that second bet can cancel out the first. In 2025 the euro rose from $1.0389 to $1.1750, a gain of 13.1% (ECB). The MSCI World returned +21.1% in dollars, but +6.8% in euros (MSCI, USD; MSCI, EUR). The MSCI USA returned +17.3% in dollars, but +3.4% in euros, while MSCI Europe returned +19.4% in euros (MSCI USA, USD; MSCI USA, EUR; MSCI Europe, EUR).

The effect reversed in 2026: by the end of October, the MSCI World was up 16.7% in euros against 10.9% in dollars (MSCI, EUR; MSCI, USD). Over 27 years the exchange rate has almost cancelled out: $1.1789 per euro at the 1999 launch, $1.1205 on 22 October 2026 (ECB). In between, it swung from $0.8252 to $1.5990. Currency matters little over 30 years and a lot in a single year: it is a risk to understand before you live through a bad one.

So should you bet against the US?

No: swapping one bet for the opposite bet solves nothing, and the forecasters themselves disagree.

Vanguard puts US equities slightly below the rest of the developed world (Vanguard, 30 September 2026). J.P. Morgan raised its US forecast from 6.7% to 7.5% and now sees almost no gap (J.P. Morgan LTCMA 2027). The market assumptions copio uses place the major regions between 6.8% and 8.6% a year, with volatility of 15 to 19% a year (copio CMA, Q4 2026). A gap of one or two points between regions is far smaller than what a single year can move.

The conclusion fits in one sentence: don't try to predict the next decade, build a portfolio that survives being wrong about it.

What is a return engine?

A return engine is a group of investments that rise and fall for the same reasons. copio's portfolio checker draws the line at an expected correlation of 0.6: two asset classes above it count as one engine. copio then measures how much of a portfolio's expected growth survives if its main engine stalls.

Applied to global equities, the measure gives a surprising result.

Every major equity region is above the 0.6 threshold with the others: together they form a single engine. Euro aggregate bonds (0.24 to 0.30 with equities) and gold (-0.18 to 0.05) behave as separate engines.

This distinction separates two problems that are often confused:

1. Leadership risk: betting the whole decade on one region, one sector or ten companies. Geographic diversification reduces it.
2. Crash risk: in 2008 and in 2020, every equity region fell together. Geographic diversification does not eliminate it. Other asset classes can cushion losses, but correlations can rise and no mix guarantees protection.

Cutting your US share deals with the first problem and leaves the second untouched. A 100% equity portfolio spread across four continents is still a single-engine portfolio. The diversification study shows how equity correlations rose in each crisis since 2000.

Why keep a broad core rather than pick stocks?

Because most stock market wealth comes from a tiny number of companies that are almost impossible to identify in advance. Hendrik Bessembinder studied 29,754 US stocks from 1926 to 2025: the median stock's lifetime return was -6.9% (Bessembinder, 2026). Just over half of those stocks (51.8%) lost money over their listed life, 58.8% did worse than one-month Treasury bills (Hedge Fund Alpha, June 2026), and just 46 companies account for half of the $91 trillion of wealth created.

The finding holds outside the US. Across 64,000 global stocks from 1990 to 2020, 57.4% of non-US stocks did worse than one-month Treasury bills (Bessembinder, Chen, Choi and Wei, 2023). Outside the US, 1.41% of companies created all of the net wealth.

A broad index fund owns the whole distribution, rare winners included. Picking stocks is a bet that you will find those winners before everyone else. That bet is legitimate when you have a real edge: persistent, explainable, and measurable after costs and taxes. Without one, the market return at the lowest cost remains the best option known, and convictions belong in a capped satellite pocket whose total loss you can calculate and survive.

How much should you stake on a conviction?

Less than you think, because losses weigh more than gains. A 50% fall followed by a 50% rise leaves a portfolio down 25% (Willenbrock, 2011). An investment's compound return is roughly its average return minus half its variance (Booth and Fama, 1992). At an average return of 10% a year, volatility of 20% brings the compound return down to about 8%, and volatility of 40% to about 2%.

That is the idea behind the Kelly criterion, published by the mathematician John Kelly in 1956 (Kelly, 1956). It calculates the stake that maximises the long-run growth of capital on a favourable bet. Beyond that stake, growth falls; at twice the Kelly stake it drops back to the level of a risk-free investment, and beyond that it turns negative (MacLean, Thorp and Ziemba). Overweighting the favourite, even when it really is the favourite, can therefore reduce final wealth.

Why sell part of your winners?

Because a portfolio that is never rebalanced turns into a bet on whatever has risen most. Vanguard simulated a global portfolio of 50% equities and 50% bonds from 1926 to 2014 (Vanguard, 2015). Without rebalancing it drifted up to 97% equities, with volatility of 13.2% a year against 9.9% for the version rebalanced annually. Frequency matters little: monthly, quarterly and annual rebalancing give similar results, and Vanguard suggests a review once or twice a year with a drift threshold of about 5 points.

Rebalancing can also add return when leadership rotates. From 2000 to 2009, a portfolio split half S&P 500, half long US Treasuries, rebalanced annually, returned 4.44% a year, of which 1.12 points came from rebalancing alone (Willenbrock, 2011). Selling what has risen to buy what has fallen is the only rule that forces you to buy low without predicting anything.

copio's six principles for a long-term portfolio

copio's approach to a long-term portfolio rests on six principles. Each one connects to a section of this guide.

1. Survive first. Ruin is the only irreversible outcome: an emergency reserve, no leverage on the core of the portfolio, capped positions.
2. Black swan awareness. Returns have fat tails, and crashes hit every equity region at once.
3. Geometric return. Only the compound return you live through counts, after volatility drag.
4. The real edge. Concentrating only makes sense with a persistent, explainable edge; without one, you take the market.
5. Beta over fake alpha. Without a real edge, the market return at the lowest cost beats most attempts to beat it.
6. Wrapper awareness. PEA, assurance-vie or brokerage account: the wrapper sets the after-tax return as much as the investments inside it.

Five questions to review your portfolio

1. What is your real share of US stocks, once your "World" ETFs are broken down line by line?
2. How many independent return engines do you hold, and how much of your expected growth survives if the main one stalls?
3. How much of your portfolio depends on ten companies?
4. How much is exposed to the dollar, and did you choose it?
5. Have you written down your rebalancing rule before you need it?

copio's portfolio checker runs these readings on your own holdings, including the number of return engines and the diversification grade.

This article is general information published by copio. It is neither investment advice nor a personal recommendation. Past performance is not a guide to future performance, and any investment in equities carries a risk of capital loss. Data as of 30 October 2026 for MSCI indices, including the country chart, 30 October 2026 for CAPE and 22 October 2026 for exchange rates.

Data checked on 30 October 2026, and on 5 November 2026 for MSCI; return assumptions and bond correlations corrected on 1 November 2026 from the retained inputs. The bond figures refer to euro aggregate bonds, including government and corporate debt, rather than a government-only index.