How Diversification Reduces Single-Exposure Risk

Holding more than one thing reduces how much any single one can hurt the household that holds it. Diversification does nothing else. Spreading cannot reduce the risk that everything falls together, and it cannot rescue a household that needs its money on a particular date. The risk most households actually run is the opposite one: everything they hold is the same thing.

Underneath that answer sits one question. What single failure could this household not survive? The idea is simple enough to state in a sentence, and it is almost always taught as though it were a technique for improving what comes back. No such improvement is on offer. Diversification is a way of arranging things so that being wrong about any one of them is survivable, and everything interesting about it follows from asking what one failure would not be. Every rupee below belongs to the Bhosale household, an invented one.

What does diversification actually do?

Take the sentence apart in the order it is built. DiversificationHolding more than one thing so that being wrong about any one of them is survivable. holds more than one thing. Holding more than one is done so that no single one of them decides everything. And the result it delivers is not a larger amount at the end; it is a smaller hole when one particular thing goes wrong.

Notice what is missing from that description. There is no promise about what comes back, no statement that the whole will move in any particular direction, and no claim that anything has become safe. The change is not in the size of what a household holds but in the size of the worst single thing that could happen to it. A reader who was taught this as a way of doing better will keep waiting for the second half of the sentence, and there is no second half.

The everyday version is where the idea is obvious. A tailoring counter takes work from one office building next door and from nowhere else. The counter is not badly run and the work is not poorly done. Every building can close, and this one is no exception. On the day it closes the counter takes nothing. Now picture the same counter with work coming from a school, a wedding shop and two households on the next lane. Nothing about the tailor has improved. The change is that the closing of any one of those four leaves three still walking in.

Single-exposure riskThe risk that one particular thing fails, taken on its own rather than mixed in with everything else. is the name for what the first counter carried and the second one does not. The second tailor may well earn less in a good year than the first. The office building next door was convenient and the school is a walk away. The arrangement was not free and nobody said it was.

Why does holding more than one thing help at all?

Because an event lands on a thing, not on a household. Something happens to a particular business, or a particular metal, or a particular lane, and the size of the hole it makes depends entirely on how much of the household was sitting at that point when it landed.

The hole is what a household actually feels, so follow the arithmetic of the hole rather than the arithmetic of the holding. Suppose everything a household has sits at one point and that point fails completely. The hole is everything. Now suppose the same total is split across five points of equal size and one of them fails completely, with the other four untouched. The hole is one fifth. The event was identical. The failure was total. The only difference was how much of the household happened to be standing where the event landed, and that is the entire mechanism.

The mechanism is almost embarrassingly simple. The difficulty is never in understanding it. The difficulty is that the five points have to be five actual points, and they turn out not to be more easily than anybody expects.

What can diversification not do when everything falls at once?

Now the half of the subject that gets skipped, and it is the half that decides whether a household is disappointed later. Split what a household holds into two parts. The first part is whatever is specific to one holding: the one business that loses a contract, the one metal that goes out of fashion at the jeweller, the one lane that is dug up. Practitioners call that the idiosyncraticSpecific to one thing rather than shared with everything else. Spreading removes the specific part. part, which is a long word for the part that belongs to that one thing and to nothing else.

The second part is whatever is shared. When something happens that touches everything at once, the event is not asking which particular things a household holds. The shared part has a name too, the systematicShared by everything at once rather than specific to one thing. Spreading cannot remove the shared part. part, and it behaves in the exactly opposite way.

The second part is defined as the part every holding has in common, so spreading works on the first part, does not touch the second, and no number of holdings changes that. This is not a matter of degree. The shared part does not shrink slowly and would not shrink further with more holdings. The shared part is what remains after the specific parts have been spread away, and spreading is the operation that produced it.

When something happens that reaches everything at the same time, a household holding many things and a household holding one thing are in the same weather. The one holding many things is not being punished for having spread. No arrangement of holdings sells protection against weather that reaches everything, so none was ever bought.

What can it not do for a household that needs its money on a particular date?

The second thing it cannot do is quieter and it catches more households than the first. Spreading changes what can move the whole. Spreading changes nothing about the calendar.

A date is not a thing that some holdings have and others do not, so a date is not a risk that can be spread. School fees fall in a particular month. A deposit for a rented place is needed on the day the place is taken. A hospital desk asks on the morning it asks. Whatever the whole amounts to on that morning is what is available on that morning, and no arrangement of holdings has anything to say about it.

Spreading reduces how much any single thing can move the whole, and it makes no promise whatever about where the whole happens to be standing on a date somebody else has chosen. These are different questions, and a household that has answered the first one carefully can still be caught flat by the second. The buffer and the protection are covered separately and come first: what a household needs on a fixed date is not a spreading problem at all.

What happens when the holdings move together?

Everything above assumed the five points were five points. In practice that assumption is the one that fails, so remove it.

Two holdings are independentNot moving together. Nothing that happens to one says anything about what is happening to the other. when what happens to one says nothing about the other. Two holdings are correlatedMoving together. When one goes one way the other tends to go the same way. Holding several things then achieves nothing. when they tend to move the same way at the same time. The shape of the consequence is visible without a single number for how much any two things move together.

Watch what happens to the mechanism as the moving-together increases. At the independent end, an event that flattens one holding leaves the other four exactly where they were, and the hole in the whole is one fifth. At the fully in-step end the five were never five separate things, so the event that flattens one flattens all five. The count of holdings did not change between those two ends, and the exposure went from a fifth of the whole to the whole of it.

The corners of that range are worth having in words. At the entirely independent end, the one event reaches one bar of the five and takes a fifth of the whole. A quarter of the way along, it reaches all five but only lightly, and takes two fifths of the whole. Halfway, three fifths. At the entirely in-step end it takes all of it. Five bars stand on the panel at every one of those settings, and the count is the wrong thing to be looking at.

What is the test, if the count is not the test?

Replace counting with a single question, and ask it out loud rather than in the abstract. Name one thing that could actually happen. Then ask whether that one thing would move more than one of these at the same time.

An event of that kind has a name worth knowing. A common causeThe one event that would move several holdings at the same time, whether or not anybody noticed the connection when they were acquired. is any event that reaches several holdings at once, and it does not care whether anybody spotted the connection when the holdings were acquired. The connection does not have to be obvious, and it very often is not. Things acquired at different times, from different people, for different reasons, can still be sitting on one thing happening.

The test is not how many lines there are; it is what single event would move several of them at once, and the count never enters the question. A household that can answer easily has fewer holdings than it thinks. A household that has to work at the answer, naming events and finding that each one reaches only one line, has genuinely got several.

What is concentration, and why is it the risk most households run?

ConcentrationDepending heavily on one thing, whether or not anybody chose to depend on it. is the other side of the same coin, and it is worth stating separately because it is the side that describes most real households. Concentration means depending heavily on one thing. The word carries no suggestion that anybody decided to depend on it, and that is the part that gets missed.

Most concentration is not chosen. Concentration accumulates. Somebody takes a job near where they live and a small trade opens in the same locality because that is where the household already is. Something arrives at a wedding and is never sold because selling it would be odd. A scheme was opened years ago because a relative said to and nothing has been added since. Not one of those was a decision about concentration, and together they can produce a household whose whole position rests on one or two things happening.

The word belongs to exposure rather than to the choosing of things. A household that has never chosen a single holding can still be as concentrated as one that chose badly, and the arithmetic does not distinguish between them. An event does not ask how a position was arrived at. An event arrives, reaches whatever is standing at that point, and makes a hole the size of what was standing there.

Can a household be concentrated while holding nothing anybody would call risky?

Yes, and the answer is worth being slow about.

Ask where a household's money comes from before asking what it holds. For the Bhosale household in year two, Rs 4,77,600/- came from one salary and Rs 52,800/- came from a tailoring counter in a market lane, making Rs 5,30,400/- in all. Two sources. Two entirely different kinds of work. Nobody, looking at that, would use the word concentrated.

The test from two blocks ago applies here. One event: the market lane is dug up for drainage work. Does that one event move more than one of these at the same time? The drainage work closes the counter, and it thins the local economy that the salaried employer sits inside as well. Two lines in the record, one thing underneath them. The Bhosale household's income was concentrated on one local economy the whole time, and the only reason nobody called it that is that neither of the two sources was an investment.

The drainage work took five months. Ashok Bhosale's counter took Rs 52,800/- across year two against Rs 96,000/- in year one, a fall of Rs 43,200/-, and every rupee of that fall came out of a household whose ordinary month costs Rs 42,770/-. Against that stands the gold. Nobody in the household has ever valued the gold against anything, and no movement in it has ever taken Rs 43,200/- out of a year. The concentration that actually cost this household money was in where the money came from, and it was not in anything the household holds.

What does diversification cost, and who pays it?

Spreading is not free, and an account that presented it as costless would be selling something. The price is worth stating plainly.

Spreading gives up the chance of having everything in whatever turns out best. The chance given up is not a small thing and it is not a technicality. If a household could know in advance which single thing would do best, spreading would be straightforwardly the wrong arrangement, and it would be wrong by exactly the amount that thing beat everything else by. The cost of spreading is the whole of the difference between the outcome of the best single thing and the outcome of the set, and it is a real cost paid in real money.

Two things make that cost easy to underweight when it is being described and impossible to ignore when it is being paid. The first is that it is invisible in advance. Nobody knows which one it will be, so before the fact the cost is a shape rather than a number. The second is that it is perfectly visible afterwards, and it arrives in the form of somebody at a wedding describing what one thing did while the household holding several did something duller. Most spreading is abandoned in that conversation, after the outcome is known and never before.

There is a second cost, smaller and more practical. More things to hold means more to keep track of, more paperwork, more places a nomination can go out of date, and more small charges wherever charges exist. For a household counting its months of buffer in fractions, that friction is not nothing.

Whose idea is the formal version of this?

The everyday version of this idea is old and belongs to nobody. Not putting everything in one place is advice that appears in proverbs in most languages, and it did not need an author.

The formal version does have one. Harry Markowitz set it out in a paper called Portfolio Selection in 1952, and what he added was not the advice. The addition was a shift in the question. Before that paper the natural question was which single thing is the good one. Markowitz insisted instead that a holding cannot be judged on its own at all. Its effect alongside everything else already held is the thing that matters. The same thing is a sensible addition to one set and a poor addition to another, and nothing about the thing itself changed between those two cases.

Asking what one event would reach, rather than talking about a holding on its own, is the move Markowitz made. His paper is findable through Journal Storage (JSTOR) or through any university library.

What does the Bhosale household's own concentration look like?

Now the case, read plainly and in the order that matters least first. At the end of year two the Bhosale household holds Rs 3,67,887/-, and that total is made up as follows. Every figure is this invented household's own record.

Read the last column rather than the second. Four of the seven lines are amounts somebody owes the household. The two-wheeler is a thing it rides, and a thing in use loses value rather than being priced by anybody. One line is left whose value is set by a market. Of everything the Bhosale household holds whose value a market decides, the gold is 100 per cent, and there is no second thing.

Two facts about that position are worth holding side by side. The position is total. A single metal in a single form accounts for the whole of the household's exposure to any market anywhere. And the position was arrived at without one decision. The two bangles and the chain came at a wedding, and nobody has ever valued them against anything. Nobody chose to be entirely concentrated. The household simply is, and the arithmetic does not care which of those it was.

How does somebody on the other side of a desk read a concentration?

The same reading gets used by people the household never meets, and knowing how they use it explains questions that otherwise sound intrusive.

A lender writing a loan that runs for years is not really asking what a household earns. The lender is asking what would have to happen for the repayment to stop arriving. Two applicants showing identical annual figures are not in identical positions if one has money arriving from two places that share a lane and the other has money arriving from two places that share nothing. The total is the same on the form and the answer to the only question the lender cares about is different. The third and fourth questions on an application are so often about where the work is and how long it has been there for exactly that reason.

An analyst looking at a small business does the same thing under a different name, and calls it customer dependence. A stall outside one office building has a revenue line that looks perfectly healthy right up to the week the building empties. The number in the accounts never showed the exposure; only the question about what one event would reach ever did.

A household can run the same reading on itself in ten minutes with no statement and no arithmetic. Write down every place money arrives from. Then name one thing that could happen in the locality and ask how many of those lines it would touch. The check asks about where money comes from rather than about anything held, so anybody can run it, including a household with nothing set aside at all. What to do about the answer is covered separately.

What is left for a household that holds nothing to spread?

Most households actually stand here. If a household has no shares, no fund, no monthly investment plan and nothing set aside beyond a thin buffer, then every mechanism above describes an arrangement it does not have and cannot make.

One reading is left that does not require holding anything. Write down where the money arrives from. Then name one thing that could happen and count how many of those arrivals it would reach. The largest concentration most households run is in their income and not in their holdings, and that reading is available in full to somebody with nothing set aside at all.

The Bhosale household is the demonstration. Its holdings are entirely concentrated on one metal in one form, and that concentration has never cost it a rupee. Nobody has ever valued the gold against anything or needed to sell it. Its income was concentrated on one local economy that nobody had written down anywhere. The income concentration took Rs 43,200/- out of a single year, against ordinary months costing Rs 42,770/- each. The exposure that mattered was in the smaller-sounding place, and it was there before any question about holding anything arose.

The right response depends on things no general account can see: what work is available, who else is in the household, what a locality offers and what a person can actually do. The reading itself is worth having regardless. A household that knows what one event would reach is in a better position than one that does not, and knowing costs nothing.

One boundary of a different kind. The mechanism described here is universal. The mechanism is arithmetic about where an event lands and not a rule anybody wrote, so it works identically in any country, any currency and any century. No threshold, period, rate or conduct requirement enters into it, so there is nothing in the mechanism that a jurisdiction could change. Everything about how things are sold, disclosed and registered is jurisdiction bound, and that is covered separately.

The Bhosale household, Meghna Bhosale, Ashok Bhosale and Ira Bhosale are invented.
Educational material. Not advice on any investment, tax, budget or market position.