A Liquidity Assumption and a Liquidity Fact Compared

Two sentences about selling can look alike and behave nothing alike. One of them reports something that already happened, and it prints where a reader can go and see it, on what day. The other is supplied by the writer, and no reader can get behind it. Whether a stranger could check it without asking the writer is the whole of the difference.

Underneath that sits one awkward fact. Both sentences point at the same future event. Somebody has to agree to take the other side of a trade. The event has not happened yet, so neither sentence can see it. The separation lies in where each sentence was built from. One was built out of something already recorded, the other out of what the writer expects. Grammar does not carry that difference. A reader looking at the finished note cannot tell which is which, so the writer has to make the decision and mark it.

Most of the trouble on this subject comes from a firm grip on one of the two terms and a vague grip on its partner. Each one is worth setting out fully and separately before either is put against the other.

What does a liquidity fact have to carry?

A liquidity fact is an observation about dealing. Not a feeling about dealing, not a memory of dealing, and not a reasonable expectation of dealing. Something happened, somebody recorded it, and the record is somewhere a stranger can reach.

Three things have to be inside the sentence before an observation about dealing counts as a fact: what was observed, when it was observed, and where a reader can go and see it. Drop any one of the three and what is written is something else, however true it happens to be.

Take the first slot, the what. The what is the thing observed, named tightly enough that two readers would count it the same way. A quantity dealt on a named day is one. A quoted difference between what a buyer was offering and what a seller was asking, captured at a stated time, is another; that pair of quotes is what people mean by a bid and offerThe two live prices sitting either side of a market at one moment: the price somebody is willing to buy at, and the higher price somebody is willing to sell at.. A third is a count of dealing days: how many days inside a stated period saw any trade at all. Each of those names a countable thing. Compare a phrase such as reasonable activity. Two readers will count that phrase in two different ways, so it cannot be checked even when the underlying trading was real.

The second slot is the when. An observation belongs to a moment, and the moment is part of the observation rather than housekeeping attached to it. Writers treat the when most casually, and the cost of that casualness appears further down.

The third slot is the where. A reader who does not trust the writer has to be able to reach the record unaided. Reaching it means naming the published place, not describing it. The where is also the slot that most often turns out to be unfillable. Whether a record of dealing in a particular debt instrument is published at all, in what form, and by whom, is settled by a keeper rather than by the writer. What a trading venueA regulated place where buyers and sellers are brought together and trades are recorded. Which venues exist, and what each has to publish, is set by the keeper rather than by anyone writing about them. must publish about dealing in a debt instrument is set by the Securities and Exchange Board of India (SEBI), at sebi.gov.in, and government securities sit with the Reserve Bank of India, at rbi.org.in. Both keepers revise what they require, and a block further down carries the affected lines, one to a row.

Here is the everyday version. A shopkeeper says the small flat above the sweet shop is easy to let. The shopkeeper's sentence is one kind. The rent agreement register showing that flat let on four occasions in the last three years, with the dates, is a different kind of sentence, and anyone can go and look at it whether or not they like the shopkeeper. The register does not say the flat will let tomorrow. The register says what happened, when, and where it can be checked, and those are exactly the three slots.

Notice what the three slots do not include. The slots do not include how sure the writer is. Nor do they include how much experience sits behind the sentence. A fact written by somebody with two weeks of experience is still a fact, and a beautifully judged sentence written by somebody with thirty years of it is not made into a fact by the thirty years. The slots are about what the sentence carries, not about who wrote it.

What is a liquidity assumption, taken on its own terms?

A liquidity assumption is a statement about what could be done with a holding, supplied by the writer. The definition stops there, and everything interesting about an assumption follows from the word supplied.

An assumption may be extremely well founded. The sentence may come from twenty years of watching one kind of instrument change hands, in good weather and bad. In practice it may be more reliable than a single published observation from a thin day. The writer may turn out to be exactly right. The twenty years live inside the writer and the reader cannot get at them, so none of it turns the sentence into a fact.

The reflex that sentence fights is a strong one. People hear the word assumption and hear a weaker word, a placeholder, something written when the real work could not be done. On this subject that reading is simply wrong. Hardly any of the questions readers arrive with about selling have ever been recorded anywhere, so most useful statements about selling are assumptions and have to be. If a writer removed every sentence that could not be checked, what remained would be shorter, safer and close to useless.

So the work is not to eliminate assumptions. The work is to mark them, and to write them in a form the reader can argue with. An assumption written properly carries three things of its own: the word assumed, sitting in the sentence rather than implied by a careful tone; the reason this assumption was chosen rather than a different one; and what moves in the answer if the assumption turns out to be wrong. The third item is the one that gets dropped, and it is the one that makes the sentence worth printing.

Back to the flat above the sweet shop. An uncle who has lived on that street for thirty years says the flat will let inside a month. He is probably right. He is also not offering anything anyone can check, and a writer who sets his sentence down as though it came from the register leaves the reader of the note unable to tell the two apart. The honest version keeps his knowledge and marks its origin: it is assumed the flat lets inside a month, because that street has never had one stand empty in living memory, and if it takes three months instead, the sums change by this much.

An assumption is not a lesser kind of fact, and reaching for one is not a failure of research. It is a different kind of sentence with a different set of obligations, and a note built entirely from checkable observations would answer almost none of the questions readers actually arrive with.

What single question separates the two?

Everything above reduces to one line, short enough to hold in mind while drafting. Could a reader who does not trust the writer check this sentence without asking the writer?

A yes makes it a fact, and a fact owes its source and its day. A no makes it an assumption, and an assumption owes the word assumed. There is no third outcome. People resist the absence of a third outcome, so it is worth saying flatly: there is no bucket for a sentence that is probably true, no bucket for one that everybody in the market believes, and no bucket for one that rests on long experience. Sentences of that kind are assumptions with good reasons behind them. The good reason is welcome. The reason goes in the sentence, next to the word assumed, where the reader can weigh it.

The wording of the test matters in two places. The first is the phrase who does not trust the writer. A writer checking their own sentences will unconsciously grant themselves credit for what they know, and the phrase is there to block that. The question is not whether the sentence is true. The question is whether somebody with no reason to believe the writer could establish it for themselves. The second is the phrase without asking the writer. Plenty of sentences become checkable the moment the writer explains where the number came from. The reader will not be in the room, so the explanation belongs in the sentence.

Run the test on a handful of ordinary lines and it sorts them quickly. This instrument changed hands on nine days out of the twenty ending on a stated date, taken from a named published record: a stranger can go to that record and count, so it is a fact and it already carries what it needs. The market for this instrument is deep: nothing there can be checked, no source could settle it, and the word deep would be counted differently by any two readers, so it is an assumption and it needs marking. A sale is assumed possible inside a week at a price near the last recorded one: assumption, and a good one, because it is specific enough to be wrong.

One further point deserves its own sentence. The test does not reward vagueness. Making a sentence vaguer never turns it into a fact. Vagueness only makes the assumption harder to argue with, and that is worse for the reader, not better.

Why do the two look identical once they are written down?

Put these two lines next to each other. These bonds trade regularly. This instrument changed hands on nine days out of the twenty ending on a stated date, from a named published record.

The first line could be a faithful summary of the second. The same line could equally be an impression somebody formed over lunch. Nothing inside it tells the reader which, and that is the whole difficulty. The short line is not obviously worse than the long one: it is easier to read, it moves faster, and it will survive editing precisely because it is cleaner.

The fault is manufactured at exactly this point, and it is manufactured by people trying to help. An editor reads a long clause bristling with counts and dates, decides it is heavy, and shortens it. Nobody involved has done anything careless. But the parts that were removed are exactly the parts that made the line checkable, so what leaves the room is a different kind of sentence from what entered it. Compression is where a fact quietly turns into an assumption.

Two habits make this survivable. The first is to treat the counting parts of the sentence as load bearing rather than as decoration, and to say so when somebody proposes cutting them. The second is to keep a short version and a checkable version as separate sentences rather than as one sentence somebody has to choose between. Nothing stops a note from reading well and then, in the very next line, printing what the reading rests on.

The household version is familiar to anyone who has passed a message along. A neighbour reports that the electricity was off from two until half past four last Thursday. By the time that reaches the third person it has become the power keeps going off. Both statements come from the same event. Only one of them can be checked against a meter reading, and the one that survived the retelling is the one that cannot.

How does each one get written into a research note?

Both of them are about two lines long. A reader skimming for the argument reads the argument and not the apparatus underneath it, so neither of them belongs in a footnote.

A fact goes in like this. The thing observed is named, the period or day it belongs to is named, and the published record is named, all inside the running sentence. If the record has a title, the title is used. If it does not, the sentence says plainly what kind of record it is and who keeps it. The aim is a line a hostile reader could act on immediately without writing back.

An assumption goes in like this. The word assumed appears. Then why this assumption rather than a neighbouring one, in a clause, not a paragraph. Then what changes in the answer if the assumption fails. The third clause turns throat clearing into something a reader can price for themselves, and it is the part almost everybody leaves out.

Neither of those two shapes is a hedge. A hedge is a sentence written so that nothing could ever contradict it, and being uncontradictable is precisely what makes it worthless. Consider the line that this instrument may be difficult to sell in adverse conditions. No observation could ever prove that line wrong. The line costs the writer nothing and hands the reader nothing. Now consider the line that a sale is assumed possible inside a week at a price close to the last recorded one. The second line names a period and a price condition, so it can fail, and a reader who disagrees with it knows exactly what they are disagreeing with. Both a fact and a marked assumption are written so that they can be wrong. A hedge is written so that it cannot be.

One more habit, small and worth more than it looks. Put the two kinds of sentence in different places, or mark them typographically, or at minimum keep them in separate paragraphs. A reader moving quickly through a research noteThe written document an analyst produces on an instrument or an issuer. How one is put together end to end is covered separately. should be able to see, from the shape of the note alone, which parts rest on records and which rest on the writer.

What is an untestable assumption actually good for?

More than most people expect. A number nobody can check still buys something, and what it buys is the size of the hole where the observation should be.

Start with the ordinary position. For a great many debt instruments not one observation of dealing exists at all: no quantity dealt, no captured pair of quotes, no count of dealing days. So the cost of getting in and out of the instrument cannot be looked up. The cost has to be declared, and every figure grown from it has to say so in the same breath rather than once at the top where a reader stops noticing it.

One quantity does exist instead, and that is the spread. The spread is arithmetic on two recorded levels rather than a term in anybody's document.

On a face amount of Rs 1,000.00/-, the same face amount used elsewhere in this set, one year of spread incomeOne year of the extra amount a borrower's debt carries above a government level at the matching maturity. The reason the extra is paid at all is settled elsewhere. is 2.20 percentage points of Rs 1,000.00/-, which is Rs 22.00/-. The Rs 22.00/- is not observed either, in the sense that no cash has moved. Anybody with a calculator can reproduce it from two recorded levels, and that is a different situation entirely.

Now declare a cost and see what it eats. Declare a round tripBuying a holding and later selling the same holding, so the cost of dealing is counted once on the way in and once on the way out. cost of dealing of 30 basis points of price, which is Rs 3.00/- on that same face amount, and one round trip has taken 13.6364 per cent of a year of spread income. Set the declaration at 60 basis points instead: Rs 6.00/- comes off the same face amount, and that takes 27.2727 per cent of the same year. Neither figure was observed. Both are useful.

Both are useful because the cost is not what the arithmetic delivers. The arithmetic delivers the sensitivityHow far an answer moves when an input is changed. Worth measuring precisely when the input is one that nobody has recorded. of the answer to a number nobody has. A reader now knows that a cost of dealing they might have dismissed as small takes a seventh of a year of spread income, and that a cost twice that size takes better than a quarter of it. A reader can decide for themselves how much the missing observation matters to the question they came with, without ever being handed a cost that somebody made up.

The everyday version sits in every household that has ever thought about selling some gold. Nobody knows what the jeweller will actually take off the price. But what a two per cent deduction does to the plan and what a five per cent deduction does to it can both be worked out, and if the plan survives both, the missing number has stopped mattering. If the plan only survives one of them, what has been learned is something important about the plan rather than about the jeweller.

What do the two columns look like for one invented issuer?

Palash Cements Limited, invented, has five year debt at a coupon rate of 9.10 per cent a year. The five year government SPOT rate against which that sits reads 6.90 per cent a year on the invented curve this platform uses, compounding once a year. The gap between them was worked out above.

Set out the two columns for that instrument honestly and this is what appears.

Read the two columns beside each other and the point is visible without being argued: one column is empty and the other holds a number somebody supplied, and a note that ran them together would look considerably more authoritative than either column deserves. Nobody has to be careless for that to happen. The merged version reads more smoothly and no part of it looks wrong, so merging the columns is the ordinary outcome.

When does a liquidity fact stop being one?

Careful people get caught here, and being careful is part of how it happens.

An observation is a fact about the moment it was observed. The restriction is not a technicality. The restriction is the entire content of the word observation. The sentence saying that an instrument changed hands on nine days out of twenty, taken from a named record, is true of that stretch of twenty days and says nothing whatever about any other stretch.

Now quote the same sentence six months later, in the present tense, with the source still attached. Nothing has been fabricated. The record still says what it said. But the sentence has become a claim that conditions have not changed since, and nobody observed that. The result is an assumption wearing the source and the date of a fact, which is considerably harder to spot than an assumption with nothing attached to it at all.

A sentence in that condition is staleStill true of the day it belongs to and no longer true of today, without anybody having altered a word of it., and staleness is not a defect of the original writing. The original sentence was correct. The distance between the day the sentence describes and the day it is read on has changed, and nothing inside the sentence tracks that distance.

The fix is one word of practice, and it is small enough to adopt this afternoon. Write the date into the sentence rather than beside it. A line reading that the instrument dealt on nine days out of the twenty ending as atA phrase pinning a statement to one named day, so that the statement stays attached to that day however long afterwards it is read. a stated day cannot quietly become a claim about this morning. The day is grammatically part of what is being asserted. A date parked in a footnote is a date nobody applies, and a date buried in a distant column header is worse.

Why is a stale fact harder to catch than a plain assumption?

Because of what is attached to it. Strip a sentence of every source and every date and a reader treats it warily by reflex. Attach a named record and a day to the same claim and the reader relaxes. Relaxing is the correct response to a fact and exactly the wrong response to a stale one.

The furniture of checkability is doing the opposite of its job: it is making an unverifiable claim look more solid than a bare assumption rather than less. A plain assumption announces itself. A stale observation is camouflaged by its own apparatus.

What are the two sides actually for?

A liquidity fact tells a reader what happened. A liquidity assumption tells them what the writer believes, marked as belief, with its consequence worked out where they can see it. The two are different services and a serious note owes both.

Facts alone will not answer the question the reader arrived with, so a note carrying facts and assumptions side by side, each one labelled, is worth more than a note carrying facts alone. Somebody wanting to know whether a holding could be reduced next month is not asking a question any record can settle, since no record contains next month. Refusing to address it is not rigour. Refusal is just an unanswered question dressed up as discipline.

So the count of assumptions in a note is not the measure of it. A note with eleven assumptions and two facts, every one of them labelled, may be far more useful than a note with two assumptions that cannot be told apart from the observations around them. The failure is never having too many assumptions; the failure is having assumptions a reader cannot separate from observations.

Who actually uses this distinction, and for what?

Four readers, with four different reasons, and it is worth seeing that none of them is looking for a right answer.

An analyst reading somebody else's note uses the distinction as a filter. The analyst goes through it once, marking which sentences carry a record and which carry the writer, and ends up with a map of where the note stands on somebody else's ground and where it stands on its own. The map, rather than the conclusion, tells them how much of the note they can lift into their own work.

A lender sizing a facility against a holding of debt instruments uses it to decide what to stress. A recorded observation about dealing gets tested against a bad month. An assumption about dealing has no distribution to lean on, so it gets tested against being wrong altogether, and that is a harsher test. Getting the two the wrong way round produces a stress test that looks thorough and tests the wrong thing.

Somebody holding the instrument uses it at the moment they want out. The holder's use is the sharpest of the four, and it is the moment when the difference stops being a writing convention. A reader who was told that dealing is regular, and who took that as an observation, has planned around a fact. A reader who was told that a sale is assumed possible inside a week at a price near the last, with the reason and the consequence beside it, has planned around a belief and knows it. Both may be disappointed. Only one of them was warned.

And a household, meeting the same problem without any of the vocabulary. The neighbour who says the flat lets easily, the relative who says gold always sells, the friend who says that scheme is very liquid: all three sentences are assumptions, all three may be well founded, and none of them carries what a household would need to plan against. The right response is not to disbelieve the neighbour. The right response is to ask what the sentence rests on and what would change if it were wrong. Those two questions are exactly the two clauses a marked assumption already contains.