The Only Free Lunch: Why Correlation Pays You
Key takeaway
- Risk is a property of the portfolio, not of each holding. A volatile asset can lower total risk if it does not move with the others.
- For any level of risk there is a best available mix. Most real portfolios sit below that line — taking full risk for less than the available return.
- Diversification works fast and then stops. Past roughly 15–20 uncorrelated holdings you have removed what can be removed; the rest is market risk.
Learning pathThe index investor's whole jobStep 9 of 12
Read before this:Compounding Is Boring for a Long Time, Then It Isn't
Based on Portfolio Selection — Harry Markowitz, 1952
The result that sounds wrong
Two assets, each of which swings violently, can combine into something that barely moves.
Harry Markowitz's 1952 paper is the origin of the idea that a portfolio should be assessed as a whole rather than as a list of individually-judged holdings. Before it, portfolio construction largely meant picking good securities one at a time; afterwards, the interaction between them became the main event.
The word for it is the reason it matters: this is the only benefit in investing available without a prediction. Beating the market requires being right about something. This requires only that your holdings do not all fall for the same reason.
What the efficient frontier actually says
The practical reading is not the curve itself — you will never compute your own — but the space beneath it. Being under the line is not a mild inefficiency; it means you are being paid less than the market was willing to pay for the risk you took.
Where diversification stops helping
The first ten holdings do most of the work. The hundredth does almost none.
Two conclusions follow, and they point in opposite directions from where beginners usually land. First, holding three stocks is not diversified, however good they are. Second, holding sixty is not meaningfully safer than holding twenty, and it costs you the ability to know anything about what you own.
The floor also explains why a broad index fund cannot protect you in a general market fall. It was never supposed to. It removes the risk of being wrong about one company, and leaves the risk of being invested at all — which is the risk you are paid for.
The catch: correlation is not a constant
This is the part the theory handles least well in practice, and it is the part that matters most in the weeks when you need it.
- Correlations rise in crashes. Assets that spent a decade moving independently often fall together in the same month, exactly when the diversification was supposed to pay.
- Historical correlation is backward-looking. The number you optimised on came from a period whose conditions may not repeat.
- Apparent diversification is not diversification. Twelve technology stocks across four countries is one bet, wearing twelve names.
None of this makes the idea wrong — it makes precision misplaced. Optimising to two decimal places on estimated correlations is false comfort; holding genuinely different kinds of things, and not too few of them, captures nearly all of the benefit and survives the estimates being off. Home bias and global diversification covers what "genuinely different" tends to mean in practice.
Try this week
- List your holdings and group them by what would make them fall together — sector, country, currency, interest rates.
- Count the groups, not the holdings. That number is closer to your real diversification.
- Find your largest group. If it is more than half the portfolio, that is your actual bet.
- Check whether anything you own would plausibly rise in a year when equities fall.
Common questions
What is the efficient frontier?
The set of portfolios offering the highest expected return for each level of risk. Portfolios below the line take the same risk for less return, which is the practical point rather than the exact shape of the curve.
What is modern portfolio theory in simple terms?
The idea that a portfolio should be judged as a whole rather than holding by holding, because how the holdings move relative to each other affects total risk as much as how risky each one is alone.
How many stocks do I need to be diversified?
Most of the reducible risk disappears by around fifteen to twenty genuinely different holdings. Beyond that the curve is nearly flat, and what remains is market risk that no amount of diversification removes.
Why is diversification called the only free lunch in investing?
Because it reduces risk without requiring any forecast. Every other route to better risk-adjusted returns depends on being right about something; this one depends only on your holdings not all falling for the same reason.
Does diversification still work when markets crash?
Less well than the historical numbers suggest. Correlations tend to rise sharply in severe falls, so assets that normally move independently can drop together — which is an argument for owning genuinely different asset types rather than many versions of one.