Diversification: The Only Free Lunch

Two kinds of risk, only one of them paid

Every stock's fate mixes two forces. Market risk (systematic risk): the tide that moves everything, recessions, rate shocks, panics. Company-specific risk (idiosyncratic risk): the fraud, the failed product, the disrupted business model, the lost lawsuit.

The foundational insight of modern portfolio theory, the one that earned Markowitz the "free lunch" line, is that these behave completely differently in a portfolio. Specific risks are largely independent across companies: one firm's accounting scandal doesn't cause another's factory fire. Hold enough of them and the disasters and windfalls offset; the specific risk diversifies away toward zero. Market risk doesn't; everything rides the tide together. And here is the punchline the market builds in: you are only compensated for the risk you can't diversify away. Expected returns pay for bearing the tide, not for volunteering to carry one company's banana peels. Concentrated specific risk is, in expectation, unpaid risk. Removing unpaid risk for free is the lunch.

How brutal single-stock reality actually is

The theory would matter less if individual stocks merely wobbled. The long-run data says something harsher, and it's worth stating with numbers (Bessembinder's landmark studies are the source: the 2018 study of the full US market since 1926, and the global extension covering roughly 64,000 stocks):

• The typical stock is a poor investment. Around half of all US stocks ever listed delivered negative lifetime returns; the median stock underperformed one-month Treasury bills over its life.
• The market's entire net wealth creation above cash came from roughly the top 4% of companies. A few percent of extreme winners, compounding for decades, paid for everything; the majority collectively achieved nothing beyond T-bills.
• Globally the concentration is even starker: in the 64,000-stock global sample (1990–2020), roughly 2.4% of firms account for all net global stock-market wealth creation above bills.

Read those three lines together and the case assembles itself. Stock-picking isn't a game of avoiding a few losers; it's a game of needing to hold a tiny, unidentifiable set of decades-long superwinners, where missing them means trailing cash. The index holds them by construction: whoever the next transformative company is, a cap-weighted broad fund owns it early and rides it all the way up (Chapter 5's self-maintaining machinery). Diversification isn't only defense against blow-ups; it is guaranteed attendance at the winners' compounding.

In brick terms, and in one honest limit

Pull one brick from a thousand-piece build and the structure stands; a one-brick tower has no such luck. But note what the boxed set does not protect against: the shelf collapsing. Diversification within equities removes company risk, sector concentration, single-country risk. It does not remove market risk: in 2008 and March 2020, nearly everything fell together, and correlations between risky assets famously rise exactly when you'd want them low. The tool for the tide is a different one, asset allocation across things that respond to different forces (Chapter 19, next). Keep the division of labor clean: diversification handles "which holdings," allocation handles "how much risk at all."

The two ways ETF holders still get this wrong

1. Fake diversification: many funds, one bet. Holding an S&P 500 ETF, a Nasdaq-100 ETF, an MSCI World ETF and a tech-sector ETF feels like four-fold prudence and is mostly one large, overlapping position in US mega-cap technology. Diversification is measured at the level of underlying exposures, not line items. The look-through test: list your funds' top-10 holdings side by side; if the same names dominate every column, you own one bet in four costumes.

2. Narrow baskets wearing the ETF halo. Chapter 1 said it and it bears repeating with this chapter's vocabulary: a single-theme, single-sector or single-country ETF has diversified away some specific risk (individual companies) while deliberately retaining a concentrated bet (the theme). That can be a legitimate satellite; it is not what this chapter's free lunch refers to, and the word "ETF" on the label confers none of it automatically.

A worked example

Three investors put €50,000 to work in 2015-style conditions and hold for ten years.

• Anna, one stock, a beloved national champion: outcomes historically range from fortune to near-zero; the median such story, per the data above, trails T-bills. Her result is a lottery ticket with a poor median, and no rebalancing or patience fixes the shape.
• Ben, twenty hand-picked stocks: most specific risk is already diversified (the curve in Visual 1 flattens fast), but he holds ~20 slots from a universe where ~4% of names drive everything: decent odds of missing every superwinner, plus home-country and sector tilts he never chose consciously.
• Cara, one all-world ETF (~4,000 stocks): owns every future superwinner at market weight, every failure diluted to irrelevance, both facts guaranteed by construction rather than by skill. Her only remaining risk is the tide, which is the risk she's actually paid to bear, and the one Chapter 19 teaches her to size.

Ben's approach is the interesting middle: not reckless, genuinely diversified in the textbook volatility sense, and still structurally likely to miss the concentrated engine of long-run returns. That subtlety, volatility-diversified yet winner-exposed only by luck, is the modern argument for breadth that the old "hold 20–30 stocks" rule of thumb predates.