Style Drift: When a Portfolio Stops Matching Its Claim

Style drift is a portfolio ceasing to match the description it was set up under, whether or not anybody decided to change it. Drift arrives two ways: through trading, and through prices moving the weights while nothing at all is traded. Detecting drift needs the same measure struck on the same base at two dates. Most records lack exactly that.

A portfolio that has moved is not thereby a worse portfolio, and the person running it was not thereby careless. A portfolio can move a long way from its description in a year in which every decision taken inside it was sensible, and it can also move because no decision was taken at all. Drift is a fact about a portfolio, not a verdict on anybody, and a committee's job here is to notice it and write it down.

The running example is the Anantara Multi-Asset Portfolio, an invented discretionary mandate of Rs 500 crore 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. Inside the equity sleeve sit 28 names, the largest of them Rs 23 crore. The mandate caps any single holding at 5 per cent of the portfolio, or Rs 25 crore.

Two subjects covered separately set the ground style driftA portfolio no longer matching the description it was set up under, whether that description is a written mandate, a stated approach or a label somebody attached to it later. stands on: what the value, momentum, quality, size and low volatility sorts actually sort on, and how a placement on a style map is produced. Style drift starts one step later: the placement is taken as given, and the thing placed stops being the thing described.

What is a portfolio drifting away from?

Not from its benchmark, and not from good sense. A portfolio drifts away from its own descriptionThe statement of what a portfolio is meant to be. The description may be a written mandate, a stated approach, or simply a label somebody put on the portfolio in a report., which is a separate object with a separate history. Somebody wrote it: a formal mandate with numbers in it, as the Anantara Multi-Asset Portfolio has, or a stated approach set out in a letter, or a label that appeared in a report three years ago and was never questioned since. Without a claim there is no distance to measure, so the first question in any drift discussion is not what the portfolio holds but what it claimed.

A portfolio described in words that cannot be turned into a number cannot be shown to have drifted, however far it has moved. If the description says the portfolio takes a careful, long term view of high quality businesses, every possible set of holdings satisfies it and none violates it. Careful, long term and high quality are not wrong words. Not one of them is a measurement.

The Anantara mandate is the measurable kind. Equity between 50 and 70 per cent. No single holding above 5 per cent of the portfolio. No unlisted holdings, and a minimum credit standing on the fixed income sleeve stated as a policy rather than as a rating symbol. Each turns into a number on any given day, and a number on one day can be set beside the same number on another day. Comparing one date with another is the entire machinery.

Notice what it does not say: nothing about what kind of business the equity sleeve holds, where it sits on a style map, or whether any factor exposure stays steady. A great deal could change inside that sleeve without a mandate line being touched. Constraints and descriptions are not the same document, so a portfolio can sit inside every constraint it was given and still stop resembling what everybody in the room believes it to be.

How does drift arrive through trading?

The first route is the obvious one. Holdings are sold, others bought, and the character of what is held changes with them. Sell six of the 28 names and buy six different ones, and the sleeve is measurably a different sleeve afterwards. Trading is the route everybody pictures and the route that gets discussed in meetings.

The trading route is easy to see. Somebody placed each order, somebody executed it, a custodian settled it, and the sequence leaves a dated trail. Drift that arrives through trading is visible in a record somebody is already required to keep. The easy route is not the one that causes the trouble.

Trading is also deliberate. Somebody chose each order and can be asked why, so there is a person, a date and a reason, and a committee can hold all three. The route that causes trouble is the one with no person, no date and no reason attached to it at all.

How do the weights move when nothing is traded?

A weightA holding's value divided by the total it is being measured against. Because it is a ratio, it changes whenever either the holding or the total changes, with or without any trading. is not a quantity a portfolio has but a ratio: a holding's value divided by a total. The largest holding in the Anantara equity sleeve is Rs 23 crore in a Rs 500 crore portfolio. Its weight is 23 divided by 500, or 4.60 per cent. Both numbers are prices multiplied by quantities, prices move every day the market is open, and quantities move only when somebody trades. So the numerator and the denominator of every weight move on days when nobody in the room has done anything, and the shape of the portfolio changes without any decision being taken.

A household keeps Rs 6,00,000/- in a deposit and Rs 4,00,000/- in an equity plan, sixty and forty out of Rs 10,00,000/-. Over a year the equity plan rises by half to Rs 6,00,000/-, nothing is saved and nothing is moved, and on a total of Rs 12,00,000/- the split is now fifty and fifty. Nobody in that house decided to take more market exposure.

Weights are struck against a shared total, so they must always sum to one hundred, and every weight that rises is paid for by weights that fall that nobody decided either. The illustration below uses constructed units, chosen so the total does not move and the effect cannot be blamed on the total.

The three groups hold 200, 300 and 500 units on date one. Group one rises by half to 300 and group three falls by a fifth to 400, so the total is still exactly 1,000, nothing was bought or sold, and the weights still moved ten points.

The consequence that catches committees is this: a portfolio can end a year materially different from how it started while its trade record shows almost nothing, and every conversation that starts from the trade record will miss it entirely. Drifting is not a rare pathology but what a portfolio does by default when left alone, and the word drift was chosen for exactly that reason.

Can a cap be breached with nobody deciding anything?

Yes, and the arithmetic is worth doing slowly. The mandate says no single holding above 5 per cent of the portfolio. On Rs 500 crore that is a cap of Rs 25 crore, and the largest holding is Rs 23 crore. Every report anybody has read shows the position comfortably inside its limit at 4.60 per cent.

Now ask how far that holding can rise before it breaches. The tempting answer is that Rs 23 crore has to reach Rs 25 crore, a rise of 8.70 per cent, and it is wrong. The cap is not written in rupees but as a share of the portfolio, and the holding is inside the portfolio, so when the holding rises the total rises with it and the cap in rupees rises too. The target is moving.

Let the price rise be x, expressed as a decimal. The holding becomes 23 times one plus x. Every other holding is held still, so the portfolio becomes 500 plus 23x. Set the share equal to 0.05 and solve.

A rise of about nine per cent in one holding, with no order placed anywhere in the portfolio, walks straight through a constraint everybody believed had Rs 2 crore of room in it. A constraint written as a share is a constraint on an outcome, not on an action. Compliance with it is therefore measured on a date rather than inferred from a trade record.

The 0.45 point difference is small, and what matters is its direction. Computing headroom the quick way gives an answer that is too tight. Too tight is the safe direction. The position has a distance to its own limit, that distance is 9.15 per cent of price movement, and it is a number a committee can be told and can watch.

A limit crossed that way is a passive breachA limit written as a share being exceeded because prices moved, with no order placed and no decision taken by anybody., and it deserves its own name. There is nobody to ask why and no order to point at. The position report on the morning after looks exactly like the one on the morning before, except that one number in it has crossed a line.

How fast could this sleeve have changed in one year?

The other question a committee asks about drift is one of speed: how much can the portfolio change between two reviews? TurnoverA measure of how much of a portfolio was replaced over a period, usually stated as a share of the portfolio rather than of any one part of it. answers it, and the Anantara record carries one, portfolio turnover of 34 per cent over the stated twelve month period.

Scale it. Thirty four per cent of Rs 500 crore is Rs 170 crore replaced during the year, and replacing means selling one thing and buying another, so about Rs 340 crore of transactions passed through the market. A turnover figure that sounds like a third of the portfolio is two thirds of it in transactions, and every one of those carried a cost that no return figure anywhere in the record shows.

The 34 per cent is struck against the portfolio, and the equity sleeve is Rs 300 crore. The record does not split turnover by sleeve, so the honest answer about how much of the equity sleeve changed is not a figure but a range.

The lower end is nothing: every one of those trades could have happened in the fixed income and cash sleeves. The upper end assumes every trade happened inside equity, and Rs 170 crore of a Rs 300 crore sleeve is 56.7 per cent. The record does not say which, so the equity sleeve's own turnover is bounded rather than known, and the only defensible statement is that it lies between nothing and 56.7 per cent of the sleeve.

The mandate says equity between 50 and 70 per cent and nothing about which equity. More than half of that sleeve could have been replaced without a single line of the mandate being touched.

Break it down by quarter. Spread evenly across four review periods, Rs 170 crore is Rs 42.5 crore a quarter. Even spreading is an assumption the record does not support, and the picture below is labelled constructed for that reason.

A committee meeting quarterly is therefore looking at a portfolio that could, at the outside, have replaced roughly a seventh of its equity sleeve since anybody last looked. A seventh of the sleeve is a speed limit the record supports rather than a claim about what happened, and treating a bound as an estimate would be the same error in a different coat.

What does detecting drift actually require?

DetectionEstablishing that a portfolio has moved away from its description, by computing the same quantity twice and taking the difference. is subtraction. Subtraction requires two numbers, and two numbers are comparable only if they were produced the same way.

Four things have to be held identical between the two dates, and each has been quietly changed by somebody at some point: the measure, meaning the quantity chosen for computation; the baseThe total a share is struck against. Divide the same rupee holding by the portfolio or by one sleeve of it and the per cent differs. it is struck against, which for this portfolio is either Rs 500 crore or Rs 300 crore and never a vague gesture at both; the boundaries, wherever the measure sorts holdings into groups; and the weighting basis, meaning what each holding was weighted by.

Style map construction makes the third of those concrete: where the lines between groups sit is a choice, and moving a line moves holdings across it without anything happening in the portfolio. If the line moved between the two measurement dates, part of the computed difference is the line moving and part is the portfolio moving, and nothing in the arithmetic separates them afterwards.

Changing any one of the four between measurements produces a drift figure that is measuring the measurement, and in practice that is a more common failure than drift itself. It is common because it is invisible: a number arrives with a decimal point in it, nobody asks whether the two halves of the subtraction were computed the same way, and it looks like a fact about the portfolio because that is what it is labelled as.

Is a manager changing their mind the same thing as drift?

No, and this is the distinction that decides what kind of entry goes in the minutes. Two portfolios can look identical at the second date and be two completely different findings, depending on what happened to the description in between.

In the first case the manager reconsidered and said so. Faiz Ahmad Ansari comes to the committee, explains that the approach has been revised, and the revision is written down and agreed. Both objects moved, so there is no gap. A decision exists instead, taken by a named person on a stated date.

In the second case nothing was said. The description sat unchanged in a file while the portfolio moved underneath it, by either route. The gap between an unchanged description and a moved portfolio is drift. The first is a decision to be recorded and discussed and the second is a distance to be measured and recorded, and treating the two as the same thing produces either paralysis or blindness depending on which way round they are confused.

Keeping them apart has a practical reason. A committee that calls every change drift questions every decision its manager takes, and eventually stops the manager taking any. Whatever is held can always be described afterwards as what was intended, so a committee that calls every drift a change of mind accepts any portfolio at all.

Can the Anantara portfolio's drift be computed from one date?

No, and saying so plainly is the most useful thing that can be done with the record available. The record gives 28 equity names in a Rs 300 crore sleeve, a largest holding of Rs 23 crore, a cap of Rs 25 crore, a top ten of Rs 155 crore and a turnover figure, all struck at one date. There is no earlier date, and no style measurement at either end.

A snapshot is what the record does support, and a snapshot is what a first review writes down for a second review to subtract from. Rs 300 crore across 28 names gives an equal weight of Rs 10.71 crore, and the largest holding at Rs 23 crore is about 2.15 times that. Each of those is a rupee amount, and becomes a share only when a base is named beside it.

The same discipline applies to the top ten. Rs 155 crore is 31.0 per cent of the Rs 500 crore portfolio and 51.7 per cent of the Rs 300 crore equity sleeve: one holding group, two numbers, each meaningful only with its base attached. Underneath both sits a floor that is pure arithmetic: the ten largest of 28 holdings can never be less than ten twenty eighths of the sleeve, whatever anybody holds. A concentration figure below its own arithmetic floor is not a low reading, it is a wrong reading, and checking a figure against its floor costs one division.

So the refusal is the finding. The record fixes what the portfolio looked like on one date, and the arithmetic of a passive breach is settled. A movement needs two positions and only one was recorded, so how far the portfolio moved cannot be said at all.

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

A committee brings two documents to the review instead of one. Two documents are the whole practitioner answer.

The first document is the trade record, and almost every committee already has one. The trade record answers what was done since the last meeting: the orders, the dates and the sizes. The price route involves no orders, so the trade record shows nothing about it, and a quiet quarter in that record is entirely consistent with a portfolio that moved a long way.

The second is the measurement: the same measure, struck on the same base, at this date, beside the identical figure from the last review. Preparing it is boring work. The two together produce a finding, and either one alone produces a conversation that feels like a review and is not one.

An analyst outside the arrangement runs the same logic with less information. Given a holdings list at two dates, they can compute anything they like, provided they compute it identically on both. Given one date and a description in adjectives, they can compute nothing, and the honest output is a note naming the missing input.

A lender or a household runs a smaller version of the same job. Somebody who set out to keep half their savings safe and half at risk, and who has not moved a rupee in three years, still needs to open the statement and divide. The split they set is not the split they now have. RebalancingTrading back towards a chosen set of weights after prices have moved them. A portfolio's response to drift is a separate matter, covered in the monitoring work. is what some holders choose to do about it, and what any particular holder should do is not answered here.

What has to be written down so drift can be seen later?

Four lines, written at the first review rather than the second. Nobody needs them at the moment they are written, and writing them early is what makes them hard to remember. The four lines are needed a year later, by somebody trying to subtract.

Notice what is not on that card: no view about whether the portfolio is good, no comparison against anything, no instruction to anybody. The card exists to make a difference computable later, and a difference is the only thing drift ever is. Everything a drift review can honestly produce is a number and the four conditions under which it was struck.

The selection work closes at this point. The work began by fixing an equity sleeve at Rs 300 crore, handed down from the allocation work rather than chosen inside the sleeve, and everything after that concerned what happens inside that sleeve. The work ends where somebody opens the file a year later and asks whether what was built is still what is held.

What is style drift not?

Style drift is not a mistake. A portfolio that has drifted may be holding exactly what the person running it intends, and the finding is only that the description no longer matches. Drift is not a judgement about performance either. A portfolio that moved away from its description could have done anything afterwards.

The two meet in the passive breach case, but drift is not the same as a breach. The mandate constrains the equity share, the size of any one holding and the credit standing of the fixed income sleeve, and says nothing about the character of the 28 names, so a portfolio can drift a long way without breaching anything.

A holder's response when a portfolio has drifted, whether anything is traded back, and how a breach is handled, all belong to the monitoring work covered separately and depend on the mandate. Drift is a fact to be noticed and recorded.