4. Diversification Protects Against Ignorance

🧩 Before you read: a problem to solve

You inherit $500,000. A good friend who works at a major technology company gives you impassioned advice: put it all in the five biggest tech stocks. “They’re not going anywhere,” he says. “You’ll make way more than some boring index fund, and those five companies basically are the market anyway — they’re most of the S&P 500’s return.”

You happen to know that 4% of listed stocks have accounted for all net wealth creation in the U.S. market over the past century. Your friend doesn’t know this fact. Does it argue for his strategy, or against it?

“The best-performing 4% of listed firms accounted for the net wealth creation of the entire U.S. stock market.”

Diversification is the most widely endorsed principle in investing — and the most misunderstood. It is not a promise of safety, a guarantee of higher returns, or a magic shield against market crashes. It is a specific response to a specific problem: we do not know which companies, sectors, or countries will produce the returns that matter.

This chapter explains what diversification actually does, what it cannot do, and why broad global equity ownership is the default growth answer — with explicit acknowledgment of its limits.

The skewness problem

Hendrik Bessembinder’s 2018 study of the CRSP universe of U.S. common stocks from 1926–2016 contains the single most important diversification fact most investors never learn:

• The best-performing 4% of listed firms accounted for the net wealth creation of the entire U.S. stock market over the full 91-year period.
• Most individual stocks — 58% of the total — had lifetime buy-and-hold returns below one-month Treasury bills.
• The distribution of individual stock compound returns is massively positively skewed: the median stock underperformed T-bills; the mean was lifted by a tiny fraction of extreme winners.
• Concentration: the top 86 stocks (0.33% of the total) accounted for over 50% of net wealth creation.

This is not a forecast. It is a description of what happened in one market over one long period. But the mechanism — positive skewness in long-horizon individual stock outcomes, driven by the compounding effect of modest return differences over decades — is structural. Most firms fail or stagnate. A few become enormous. The aggregate market return over the long run is driven almost entirely by the extreme right tail.

The consequence for portfolio construction: a concentrated portfolio (10, 30, even 100 stocks selected without perfect foresight) has a non-trivial probability of excluding one or more of the extreme winners that will drive the market’s net return. A broad index captures them all — at the cost of holding many mediocre companies alongside the winners.

The U.S.-specific limitation. The Bessembinder evidence is U.S.-only. The mechanism (skewed lifetime stock returns driven by a small fraction of extreme winners) is likely to transport to other developed equity markets, but the magnitude and concentration parameters may differ. Global corroboration would strengthen the transport claim; its absence is an acknowledged limit, not a fatal flaw. The structural property that broad ownership reduces winner-exclusion risk does not depend on the U.S. institutional context.

Additional diversification evidence

The home-bias puzzle

French and Poterba (1991) document persistent home bias across six major markets over 1975–1989: investors held far more domestic equity than global diversification would justify. They calculate that an investor who weighted countries by market capitalization and hedged FX using three-month forward contracts would have achieved meaningful diversification benefits.

The deeper finding came from Cooper and Kaplanis (1994): they tested whether observable costs — currency hedging, international taxation, capital controls — could explain the magnitude of home bias. They found the costs were too small. The observed bias is far larger than any rational cost-based explanation can justify, pointing to informational frictions, behavioural causes, or perceived (but not actual) foreign-investment risk.

What this means for the framework: the baseline should be global market weights. Even the commonly cited rational grounds for home bias (currency matching, tax, familiarity) are empirically insufficient to justify the magnitudes typically observed. Deviations require named reasons — but those reasons should be scrutinized more carefully than most investors assume.

What diversification does

Diversification performs four distinct jobs:

1. Reduces omission risk. A broad portfolio captures the extreme winners. A concentrated one might not. This is the Bessembinder argument.

2. Reduces idiosyncratic risk. Company-specific events (fraud, product failure, regulatory action, management error) that would devastate a concentrated position become noise in a broad portfolio. This is the classic variance-reduction argument — and it works, within limits.

3. Reduces country and currency concentration. A globally diversified portfolio is less exposed to the policy errors, institutional failures, and currency depreciation of any single country. This is the French–Poterba home-bias argument.

4. Reduces the need to forecast. A diversified portfolio makes fewer implicit bets on which outcomes will occur. It admits ignorance and prices it into the structure.

What diversification does not do

Diversification cannot:

• Prevent systematic losses. When the entire equity market falls — 2008, 2020, 2022 — a diversified equity portfolio falls with it. Diversification within equities does not diversify away equity risk. The ~50% peak-to-trough drawdown of a global equity index in 2008–09 happened to investors holding 8,000+ stocks.

• Guarantee a positive return. A diversified equity portfolio can underperform bills over long periods (Japan post-1989, U.S. 1929–1949, multiple other country cases). Diversification does not eliminate valuation, earnings, or regime risk.

• Eliminate concentration. A global cap-weight index concentrates in the largest countries, sectors, and companies by design. As of mid-2025, the U.S. is ~60% of global free-float market cap. The top 10 companies are a material fraction of the total. This describes the market’s composition; it does not predict a crash — but it also does not deliver equal geographic or sectoral exposure.

• Protect against currency mismatch. If your liabilities are in Turkish lira and your assets are in a global index denominated in USD/EUR/JPY, you have reduced company concentration and added currency concentration. Diversification across assets is not the same as matching assets to liabilities.

The illusion of the large number

Consider two investors at the end of 1989. The first holds 50 U.S. blue-chip stocks — a “focused” portfolio by any conventional measure. The second holds 225 Japanese stocks — the entire Nikkei 225 index — spanning manufacturing, finance, technology, and utilities across the world’s second-largest economy. By the standard “how many stocks do you own?” test, the second investor is dramatically more diversified.

In reality, the second investor has just made the single largest concentrated bet of the era. Every one of those 225 companies is Japanese, denominated in yen, and tied to the same macroeconomic outcomes. Over the next two decades, as Japanese equities lost more than 80% of their value, that “diversified” portfolio was destroyed — not despite its 225 holdings, but because all 225 shared the same country, the same currency, and the same fate.

The number of holdings concealed the concentration. It always does. True diversification is not a count of securities. It is an honest accounting of the common shocks they share.

The diminishing marginal benefit

Critics of broad diversification correctly observe that most of the variance-reduction benefit is captured in the first 30–50 stocks (across sectors). Adding stocks 51 through 9,000 reduces idiosyncratic risk further, but the marginal benefit is small.

This is a valid observation — and it misses the point. The Bessembinder skewness argument is not primarily about variance reduction. It is about avoiding the exclusion of extreme winners. A 50-stock portfolio selected at random from a global universe has a non-trivial probability of missing the 4% of firms that drive net wealth creation. A 9,000-stock index does not. Whether that matters in practice depends on how close the 50-stock selection comes to capturing the market return — and there is no way to know that in advance.

For investors who want fewer stocks, a cap-weight index fund achieves the breadth at no additional operational complexity. The fund does the work.

The global default

The framework’s default growth implementation is broad, low-cost global public equity — not a domestic-only index, not an equal-weighted alternative, not a factor-tilted portfolio. The reference implementations are indices like MSCI ACWI or FTSE Global All Cap: free-float market-cap-weighted, covering developed and emerging markets, large/mid/small where feasible.

This is a conditional default, not a universal law. It is the right starting point for an investor who:

• Has a genuinely long horizon (capital not needed for a decade or more);
• Can bear deep and prolonged loss without panic selling;
• Has no specific liability or currency reason to deviate;
• Wants maximum simplicity and minimum turnover.

For investors who do not meet these conditions, deviations are available — but they are deviations, requiring a named reason, mechanism, and failure mode.

The Japan problem — a story

In December 1989, the Nikkei 225 index hit 38,957. The Imperial Palace grounds in Tokyo were, by some estimates, worth more than all the real estate in California. Japanese companies dominated global market capitalizations: of the world’s top ten companies by market value, seven were Japanese. A global cap-weight investor at that moment would have held approximately 45% of their equity portfolio in Japanese stocks.

It was not irrational at the time. Japan had spent four decades delivering an economic miracle — postwar reconstruction, export-led growth, world-beating manufacturing. The companies were profitable. The market had risen for years. The weight in the global index reflected genuine economic transformation, not a statistical error.

Then, over the next two decades, the Nikkei fell by over 80% from its peak. Japanese equities delivered negative real returns for more than twenty years. An investor who mechanically held global cap-weight through that entire period experienced a severe and prolonged drag from the single largest country weight in their portfolio.

Now consider the alternatives that were available at the time — and the problems with each:

• Underweight Japan because it “looked expensive.” An investor who made this call in 1985 or 1986, when Japan’s weight was already large and valuations were already elevated, would have missed several more years of extraordinary returns before the eventual decline. Timing is hard even when the diagnosis is right.
• Overweight Japan because the growth story was compelling. Many did. They were wiped out.
• Equal-weight countries. This would have reduced the Japan drag — but would have introduced active bets against the U.S. and other markets at other times, with higher turnover and cost. The cure is not free.
• Do nothing; hold global cap-weight and endure. The investor who contributed steadily through the Japan drawdown, reinvested dividends, and held for the full cycle eventually recovered — but only if they did not panic-sell during two decades of negative returns. The behavioural challenge of holding an asset that has underperformed for twenty years is severe.

What the Japan story teaches us. Cap-weight indexing is not a promise of safety or optimality. It is a transparent rule: own the market in proportion to what the market thinks each component is worth. That rule will sometimes concentrate your portfolio in overvalued assets. All alternatives — equal weighting, GDP weighting, factor weighting, valuation-driven timing — replace that transparent rule with an active bet, and each of those alternatives would have had its own Japan moments in different eras.

The framework’s response is not to solve the Japan problem. It is to acknowledge it honestly, and to give investors who want to deviate from cap weight a defined path — with the understanding that deviation is an active bet, not a repair.

The rule

The diversification constraint. Do not make the portfolio depend on identifying a small set of future winners, one country, one employer, or one macro outcome unless a deliberate and supportable edge or liability justifies it.

Global cap-weight equity as the conditional default. Use broad global public equity as the default long-horizon growth building block for capital that can bear deep and prolonged loss. This is a conditional implementation of the diversification constraint, not a separate core principle. Weights, home bias, factor tilts, and currency hedging remain adaptation decisions.

Key idea: Diversification is not a guarantee. It is an admission of ignorance — and ignorance, honestly acknowledged, is a better foundation for portfolio construction than false confidence in identifying future winners.

Diversification protects against ignorance about which assets will succeed. But it does not protect against a more immediate problem: needing money at the wrong time. The next two constraints — liquidity and survival — address the preconditions that must hold before long-horizon diversification can work.