Why is diversification called the only free lunch in finance?
Markets are relentless about charging for benefits. Want higher expected return? Accept higher risk. Want safety? Accept lower expected return. Liquidity, certainty, simplicity — each has a price. Diversification is the one documented exception, and the phrase attached to it — widely credited to Harry Markowitz, whose 1952 paper on portfolio selection led to a Nobel Prize in 1990 — is "the only free lunch in finance."
What "free" precisely means
It does not mean risk-free, costless or guaranteed. It means something narrower and stronger:
You can lower a portfolio's volatility without lowering its expected return.
That is it. And it follows directly from Unit 1's two arithmetics. Expected return is a weighted average — it depends only on the holdings' expected returns and the weights. Risk depends on those plus correlation. So correlation is a dial that moves one side of the trade-off without touching the other. Every other decision in investing moves both.
The same numbers, one more time
Two holdings, each with 20% volatility and each with the same 8% expected return, split 50/50, correlation zero:
- Expected return: 8% — unchanged.
- Volatility: 14.1% — down from 20%.
Return per unit of risk went from 8 ÷ 20 = 0.40 to 8 ÷ 14.1 = 0.57. That is roughly a 42% improvement in the risk-adjusted result, paid for with nothing but the decision to hold two different things instead of one.
Where the lunch is actually served
Two footnotes, so the phrase doesn't harden into a slogan.
Correlations below +1 are the whole meal. At ρ = 1 there is no lunch — the arithmetic returns exactly the weighted average. The benefit is proportional to how different the holdings are, which is why Unit 2's overlap hunting matters more than any holding count.
Costs are real. Spreading across more instruments can carry transaction costs, fund fees, tax friction and attention. The lunch is free in the risk-return arithmetic; the restaurant still charges for the table. Where those costs are significant, part of the benefit is eaten before it reaches the plate.
What it does not promise
Diversification improves the distribution of outcomes; it does not deliver a specific one. A diversified portfolio can still fall a long way — it will fall with the market, which is precisely the risk Unit 4 shows it cannot remove. And in any single year, a concentrated portfolio may well outperform it. This is a claim about the average across many possible futures, not a guarantee about the one that happens.
Try it now
- Compute a rough return-per-unit-of-risk for the single holding below — five years of daily closes. Read the start-to-end change as the return and the width of the swings as the risk, then divide one by the other.
- Do the same for a broad basket over the same five years. Which ratio is larger, and by how much? The basket contains that company and hundreds of others; nothing was forecast to get the difference.
- Say the precise version rather than the slogan: diversification lowers risk without lowering expected return — when the holdings are genuinely different, and net of what it costs to hold them.