Equity Research vs Security Analysis: Job and Method

Equity research is a job: producing written views on listed companies and sending them to people who allocate money, on a calendar, under conduct obligations. Security analysis is the method that makes any such view defensible. The job employs the method, the method exists without the job, and mistaking one for the other turns a craft into a role.

Almost everybody who has spent a year in the trade uses the two phrases as though they were the same phrase. In conversation that costs nothing. The person listening knows which one is meant. The confusion starts costing something the moment somebody is deciding what to learn, what to practise, or what a written view in front of them is actually worth. At that point the two words are pulling in different directions and the speaker has stopped noticing.

Three things settled earlier are assumed here rather than rebuilt. The shape of the research job across a working week, hour by hour, has been described already. The method's demands on a view, before that view is defensible, are set out on their own. And what a security is, in the sense the law uses rather than the sense a trading screen uses, has been settled too. Each definition has to be built without leaning on the other before the exact seam along which they separate can be located. The seam turns out to be narrow, entirely non-analytical, and responsible for a surprising share of what beginners get wrong about the work.

The difference between a job and a method might be expected to be a difference in what somebody thinks about. It is not. The thinking is very nearly identical. The difference is everything wrapped around the thinking: who asked for it, when it has to be finished, who reads it afterwards, and what happens if there is nothing to say. Hold on to that while both definitions are being built. The temptation to define each one by contrast with the other is strong, and a definition built by contrast is only half an answer.

What is equity research, taken entirely on its own terms?

Equity research is the production and distribution of written views on listed companies, for readers who allocate money. Every word in that sentence is doing work, so take them one at a time.

Production. There is an output and somebody is waiting for it. A view held in somebody's head is not research in this sense, however good it is. The activity is not complete until a document exists.

Distribution. The document goes somewhere, leaving the desk and reaching people who did not write it. Leaving the desk pulls in a whole apparatus of obligations. The sharpest difference of all lives there, and it returns below.

Written views. The output is an argument rather than a number. A model on its own is not the output. A figure on its own is certainly not the output. A claim travels, with the evidence attached, the assumptions named, and the case against it stated where a reader can find it.

On listed companies. The job is defined by its subject matter. Somebody whose whole working life is spent on privately held businesses is doing something with the same intellectual content and a different name for the job.

For readers who allocate money. The audience is not general. The reader has a decision to make and a very short amount of attention to spend making it, and that shapes the length of a note, the order of its paragraphs and the point at which the writer stops.

Where does this job sit? Three places, mostly. Inside a broking house, on the sell sideThe side that makes written work and execution for people who invest, earning from producing and servicing rather than from committing any money of its own., where the work goes out to institutions that trade through the house. Inside an asset manager, on the buy sideThe part of the market that actually commits money, whether somebody else's or its own, and which reads far more work than it publishes., where the work goes to a small internal audience with capital in their hands. Or in an independent practice, where the work is sold to whoever subscribes. The three differ in who reads the output, how many names one person carries, and how the work is paid for. All three share one thing: in every one of them a written output goes to a reader on a schedule the writer does not set.

The schedule is invisible until it is met, so it is worth being concrete about. A listed company puts out results on dates fixed by its reporting calendarThe fixed dates on which a listed company has to publish its results and its filings. Everybody covering the company works to that clock, and nobody covering it sets that clock., not by anybody covering it. Sarvani Coatings Limited reports, and within a known and quite short window the people who cover Sarvani Coatings are expected to have something in front of their readers. Nobody in that chain chose the date. It arrived.

What is security analysis, taken entirely on its own terms?

Security analysis is the practice of reaching a view about an instrument from evidence anybody could check, with the assumptions written down and the case against the view examined honestly. Security analysis is a method, in the ordinary sense that a way of doing surgery or a way of surveying land is a method. The method has requirements. A person can practise it well or badly. Anybody can be taught it, and anybody can practise it alone in a room with no employer, no reader and no deadline anywhere in sight.

The absences in that definition are worth naming. There is no employer in it. There is no audience in it. There is no output format in it, no fixed length, no calendar and no listed company. The definition holds a standard instead: evidence that can be checked by somebody who does not trust the analyst, assumptions that are visible rather than buried, and a deliberate search for the reading that would make the analyst wrong. Every requirement in the method is a requirement about the quality of a claim, and not one of them is a requirement about the circumstances in which the claim was made.

The origin matters here, and it is worth naming rather than assuming. Benjamin Graham and David Dodd published a book called Security Analysis in 1934. The book set the practice down in writing as a method with rules, rather than as a description of a role somebody held. The name of the discipline and the title of that book are the same words. The coincidence is one small reason the confusion has survived as long as it has.

The method survives a trip out of finance entirely. Take it out for a moment and look. A doctor looking at a set of test results is doing something with the same structure: a claim about what is happening, evidence anybody with the same results could check, assumptions stated, and a deliberate look at what else would produce this picture. A structural engineer inspecting a building does it too. None of them calls it research, and none of them needs an employer for the standard to apply. The standard is the practice.

Now bring it back. A person working out whether a coatings maker's improved gross marginGross profit measured against revenue, so it is the part of each rupee sold that survives the cost of making the thing that was sold. How it is built and where it sits belongs with the statements. came from its own pricing or from a pricing environment everybody in the trade enjoyed is doing security analysis, whether that person is an analyst at a broking house, somebody managing their own money, or a student with nothing at stake at all. The instrument does not have to be a share, either. The same method points at a debt securityAn instrument that names an amount and a date, so failing to pay it is a default with a remedy attached. A share names neither, which is the difference the two are usually compared on., at a preference share, at a lease, at a whole sector, at a privately held business with no quoted price anywhere.

Where do the two genuinely overlap?

Across almost all of the intellectual content, which is exactly why the confusion is so durable. Sitting behind somebody doing either one, for most of the working day the two are indistinguishable.

Both read the same disclosures, and read them the same way: the notes before the face of the statement, the segment tables before the summary, the cash flow before the profit ladder. Both build a model, and for the same reason. A model carries a set of assumptions through to whatever those assumptions imply, and there is no other way to see what has quietly been committed to. Both frame the question before answering it. Work that has not decided what it is asking will answer whatever is easiest to answer. And they both test the claim, deliberately, against the evidence that would break it.

Take one concrete stretch of work. Between year two and year three, Sarvani Coatings Limited's gross margin went from 44.0 to 46.0 per cent. Anybody doing either activity would refuse that at face value and would ask first what it is arithmetically. Since the only thing deducted above the gross line is the materials bill, a gain of two points in margin and a drop of two points in what materials absorb, 56.0 down to 54.0, are not two separate findings. The two findings are one number written out twice. Restating a figure in a different form feels like progress and explains nothing, and catching that is the first thing either activity does.

The second stretch is the same in both too. Revenue rose 13.9 per cent while volume rose 6.0 per cent, so realisationThe money a company gets for each unit it sells, which moves both with price and with which products happened to go out of the door. Where it is decomposed properly is the earnings material. per unit climbed about 7.5 per cent. The materials bill climbed 9.9 per cent on that same 6.0 per cent of extra volume, so each unit made absorbed roughly 3.6 per cent more material cost than it had a year earlier. Input costs per unit went up rather than down. The margin improved because what Sarvani Coatings got for each unit outran what each unit cost it, and for no other reason at all. The decomposition is the same work whoever does it and whatever they call themselves.

So if the reading is shared, the model is shared, the framing is shared and the testing is shared, what exactly is left over? The leftover is the whole question, and the honest answer is that it is small, is not analytical, and changes the output more than the analytical part does.

What does the job carry that the method does not?

Five things, and it is worth saying at the outset that not one of them is analytical. The five are the surroundings, and the surroundings are the whole of the difference.

A calendar and a coverage list. A person in the job carries a set of names, and outputs are expected on each of them on a rhythm the results calendar sets. Neither the names nor the rhythm is chosen by the person doing the work, and neither is chosen by the evidence.

An audience with its own decisions to make. The reader is not reading to learn. The reader is deciding, has three other notes open, and will give the first paragraph more attention than the next four combined. A deciding reader reorders the document: the claim goes in front, the background goes at the back or nowhere.

Distribution, and therefore regulation. The moment the document leaves the desk it is a distributed view, and a set of obligations attaches to it that has nothing to do with whether the argument is any good. A conflict disclosureThe statement of interests that has to travel alongside a distributed view, saying what the writer holds and what connections exist. Its contents are fixed by regulation and not by whoever is doing the writing. is the visible edge of this, and there is more behind it.

A house style. There is a length, a template, a place the claim sits, a way tables are laid out and a vocabulary the house uses. None of it is analytical and all of it is compulsory.

Relationships with the companies covered, and the conflicts that follow. The person in the job talks to the company, and needs to keep talking to it. The person's employer may have other business with that company. Neither fact establishes that the work is compromised, and both are exactly why disclosure exists rather than trust.

Look at the shape of that list. A calendar, an audience, a rule about sending, a template and a set of relationships. Every single one is a fact about the circumstances rather than about the argument, and every single one changes the output that reaches a reader. The uncomfortable part is that the non-analytical surroundings shape the published record more visibly than analytical ability does.

What does the method carry that the job does not?

Two things, and the second one is the sharpest single difference between them.

The first is reach. The method applies to any instrument, and to things that are not instruments at all. Point it at a bond and it works. Point it at a preference share, a lease, a supplier's receivable, a whole sector, a privately held business with no quoted price, and it still works. None of its requirements mentioned a listing. The job cannot follow it there. A person paid to cover listed coatings makers may not spend the quarter on an unlisted one, however interesting the unlisted one is. The method has no boundary of subject matter and the job is defined by one.

The second is the right to say nothing. The method permits three answers a coverage list does not: nothing has changed since the last time, so there is nothing worth writing; this cannot be analysed with the evidence available, so no view is available; and this sits outside what I can competently judge. Each of those is a legitimate conclusion. Each is sometimes the only honest conclusion. The obligation runs to the name and not to the evidence, so a coverage list cannot accept any of the three. The asymmetry between them is the sharpest single difference between the job and the method.

Sit with what that does over years rather than over one quarter. Twenty names, four quarters, and eighty outputs have to exist. In how many of those eighty did something genuinely change? Nobody knows, but it is not eighty. The gap between the number of outputs and the number of changes is filled by writing, and the writing is honest, and it is still filling a gap that the evidence did not create. The gap is a structural feature of the published record, and it has nothing to do with anybody's ability.

There is an everyday version of this that makes it obvious. Think of a school that requires every teacher to write a paragraph on every child every month. Some months a child has genuinely changed and the paragraph writes itself. Most months nothing much has happened, and the paragraph still has to exist, so it gets written carefully out of very little. Nobody in that school is lying. The requirement simply does not match the supply of things worth reporting, and a reader who does not know about the requirement will read every paragraph as though it were prompted by something.

When does each one decide the work is finished?

The stopping rule is the quietest of the differences and the most damaging of them. The rule operates every single week and almost nobody names it.

The method has a stopping rule and it is about the question. The work is finished when the question the claim depends on has been settled, or when it has been established that the question cannot be settled with the evidence available. Both of those are genuine endings, and the second one is an ending most people are reluctant to accept as an ending. Neither of them appears on a calendar anywhere.

The job has a stopping rule too, and it is about the clock. The work is finished when the window closes. One rule looks at the question and the other looks at the calendar, and there is no reason at all why the two should land in the same place.

Most weeks they land close enough together that nobody notices the difference. Look at Meghna Iyer's week again. By Friday the volume and realisation split is settled, the three candidate explanations are written down, and the one question that would separate an environment in which every maker could price ahead of its inputs from Sarvani Coatings' own pricing is not settled and cannot be settled from published statements at all. So the method's instruction is to state what is settled, name what is not, and stop. The job's instruction is identical, with three words added: and send it. The two instructions sit together perfectly comfortably.

The uncomfortable week is the one where the unsettled question is the question the whole claim rests on. The method then says plainly that no view is available yet. The window closes anyway. The honest part is worth stating carefully. A good research seat handles that exactly the way the method does, by publishing something that says which question is open and why it is open. The pressure that exists is not a pressure to lie. An open question reads to a hurried reader as weakness rather than as precision, so the pressure is to write as though an open question were a closed one. The pressure comes entirely from the audience, and the audience is a feature of the job.

There is a version of this outside finance that everybody has met. A mechanic who has found the fault and a mechanic who has to hand the car back at five o'clock are working to two different stopping rules. Most days the fault is found before five and nobody ever learns that the two rules were different. The day the fault is not found by five is the day the rule the garage actually runs on becomes visible. A model built to a deadline is not automatically a worse model, but it is a model that stopped where the clock was rather than where the question was, and that is equally true of a three line margin bridge and of a full discounted cash flowA valuation method that converts a stream of expected future cash into a single present figure using a rate. How one is built is covered where valuation method is taught..

Why has the confusion between the two lasted so long?

Because three separate things push in the same direction, and none of them is anybody's mistake.

The first is the job title. The person holding the job is called an analyst. Not a researcher, not a writer, an analyst. The word for the method is sitting inside the name of the role, so every conversation about the role uses the word for the method a dozen times a day without anybody noticing they have.

The second is the founding text. Graham and Dodd called the book Security Analysis, and the subject matter of the book is very largely the subject matter of the job: listed companies, their statements, and what a price is assuming. So the book that established the method as a method reads, to somebody skimming, like a book about the job. The name and the contents point at each other.

The third is how people meet it. Almost nobody encounters the method first and the job second. People see the job first, in a description of what analysts do, in somebody's account of their week, in a note somebody forwarded. The method is inside all of that, but it is inside it, and what is visible from outside is the artefact and the routine. People learn the container because the container is what they can see, and the contents are what actually transfers.

The cost of the confusion is not that anybody says the wrong word. The damage is that somebody entering the field spends a year learning a role when a method would have served them. The role is not portable: change employer and the template changes, the coverage list changes, the house vocabulary changes, and half of what was learned has to be relearned. The method is completely portable, and it is the part that would have survived every one of those changes.

What is the real difference between holding a job and practising a discipline?

A job can be held and it can be lost. Somebody grants it, somebody can withdraw it, and holding it means being inside an arrangement with a name, a seat and a set of expectations. Nothing about that is a comment on the quality of the work. Seats close for reasons that have nothing to do with the person sitting in them.

Nobody granted a discipline, so it cannot be held or lost. A discipline can only be practised, and practised well or badly. There is no ceremony, no appointment, and nobody to withdraw it. The method is what remains when the job changes, and that is the entire reason it is worth learning first.

Follow one working life across three seats. First a broking house, where the readers are institutions and the coverage list is long. Then an asset manager, where the readers are four colleagues and the list is short. Then an independent practice, where the readers are whoever subscribes. Every one of the surroundings changed at each move. The template changed, the audience changed, the obligations changed shape, and the rhythm changed. None of the method's requirements ever mentioned any of those things, so the method did not change at any point.

What does one results week look like when every activity is labelled?

Abstractions about jobs and methods stay slippery until they are set against a week somebody actually worked. Meghna Iyer's week around Sarvani Coatings Limited's year three results serves for that: each activity is marked as belonging to the job, to the method, or to both, and the seam shows where it falls.

The test for the marking is one question, and it is worth memorising because it settles almost every case. Would this activity still be worth doing with no employer and nobody at all to send the work to? If yes, it belongs to the method. If no, it belongs to the job. The test says nothing about difficulty, nothing about how much time the activity takes, and nothing about whether it is enjoyable. Recording holdings takes four minutes and belongs squarely to the job. Deciding which evidence would settle the question can take a day and belongs squarely to the method.

Here is the week, activity by activity. Nine of them, in the order they happened.

Count them. Four belong to the job alone, three to both, and two to the method alone. Now count only the five that carry any analytical content at all, being the first five in the table. Three belong to both and two to the method, and the job's own share of the analytical work is zero. The job contributes not one analytical activity of its own to the week, and it decides when the week ends.

The last row is where the two most visibly pull apart. Sit with it. Next quarter, suppose genuinely nothing has changed: no move in input costs, no shift in the segment split, no new disclosure. The method has a clean answer available. Nothing new is worth reporting, and the previous work stands. The coverage list does not permit that answer. Something will be written, and it will be written honestly and competently out of very little, and a reader who does not know a list exists will read it as though it had been prompted by something.