Diversification: the only free lunch

There's a line every professional investor knows, usually credited to the Nobel laureate Harry Markowitz: diversification is the only free lunch in investing. It's a strange thing to say about markets, where almost nothing is free — more return normally demands more risk. Diversification is the rare exception, and it's worth understanding exactly why.

The lunch, and why it's free

Markowitz's insight, which became the backbone of Modern Portfolio Theory, was mathematical. Hold a single stock and you carry two kinds of risk: the risk of the whole market moving, and the risk specific to that one company — a fire, a fraud, a failed product. The market pays you, on average, for bearing market risk. It does not pay you for the company-specific kind, because you could have erased it for free by simply holding more names.

That's the free lunch: diversification removes a risk you were never being compensated for. Spread money across enough assets that move differently and the company-specific shocks begin to cancel — one firm's disaster offsetting another's surprise — while expected return stays intact. Lower risk, same reward. In a world where you usually pay for everything, that combination is close to magic.

The key word is differently. Two stocks that rise and fall in lockstep give you nothing; two that zig and zag on their own schedules smooth each other out. And here the maths does something quietly remarkable. A portfolio's expected return is just the weighted average of its holdings — blend a 6% asset and a 10% asset evenly and you expect 8%. Its risk refuses to average that way. A portfolio's volatility depends not only on how much each asset swings, but on how much they swing together; when they don't move in perfect step, part of the risk cancels in the arithmetic, and the portfolio's volatility lands below the weighted average of its parts. Returns add up; risk under-adds. That asymmetry is the whole free lunch.

And it strengthens with numbers. Spread evenly across many genuinely independent holdings and the company-specific risk falls roughly with the square root of how many you hold — a hundred such names carry only about a tenth of the idiosyncratic risk of one — while the return you expect doesn't drop by a cent. You shed risk without paying for it; that is what "free" actually means here.

A small, almost unbelievable example

A professor's favourite demonstration makes the point better than any formula. Picture two wildly volatile stocks, each of which ends a two-year stretch exactly where it began:

• Stock A doubles in year one (+100%), then halves in year two (−50%): $100 → $200 → $100.
• Stock B halves in year one (−50%), then doubles in year two (+100%): $100 → $50 → $100.

Hold either one on its own and you have made precisely nothing. Now split $100 between them and do one boring thing — rebalance back to 50/50 at the end of year one:

• Start: $50 in A, $50 in B.
• End of year one: A has grown to $100, B has shrunk to $25 — $125 in total. Rebalance to even: $62.50 each.
• End of year two: A halves to $31.25, B doubles to $125 — $156.25 in total.

Two assets that each went nowhere, combined into a portfolio that grew about 56%. Nothing was added but a rule to trim the winner and top up the loser. That is diversification and rebalancing turning raw volatility — normally a drag on returns — into return itself.

Held alone, each stock's big up year was cancelled by its big down year. Held together and rebalanced, those same swings became the engine.

The opposite instinct: the one big bet

Now look at the other end of the spectrum. Open any corner of the internet where people trade — the WallStreetBets crowd is the loud example — and you'll find the opposite philosophy: everything on one name, often with leverage, swinging for the life-changing score.

The appeal is obvious. Concentration is how fortunes are made: the person who put everything into the right stock at the right moment looks like a genius. What the screenshots never show is the denominator. For every concentrated winner posted online, a crowd of identical bets quietly went to zero and posted nothing. We see the survivors and infer skill; we never meet the failures, because they don't tweet.

It's a trap dressed up as ambition. Concentration is at once how people get rich and how people go broke — and you don't get to know in advance which one you'll be. Diversification gives up the bragging-rights jackpot in exchange for something duller and far more durable: staying in the game.

Is the lunch really free?

Honest answer: free of cost, not free of discomfort. Markowitz's lunch comes with a bill paid in feelings rather than dollars.

Because your holdings move at different times, something in a diversified portfolio is always lagging. You will never get the clean satisfaction of being all-in on the year's best performer; the explosive highs of one asset are forever tempered by the quiet middle of another. And by design you give up the tails — diversification protects you from catastrophe, but it also caps the fantasy of concentrated, generational outperformance. For most people that is a trade well worth making. It simply doesn't feel free in the moment.

Where this meets our work

Diversification is central to how we think — but the word gets used too loosely, and that's where we try to be careful.

It's easy to pull a correlation matrix, find two things that historically zigged when the other zagged, and call it diversification. The problem is that correlations measured from the past are not promises. Relationships that exist by coincidence tend to break exactly when you need them most — in a genuine panic, scattered assets that looked unrelated can fall together, their diversification evaporating at the worst possible moment.

So we try not to stop at the number. We want to understand why two return streams are independent — what different economic force drives each one — because a difference you can explain is far more likely to survive a crisis than one that merely showed up in the data. Real diversification isn't a low correlation; it's a low correlation with a reason behind it.

That's the free lunch worth ordering — and the discipline to make sure it's actually on the menu.