Rating Actions: Upgrade, Downgrade, Outlook and Watch

A rating action is any published change to what a rating agency says about a borrower. Two of the four kinds move the assessment itself to a different position on the agency's own scale, and those two are an upgrade and a downgrade. The other two move nothing. An outlook speaks about a direction, a watch says a review is running, and after either one the position is where it was.

What does a rating agency actually publish, and how many kinds of thing are there?

The confusion comes before the definition. A rating agency puts out documents. Some of those documents change the agency's own opinion of a borrower. Some of them talk about that opinion without touching it. Both arrive under the same masthead, both carry the same borrower's name in the headline, and readers routinely file the second kind as though it were the first.

A credit ratingA rating agency's opinion about a borrower's ability and willingness to pay what was promised, expressed as a position on the agency's own scale. is a rating agency's opinion about a borrower's ability and willingness to pay what was promised, expressed as a position on a scaleThe ordered set of positions an agency defines and maintains. Each agency writes its own, and each writes its own definition for every step of it. that the agency itself defines and maintains. The definition just given is the whole of what is needed here. Claims a credit rating makes, silences it keeps and weight it can carry are covered separately and at far greater length.

Notice what that one sentence does not do: it does not print the scale, it does not name a single step on it, and it does not say what any step means. A scale, and the definition an agency attaches to each step of its own scale, are set by the rating agencies in their published method documents and by the Securities and Exchange Board of India (SEBI) at sebi.gov.in, and they are revised. A recital written from memory is not stale on the day it changes; it is simply wrong, and it is wrong in a way a reader cannot detect.

So here is the shape to hold. The agency publishes an assessment. An assessment is a position. The agency also publishes commentary on that assessment, and commentary is talk about the position. One word, rating, gets stretched across both, and almost every misreading in this material comes from that stretch.

An everyday version sits closer than it might seem. A school sends home two envelopes. One holds a report card with a grade on it. The other holds a note from the class teacher saying she has some concerns and will be looking at the work again next month. The note may be more alarming to read than the report card, and it may well end up changing next term's grade. But it is not a grade. Nothing in the school's record moved when it arrived. A parent who files the note as a grade has recorded something that did not happen, and a parent who files a genuinely lowered grade as a note has done the opposite and worse.

The two envelopes hold the whole subject in household form. Everything else is the finance names for those two envelopes, and the work of telling which is which.

Which announcements move the assessment, and which only talk about it?

One test sorts every announcement that can arrive, and it is short enough to apply before the eye reaches the second paragraph of the document. Ask whether the position on the scale moved. That is it. There is no second question at this stage and no weighing of how grave the language is.

The five things an agency can send all run through it. An upgrade moved the position. A downgrade moved the position. A rating outlookA statement about the direction an assessment may take over a longer period, published while the assessment itself stays exactly where it is. did not. A rating watchA statement that an assessment is under review, usually because of a specific event, published while the assessment itself stays exactly where it is. did not. A research updateA publication in which an agency sets out its reasoning about a borrower while changing no assessment, no outlook and no watch status. did not. Two of the five move it, three of them do not, and the split does not follow how worrying any of them sounds.

The reason the test is worth having, rather than merely being tidy, is the order it forces on the reading. Reasoning is persuasive. A well argued document about a borrower's cash generation will leave almost any reader with the feeling that something has been decided. Good argument does that to people. Settling whether anything moved before the argument gets to work leaves the argument to be read as argument. Settling it afterwards means reading an impression back into the document and calling it a finding.

Keep one more thing in view before the individual definitions arrive. A change to the outlook, or the placing of a watch, is still a rating actionA published change to what an agency says about a borrower, which includes a change to the assessment, to the outlook or to the watch status.. The agency has changed something it publishes as its position. The assessment itself is simply untouched. So a rating action is the wider box and a change of assessment is the narrower box inside it, and the two are not the same size. The distinction between the wider box and the narrower one does more work later than it looks capable of doing here.

What is an Upgrade, and what exactly has moved?

An upgradeA rating action that moves the assessment to a better position on the agency's scale. The scale, and what a better position means on it, belong to the agency. is a rating action in which the agency moves its assessment of a borrower to a better position on its own scale. Two statements are both true here and readers usually hold only one of them.

The first is that something genuine has changed. The agency's opinion is not what it was yesterday. Whatever documents, mandates and internal rules point at that opinion are now pointing at a different value, and anything downstream that reads the assessment as an input has a new input.

The second is that the borrower did not change on the day the announcement appeared. Whatever the agency saw, it saw over months. The cash generation improved, or the amount borrowed came down, or a contract that had been in doubt was signed, and none of that happened at nine o'clock on the morning of publication. An assessment follows a view that formed over a period and is then published at a single moment, so the announcement is the first time the change is visible and is almost never the first time the reasons were.

Feel that with something ordinary. A shopkeeper who has been letting a customer buy on credit watches that customer settle every month for a year and a half, and at some point decides to raise the limit. The decision gets written in the ledger on a Tuesday. The customer's reliability was not created on Tuesday. The reliability accumulated, and Tuesday is only when the shopkeeper's book caught up with it.

The gap between the two dates matters for a practical reason and not a philosophical one. In working out whether an upgrade carries anything that was not already known, the question is not what changed at the borrower. The question is what the agency knew and when, against what everybody else knew and when. The comparison is genuinely hard and is settled elsewhere. The smaller point holds regardless: the date of the announcement is not the date of the change.

What is a downgrade, and what does it carry that an upgrade does not?

A downgradeA rating action that moves the assessment to a worse position on the agency's scale, in the same way an upgrade moves it to a better one. is a rating action in which the agency moves its assessment of a borrower to a worse position on its own scale. Structurally it is the mirror of the upgrade: the assessment has genuinely changed, and the borrower did not change on the morning of the announcement.

One thing is different, and it is not sentiment. A downgrade can set off consequences that were written into documents long beforehand. Loan agreements, trust deeds, the mandates that tell a pooled vehicle what it may hold, and the rules about who may hold what, can all name an assessment as a trigger. When the assessment moves, those clauses read the new value and do whatever they were written to do. Documents are written to protect against deterioration rather than to react to improvement, so an upgrade has no comparable machinery attached to it as a rule.

The consequences themselves belong to the documents and to the rules on who may hold what, and SEBI sets those rules at sebi.gov.in. The rules change and they differ between kinds of holder, so a summary written from memory would be a confident sentence with nothing behind it.

There is a dignity point worth stating plainly too. A downgrade is not a verdict on anybody's competence, and a borrower whose assessment fell is not thereby a borrower that should obviously have avoided it. Businesses meet weather they did not order. The arithmetic is the subject; the judgement is not.

What is a rating outlook, and what has not moved?

A rating outlook is the agency's statement about the direction its assessment may take over a longer period. An outlook is published alongside the assessment, and it is about the assessment rather than being part of it.

Two facts carry this whole section: the assessment has not moved, and an outlook is not a commitment to move it. Both halves get lost. The first gets lost because an outlook reads like news. The second gets lost because a direction, once stated, feels like a plan, and a reader starts treating the eventual move as scheduled rather than as possible.

The horizon an agency attaches to an outlook, and the words it uses for each direction, are set out in that agency's own published method document. The horizon and the wording are not general facts about the world and they are not identical between agencies. How long an outlook is supposed to look ahead is therefore answered in that document, by the agency that published it, and the answer should be read there rather than recalled from elsewhere.

An outlook does give something cheap and real. It says which way the agency currently leans, on its own account, in its own words. An outlook can also lapse. The period passes, the concern that produced the lean fades, and nothing is ever done about it. A lapsed outlook is a perfectly ordinary ending and it is not a failure of the outlook.

What is a rating watch, and why is it not a decision?

A rating watch is the agency's statement that its assessment is under review. Usually a specific event sits behind it: something has happened, or is expected to happen, whose effect on the borrower the agency has not yet worked through.

The same two facts hold, word for word: the assessment has not moved, and a watch is not a decision. A watch is the announcement that a decision is being worked on. The distance between those two sentences is the distance between a court listing a case and a court delivering a judgement, and nobody would confuse those two, yet the equivalent confusion happens constantly here.

The horizon a watch is expected to run for, and the conditions that trigger one, are the agency's own and are set out in its method document alongside everything else. Named here, not reproduced, for the same reason as before.

A watch says one thing an outlook never does. Somebody is actively working. A review that is expected to resolve is a review with an end. An ending is a different kind of information from a lean. Setting the two side by side beats leaving a vague sense that they are two words for concern.

Rating Outlook vs Rating Watch: who is looking, and who is leaning?

Both leave the assessment exactly where it was, and the shared fact is why they get merged. The pair are not, however, two intensities of the same thing. Outlook and watch differ in what caused each one and in what is expected to follow it.

An outlook attaches to a general direction over a longer period. An outlook can be acted on and it can simply lapse, and neither ending is surprising. A watch attaches to a specific review that is expected to reach a conclusion. Something set it running, and something is expected to stop it.

The shortest way to hold the pair is this: a watch says somebody is currently looking, and an outlook says somebody currently leans. Looking has an end. Leaning does not need one.

One trap to name while the pair is in view. Neither is a measure of severity. A watch is not automatically more serious than an outlook, and an outlook is not automatically the milder of the two. The two answer different questions. Ranking them by alarm is a habit worth losing early.

Downgrade vs Rating Watch: which one has already changed the position?

Downgrade and watch are the pair that costs the most to confuse, and the confusion has a different source. An outlook and a watch get merged because they resemble each other. A downgrade and a watch get merged because they feel the same, and feeling is a poor guide here.

A downgrade is a completed action. Whatever the position on the scale was, it is not that any more. A rating watch is an announcement that a review is under way. Whatever the position was, it is exactly that.

One fact matters downstream, and it is whether the position on the scale is now different. Two documents can read almost identically field for field and still differ on that one fact. Both name the same borrower. Both come from the same agency. Both set out reasoning about the same pressures. On both, the least dramatic field in the document is the one that separates them.

Both directions of this error cost something, and they cost different things. A watch recorded as a downgrade puts a change that did not happen onto the record, which means anything a process attaches to a real change starts moving on the strength of a review that has not finished. A downgrade recorded as a watch files a completed change as a possibility. The second error is worse in a quieter way: the thing has happened, the consequences written into documents are already live, and the note says it might.

Research Update vs Rating Action: did anything the agency publishes move?

The third pair is the one people forget exists. A research update is a publication in which an agency sets out its reasoning about a borrower and changes nothing: not the assessment, not the outlook, not the watch status. A rating action changes at least one of those three.

Apply the same test and a research update branches away at the first question and never comes back. The detail in it, the new information it carries and how uncomfortable it is to read make no difference to the sort. Nothing the agency publishes as its position moved, so what arrived is reading material.

Reading material and an action are different objects even when the reading material is the more useful of the two. That is worth saying because the test can look dismissive, and it is not. A research update can be the most informative thing an agency publishes all year. Sorting it correctly does not mean valuing it less; it means not recording it as a change to something that did not change.

Note the asymmetry the right branch hides. Most readers miss it. A change to the outlook, with the assessment untouched, is a rating action. The outlook change goes down the yes branch. So the right hand box is not a synonym for the assessment moving, and the earlier distinction between the wider box and the narrower one inside it is doing its work here.

What can be filled in about Palash Cements Limited, and what stays empty?

Palash Cements Limited, an invented cement maker and the only issuer in this record, has no credit rating and has never been the subject of a rating action. The worked instance is therefore an absence where an announcement would sit.

Inventing a rating would mean inventing a scale to express it on, inventing a step on that scale, and inventing a definition for the step, and all three of those belong to the rating agencies and to SEBI at sebi.gov.in. A default frequency comes from counting real defaults across many borrowers over many years, and a recovery rate from studying what creditors actually got back after real failures. Neither exercise is arithmetic performed on one bond's price.

So the worked instance is a blank announcement with its fields labelled and left empty, and the emptiness is the teaching rather than a gap in it. Nine fields sit on a rating announcement: the agency, the issuer, the instrument, the assessment before, the assessment after, the outlook, the watch status, the type of action and the reason. The issuer and the instrument are already fixed, so two of the nine can be filled. Seven cannot.

Beside that blank document sits the single row that can actually be computed, and it is worth seeing what kind of row it is. The row is not a rating. The row is a price.

The subtraction can be repeated by anybody holding the two rates, and it required no agency, no scale and no opinion. Now push it one step further and watch a second kind of number appear, one that needs an assumption before it will exist at all.

One filled row and eight empty ones, and the contrast is the entire worked instance. A price implies something anybody can compute. Nobody has given an agency opinion, so there is nothing of that kind to compute.

What does a rating action say about the price?

Every reader arrives with this question, and the answer is a refusal with reasons attached rather than a shrug. Answering it takes a rating, a rating action, a series of spreads recorded over the days around that action, and enough separate cases to tell a pattern from a coincidence. Palash Cements Limited supplies none of the four. Producing a number would mean inventing one, and an invented number about how prices respond to announcements is exactly the sort of thing a reader carries away and repeats.

One thing can be said structurally, and it is worth more than it sounds. A price already contains whatever the people dealing in the bond believe. An announcement changes a price only to the extent it contains something those people did not already believe, and nothing about the announcement itself says how much of that there is. A document can be full of grave reasoning that everybody had already worked out for themselves, and a document can be brief and contain one fact nobody had.

Three limits travel with the implied default rate wherever it goes, and none of the three is optional.

First, the 40 per cent recovery is an assumption. The figure was supplied because the arithmetic needs a recovery rate, not because anybody measured one. Holding the 220 basis point spread perfectly still and moving only the assumption moves the answer with it. The table below is worth reading as a warning rather than as a reference.

Same price, four answers, and nothing in the arithmetic prefers one of the four over another. The assumption is doing that much of the work.

Second, the whole spread has been treated as compensation for credit. In a real market some part of what a borrower pays covers the difficulty of selling the bond on demand, and every basis point of that counted as credit pushes the implied rate too high. Carve 0.40 percentage points out of the 2.20 and treat them as payment for something other than the borrower failing, and 1.80 points are left as credit. At the same loss given default of 0.60, 1.80 points imply 3.0000 per cent a year instead of 3.6667. Separating the two parts takes evidence about how easily the bond changes hands, and the arithmetic never had it.

Third, an implied default rate is what the price says. An implied default rate is not a forecast and not a measured frequency of anything. Nobody counted defaults to produce 3.6667 per cent a year. The number was solved backwards out of one spread and one supplied assumption, and reading it as the probability that Palash Cements Limited fails misreads the arithmetic that produced it.

How to interpret a Credit-Rating Action: which questions, and in what order?

Five questions, in a fixed order, and the order is the part people skip. Asked in sequence they take under a minute on any announcement.

What moved. The assessment, the outlook, the watch status, or nothing at all. Everything else reads differently depending on the answer, so the sorting test comes first.

On whose scale. Which agency published it. A scale belongs to the agency that publishes it and no two are interchangeable, so a position without an agency attached is not a usable fact. Two agencies can hold different views of the same borrower at the same time, and neither is thereby wrong.

What did not move. Name the unchanged parts out loud. If the outlook changed and the assessment did not, say both. The unchanged parts are half the information, and they are the half a reader silently discards.

What triggered it. A specific event, or a review that comes round on a cycle. The announcement itself usually states which, and the two mean quite different things about how much the agency has newly learned.

And what had the price already done. Asked last, and the position is deliberate. A reader who asks this first has an answer in hand before the other four questions are put, and will read that answer back into all of them. Ask it fifth and it stays what it is: one more observation about a price, not a verdict on the announcement.

So what is a rating action honestly good for?

Less than a reader arriving cold usually hopes, and still enough to be worth the minute it takes. A rating action says that an agency changed something it publishes as its position on a borrower, and it says precisely which of the three things it changed. Which of the three moved is a fact about the agency's published opinion, dated and attributable, and documents can be written around a fact of that kind.

A rating action does not give what a bond is worth, whether a spread will widen or narrow, or how likely any particular borrower is to fail. A rating action is not a measurement. A rating action is an opinion with a publication date, expressed on a scale its author maintains, and the whole of its usefulness sits inside that description rather than beyond it.

The one habit worth carrying away is smaller than any of that: before a view about an announcement is formed, find the field that says what the assessment was and what it is, and write both down even when they are the same. Everything else is scaffolding around that sentence.