Allocation Dimensions: Sector, Country and Currency

An allocation dimension is a way of cutting the same set of holdings so that a shared exposure becomes visible. Sector, country and currency are three such cuts, and every rupee in the portfolio sits inside all three at once. Currency is the only one that can be held apart from the asset, so currency is decided on its own rather than inherited from a choice already made.

The big shape is already settled: the money is split across classes, and the split sits inside a mandate that says how far it may move. Then someone asks a question the class weights cannot answer. How much of this portfolio depends on one industry having a good year? On one country not changing its rules? On one currency not weakening? A class weight counts rupees by class and nothing else, so the class weights are silent on all three by construction.

The running example throughout is the Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 5,00,00,00,000/- run by Faiz Ahmad Ansari for an invented charitable endowment whose investment committee is chaired by Rukmini Deshpande. Its stated shape is equity 60.0 per cent at Rs 300 crore, fixed income 30.0 per cent at Rs 150 crore and cash 10.0 per cent at Rs 50 crore. Each of those three is a policy weight, the shape the holder chose in advance, and actual weights drift away from the policy weights between rebalancings. The equity sleeve is held across 28 invented names, the mandate caps any single holding at 5 per cent of the portfolio, and it keeps equity between 50 and 70 per cent.

The record for the Anantara portfolio carries no sector split, no country split and no currency split, so no such figure for this mandate can honestly be stated at all. The absence is the subject itself rather than an omission. The cuts can still be taught and every part of them the record does lock can still be computed, and the rest stands as NOT SUPPLIED.

What is an allocation dimension, and why is one never enough?

An allocation dimensionA characteristic used to sort holdings into groups: class, industry, country, currency, and others besides. is a way of grouping holdings already held. The grouping makes the total behind one shared characteristic readable as a single figure. Asset class, sector, country and currency are each one of them, and each takes the identical set of holdings and sorts it into different piles.

The first meeting with a dimension is where people go wrong. A dimension is not a new decision layer stacked on the allocation already made, and it adds nothing to the portfolio. A dimension is a cut of holdings already in place, so the same rupee appears in the sector cut, the country cut and the currency cut simultaneously, and adding the three cuts together would count that rupee three times.

The everyday version costs nothing to picture. A household lists its monthly outgoings by category: rent, school, food, transport, medical. Now cut the identical list by who is paid: one landlord, one school, four shops, two drivers, one clinic. The rupees have not changed. The second cut answers a question the first cannot: how much of the month depends on one counterparty staying in business. Both describe the same spending and neither is more true.

Back on the mandate, the Anantara portfolio reads 60, 30 and 10 across equity, fixed income and cash, and that cut says nothing about industry, about country or about currency. A class cut was never built to carry those answers.

One dimension is never enough, and the reason is structural rather than a shortcoming of anybody's reporting. Each cut is a projection: it keeps the characteristic it sorts on and throws the rest away. Keeping industry, country and currency would stop a class cut being a class cut, so a class cut discards all three. Recovering those characteristics takes another cut.

What is Sector Allocation, and what exactly is it a cut of?

Sector allocationSector and industry are used interchangeably here, and how fine the groupings are is set by whichever classification scheme is named. is the distribution of a sleeve across industry groupings. For the Anantara equity sleeve it would report how the Rs 300 crore sits across whatever groupings the reporting scheme uses, each shown as rupees and as a share of a stated base.

Why bother, when every one of the 28 names has already been reviewed? Because what a sector cut finds is invisible in any single name. Two holdings that look entirely unrelated can share an input cost, a customer or a regulator. A sector cut groups on the shared driver rather than on the business description, and no cheaper instrument finds that overlap.

The street version. Ten shops in one shopping centre sell ten different things: shoes, phones, sweets, spectacles, luggage. The ten share one thing, footfall through a single entrance. Cut them that way and ten unrelated small businesses become one position wearing ten signboards.

The arithmetic the record does lock is the denominator: Rs 300 crore if the sector cut is taken on the equity sleeve, Rs 500 crore if it is taken on the whole portfolio. Every numerator is absent, so the honest picture is a fully drawn denominator with an undrawn numerator, and that is a computed refusal rather than an asserted one.

Who decides the industry groupings, and does the choice change the picture?

Somebody does, and the choice undoes a lot of comparisons. Somebody decided how many groupings there are, where the boundaries fall, and which one a business belongs to when it does three things at once. Change the scheme and the picture changes without a single holding moving. The rules for these classifications are published by the exchanges at nseindia.com and bseindia.com, and a sector figure quoted without saying which scheme produced it is not comparable with anybody else's.

What is Country Allocation, and why is the honest answer uncomfortable?

Country allocationAlso called a geographic breakdown or a regional split, and the same ambiguity runs under all three names. is the distribution of a portfolio across countries. Country allocation sounds like the simplest of the three and is the messiest. A business does not sit in one country the way a rupee sits in one account. A business is incorporated somewhere, listed somewhere, taxed somewhere and paid by customers somewhere, and those places are frequently different.

So two honest cuts exist and they disagree. A cut on the listing baseThe venue where a security is admitted to trading. The venue is a fact about the security rather than about the business behind it. assigns each holding to the country where its shares trade. A cut on the revenue baseSales broken down by geography in a business's own reporting. A revenue cut takes its proportions from that breakdown. assigns each holding across countries in proportion to where its sales arise. Run on one portfolio, the two produce two different country tables.

The two cuts answer different questions, so neither of them is wrong. The listing cut answers where the market for these shares sits and the revenue cut answers where the earnings behind them are generated, so a portfolio described as having a country exposure without naming the cut has said nothing anybody can check.

The two are not equally easy to get. A listing cut comes off a custody statement on any day; a revenue cut waits on disclosure from each business, so a holder who wants one has to decide how stale a figure they will accept, and that decision is part of the reported number.

On the record the position repeats. A country cut on the whole Anantara portfolio has a base of Rs 500 crore and one on the equity sleeve a base of Rs 300 crore, and beyond those two denominators there is nothing.

What is Currency Allocation, and why is it the odd one out?

Currency allocationDenominated means the currency a holding is priced and settled in. The holder may measure in a different currency altogether. is the distribution of a portfolio across the currencies its holdings are denominated in. On the face of it that is a third cut just like the other two, and it is not.

A sector exposure arrives welded to the holding: hold a business in one industry and the only way to drop that industry exposure is to sell. A country exposure comes attached the same way and leaves only when the asset leaves. A currency exposure is different in kind. The currency exposure can be separated from the holding and dealt with on its own, and that makes currency a decision in its own right rather than a consequence of a decision already taken.

The everyday version again. A shopkeeper cannot stop having a wholesaler as a supplier without changing supplier. But if that wholesaler bills in a currency the shopkeeper does not earn in, the currency question and the supplier question have come apart, and the shopkeeper can keep the supplier and still address the currency on its own.

Hedging a currency exposure covers the instrument that separates one, how it is executed and what it costs. Structurally the exposure and the asset can be held apart, so the currency question gets asked and answered on its own.

Does a foreign return and a currency move add, or do they do something else?

The two multiply, and the difference matters more than it looks. The Anantara record carries no foreign holding of any kind, so the moves used below are assumed sizes chosen to expose the arithmetic.

Suppose an asset rises 10.0 per cent in its own currency, and that currency weakens 4.0 per cent against the holder's currency over the same window. The tempting sum is ten less four, or six. The actual calculation multiplies one and one tenth by nought point nine six. The product is 1.0560, so the return is 5.6 per cent, and the naive answer overstates it by 0.4 percentage points.

The missing 0.4 points has a name. The gap is the cross termThe part of a combined result that comes from two moves acting on each other rather than from either one alone., the asset move multiplied by the currency move: nought point one zero times minus nought point nought four is minus nought point nought nought four, or minus 0.40 points. At small moves it looks like a rounding difference and people stop worrying about it, and that habit is where the trouble starts.

Run it again at twenty per cent each way. The asset rises 20.0 per cent, the currency falls 20.0 per cent, and the tempting sum is zero. One point two multiplied by nought point eight is 0.9600, so the return is minus 4.0 per cent. The cross term grows with the product of the two moves and not with their sum, so it reads minus 4.00 points here rather than minus 0.40. A currency surprise is a surprise for exactly that reason.

Read the fourth row slowly. An asset that rose a fifth, in a currency that fell a fifth, produced a loss of 4.0 per cent for a holder measuring in the other currency. Nothing went wrong. The two moves were never being added, so they did not cancel.

Why must the base of every dimension figure be stated?

Because every dimension figure has at least two honest values, and on this mandate the two differ by a factor of nought point six. Working slowly through the mandate's own numbers shows what skipping the base costs.

The largest single holding in the Anantara equity sleeve is Rs 23,00,00,000/-. Against the whole Rs 500 crore portfolio that is 4.60 per cent. Twenty three divided by three hundred is nought point zero seven six six seven, so against the Rs 300 crore equity sleeve the same rupees are 7.67 per cent. The mandate's single holding capA mandate rule stating the most any one holding may be, always written against a stated base. is 5 per cent of the portfolio, or Rs 25,00,00,000/-, and that same cap is 8.33 per cent of the equity sleeve.

So one holding carries the figures 4.60 and 7.67, one cap carries 5.00 and 8.33, and those four numbers describe two things rather than four.

The factor is not the same for every sleeve either. People who learn the divide-by-nought-point-six rule and then apply it everywhere get caught by the difference. The conversion factor is the weight of whichever sleeve the figure is moving to, so the same Rs 25 crore cap reads 8.33 per cent against the Rs 300 crore equity sleeve and 50.00 per cent against the Rs 50 crore cash sleeve. Same rupees, one answer for each sleeve.

Worse, the sleeve base is not even a fixed number. The mandate keeps equity between 50 and 70 per cent, so the equity sleeve ranges from Rs 250 crore to Rs 350 crore without anything unusual happening. Against a Rs 250 crore sleeve the same Rs 23 crore holding reads 9.20 per cent; against Rs 350 crore it reads 6.57 per cent. Three sleeve figures for one holding, all correct, all inside a mandate that has not been breached.

The headroom supplies a check anyone can run. The holding is Rs 23 crore and the cap is Rs 25 crore, so the gap is Rs 2 crore however it is expressed: 0.40 percentage points on the portfolio base, 0.67 points on the sleeve base. A figure and its headroom that change together in that way sit on one base; if only one of them changes, somebody has mixed two.

What happens when the dimensions overlap?

The cuts are not independent of one another, and that has a consequence most limit frameworks are never tested against. A business in one industry is often listed in one country and reports in one currency, so a rupee that lands in a particular sector bucket is frequently landing in a particular country bucket and a particular currency bucket at the same time. The cuts move together.

Which means a set of overlapping limitsA limit is a ceiling written into the mandate, and limits overlap when they are written on cuts of the same holdings. behaves in a way people do not expect. Write a limit on each dimension, satisfy every one of them, and total concentration is still unlimited. Per dimension limits never compose into a limit on the portfolio as a whole, and that is a structural property of overlapping cuts rather than a failure of anybody's drafting.

The household version is quick. A person caps their savings at a third in any one bank, a third in any one city and a third in any one kind of instrument, and keeps all three rules. If that bank sits in that city and holds that kind of deposit, the three thirds are largely the same third, and the rules that felt like three protections were mostly one.

What does a cap on any single holding actually prevent?

A cap prevents one thing precisely, and the one thing is worth computing rather than assuming. A cap of 5 per cent of the portfolio means no single name can exceed Rs 25 crore, and that forces a minimum count. One hundred divided by five is twenty, so a fully invested portfolio needs at least 20 names. Inside a 60 per cent equity sleeve, sixty divided by five gives 12. The Anantara sleeve holds 28 names, comfortably above both floors, so the cap is not what set the count.

The limit of that result is sharper than it sounds. Twelve names at exactly Rs 25 crore each come to Rs 300 crore, the whole equity sleeve to the rupee, so the same twelve the cap forces as a minimum could put the entire sleeve into one industry, one country and one currency without the cap being touched once. The cap counts names and the dimensions cut across names, so a limit on how big one holding may be is not a limit on any dimension whatsoever.

The same point runs the other way for the 28 names actually held. Split the sleeve equally and each name is about Rs 10.71 crore, or 2.14 per cent of the portfolio and 3.57 per cent of the sleeve, both far below the cap. The equal split shows the cap is inactive at these sizes and nothing at all about whether the 28 names share a driver.

What does the Anantara record hold, and what does it not?

The record for the Anantara Multi-Asset Portfolio locks the total, the policy split, the rupee value of each sleeve, the count of 28 equity names, the largest holding, the cap, the equity band and the ban on unlisted holdings. Every figure computed above came out of that set.

The record does not lock any division of any sleeve by industry, by country or by currency. Not a share, not a rupee amount, not a count of names. A manufactured split would look exactly like a measured one, so naming the gap is the correct response.

An interval is left to compute, and the interval is real arithmetic. Any one industry grouping holds between Rs 0 and Rs 300 crore of the sleeve, running 0 to 100 per cent of the sleeve and 0 to 60 per cent of the portfolio at the policy weight, rising to 70 per cent if equity sits at the top of its mandate band. The record fixes that interval and locates no point inside it.

How does a committee use this in a room, on a Tuesday?

Three habits, and none of them takes long. The first is that every dimension figure arrives with its base beside it, on the same line, not in a footnote. A pack carrying "largest holding 4.60 per cent of the portfolio" and "largest holding 7.67 per cent of the equity sleeve" in adjacent columns costs nobody anything and removes the whole class of error a missing base produces.

The second is that the word concentrated is not used about anything until the cuts have been run. An investment committee chaired by someone like Rukmini Deshpande can make that a procedural rule: the pack may state name sizes, and it may state that no cap has been breached. Concentration needs the cuts, and neither of those two is a statement about concentration, so a report that gives them and then concludes on concentration has answered a question it did not ask.

The third is that where a cut cannot be produced, the pack says so in the place the cut would have gone. A blank labelled not available is usable information: it tells the committee what to commission next.

The three habits scale down without changing. A household writing out where its savings sit should say whether a share is of total savings or only of the part held in market instruments, and should write not known where it does not know. A lender building a credit file separates what a borrower has disclosed from what it has assumed for the same reason.

What is an allocation dimension not?

A dimension is not a decision. Cutting a portfolio by sector changes the portfolio no more than sorting a drawer changes what is in it: the decision was taken when the holdings were bought, and the cut only makes visible what was already there. A report presenting a sector table as an allocation choice leaves the reader unsure whether anybody chose that shape or simply arrived at it.

A dimension is not a ranking either. A dimension reports the size of each grouping and says nothing at all about which grouping is better. A dimension says only that the cut exists, that it has a base, and that the base has to be named.

And it is not complete. Three cuts are three projections and there are others: by liquidity, by size, by how a holding earns. Each additional cut adds a way of seeing and none of them adds a limit, so the number of cuts a holder runs is a choice about how much they want to know rather than a control on anything.