Value, Momentum, Quality, Size and Low Volatility

Value, momentum, quality, size and low volatility are five sorting rules. Each ranks a stated set of holdings on a measurable characteristic rather than on a judgement about any business, and each has to name four things: the set it runs on, the measure it ranks by, how the ranking becomes weights, and how often it is redone. Leave one out and nothing can be checked.

Five words get thrown around a great deal, and most of the time they are being used as adjectives. Somebody says a portfolio is a quality portfolio the way they might say a restaurant is a good restaurant. A description cannot be argued with. Being unarguable is what makes it useless for learning. All five are rules, written down in advance, that anybody else could run on the same starting set and either reproduce or fail to reproduce.

Two things are settled elsewhere. What a loading is belongs to the factor model. The difference between a rule that defines a holding set all at once and a view that defines it one case at a time is covered separately.

Everything is worked on 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 equity part is Rs 300 crore, or 60.0 per cent of the Rs 500 crore whole, and it is held across 28 names. The single holding limit written into the mandate is 5 per cent of the portfolio, or Rs 25 crore. A sort has to survive those four numbers before it becomes a set of holdings, so the four do more work than any of the five sorts.

A rule that ranks the world and says hold the top eight cannot be run here at all: eight names at the Rs 25 crore limit come to Rs 200 crore against a Rs 300 crore sleeve. The arithmetic of the limit arrives before anybody argues about which of the five sorts is interesting.

What makes a sort a factor rather than a preference?

Four things stated in advance. A style factorStyle factor and factor are used interchangeably in this setting, and both name a ranked characteristic rather than a company. names the universeThe complete set of candidates a rule is allowed to look at before it ranks anything. it runs over, the metricA single measured number, chosen so that it exists for every candidate and can be compared across them. it ranks on, how that ranking turns into weights, and the rebalancing periodThe gap between one running of a rule and the next, also quoted the other way round as a frequency. at which the whole exercise is repeated. A rule that states all four can be handed to a stranger who will produce the same list; a rule missing any one of them is a preference wearing a technical name.

The everyday version is closer than it looks. A wedding caterer asks for a shortlist of vegetable suppliers. Bring me the good ones, and three people carrying it out come back with three different lists, none of them wrong. Take every supplier within twenty kilometres who delivered at least fifty times last year, rank them by the share of deliveries that arrived before nine in the morning, take the best twelve, split the order equally between them and redo it every quarter, and three people come back with the same twelve. The second instruction is a sort and the first one is a preference, and the difference is not care or expertise but whether anyone else can run it.

Change the universe and the identical metric returns a different list. Ranking the same measure over everything listed on an exchange, then over only the largest three hundred listings, then over only what a particular mandate permits, gives three answers from one rule. The universe is a decision taken before any measurement happens, and it is the part of a style claim that most often goes unstated because it feels like background rather than a choice.

Turning a ranking into weights is a separate decision from making the ranking, and there are several honest ways to do it: give every survivor the same amount, give more to the ones that ranked higher, or scale each one by how far it sits from the middle of the measured distribution. All three are defensible, all three produce different portfolios from the same ranking, and a claim that stops at holding the cheap ones has not said which was used.

The period, the last of the four, is what stops a sort from being a photograph. Metrics move: a holding that ranked in the cheapest tenth in January may rank in the dearest third by December, and whether the portfolio still holds it depends entirely on whether the sort is rerun monthly, quarterly or once a year. The rebalancing period is the difference between a rule and a memory of having once applied a rule, and it is the part of the specification that turns directly into trading and therefore into cost. The trading that follows is covered under trading cost.

What does the value sort actually sort on?

The value sort ranks a price set against something measured in the accounts. Take a stated universe, compute for each candidate the ratio of its market price to some accounting quantity belonging to it, put the ratios in order, and hold the end where the price is low relative to that quantity. The sort itself is arithmetic that a spreadsheet performs without an opinion; every judgement in it sits in the choice of what goes on the bottom of the ratio.

The denominator is where the argument lives. Price can be set against what a business earned, against what it holds on its balance sheet, against the cash it generated, against what it sold. Each choice produces a different ranking of the identical universe, and each is defensible on its own terms. Choosing the denominator is the only genuinely contestable step in a value sort, and a claim that names no denominator has hidden the one part anybody would want to argue about.

The research tradition behind the size and value sorts belongs to Eugene Fama and Kenneth French, whose 1993 work set out how a characteristic-ranked portfolio is built and studied. The construction shape they described is what carries over: a set, a measure, an order, a cut.

Deciding whether a business is cheap is a separate matter. Whether a low ratio means a bargain or a warning is a question about the business, and forming a view on a business is covered separately. A value sort ranks a ratio; it does not think about what the ratio means, and pretending otherwise is how a sorting rule quietly turns back into an opinion.

What does the momentum sort use, and over what window?

The momentum sort ranks each candidate by its own return over a stated past window and holds the stronger end, and almost every version of it leaves out the most recent stretch before today. Momentum is the one of the five that takes nothing at all from the accounts. It needs only a price series, so it can be run over a universe where accounting comparability is poor. The research work naming and testing this ranking is attributed to Narasimhan Jegadeesh and Sheridan Titman, writing in 1993.

Why deliberately throw away the freshest information? The construction reason is that very recent moves behave differently from the stretch before them, and a rule that wants to rank the longer stretch does not want the most recent days doing the ranking for it. Whether that reason convinces or not, the skip is written into the rule, so two sorts with the same window and different skips are two different rules.

What is the quality sort trying to capture, and why is it the least settled?

The quality sort ranks on how a business is financed and how steadily its results behave, and it is the only one of the five where reasonable people cannot agree on the measure. Borrowing against the assets, steadiness of the year to year results and how much of the result arrives as cash are three defensible readings of the same English word.

Two rules both truthfully described as quality sorts can be run over one identical universe on one identical day and hold substantially different names, and neither of them is cheating. Such a split does not happen with size, where everybody measures the same thing, and it barely happens with momentum, where the disagreement is about the window rather than about what is being measured.

No originator can be named for the quality sort. Value, size and momentum have work that can be pointed at and confirmed against the actual text; quality as a ranked characteristic emerged across enough separate strands that attaching one name to it would be a guess dressed as a citation. A confident wrong attribution is the kind that gets repeated, so saying that no name can be given with confidence is more useful than naming one.

What does the size sort measure, and what does it not measure?

The size sort ranks candidates by their market valueWhat the market is currently asking for the whole of something, worked out from its price and how many units of it exist. It moves whenever the price moves. and holds the smaller end. The size sort is the simplest of the five to compute and the easiest to misread. The word size makes people think of the business rather than of the price the market is asking for it. The size sort measures a price total, not an operating scale, and those two quantities come apart exactly where the sort becomes interesting.

The bigger business is the stall. The stall moves more money through the till in a year. The showroom's price also carries the location, the stock and whatever anybody thinks the next few years look like, so the showroom is the one that would cost more to buy outright today. A size sort ranks on the second question and stays silent on the first, and every misreading of the sort comes from answering the first question and treating the second as answered.

What does the low volatility sort do, and what is odd about it?

The low volatility sort ranks candidates by how much their own past returns moved about and holds the steadier end. Computing that ranking is the least controversial step of the five. Measuring dispersionHow widely a series of numbers spreads out around its own average. A steadier series has low dispersion and a jumpier one has high dispersion. in a return series is a settled exercise, taken as already familiar. The oddity is not the arithmetic but where the low volatility sort sits beside the other four.

Line the five up and the asymmetry appears. Value, momentum, quality and size each read a fact about the candidate or about what the market asks for it. Low volatility reads a property of the return series itself. Dispersion in that series is the quantity the rest of this subject area treats as risk. So a claim that ranking on low dispersion earns something extra sits directly against the usual account, in which dispersion is what a holder is being compensated for carrying rather than something to be avoided for gain.

The question stays open. The discomfort does not prove the sort is empty, and the sort does not resolve the discomfort. No evidence settles it either way, so the honest position is to state the tension exactly and stop. Stopping there is a different thing from hedging. A hedge would be to say results vary; naming the structural reason the claim is awkward is a finding that can be carried into somebody else's argument.

What must a factor claim state before anybody can reproduce it?

Five items, gathered here in one place. The four from the specification, plus the window over which somebody measured whatever is being claimed about the sort. The fifth matters because a claim about how a sort behaved is a claim about a period, and a period that is not stated cannot be checked, compared or argued with.

Reproducibility is the only mechanism by which a claim of this kind can ever be checked. There is no laboratory and no referee. If a stranger can run the rule and arrive at the same list, then a disagreement about what it means is an argument that can go somewhere; if the rule cannot be run at all, the two sides are describing their tastes at each other.

What does the mandate do to a sort before it reaches the market?

The mandate cuts it, and the arithmetic is the same whichever of the five is being run. The mandateThe written instruction a holder gives a manager, setting out what may be held, in what proportions and inside what limits, before any decision is taken. here permits no single holding above 5 per cent of the Rs 500 crore portfolio, or Rs 25 crore. The equity sleeveThe part of the portfolio set aside for equity holdings, measured against the portfolio it sits inside. is Rs 300 crore. Rs 25 crore out of Rs 300 crore is 8.33 per cent, so no ranking rule, however strongly it prefers one name, can put more than 8.33 per cent of the sleeve into that name.

Rs 300 crore divided by Rs 25 crore is 12. A sort that returns fewer than twelve names cannot fill this equity sleeve without pushing at least one name past the Rs 25 crore limit, so twelve is a floor on the sort itself, imposed by the mandate long before anybody looked at a ranking. The floor of twelve is invisible to anybody who thinks about factors only as research.

The bottom row is the sleeve as the record carries it: Rs 300 crore across 28 names, or Rs 10,71,42,857/- each at an equal split. The record does not hold it equally. Its largest holding is Rs 23 crore, or 4.6 per cent of the Rs 500 crore portfolio and 7.67 per cent of the Rs 300 crore sleeve. Both percentages are correct and they answer different questions, and the 5 per cent limit is written against the portfolio, so the holding sits inside it.

The equity sleeve is Rs 300 crore of a Rs 500 crore portfolio, so anything a sort achieves inside the sleeve reaches the whole at 0.60 of its size. A style decision that moves the equity sleeve by one point moves the Rs 500 crore portfolio by six tenths of a point, and quoting the sleeve figure as though it were the portfolio figure overstates the effect by two thirds.

What does the Anantara record show about these five?

One thing, and it is negative. For the one stated twelve month period the record carries, the Anantara portfolio's volatility was 11.8 per cent against the composite benchmark's 10.4 per cent, a ratio of 1.135. Its betaHow much a portfolio tends to move for each unit the benchmark moves. Above one it moves more than the benchmark, below one it moves less. against that same benchmark was 1.08. Both figures point the same way.

So whatever the Anantara equity sleeve was doing over that year, a tilt towards steadier holdings does not appear in the only dispersion figures the record holds, and that is a finding rather than a hedge. Precision matters about what has and has not been shown. The record does not say the sleeve avoided a low volatility ranking; it says that if such a tilt were present and strong, the portfolio would be expected to move about less than its benchmark, and it moved about more.

The record carries no price relative to any accounting measure, no past return window, no financing or stability measure and no market value for any of the 28 holdings, so value, momentum, quality and size cannot be measured from it at all. Not weakly. Not approximately. The inputs are absent.

One limitation attaches even to the figure that does exist. The 11.8 per cent and the beta of 1.08 belong to the whole Rs 500 crore portfolio, not to the Rs 300 crore equity sleeve where a style ranking would actually operate. Reading a portfolio level dispersion figure as though it described the equity sleeve is the same base error this subject area exists to teach, so the negative finding is stated about the portfolio and left there.

One shortcut is easy to take without noticing, so it is worth naming. Somebody looks at a return figure for the year, decides it looks like the sort of year a particular style has, and reports a tilt. Reporting a tilt that way runs the argument backwards: the tilt was supposed to be what explained the return, and using the return to infer the tilt makes the explanation unable to be wrong in one move.

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

An investment committee like Rukmini Deshpande's does not use these five words to decide anything. The committee uses them to convert a description into something that can be checked at the next meeting. When a paper says the sleeve is quality focused, the useful response is not agreement or disagreement but five short questions, asked in order, that change what the paper has to say next time.

A lender does a version of this without using any of the vocabulary. Asked to accept a portfolio as security, a lender wants the rule that produced it rather than the view behind it. A rule can be reapplied next quarter and a view cannot. An analyst reading somebody else's holdings does the same in reverse: if the stated rule and the actual list disagree, that gap is the finding, and where a portfolio has stopped being what it claimed is covered separately under style drift.

A household version exists too. Somebody says they save carefully. Asking what carefully means might produce anything. Asking instead which accounts were considered, what number was compared across them, how much went into each and how often the exercise is repeated either produces a rule or makes clear that a habit has been described as a policy. The five questions are not finance machinery; they are what turns any repeated choice into something the person making it can learn from.

What is a style factor not?

A style factor is not a judgement about a business. A sort is a rule applied to a measurable characteristic, and forming a view on whether a business is well run, well positioned or fairly priced is a different exercise, covered separately. The two can reach the same holding and they are not the same activity, and describing a preference in the vocabulary of the first is how the difference gets lost.

A style factor is also not a claim about what any of these five will do next. None of the five sorts states a return, a premium or a spread between ranked groups; a sort produces a list and stops. A sort is not an argument that any of the five is worth carrying either. Whether a holder should want a tilt at all is a separate question.

The result is small and durable. Five words that get used as adjectives are actually five ranking rules. A rule states four things or it is not a rule. A claim about a rule states a fifth, the window. And when somebody offers a style description, the useful move is neither belief nor scepticism but a question about the metric. Asking about the metric is the cheapest diagnostic in this entire subject area.