2. Concentration: a virtue or a bet in disguise

Risk 5: Building the portfolio

The previous module ended by saying that twenty or thirty positions are enough and that going beyond that is replicating the index by hand. The symmetrical conclusion would be that fewer positions are better still. They are not: they are something else. Concentrating changes the nature of the result, not only its size, and it is worth knowing exactly what is being changed.

The distribution underneath all of this

Hendrik Bessembinder published a paper in 2018 that anybody considering holding few positions ought to read. He analyzed the roughly twenty-six thousand companies that traded on the American market between 1926 and 2016 and found this:

• Only around 42% beat the return on a simple one-month Treasury bill over their entire listed life. Most shares did not beat the most boring asset in existence.
• All the net wealth created by the American stock market over ninety years, tens of trillions of dollars, comes from approximately 4% of the companies. The other 96%, taken together, contributed net creation of nothing at all.
• Fewer than a hundred companies, some 0.36% of the total, account for around half of that wealth creation.
• The most frequent outcome for an individual share across its whole listed life is total loss.

This is not a statistical curiosity: it is the reason the arithmetic of concentration works the way it does. A share's return is not spread around a mean, it is spread with an extremely long right tail and a mode close to zero.

What that shape implies for a portfolio of few names

Suppose that 4% of the stocks in the universe are the ones that generate the result. If you pick twelve names with no real predictive ability, the probability of not holding a single one of them is approximately 61%. With twenty names it is still 44%.

Out of that comes the most important property of a concentrated portfolio: its mean resembles the index's, but its median sits clearly below it. Most concentrated portfolios return less than the market, and a few crush it. The average holds up because those few drag it along. If you turn out to be the median case, and by definition that is the most likely thing, concentrating will have cost you dearly.

The same asymmetry explains why the concentration stories told in public always end well: the ones that end badly are not told, they are wound up.

Concentrating out of conviction and concentrating out of carelessness

A defensible concentrated portfolio meets three conditions at once, and all three are demanding:

• An edge you can articulate. You know something the price does not reflect, and you can explain what it is and why the market has not incorporated it yet. If the answer is 'it is an excellent company', everybody knows that and it is already in the price.
• The ability to hold on. A 25% position that falls 60% takes 15 points off your portfolio in one go. The question is not whether you can bear it on a spreadsheet, but whether you will keep the thesis when you are eighteen months into losing and the press is telling you daily that you are wrong.
• A compatible horizon and liquidity. If the money may be needed before the thesis matures, conviction is irrelevant.

There is an implicit fourth condition, and it is the one most often broken: that the size of the position should come out of those three and not out of how good the explanation sounds when you say it out loud. Conviction is a state of mind and cannot be measured. The weight in the portfolio can be measured, and it is the only thing the market is going to charge you for.

Against all that, most concentrated portfolios in existence are concentrated for none of those reasons. They are concentrated by drift: a position rose a great deal and nobody trimmed it, until an initial 8% had become 40%. Or by overlap: fifteen names that look different and are the same sector bet. The first is concentration by inertia; the second, concentration without knowing it. Neither comes from a decision.

The arithmetic of contribution, in both directions

Take an index weighted by market capitalization in which the ten largest positions carry 35%. If those ten rise 40% and the rest 4%, the index rises 16.6%: 14 points come from ten companies and 2.6 from the other several hundred. Put another way, 84% of the result was produced by 2% of the names.

That arithmetic works just as well downwards, and that is the point people forget. The same block that contributes 14 points in a good year subtracts a similar amount in a bad one, because its weight does not change sign. Concentrating in the winners of the last cycle is not identifying quality: it is accepting an exposure that amplifies in both directions.

2022 and 2023: the same handful of stocks, two opposite years

The contrast is recent and it comes with figures. In 2023 the S&P 500 delivered a total return of close to 26%, while the same index equally weighted came in near 14%. Twelve points of difference between two portfolios holding exactly the same five hundred companies: the whole gap is explained by the weights, not by the selection. A small group of large technology companies contributed something like two thirds of the index's rise.

The year before, those same companies had done the opposite. With the S&P 500 falling around 18%, several of them lost between 50% and 65% over twelve months. Whoever was concentrated in them lived through both extremes twenty-four months apart, without their thesis having changed by so much as a comma. That is the signature of concentration: it does not improve the quality of the calls you get right, it multiplies their effect.