Leverage Ratios: What Each One Can and Cannot Tell You
The instrument: nine amounts off two documents, and the basis each reading was taken on
Nine amounts go in, read straight off a balance sheet and an income statement, and the panel assembles them into a numerator and a denominator. The panel loads with a declared set of inputs attached to no borrower anywhere. Every amount is supplied by the reader, and a reading assembled from supplied amounts is worth exactly what the date and the period written beside them are worth.
| Numerator, assembled line by line | Direction | Amount |
|---|---|---|
| Long-term borrowings | adds | |
| Current maturities of those borrowings | adds | |
| Short-term borrowings | adds | |
| Gross debt | ||
| Cash and bank balances | ||
| Investments in liquid schemes | ||
| Restricted part | ||
| Cash counted | ||
| Net debt | ||
| Numerator used |
| Denominator, assembled line by line | Direction | Amount |
|---|---|---|
| Earnings before interest, tax, depreciation and amortisation, as reported | ||
| One-off items | ||
| Denominator used | ||
| Interest cost, the denominator of the coverage reading |
And, printed in the same size type, the ten questions these answers open
Educational illustration. The nine amounts that load by default were built for teaching and describe no company at all. The reader confirms, rather than the panel checks, that the two levels sit at one date and the two flows across one period. No tax. The illustrative test level is not a market level, not a standard and not anybody's covenant: what such a level is set at, and what follows from crossing one, varies from one borrowing agreement to the next and is covered separately. Whether a reading is comfortable, high or safe is a comparison, and a comparison needs evidence that no division carries. A reading beyond the illustrative level is a distance, not an error.
Left where it loads, the panel reproduces the worked position exactly. Long-term borrowings of Rs 560 crore, current maturities of Rs 90 crore and short-term borrowings of Rs 150 crore make gross debt of Rs 800 crore at one date. Cash and bank balances of Rs 110 crore and liquid scheme investments of Rs 190 crore make Rs 300 crore of cash at that same date, Rs 60 crore of it restricted. Earnings of Rs 200 crore, carrying Rs 15 crore of one-off items, and interest of Rs 40 crore, both across those same twelve months. Gross leverage 4.00 times one year of earnings; net debt Rs 500 crore, so net leverage 2.50 times that same year; interest coverage 5.00 times one year of interest; the two leverage readings 1.50 turns apart.
What does this tool compute, and what does it refuse to do?
The panel above assembles nine supplied amounts into a numerator and a denominator and prints three multiples. Everything it will not do is set out below.
| What it will not do | Why it cannot |
|---|---|
| Say whether a reading is good, high, safe or comfortable | A judgement is a comparison, and there is nothing on this platform to compare a reading against |
| Set one borrower beside another | One invented borrower exists here and there is no second one anywhere on this platform |
| Hold or recall a figure for any borrower | This platform keeps accounts for nobody, so no figure loads itself and no blank fills itself in |
| Check the nine amounts entered against anything | It has no source to check them against, and it cannot see where any of them came from |
The last row is the honest limit of every calculator ever built. Two figures that have nothing to do with each other divide just as cleanly as two off the same sheet of the same accounts, and the answer looks identical either way. A machine handed Rs 800 crore and Rs 200 crore cannot ask whether they describe the same company, the same moment or the same twelve months. The machine divides, and prints to two decimals.
Every division in this guide can be done on paper in under a minute. The work sits at the two ends instead: in what goes in, and in what the answer is then used for.
Debt is measured at a single date and earnings across a stretch of time. Predict what goes wrong when the date and the stretch do not line up.
What four figures go in, and what has to be true of each?
The nine amounts collapse into four things, and each carries a condition that has nothing to do with arithmetic.
Debt. The total of what the borrower has agreed to repay on interest bearing borrowings, read at one stated date and assembled from every line carrying such a borrowing. Borrowings sit on three separate lines of the accounts, so the panel asks for three. Not an average across the year. One date, written down.
Cash. What the borrower holds in cash and equivalentsMoney in hand and in bank accounts, plus holdings so short and so easily turned into money that treating them as money changes nothing. The accounting basis decides what qualifies, and the accounting standard-setter sets and revises that basis., at the same date as the debt. Matching the two dates is the condition readers drop most often. Cash moves every hour of every working day, so subtract a June cash balance from a March debt balance and the amount that comes out was never true of anybody on any day.
Earnings. Earnings before interest, tax, depreciation and amortisation, gathered across a stated period of twelve months. How that figure is arrived at on a set of accounts is finished work from a layer below, so the panel asks for the figure and never builds one.
Interest. The interest cost for the same twelve months the earnings were measured over. Not the interest that would be payable on today's debt at today's rate: that is a forecast rather than a measurement. And not the coupon on one bond among several.
The split in that drawing is the whole discipline of this guide. Two of the four are a levelA quantity that exists only at a stated moment and has no duration. A photograph of it captures all of it. A bank balance is one; so is what remains owed on a loan. and two are a flowA quantity that only exists by accumulating across a stretch of time. There is no such thing as the flow at an instant, only the flow across a named window. Sales across a year, or interest across a year., and they are different kinds of thing.
The household version: a home loan balance is a level, one number on any given morning, and asking what it was between Tuesday and Friday is a question with no answer. Household spending is a flow, and a flow exists only across a week or a month. Divide one by the other and the multiple means one thing if the spending is a year of spending, something else if it was one month, and nothing at all if the period cannot be named.
The two dates must be one date and the two periods must be one period, and a reader who cannot say which date and which period has supplied a number rather than a figure. A number is a quantity; a figure is a quantity carrying a claim about what it measures and when. The panel divides numbers, and only the person supplying them can turn them into figures.
| D | Debt, a level, read at one stated date |
| C | Cash and equivalents, a level, read at that same date |
| E | Earnings before interest, tax, depreciation and amortisation, a flow, across a stated twelve months |
| I | Interest cost, a flow, across those same twelve months |
| G, N, K | Gross leverage, net leverage and interest coverage, in that written order and not in reading order |
Of the four figures the three divisions rest on, which two are measured at a single date?
What comes out, and what is the base of each answer?
A multiple is a compressed sentence worth uncompressing. When the panel prints 4.00 times, the sentence underneath reads: the whole of one year of earnings would have to be handed over four times to clear what is owed. The multiple states a scale and nothing else, and says nothing about when the handing over would happen, whether the earnings recur, or whether anybody agreed to any of it.
Take the declared set of inputs above, attached to no borrower anywhere: debt of Rs 800 crore and cash of Rs 300 crore at one date, earnings of Rs 200 crore and interest of Rs 40 crore across those same twelve months.
| 4.00 | times one year of earnings. Debt of Rs 800 crore against earnings of Rs 200 crore |
| 2.50 | times that same year of earnings. Rs 500 crore remaining after the cash is taken off, against the same Rs 200 crore |
| 5.00 | times one year of interest. Earnings of Rs 200 crore against interest of Rs 40 crore |
So the panel prints the baseThe quantity a multiple is counting. Four times what? The what is the base, and it does not travel with the multiple unless somebody deliberately attaches it. inside every answer rather than in a footnote. A multiple travels and its base does not travel with it. A reading that leaves the screen as a bare 4.00 will be handed to somebody who supplies a base of their own. It gets written into a note, the note gets summarised, and three weeks later somebody applies it to profit after tax, which looks just as plausible on a slide.
The gap of 1.50 turnsThe unit a distance between two readings of one multiple is stated in. Going from 4.00 times down to 2.50 times covers 1.50 turns, the way a drop from four kilograms to two and a half covers one and a half kilograms. is not new information about the borrowing: it is the cash over the same earnings, Rs 300 crore over Rs 200 crore, an identity holding on any inputs. The gap therefore shows how large the cash is against a year of earnings, and nothing about whether it can be reached. Why the contrast matters is covered separately.
The tool returns 4.00 times, 2.50 times and 5.00 times. What are the three bases?
What does each answer settle for a lender?
Less than it looks in every case, and the narrowness is the point.
Gross leverage settles the scale of the obligation against the earnings that service it, and settles nothing else at all. At 4.00 times, four whole years of earnings would be consumed clearing what is owed, if nothing else claimed a rupee of them, which of course everything does. Gross leverage is a size, with no opinion about when, about what the borrowing costs, or about next year.
Net leverage settles the same scale on an assumption: that the cash on the balance sheet is available to repay the debt. Writing 2.50 times makes that assumption, noticed or not. The reading is stated here and not relied on. Taking the restricted part of the cash back out in the panel above moves the reading to 2.80 times, the cheapest way to see what the assumption was carrying. Why netting changes what the number asserts is covered separately.
Interest coverage settles something different in kind. Coverage asks whether a year of earnings carries a year of interest, and at 5.00 times the answer is that it does, five times over. Debt of Rs 800 crore is a fact about a moment; interest of Rs 40 crore is a bill arriving across a year. Coverage is therefore the only one of the three speaking about the next twelve months rather than a balance sitting still.
Two of the three are about size and one is about weight, and they are not interchangeable. A household with a large home loan at a modest rate owes far more, and pays far less each month, than one with a small borrowing on a card. A reader who wanted to know about weight and looked at a leverage reading has looked at the wrong output, and no amount of care in reading it will supply the missing answer.
Coverage read backwards, the most useful form of the three
Interest coverage has a second form, and it is the one a lender can act on. Coverage of 5.00 times says the earnings are five times the interest; turned round, it says the earnings can fall to one fifth of where they are before the interest stops being covered. On the declared inputs that is Rs 200 crore dropping to Rs 40 crore, a distance of Rs 160 crore, being 80.0 per cent of the starting earnings.
The reversed form is better because it names the earnings, the one quantity that actually moves. Earnings move by trading, month after month, and a lender can form a view about how far. Nobody forms a view about a multiple.
Net leverage comes back at 2.50 times one year of earnings. Predict how many further questions that single answer opens about the cash.
What does each answer leave open?
The list of what an answer leaves open is the part every calculator on the internet leaves off. Three answers came back and opened ten questions that none of the amounts producing them can answer, and all ten print beside the numbers rather than in a paragraph somebody has to scroll to.
Gross leverage opens three. When is the Rs 800 crore due, and how much falls in any single year? A lump due in eighteen months and an even spread across twelve years produce the identical 4.00 times, and are not the same situation in any sense a lender would recognise. What does the debt cost, and does all of it cost the same? And is the Rs 200 crore of earnings repeatableThe property an earnings figure needs before a multiple built on it says anything about a future year. A figure swollen by a one-off sale or a single unusual contract is a fact about last year and a poor guide to the next one., or did something one-off sit inside it?
Net leverage opens four, and all four are about the cash. Where does the Rs 300 crore sit, in which company and in which country? Whose is it, in the sense of whether the borrower that signed the borrowing can spend it? Is any of it already committed to something standing ahead of the debt? And can it be moved to where the debt is, without a tax, a consent or a delay? Every one rides on a subtraction that took less than a second, and the conditions themselves are covered separately.
Interest coverage opens three. Is the Rs 40 crore fixed, or does it reset on a date? Is any part of it being added to the debt rather than paid in cash, in which case the reading describes a payment that is not being made? And what happens when the borrowing is refinancedReplaced by a new borrowing, usually because the old one has reached its repayment date. The amount can stay the same while the cost changes completely, so a coverage reading has a shelf life. on whatever terms are available then?
Three answers, ten questions. Three into ten is not a failure of the panel: it is what compression does. Nine amounts were squeezed into three sentences, and the ten questions are a list of what the squeezing threw away, written down so somebody can go and get it.
The household version: somebody reports that a neighbour has borrowings of four times their annual income. The neighbour's four times is one thing known and about ten questions acquired. Home loan across twenty years or personal loan across three, at what rate, does the income include an unusual bonus, whose is the savings account. The single fact was worth having, and what it mostly did was settle what to ask next.
Interest coverage reads 5.00 times one year of interest. The earnings were Rs 200 crore and the interest Rs 40 crore across one twelve month period. State that same answer as a distance.
In what order should the three answers be read?
The order is specified rather than left open, and the reason is uncomfortable: identical figures leave a different impression depending on which of the three arrives first. Whichever number arrives first becomes the frame, and the other two get read as adjustments to it.
Take interest coverage before anything else. A borrower that cannot carry a year of interest has a problem no leverage reading can describe. If coverage comes back at 1.10 times, the size of the borrowing has stopped being the interesting question: something is going to happen inside the year whether gross leverage reads 2.00 or 6.00.
Gross leverage rests on no assumption about anything, so take it second. The debt is what was agreed and the earnings were measured; nothing has been netted, adjusted or added back. Gross leverage is therefore the one reading both sides of a conversation can agree on before the arguing starts.
Take net leverage last, and as a question rather than an answer. Net leverage is the same obligation restated on a claim about cash, and the claim has to be tested before the reading is used for anything. A reader who starts here starts with the most flattering of the three, on the one assumption the panel has not examined, and spends the rest of the enquiry defending a subtraction instead of examining a borrower.
The second reason has nothing to do with impressions. The three readings route the analyst to different evidence, so which one comes back uncomfortable decides where the next hour goes. The panel's real output is not a score but a direction.
None of that is the same question as measuring headroomThe distance between a reading and a level somebody has agreed it must stay inside, measured in turns of the same multiple. Setting the level, and what follows from a breach, is covered separately. against an agreed test level. The panel produces three readings and a direction; what a borrower promised to keep them inside is a different subject, covered separately.
Which of the three readings is taken before the other two, and on what ground?
How a lender actually uses these three, on an ordinary Tuesday
Nobody in a credit team opens a calculator, types nine amounts and forms a view. The labels come first, so an analyst pulls the statements, writes the date and the period at the top of the input sheet, then fills the amounts under them. Coverage gets computed and read. If it is thin, the afternoon goes to the interest terms and the repayment dates, and the leverage readings become background.
If coverage is comfortable, gross leverage goes into the file with its base and its basis attached. The reading has to survive being quoted by somebody who was not in the room. Then net leverage gets a question mark beside it, along with the four cash questions, and those go on a list for the borrower rather than into the model.
Then the ten questions get divided up. The repayment dates and the interest terms sit in the borrowing agreement. Some need a schedule the borrower has not sent, and some a conversation, particularly the ones about where the cash sits and what it is promised to. The three multiples take a minute and the ten questions take a fortnight, and the fortnight is a fair description of what credit work actually is.
An investor reading a bond does a shorter version, and a household the shortest of all: before guaranteeing a relative's loan, what are the repayments, is the income steady, and are the savings everyone is counting on actually reachable. Same three shapes, no spreadsheet.
The reading that goes wrong, and what it costs
Here is the most common way a perfectly correct calculator produces a completely wrong sentence, and it is invisible because nothing about the output looks unusual.
An analyst is working from the most recently published statements, and those statements cover six months rather than twelve. Working from the freshest statements is ordinary and entirely reasonable: the half year figures are the best evidence available, and using stale figures has its own costs. The earnings line reads Rs 100 crore. The debt is Rs 800 crore, read at the half year date. Both amounts are correct, both are properly sourced, and both go into the tool.
The tool returns 8.0000 times. The arithmetic is faultless. Rs 800 crore divided by Rs 100 crore is exactly eight, and the tool prints the base it was handed. The error sits entirely upstream: a half year of earnings has been read as a year of earnings, so the reading is exactly twice the size of the real one.
Nothing about 8.0000 times looks strange enough to make anybody open the source again. The error therefore survives into files, notes and decisions. It does not produce a division by zero, a negative number or an absurd magnitude. The division produces a plausible multiple, and plausible multiples get recorded.
The same error runs the other way just as easily, and the other way is if anything more respectable. It involves an extra step that feels like care: taking a quarter of earnings and multiplying by four to make a year. The multiplication manufactures twelve months that never happened, and embeds an assumption about the other nine that nobody stated out loud.
The repair fits on one line and belongs on the input sheet rather than in a review afterwards: the date goes beside every level and the period beside every flow, before anything is divided. The panel above therefore will not compute with either field left empty, and it carries a period selector. Set that selector to six months and every leverage reading relabels itself a multiple of half a year, the one thing the analyst above never wrote down. The panel cannot check whether the date is right. Insisting that a date is stated is the whole of what it can do.
The same shape of error has a second face, and the panel will produce that one too. Set the illustrative test level to 3.50 times and read the answer on gross debt against reported earnings: 4.00 times, beyond the level. Change the debt basis to net: 2.50 times, comfortably inside. Take the earnings adjusted for the one-off items instead: 3.72 times on gross debt, 2.33 times on net. Four readings, one borrower, one set of nine amounts, and the only thing that moved was the basis somebody chose to compute on. A reading is comparable with a level only when both were struck on the same basis, and what a particular agreement sets its level at, and on what basis, differs from one agreement to the next and is covered separately.
A reader enters Rs 100 crore of earnings taken from a six month statement, against debt of Rs 800 crore. What comes back, and where is the error?
Why is there no verdict printed at the end of it?
Nine amounts, three readings, ten questions and a stated order to take them in. No word follows at the bottom saying whether any of it is fine. The absence is not caution or lawyering. The absence is evidence.
To say that 4.00 times one year of earnings is a lot requires one of three things behind the statement: the same readings for a borrower doing something similar, so a comparison exists; the same readings for this borrower across several years, so a direction exists; or a record of what became of borrowers sitting at these readings before. Not one of the three exists here, and a tool that produced a verdict anyway would be manufacturing the missing evidence rather than reading it.
Instead there is one invented borrower, Palash Cements Limited, with a single five year bond and no accounts of any kind. No second borrower to set beside it, no run of years because it has no first year, and no study of what became of borrowers at 4.00 times. The declared set of inputs the panel loads with is not even attached to Palash Cements Limited: it was built for teaching and describes no company at all.
So the enquiry ends where the evidence ends. The panel divides, prints the base and the basis inside each answer, prints the ten questions beside them, and stops. A reader who wanted the word comfortable has to go and earn it somewhere the evidence lives.
The everyday version is a bathroom scale. The scale gives a weight to two decimal places, instantly, and has never once been able to say whether that weight is a problem. Answering that needs height, build, what the number was last year and what happens to people at it. The scale holds none of it and is not a worse instrument for saying so. A second display flashing a judgement it had no basis for would make the scale worse. The second display would start to be believed.
Predict what this tool says about whether 4.00 times one year of earnings is comfortable.
What is worth walking away with?
Not the three numbers. Any calculator divides as well as this one does. A reading of 4.00 times without a base attached is worth less than nothing. Somebody downstream will attach the wrong one. The shape of the enquiry the panel leaves behind is worth carrying.
The shape has five parts, none of them arithmetic. Write the date beside every level and the period beside every flow, before anything is divided. The same nine amounts read 4.00 times on gross debt and 2.50 times on net, so write the basis beside the answer. Take the readings in the order set out above: coverage, then gross leverage, then net leverage read as a question. Attach the base to every reading the moment it leaves the screen. And treat the ten open questions as the work list. Each of them can move a view further than any of the three multiples can.
Three answers went in as compression and ten questions came back out as expansion, and the expansion is the useful direction. A reader who ends with three numbers has finished. A reader who ends with ten questions has started.
What is the output of this tool that is actually worth carrying away?
Five items routed elsewhere and left blank
- The accounting basis the four input figures are measured and presented on, before any of them reaches a division. Kept by the Institute of Chartered Accountants of India at icai.org.
- The basis on which an expected credit loss is measured and reported. Kept by the Institute of Chartered Accountants of India at icai.org.
- What an issuer of corporate debt has to publish, and to whom, which is where a reader would go to find figures to type in at all. Kept by the Securities and Exchange Board of India (SEBI) at sebi.gov.in.
- The valuation norm setting the price at which a credit holding is carried. Kept by the Reserve Bank of India at rbi.org.in.
- The capital treatment attached to holding a credit exposure. Kept by the Reserve Bank of India at rbi.org.in.
Not one of those five rows is filled in above, and the arithmetic here needed none of them to run. The five are easy to walk past for that reason: amounts divide whether or not anybody knows what basis produced them.
Where this guide stops. Building a debt figure, a cash figure or an earnings figure out of a set of accounts is finished work from a layer below, covered separately. Testing a reading against a level somebody agreed to keep it inside, and measuring the distance to that level, are a separate exercise covered separately. Arguing the case for the cash subtraction is a third subject, also covered separately, and it is the entire reason the net reading is taken last rather than early.
A lender carrying credit risk is compensated by a spread struck over the government spot rate at the matching maturity, and that too is covered separately. Each of the five rule-set items is revised by whoever keeps it, so all five sit in the block above with nothing written inside them.
The five items named above, and where each is kept
| Item left blank | Kept by | Site | Opened |
|---|---|---|---|
| The accounting basis the four input figures are measured and presented on | Institute of Chartered Accountants of India | icai.org | 28 August 2026 |
| The basis on which an expected credit loss is measured and reported | Institute of Chartered Accountants of India | icai.org | 28 August 2026 |
| What an issuer of corporate debt has to publish, and to whom | SEBI | sebi.gov.in | 28 August 2026 |
| The valuation norm setting the price a credit holding is carried at | Reserve Bank of India | rbi.org.in | 28 August 2026 |
| The capital treatment attached to holding a credit exposure | Reserve Bank of India | rbi.org.in | 28 August 2026 |
Palash Cements Limited is invented.
Educational material. Not advice on any investment, tax, budget or market position.
