Diversification, explained with one falling holding

Diversification is the only thing in this subject that is close to free. Almost everything else is a trade: more expected growth for more volatility, more certainty for less growth. Spreading money across different things reduces one particular kind of damage without demanding much in return, which is why it is the nearest thing to a consensus that exists here.

But it is routinely oversold, and it is worth being precise about what it does, because a beginner who believes it prevents losses will conclude it failed at the first bad year.

One holding, four portfolios

Take a single holding that falls 60 percent. That happens to real companies for entirely ordinary reasons: a lost contract, a regulatory change, an accounting problem, a competitor. Now vary only how much of the portfolio it represented.

Nothing about the company changed across those rows. The same event happened, with the same severity, and the consequence for the household ranged from ruinous to forgettable. That is the entire mechanism, and it is why the arithmetic beats the proverb.

Note what changed and what did not. The loss on that holding is 60 percent in every row. Diversification did not make the holding fall less. It made the fall matter less, which is a different and more modest claim.

What diversification actually removes

There are two broad kinds of thing that can go wrong, and diversification addresses exactly one of them.

• Risk specific to one holding. A factory fire, a fraud, a failed product, a lawsuit. These are largely independent of each other, and holding many things means no single one of them can decide your outcome. This is the risk diversification removes, and it removes it well.
• Risk that affects everything at once. A broad economic contraction, a rate shock, a shift in the general appetite for risk. Owning fifty companies instead of five does very little here, because they are all being pushed by the same force.

This is why a diversified portfolio still falls hard in a bad year, and why that is not evidence of failure. It was never insurance against markets. It was insurance against being wrong about one thing in particular.

The four dimensions worth spreading across

Holding twenty companies in the same industry, in the same country, at the same moment, is one bet with twenty names on it. Real spreading happens along several axes at once.

1. Across companies. The most obvious, and the one a broad fund handles automatically. Twenty is dramatically better than two.
2. Across industries. Companies in the same sector face the same regulation, the same input costs and the same demand cycle. Their fortunes move together more than their names suggest.
3. Across countries. A single country can underperform for a decade for reasons that have nothing to do with the companies inside it.
4. Across types of holding. Shares and bonds are claims of fundamentally different kinds, and they respond to different pressures. This is the axis that does the most to control how violently a portfolio moves.

There is a fifth axis that is rarely listed because it is not about holdings at all: spreading across time. Contributing the same amount every month means no single purchase date determines your average price, which is the whole point of dollar cost averaging.

Where it stops helping

The benefit arrives fast and then flattens. Going from one holding to ten removes the overwhelming majority of the single holding risk, as the first table shows: at a 10 percent weight, a 60 percent collapse costs the portfolio 6 percent. Going from thirty holdings to three hundred changes very little, because the remaining risk is the kind that moves everything together.

Two practical consequences follow. First, a beginner holding one broad fund is already diversified in the way that matters most, and adding four more broad funds that hold much the same companies is duplication rather than protection. Second, the number of holdings is a poor measure by itself. What matters is whether they respond to different things, which is the idea behind the word correlation.

When it does not help at all

Honesty requires naming the failure modes, because they are exactly the moments people notice.

• In a broad fall, things move together. Holdings that behaved independently for years can decline in unison during a period of general stress, which is when independence would have been most valuable.
• It cannot help with a date. If you need the money in eighteen months, spreading it does not solve the problem, because the problem is the horizon. That money belongs in the saving job.
• It cannot fix cost. Ten expensive funds are still expensive. Charges are subtracted every year regardless of how well spread the holdings are, as shown in index funds and mutual funds, side by side.

What it costs you

There is a price, and pretending otherwise would be dishonest. Diversification removes the possibility of the extraordinary single outcome. Nobody has ever become spectacularly wealthy by holding a broad, well spread portfolio; the largest fortunes come from concentration, usually in a single business the owner also runs.

The trade is that concentration produces the other tail too, far more often, and the other tail is not a smaller balance but a household that has to start again. For money that is meant to be there in thirty years, giving up the spectacular outcome in exchange for removing the ruinous one is not a compromise. It is the point.

The falls used on this page are arithmetic examples chosen to isolate a mechanism, not predictions and not descriptions of any real holding. Nothing here recommends any allocation, holding or product.