Credit Analysis: Judging Whether the Borrower Can Pay

Credit analysis is an ordered enquiry, and the order is the point. Ask what is owed and on what dates. Ask what cash pays it. Ask how large that obligation is set against the cash. Ask how far things can move before the answer changes. Ask what stands behind the claim if payment stops. Each answer decides which question is worth asking next.

A lender is not buying a business. A lender is buying a promise that stated amounts will arrive on stated dates, and every question below is aimed at that one thing. The enquiry therefore begins with the promise rather than with the borrower. A question put in the wrong place is answered with evidence the enquiry has not yet earned, or answered after the decision has already been taken.

Here is the everyday version, and it is worth holding in mind throughout. An acquaintance asks to borrow money. The lender does not start by admiring their flat. The lender asks how much and by when. How much they are paid. How much of that pay is already committed. Whether the pay survives the work stopping. And only last, if none of it comes back, whether there is anything else within reach. A lending desk that has been running the same five questions in the same order for thirty years is doing nothing more sophisticated.

What are the five questions, and what order do they run in?

The list is the subject and the answers are only the illustration, so the five stand on their own before any of them is answered. One: what is owed, and on what dates. Two: what cash pays it. Three: how large is that obligation set against the cash. Four: how far can things move before the answer changes. Five: what stands behind the claim if payment stops.

Five questions, and the sequence is not a matter of taste. Each answer narrows what the next question is even asking about. Until the shape of the obligation is known, there is no telling which ratio is worth computing. Until the cash expected to do the work is identified, there is no telling which year of earnings to measure against. Until the size of the obligation is known, there is nothing to measure room against. And until all four have been answered, the last question is asking about a situation that has not been described yet.

Notice what is missing from that list. There is no question asking whether the borrower is a good business, whether the sector is growing, or whether the management is impressive. A share buyer asks those questions, and they matter to a lender only where they change one of the five. A lender's upside is fixed by contract, so the whole of the work is spent on the ways the stated amounts might fail to arrive.

Why does what is owed come before anything else?

Because it is the only one of the five with a fixed answer. The amount, the dates and the terms sit inside a document that has already been signed. Two people reading that document arrive at the same figures, and neither of them has formed an opinion yet. Everything after this point is a reading; this is a reading of nothing at all, it is simply what the paper says.

Palash Cements Limited, an invented issuer, is the one borrower used throughout. Its document is short. The document sets the face amount at Rs 1,000.00/-. The document fixes a rate of 9.10 per cent a year on that amount, compounding once a year like every rate in this material, names five annual dates, and repays the whole face amount on the last of them. Four terms are enough to write the entire obligation out. The last interest payment and the repayment land together, so the first four dates carry Rs 91.00/- each and the fifth carries Rs 1,091.00/-. Everything promised comes to Rs 1,455.00/-, and not one figure in that sentence required a judgement.

The shape is what question one is for, so look at the shape rather than at the total. Four small dates and one enormous one. A borrower facing that document has to survive four modest payments and then find the entire face amount in a single year. Now imagine the same Rs 1,000.00/- and the same 9.10 per cent arranged the other way, with the face amount paid down in equal slices. The obligation is unrecognisable.

The second arrangement appears in no document. It is constructed from the same two contracted terms, the face amount and the rate. Repaying Rs 200.00/- of principal a year, and charging 9.10 per cent on whatever is still outstanding at the start of each year, the balances run Rs 1,000.00/-, Rs 800.00/-, Rs 600.00/-, Rs 400.00/- and Rs 200.00/-, so the interest runs Rs 91.00/-, Rs 72.80/-, Rs 54.60/-, Rs 36.40/- and Rs 18.20/-.

A built schedule deserves the same suspicion as a quoted one, so the total is worth checking by a second route. The five opening balances add to Rs 3,000.00/- of amount outstanding across the life. Applying 9.10 per cent to that sum once gives Rs 273.00/- with no schedule involved at all. The same treatment of the bond, whose balance never falls: five years at Rs 1,000.00/- is Rs 5,000.00/- of amount outstanding, and 9.10 per cent of that is Rs 455.00/-. The two agree with the row-by-row build. The identical rate has cancelled out of both sides, so the Rs 182.00/- between them is a difference of base and nothing else. Rs 2,000.00/- more amount outstanding, charged at the same 9.10 per cent, is Rs 182.00/- more interest.

Question one has just ruled something out: treating those two arrangements as the same problem. The bond needs one very good year in year five and can survive four mediocre ones. The slice schedule needs five adequate years and forgives none of them, but it never asks for a wall of money at once. A single leverage figure, computed without knowing which of those two documents is in hand, describes neither.

What pays it, and why is that not the same as what the borrower has?

A borrowing is repaid out of cash. Not out of assets, not out of profit, not out of a valuation. Cash, arriving in an account, on or before a date. There are exactly three places it can come from, and naming which of the three is expected to do the work is the whole of question two.

Now the trap that this question exists to close. An asset is not cash until somebody has bought it. A borrower who says the loan will be repaid by selling a building has not answered question two at all; that reply supplies a fresh set of questions, and three assumptions now stand where there were none. There is a buyer. There is a price. There is a date by which both of those exist. None of the three was stated, and every one of them is now sitting inside the credit decision doing work nobody agreed to.

The household version makes it obvious. Someone with a flat worth a great deal of money and a salary that stopped last month is not in a strong position to borrow against next month. The flat is wealth, not liquidity, and a payment falling due next month has to be met in liquidity. A food stall outside one office building has a similar problem in a different form: the stove, the cart and the reputation are real, and none of them turns into cash on a Tuesday if the office has shut.

How large is the obligation set against the cash that pays it?

Question three is where the enquiry finally reaches a number, and it is worth noticing how late that happens. Two full questions have already been answered without one. Question three now asks for a comparison, and a comparison needs two quantities: what is owed, and one year of what the borrower earns before its financing and its non-cash charges are taken off.

Everything from here to the end of the arithmetic runs on a declared set of inputs. The word declared is doing real work here. The four amounts below belong to no borrower, and in particular they are not Palash Cements Limited's, whose document states its contract terms and nothing more. The declared set of inputs is this: debt of Rs 800 crore, cash held of Rs 300 crore, earnings before interest, tax, depreciation and amortisationA year of operating profit measured before the interest bill, the tax bill and the two non-cash charges have been taken off it. How it is arrived at was settled in the statements material and is used here as a given. of Rs 200 crore for one year, and interest of Rs 40 crore for the same year.

What does Gross Leverage measure, and what is its base?

Gross Leverage sets everything owed against one year of those earnings. Rs 800 crore over Rs 200 crore is 4.00 times a year's earnings. The figure is meaningless without its base, so say the base out loud every single time. The reading is not four of anything but four years' worth of one particular year's earnings, and a reader who writes down a bare 4.00 has thrown away the only thing that tells the next person what to divide by.

What does Net Leverage take out, and what does taking it out claim?

Net Leverage runs the same division on a smaller top line. Take the Rs 300 crore of cash the borrower already holds out of the Rs 800 crore of debt, leaving net debt of Rs 500 crore, and set that against the same Rs 200 crore of earnings. The reading is 2.50 times that year's earnings. Same debt, same year, same borrower, and a reading 1.50 turnsThe unit a multiple is counted in. One turn is one whole unit of whatever multiple is being quoted, so moving from 4.00 times to 2.50 times is a move of one and a half turns. lower.

So which of the two is right? Neither, and the question is the wrong one. The gap between them is not a fact about the borrowing, it is a claim about the cash. Taking Rs 300 crore off the debt asserts that the Rs 300 crore is available to repay debt, and that assertion is exactly as strong or as weak as the reasons behind it. Cash sitting in a subsidiary that cannot pass it upward is not available. Cash that the business needs to run next week is not available either. The claim the subtraction makes, and what has to be true before it holds, are covered separately. The 2.50 times reading is the one carrying that claim, so the claim at least stays visible as a claim.

Here is the arithmetic identity worth carrying away. The distance between the two readings is always the cash divided by the same year of earnings: Rs 300 crore over Rs 200 crore is 1.50 turns, and that is precisely the gap on the drawing. The gap is not an accident of these figures. Subtracting the same quantity from the top of a fraction always does exactly that.

Why is the running cost a separate question from the size?

Because owing a lot and paying a lot are different problems, and one figure cannot tell them apart. Leverage asks how big the thing is. Interest coverage asks how heavy it is to carry while it sits there, and it asks by setting one year of earnings against one year of interest. On the declared set of inputs, Rs 200 crore over Rs 40 crore is 5.00 times a year's interest.

A borrower can owe an enormous amount cheaply, and a borrower can owe very little at a punishing price, and leverage alone reads those two as the same borrower. Consider two households on the same street. One has borrowed thirty lakh at a low fixed rate over twenty years. The other has borrowed four lakh on cards and personal loans at punishing rates. Asked which one owes more, the first wins by a distance. Asked which one has more of this month's pay already committed before the groceries, the second one loses badly. Both readings are true. The two answer different questions. A borrower can be destroyed by either, so a lender needs both.

How to analyse Leverage and Coverage when neither one settles it alone

The two are read together, not one after the other. Reading them together sounds like a platitude until the meaning is set out: the pair of readings does not produce a score, it produces a destination. Each combination of a large or small obligation with a comfortable or tight cover sends the enquiry to a different next question, and that next question is the whole output.

Each of the four is a genuinely different animal, so all four are worth taking in order. A large obligation with a comfortable cover means the debt is cheap or long dated; the pressure is not in carrying it, it is in repaying it, so the question is when the maturity date falls and what is supposed to happen on it. A small obligation with a tight cover is the reverse. The amount is manageable and the price is not, so the question is what the rate is fixed against and when it next moves. Both comfortable is the pair that lulls people. A comfortable pair is a statement about today only, so the correct response is to ask what would have to change and how quickly. Both stretched has an answer nobody enjoys. At that point the calendar outranks both ratios, so the question is what falls due first.

Notice what has just happened to the numbers. The 4.00 times and the 5.00 times have not produced a judgement. The two readings have produced a location on a grid, and the location produced a question. Producing a question is what a pair of ratios is for. A reader who wants the pair to add up to a verdict is making the error set out below.

How much room is there before the answer changes?

Every reading so far is a photograph: one year of figures against one moment's debt. Question four asks the thing a photograph cannot answer: how far can the earnings fall before this reading stops being a comfortable one?

Run it. Hold the declared Rs 800 crore of debt and the declared Rs 40 crore of interest still, and take the year's earnings of Rs 200 crore progressively lower.

The bottom row is the cleanest thing in the table. Halving the earnings exactly doubles gross leverage, from 4.00 times to 8.00 times. Interest coverage exactly halves at the same moment, from 5.00 times to 2.50 times. The symmetry is not a property of these amounts. It is forced by where the earnings sit in each fraction: underneath one reading and on top of the other. Halve what is underneath and the answer doubles. Halve what is on top and the answer halves. Every other row in the table is the same arithmetic at a gentler setting.

Two more things fall out of that table, and both are worth having. First, the gap between the gross and the net readings is not fixed. The gap is always the Rs 300 crore of cash divided by whatever the earnings are. At Rs 200 crore of earnings the two sit 1.50 turns apart; at Rs 100 crore they sit 3.00 turns apart. The cash subtraction buys more turns precisely when the earnings are weakest, and weakest earnings are exactly when a lender trusts the subtraction least. Second, and this is the useful one, coverage has a natural floor and leverage does not.

Coverage reaching 1.00 times means one year of earnings exactly meets one year of interest and nothing else. Coverage of 1.00 times is a real place. On the declared amounts it sits at earnings of Rs 40 crore. The drop from Rs 200 crore to Rs 40 crore is Rs 160 crore of room, 80.00 per cent of where the earnings started, or 4.00 turns of coverage. Any of those units will serve, provided the base is stated each time. Leverage has no such place. There is no level of a leverage reading that arithmetic declares to be the edge; the edge only exists if somebody wrote one into an agreement.

Which is what gives room its second, sharper meaning. Headroom is the distance between the reading today and the level at which a covenantA term written into a borrowing that the borrower agrees to keep to for as long as the borrowing is outstanding. The tests a covenant sets, and what a breach entitles a lender to do, are covered separately. would be failed, and it is quoted in turns of the same multiple the reading is in. Headroom is a distance rather than a level, so two borrowers can both read 4.00 times and sit at completely different headrooms. The terms a borrower has agreed to, what a breach entitles a lender to do, and what a waiverA lender agreeing, in writing, not to act on something the borrower has failed to keep to. Distinct from a cure. A cure is the borrower putting the failure right. changes, are all covered separately. The distance itself is what question four measures.

What stands behind the claim if payment stops?

Last, and last for a reason. The first four questions are all about the borrower paying. Question five is about the world in which the first four have already failed. Asked first, it starts the planning of the wreck before anybody has checked whether the vehicle runs, a strange way to lend money and a very common one.

Three different things can stand behind a claim, and they are not variations on a theme. CollateralA specific asset pledged against a claim, so the claim has something identifiable to reach for if payment stops. The protection collateral adds, and how a pledge is created and registered, are covered separately. is an asset pledged against the claim, so there is a particular thing to reach for. A guaranteeA promise by somebody other than the borrower to pay if the borrower does not. A guarantee adds a second party to reach rather than a second asset. is a promise by somebody else to pay if the borrower does not, so there is a second party rather than a second asset. SeniorityWhere a claim sits in the queue when there is not enough money to meet everything owed. A higher position is met earlier out of whatever is available. is a position in a queue ahead of other claims on the same borrower, so there is neither a thing nor a party, only an order. Credit enhancement is a fourth arrangement built out of the same ideas. Every one of them is covered separately, and each answers a different question. Being told a claim is secured names which of the three is in play and says absolutely nothing about whether it will work.

The everyday test is quick. A cousin who borrows and offers their motorbike keys is one thing. The cousin's employer signing a letter promising to pay out of salary is a completely different thing. A bare assurance of being paid before their other friends are is a third. All three are commonly described as security, and a lender who cannot say which one is in hand does not yet know what is being held.

What does each question rule out?

Here is the reason the order exists at all. Skipping a question does not merely lose information. Each of the five closes off one specific wrong reading, and the wrong readings are available the instant its question is missing.

Down the right-hand column, not one of those five errors is a mistake in arithmetic. Every one of them is a mistake about what a correctly computed number is entitled to say. Mistakes of that kind survive careful people, spreadsheets and review meetings. The arithmetic never objects.

What can be known about Palash Cements Limited, and what cannot?

The worked example has a hard limit, and stating it plainly is more useful than working around it. Palash Cements Limited issues one five year bond at 9.10 per cent a year on Rs 1,000.00/- of face amount. Those terms are everything that can be known about the issuer, so every other row of a credit enquiry on it stands empty, with the reason printed inside.

Readers reach hardest for the sixth row, so it deserves its own sentence. A grade is somebody's published opinion. The firm that publishes it also publishes the ladder the opinion sits on. The publishing firm also decides what each rung of the ladder is supposed to mean, and writes its reasoning out separately. Read the ladder where it is published, and read what a regulator requires that publisher to disclose about it. Putting a rung on this issuer would be inventing an opinion and dressing it as a fact. An empty row is the better of the two.

So the honest position is this. The enquiry above is complete, teachable and correct. Applied to the one borrower available here, it produces answers for question one and empty rows for everything else. Empty rows are not a weakness of the enquiry. The enquiry is doing its job: naming exactly which things would have to be known before a conclusion is available.

Why a sequence has no dial

The subject is a sequence, and a sequence has no dial: there is no quantity whose movement redraws an order. The one control a reader might expect, something that turns a ratio into a judgement, would perform live the exact mistake set out below. The arithmetic belongs in a tool where the reader supplies every input, and such a tool is covered separately.

What can the finished enquiry still not state?

Suppose everything were available. Full statements, a schedule of every borrowing, a trusted cash position, and a history. Run all five questions and what is there at the end?

A description. The obligation is this shape, on these dates. The cash expected to meet it comes from here. The size of the claim against that cash reads this many times a year of earnings and this many times a year of interest. The readings can move this far before they change. And this stands behind the claim if payment stops. Absent from all of it is a probability, a date, or a comparison with anything the same enquiry has not been run on.

A description is a smaller claim than most readers expect, and the correct one. The output of credit analysis is a list of the things that would have to change for the answer to change. Such a list is genuinely useful: it names what to monitor, what to ask for, and what would constitute news. The five questions are not the kind of instrument that produces forecasts, so no amount of additional rigour inside them converts the list into one.

Where does the price of all this sit?

One calculation borders this subject so closely that it is worth naming here and then handing over. A lender who accepts the risk described above is paid for it, and that payment shows up as a rate above what the government pays for money over the same term. Palash Cements pays 9.10 per cent a year. The five year government SPOT rateThe rate agreed today for money placed today and returned on one named future date, with nothing paid in between. Every rate in this material says SPOT or FORWARD. stands at 6.90 per cent a year on the same once-a-year compounding clock. Subtract and the credit spreadThe gap between what a borrower pays and what the government pays for money over the same term. The risks a spread compensates for are covered separately. is 2.20 percentage points, which is 220 basis points. Note that this figure appears in nobody's document: 9.10 per cent is a line in an issuer's contract, 6.90 per cent is a market level in no contract at all, and the 220 basis points between them is arithmetic a reader performs.

Where the credit sequence takes that number is a triangle, and with the figure already on screen the triangle closes both ways rather than hanging. A spread is a price struck for an expected loss, and an expected loss is a rate of default multiplied by what is lost when default happens. Assume a recovery rateThe share of a claim that comes back after a borrower has stopped paying, stated as a percentage of the amount owed. Settled earlier in this material. of 40 per cent, so 60 per cent is lost. Then 2.20 points divided by 0.60 implies an annual default rate of 3.67 per cent. Run it the other way to check. 3.67 per cent multiplied by 0.60 gives 2.2020 points, two thousandths of a point above the spread, and that residual comes from the printed 3.67 rather than from the arithmetic. Carry the unrounded figure through instead and it closes on exactly 2.2000 points. A display figure is not an input.

Three limits sit on that 3.67 per cent and none of them may be dropped. The 40 per cent recovery is an assumption rather than a measured figure, so a change in the assumption changes the implied rate with it. The whole of the spread has been treated as payment for credit. In a real market part of it pays for the difficulty of selling, and that part pushes the implied rate too high. And an implied rate is what a price says, not a forecast and not a measured frequency. The compensation a spread carries, how the triangle is built, and how far it can be trusted are all covered separately.

Palash Cements Limited and its five year bond are invented.
Educational material. Not advice on any investment, tax, budget or market position.