How to update Credit Analysis After a Credit Event

The work runs in a fixed order, starting at the document rather than at the figure already in hand. Check the occurrence against its list of conditions. Establish who declared, and under which clause. Separate what the document moved from what the business moved. Mark the carried figures nobody measured. Set down what has become less certain. Fetch each rule the work leans on. Record what was read, and when.

The order is the whole of the method. A procedure that promises a conclusion has stopped being a procedure and become a prediction, so not one of the seven steps states what the looking will find. Every step below names a place to look and a thing to write down. Where the looking ends is not something a procedure can see, and it is said plainly at the close rather than dressed up.

Why does a view that is mostly still right need rebuilding at all?

Consider the sheet in use last week. A rate written into a contract. A government level read off a screen. The distance between the two. A default rate somebody worked backwards out of that distance. Perhaps thirty more lines under those four, all of them fed by the four.

Now something is declared. The natural move, and it is natural in a way that makes it dangerous, is to go down the sheet fixing whatever plainly needs fixing and leaving the rest alone. The trouble is not that a view built before an event is wrong; the trouble is that most of it is still right, so the parts that stopped being trustworthy do not announce themselves by looking odd. A figure that has quietly lost its support renders in the same font as one that never had any problem. Nothing on the sheet changes colour.

An order exists to force each line past the eye exactly once, including the lines that look untouched, and above all including the lines that were never measurements to begin with. Running the order costs a morning. Skipping it costs a sheet that reads as though it had been reviewed.

What are the seven steps, and why is the order the whole of the teaching?

The sequence in full, before any of it is unpacked, is best read once as a shape. Each row names where the work goes and what it comes away holding, and nothing in any row states what the looking will turn up.

Two features of that ladder are worth pausing on. Step one starts at a document rather than at a figure, and that single choice decides everything downstream. The instinct after bad news is to open the model, and the model is where the view lives and where the effort went. But a model is a description of an instrument, and what counts as an event was settled in advance, in one particular document, in words chosen long before anybody was under pressure.

The other feature is steps three and five. Neither step appears in the procedure most people carry in their heads. Step three exists to stop a borrower being credited with an arithmetic change it had no part in. A lost certainty leaves no gap on the sheet where a missing number would, so step five exists to force those losses of knowledge onto paper, and almost nobody does it.

Step one: which list is the occurrence checked against?

One list, and it is the list of defined conditions in the terms this particular bond was issued under. Not a general idea of what counts as trouble. Not what a headline has called it. Not what a colleague two desks away, holding a different instrument from a different borrower, says happened at their end.

The reason is that the list was drawn up while everything was calm. Somebody sat down before a rupee changed hands and wrote out the occurrences that would carry consequences, precisely so that nobody would have to decide the question later under pressure and in public. Whether an occurrence counts is a question with a written answer, and step one is nothing more than going and reading it.

Reading it produces one of exactly three results, and only one of the three leads on to step two.

The middle outcome deserves a moment. Finding that an occurrence is not on the list is not a wasted morning and it is not a soft answer. The finding is worth recording at step seven, and it will save the person who reads the work in six months from starting the same search from nothing.

Step two: how is it established that anybody has actually declared?

An owner lets out the ground floor of a house on a written agreement. Clause four lists what counts as a breach; clause nine says the owner may end the tenancy by giving notice in writing. One morning the owner sees something that clause four plainly covers. The seeing of it changes not one thing. The tenancy runs on precisely as before, and it goes on running until somebody entitled to give that notice gives it.

An occurrence sitting on a list stands in exactly that position. Facts on their own do not move a document; what moves it is a declaration, meaning somebody the terms entitle, acting under a clause the terms specify, stating that a defined condition has been met. The clock run across that act shows the change. On the near side there is an occurrence and no status. On the far side there is a status, and whatever the terms hung on it has come within reach of the party the terms put it within reach of.

So step two is two lookups rather than one. First, the clause naming who may act. Where thousands of holders stand behind one bond, that role is commonly handed to the trusteethe party a bond's own terms appoint to act on behalf of everyone holding it, so that a thousand holders need not act one by one. Thousands of people cannot each serve something of their own. Second, whatever record exists that the act has taken place: a default noticethe written communication by which the party entitled to act says that a listed condition has been answered is the usual shape that record takes, and its absence is as much a finding as its presence.

The same facts given to a pair of readers can leave them in contradictory places on whether a condition is met. Disagreement of that kind happens constantly, and arguing it out settles nothing. The entitlement to declare rests with one of them, and the terms say which one. The Securities and Exchange Board of India (SEBI), at sebi.gov.in, decides what an issuer of corporate debt must disclose once a dated payment goes unmade, and what falls on the trustee from that moment. The live wording sits at that address.

Step three: what moved because of a clause, and what moved in the business?

Two columns, one sheet, and every item that moved goes into one of them before another word is written.

Into the left column goes everything that moved because a sentence in a document was exercised: what is claimable, from whom it is claimable, and who may now act without needing anybody else to agree. Into the right column goes everything that moved in the business itself. The two columns move at different moments and for entirely different reasons, and a reader who lets them run together will hand a borrower the credit or the blame for an arithmetic change the borrower had no hand in that morning.

The right column of that drawing carries no figures, and the emptiness is not a gap in the drawing. Filling the right column takes a description of how a business actually trades, and a rate, a term and a face amount do not amount to one. The left column takes arithmetic on figures already in hand, and that arithmetic comes below.

What does that separation look like once there are rupees on it?

Palash Cements Limited, an invented borrower, comes down to a rate, a term and a face amount. The document sets five annual dates. On each of the first four, Rs 91/- falls due. On the fifth, that same Rs 91/- arrives alongside Rs 1,000.00/- of face. Where does the Rs 91/- come from? A contracted rate of 9.10 per cent a year, struck on Rs 1,000.00/-, produces it exactly.

One discounting period a year is what all of this is built on: an amount landing five years out gets divided by 1.0910 once for every year between now and then. Work the five dated amounts that way and they add back to Rs 1,000.000000/-. An exact return to the face amount is what puts this bond at par, and it is why the rate written into the contract and the rate the discounting uses read as one and the same number.

Now a supposition, worked forward from the figures already on the table. Suppose one scheduled payment is not made.

With nothing declared anywhere, a single payment has gone unmade and the shortfall stands at Rs 91/-. Once a declaration lands that carries accelerationa written consequence under which money that was due only at the end of the term becomes claimable straight away, the face amount joins the shortfall, so what can be claimed comes to Rs 1,091.00/-. Set that against the face amount and it reads 109.10 per cent of it. Set it against the first reading and it stands at 11.9890 times as much.

Now look hard at what did not move. Underneath both readings sits one borrower, unchanged. Whatever it does all day, it was doing yesterday. Payments already made stay made. A sentence drafted years earlier, plus somebody's decision to use that sentence, are between them the whole cause of a figure growing almost twelvefold. Take the two column sheet from step three: all of that goes on the left, and nothing at all lands on the right.

A second feature of that drawing looks like a coincidence and is not one. The accelerated claim of Rs 1,091.00/- is exactly what the fifth scheduled date was always going to pay: one coupon of Rs 91/- plus Rs 1,000.00/- of face. The two figures agree because they are the same two components added in the same way. The arithmetic is forced rather than a happy accident. The two are still different objects. One is a payment due at the end of a term. The other is an amount claimable now. A sheet that lets those two share a cell has thrown away the only distinction that matters.

Step four: which of the figures being carried did anybody measure?

Three years ago somebody in a household guessed at what is spent on medicines in a month, wrote it into the budget sheet, and moved on. The guess has been totalled, carried forward and quoted in arguments ever since. Nothing about the way it sits in the column says it began as a guess. The line looks precisely like the rent, and the rent is a figure somebody actually pays.

Step four is going down the sheet with a pen and putting a mark against every number that was supplied rather than seen. The marking is slow, it feels like clerical work, and it is where the one figure that always survives a review finally gets caught.

Take the four in order. A rate of 9.10 per cent a year is a line of text inside a document, and a finger can be put on it. The government SPOT curve, read at its five year point, gives 6.90 per cent a year, and that figure came off a screen. The 2.20 percentage points between the two, or 220 basis points, is a subtraction performed on a pair of things anybody can point at, and there is no clause anywhere that creates it.

The fourth is a different animal. An implied default ratea yearly rate worked backwards out of a price, under a stated assumption about how much comes back, rather than counted from anything that occurred of 3.6667 per cent a year exists only because somebody needed to turn the spread into a rate, and turning a spread into a rate needs an assumption about how much comes back. A recovery of 40 per cent of the amount owedthe sum a claim is measured against, which is a different quantity from whatever anybody paid to buy the holding was supplied to do exactly that. Sitting on the other side of the same assumption is 60 per cent. 2.20 percentage points over 0.60 lands on 3.6667 per cent a year. Pushed back the other way as a check, 3.6667 per cent of 0.60 returns 2.2000 percentage points.

Nothing counted that 40 per cent. No study sits behind it. The figure describes nothing whatever: not a borrower, not an instrument, not an occurrence. Applied to Rs 1,000.00/- of amount owed it produces Rs 400.00/- on one side and Rs 600.00/- on the other, and that pair is the entire stock of recovery figures available anywhere here. Both keep the assumption label every time they appear, this appearance included, and neither is ever tied to money that came back to somebody.

Step five: what is now less certain than it was yesterday?

Step five is missing from nearly every version of the procedure people carry around, and it is the one that keeps a rebuilt view honest. Steps one to four all add. Step five subtracts, and the subtraction has to be written down.

Before anything was declared, a schedule of dated payments was a perfectly reasonable base to do arithmetic on. Five dates, known amounts, known years. Everything downstream of that schedule inherited its solidity. After a declaration carrying acceleration, the schedule everything was built on is not the schedule any more. A figure whose base has been withdrawn is not resting on something out of date; it is resting on nothing, and on the sheet those two conditions look exactly alike.

Work it on the supposition already on the table. The whole promise, added up without discounting, is five payments of Rs 91/- and Rs 1,000.00/- of face, or Rs 1,455.00/-. Under the supposition, Rs 1,091.00/- of that is claimable now. The remainder is Rs 364.00/-, exactly the four coupons that were scheduled for later. Read as shares of the whole promise, Rs 1,091.00/- is 74.9828 per cent and Rs 364.00/- is 25.0172 per cent, and those two close on 100.0000 with nothing left over.

Here is the honest part. Whether those four coupons survive at all is a question for the document, and no document has been read here. Yesterday that Rs 364.00/- had a settled meaning: money promised on four dated occasions. Today it is less certain than it was, and the reason has nothing to do with the borrower.

Step five produced no estimate, no probability and no range. The output is a shorter list of things that can be stood behind and a written note of what left that list. A number carried forward with no mark against it looks exactly like a number somebody checked, and the note has to be written rather than remembered.

The same discipline catches a second class of item. If the document carries cross defaulta written link by which a borrower falling short under one agreement is treated as falling short under another one as well language, then arrangements nobody was looking at yesterday may need looking at today. Nobody yet knows what those arrangements say. Step five is where that lack of knowledge is written down.

Step six: which rules are fetched before going any further?

By this point there is a shorter view, a marked view, and a list of things no longer known. Step six establishes how much of what happens next is decided by somebody other than the analyst and the document.

The instruction is one line long: fetch each rule, do not recall it. A view built on a remembered version of a rule is not out of date on the day that rule is amended; it stopped being right at that moment, and nothing about the way it sits on the sheet will say so. A remembered rule is a harder failure than an out of date one. An out of date figure at least has a date on it.

Step seven: what is written down so that somebody can check this later?

Three things, and they take five minutes. Which version of the document was read. The date on which each rule was read. The date attaching to every figure carried in.

Recording is not filing and it does not belong in a footnote. A record of what was current at the time of writing is the only thing that separates a view that has gone out of date from a view that was wrong the day it was written, and nothing else in the world separates them. Two sets of work can look identical: same layout, same figures, same confident tone. One was correct against the rules of its morning and has since been overtaken. The other never was correct. Without the record, a later reader cannot tell which they are holding, and will usually assume the kinder of the two.

Three matters sit outside this order, and step seven is where that becomes visible. A choice about declaring belongs to whoever the terms entitle. A decision on an offer turns on wording that has not been drafted yet. Acting alone, or in company with other holders, is a legal question carrying legal consequences, and it is settled on a real document with legal advice rather than inside a procedure. So the order stops one step short of all three. The order points at the working parts, says who the terms put in charge of each, and names an address for everything else.

How does a lender or an analyst actually run this order?

Three people run it differently, and the differences are worth knowing because they show which step is expensive in which seat.

A credit officer inside a lender runs steps one, two and six hardest, and often in that order on the same morning. The reason is that a great deal of what happens to the holding on the lender's own books is decided by rules rather than by judgement: the label the holding takes, the figure it is carried at, how much of the lender's own funds get tied up behind it. All three are kept by the Reserve Bank of India at rbi.org.in and all three are fetched, never recalled. Step three matters to this reader in a specific way: the internal conversation about whether the borrower is in trouble is a different conversation from what the document now permits, and the two arrive on the same day wearing similar clothes.

An analyst's file is mostly carried figures, and almost nothing in it was measured by the analyst, so an analyst covering the instrument runs step four hardest. The point is not a criticism of analysts; it is what a file is. Recoveries, transition frequencies, an assumed severity behind a spread: each of them arrived from somewhere, most of them arrived without their labels attached, and every one of them will be reached for the moment somebody asks how bad this could get. Step four is half an hour with a pen, and it is the half hour that decides whether the rest of the work is honest.

A holder who is not an institution at all runs steps one, two and seven and can safely be slow about the rest. Read the terms. Find out whether anybody entitled to act has acted, and what a distressed exchangean arrangement in which holders are offered a changed promise in place of the one they hold, without any payment being missed would be if one were offered. Write down what was read and when. The last step costs nothing, produces nothing to look at today, and is for that reason the one most likely to be skipped.

And the everyday version is smaller than any of those. Anybody who has kept the receipt for a repair alongside the warranty card has already run step seven. The receipt is not the repair. The receipt is what lets somebody establish, later, what was true at the time.

What can this order honestly not give?

The order produces no conclusion, and no order of operations ever could. A conclusion comes out of a particular document, a particular declaration and a particular set of rules, read on a particular morning. The seven steps say where each of those three sits. Their contents, once they are opened, are beyond anything a procedure can settle in advance, and a procedure that promised otherwise would be a prediction wearing a checklist.

A reader may reasonably arrive expecting a worked case: a view as it stood, then the same view rebuilt. An order of operations with a named but unfilled space is the more useful object, and the reason is worth stating. One invented case becomes what a reader takes away, with the procedure left behind as scenery. A procedure of this kind is worth having in a year when nobody needs it.

An order of operations has no dial in it: each step names somewhere to look, and somewhere to look does not get better or worse by degrees. A control offering a choice of how badly things had gone would manufacture the trouble first and then invite the manufactured amount to be read as though it meant something. The last question below does the work such a control would have done.

Hindsight is the thing to be careful of here. Once something has occurred it takes on an air of having been obvious all along, and it was not obvious to the people who signed the document, took the rate on offer, or put money in fully expecting to get it back. Confidence of that kind is exactly why the seven steps exist: the whole apparatus was built by parties who understood that their own confidence could be misplaced. Nobody who signed, priced or bought ought to have known. A procedure is what is being taught, and a story about a borrower is not.

Where did every number in the arithmetic above come from?

Two invented figures sit under all the arithmetic above, and everything else was worked from them here rather than fetched from anywhere. A rate of 9.10 per cent a year on Rs 1,000.00/- of face is a teaching level and matches no borrower who exists. So is 6.90 per cent a year, read at the five year point of the government SPOT curve. No data service was opened and nothing was copied in from one.

The harder half of this note is about what is absent. The arithmetic above rests on a supposition rather than on an occurrence: no payment was missed, no notice was served, no promise was rewritten by agreement, and no money came back to a holder. The emptiness was chosen rather than left, and it is why the seventh step closes on a record instead of on a verdict. As for the recovery of 40 per cent that arrives at step four: somebody needed a figure so that a spread could be turned round into a rate, and 40 per cent was supplied to do that job. The 40 per cent measured nothing, and it keeps the assumption label at every appearance above, including inside the questions.

The rules this order sends an analyst to fetch are kept by other people, and they get amended. The Insolvency and Bankruptcy Board of India keeps insolvency procedure at ibbi.gov.in. Duties falling on a trustee, and on a borrower that has issued corporate debt, belong to SEBI at sebi.gov.in. How a lender carries and classifies a holding that has ceased to pay belongs to the Reserve Bank of India at rbi.org.in. The basis on which an expected credit loss is measured belongs to the Institute of Chartered Accountants of India at icai.org. The live text of each sits at the address beside it, on the day the question arises.