Risk, return and diversification

The answer

Risk is not the same as volatility. Volatility is how much the price moves. Risk is the chance of a permanent loss you cannot recover from.

Diversification reduces one kind of risk and not the other. It removes the risk of a single company failing. It does nothing about the risk of the whole market falling, and nothing at all about the risk of you selling at the bottom.

Why this costs you money

Almost every retail portfolio contains a specific illusion.

You own 14 stocks. That feels diversified. But 6 of them are banks and financial companies, 3 more are companies that lend to consumers, and 2 are property developers. You do not own 14 different bets. You own 1 bet on Indian interest rates and consumer credit, expressed 11 different ways, plus 3 other stocks.

When that bet goes wrong, all 11 fall in the same week. And they fall together precisely when you most want one of them to hold up, because you may need the money.

The number of holdings created a feeling of safety while the correlation between them — the tendency to move together — removed it. Correlation is the word that matters and it is the one nobody checks.

There is a second version of the same mistake, and it is more expensive. You own 40 stocks, which really is spread out, but you have never calculated that your largest 3 positions are 55% of the money. The other 37 are decoration. Your result over the next 5 years will be decided almost entirely by 3 companies, and you have been telling yourself for years that you are diversified.

Both mistakes come from counting names instead of measuring exposure.

How it works

Start with the trade-off, because it is real and there is no way around it.

Higher expected return requires accepting a wider range of outcomes. Anything offering a higher return with the same range of outcomes is either mispriced, which is rare and temporary, or you have not found the risk yet. In most cases you have not found the risk yet.

Now the useful part. The total risk in a single stock has 2 components.

1. Specific risk. Things that can go wrong at this company. A factory fire, a fraud, a failed product, a regulator, a bad chief executive. This risk can be removed by owning other companies, because these events are largely independent of each other.
2. Market risk. Things that go wrong for everybody at once. A recession, a large rise in interest rates, a war, a pandemic. Owning more stocks does not remove this, because it is the thing they all share.

Diversification is the operation that removes the first and leaves the second. That is its entire function. It is not a way to increase return, and any description of it as free is wrong: you give up the chance of an extreme good outcome in exchange for removing the chance of an extreme bad one.

How many stocks does it take? The research is old and consistent. Evans and Archer, in 1968, found that most of the reduction in portfolio variability happens within the first 10 to 15 stocks, with little improvement after that. Statman, in 1987, argued the number was higher, around 30, once the cost of diversifying was included. Later work found the number rises when individual stocks are more volatile.

The important point is the shape of the curve, not the exact number. Going from 1 stock to 10 removes a large amount of specific risk. Going from 10 to 30 removes some more. Going from 30 to 60 removes almost nothing and doubles the work.

But all of that assumes the stocks are independent. They usually are not. Twenty companies in one sector, one country and one currency behave like far fewer than 20 independent bets. This is why the count of holdings is the wrong measurement and the count of distinct exposures is the right one.

And there is a final property that is inconvenient. Correlations rise in a crisis. In calm markets, sectors move apart from each other and diversification works. In a severe fall, most things move down together, because the people selling are selling everything they can sell rather than everything they dislike. Diversification is at its weakest exactly when you are testing it.

What it tells you, and what it does not

Diversification tells you that no single company failure can end your investing. That is a large benefit and it is the reason to do it.

It does not tell you that you will avoid a large fall. A fully diversified equity portfolio fell by roughly a third in a month in early 2020, and by more than half in 2008. Diversification is protection against ruin from 1 name, not protection against a bad market.

It also does not fix a bad decision about how much equity to own in the first place. If you will need the money in 18 months, the correct amount of stock market risk is low, and no amount of diversification within the stock market changes that. Matching risk to the date you need the money is a bigger decision than which stocks you pick.

And diversification has a cost that is rarely stated. Beyond a certain number of holdings you cannot follow them. You end up owning companies you no longer understand, which reintroduces exactly the risk you were trying to remove, in a form you cannot see.

The decision rule

Count exposures, not holdings.

1. Add up the weight of your largest 3 positions. If it is above 40%, your

result is decided by 3 companies, whatever else you own.

1. Add up the weight of your largest sector. If it is above 35%, you own a

sector view, not a portfolio.

1. Ask what single event would hurt 5 or more of your holdings on the same day.

If you can name it in under 10 seconds, that event is your real position.

Then compare all 3 numbers with the index you would otherwise have bought. The index is not a target. It is the benchmark you must beat to justify the work.

Try this now

Five minutes, on your own portfolio, and most people are surprised by step 4.

1. Open your holdings page. Note the current value of each holding and the total.
2. Divide each holding's value by the total and multiply by 100. That is its weight in percent.
3. Sort the list by weight. Add up the top 3. Write the number down.
4. Tag each holding with its sector: banking and finance, technology, energy, consumer, healthcare, industrials, other. Add up the weight of the largest sector.
5. Look up the same 2 numbers for the index you would otherwise have bought. Most index provider websites publish the top-10 constituents with weights and the sector breakdown, free.

What you should see. Three things.

Your top-3 weight is almost always higher than you expected, because a position that has doubled has grown its weight without you buying anything. That is how concentration arrives: not by decision, but by success.

Your largest sector is usually larger than you expected, because sector labels are not how you think about stocks when you buy them.

The comparison with the index is the point of the exercise. For reference, the top 10 companies are roughly 38% of the S&P 500, and the Nifty 50 has a heavy weight in financial services. If your own concentration is far higher than the index's, that is a choice, and you should be able to say what you are being paid for taking it.

Three real cases

1. Long-Term Capital Management, September 1998 (United States) — many positions, one bet The fund was run by John Meriwether with a board including Myron Scholes and Robert Merton, who had won the Nobel Prize in economics in 1997. It held thousands of positions across many countries and instruments, and by its own models it was extremely diversified. At the start of 1998 it had about $4.7 billion of equity against roughly $124.5 billion of borrowed funds. Almost all of the positions turned out to be the same bet: that the price gap between safer and riskier assets would narrow. After the Russian default in August 1998 that gap widened everywhere at once. The fund lost about $4.6 billion in under 4 months, and on 23 September 1998 the Federal Reserve Bank of New York organised a recapitalisation of about $3.625 billion by 14 financial institutions. The lesson is not that leverage is dangerous, though it is. The lesson is that the number of positions told them nothing about the number of bets.

2. The fall of February and March 2020 (both countries) — correlations go to 1 The S&P 500 peaked on 19 February 2020 and reached its low on 23 March 2020, about 33 trading days later, falling roughly a third. Indian indices fell further and reached their low on 24 March 2020. Sectors that normally behave differently fell together. Gold fell for several days in the middle of the panic, because investors sold what they could sell to raise cash. Both markets then recovered: the S&P 500 returned to its February high in August 2020, and Indian indices during November 2020. Diversification did not stop the fall. It did mean that almost every diversified holder who did nothing was whole again within a year.

3. The Nikkei 225, 29 December 1989 to 22 February 2024 (Japan) — diversified inside one country The Nikkei 225 closed at 38,915.87 on 29 December 1989. It did not close above that level again until 22 February 2024, more than 34 years later, when it closed at 39,098.68. An investor who held every large company in Japan was completely diversified across Japanese companies and completely undiversified across countries. Owning 225 stocks was no protection at all, because the thing that went wrong was the price of Japanese equities as a whole. This is the clearest demonstration available that diversification is about distinct exposures, not about counting names.

The question that resolves it

A novice looks at a portfolio and asks: how many stocks do I own?

An expert asks: how many different ways can I be wrong, and what is the largest single one?

A portfolio of 25 stocks with 3 distinct exposures is less diversified than a portfolio of 8 stocks with 8 distinct exposures. Only the second question sees this.

What would make this wrong

If diversification removed risk in general, then diversified portfolios would not have fallen sharply in 2008 or in March 2020. They did, everywhere, by similar amounts. That is the evidence that diversification removes company risk and not market risk, and any description of it that promises more than that is wrong.

The honest limits.

Diversification is not free in the way it is usually described. Spreading across 30 companies guarantees you will own some poor ones. Somebody who genuinely can analyse companies well is paying a real price for the safety. This article does not settle the argument between concentration and diversification; it insists that whichever side you take, you measure it.

Historical correlation is also a poor guide to future correlation, and it is worst exactly when it matters most. Assets that behaved independently for 10 years can move together for the 3 months that decide your decade.

And diversification cannot fix leverage. A borrowed position can be closed by somebody else at the worst moment regardless of how many other things you own. That is what happened to Long-Term Capital Management, and it is why the next article in this cluster is about size.

Carry this

• Risk is permanent loss. Volatility is movement. They are not the same thing.
• Diversification removes company risk. It does not remove market risk and cannot.
• Count exposures, not holdings. Top-3 weight and largest sector weight are the 2 numbers.
• Correlations rise in a crisis, so measure your concentration before you need it to be low.

Knowledge check

Related

• Position sizing, and how many stocks to own
• Analysing a company, end to end
• The kill switch: market psychology and your own mind
• ← Becoming a confident investor