Accounting Policies, Estimates and Errors Explained

An accounting policy is the method a business chooses for measuring something, such as valuing its stock at weighted average cost. An accounting estimate is a judgement made inside that method, such as how long a van will last. An accounting error is a mistake: something recorded wrongly or left out. Changing an estimate affects this year and the years ahead. Correcting an error means going back and restating what was already reported.

Here is what sits underneath. A set of accounts looks like arithmetic, and it is not. Beneath every reported profit lie a set of chosen methods and a set of judgements about a future nobody has seen yet, and neither of those can be checked by adding the column up again. An addition can be audited. An opinion about how long a van will run cannot. So the whole skill is separating the method from the judgement from the mistake, and that separation is the only thing that lets an honest change be told apart from a correction. The two are handled in opposite directions in time.

This guide sorts the three with a single question, follows each kind of change to the year it belongs in, and ends by putting a rupee figure on how much of one small printer's reported profit rests on judgement rather than on any document at all.

What is an accounting policy, and who chooses it?

Anjani Stationers Private Limited, an invented printer of school notebooks and exercise books, works in one city with one small printing unit and one delivery van. Anjani Kulkarni started it and still signs the cheques. Meera Rao keeps the books three days a week. Anjani Stationers sells only to schools, all on credit, and its largest single customer is the invented Sunrise Public School group. Year one runs from 1 April to 31 March.

An accounting policy is the method a business picks for measuring a thing, and once picked it is applied the same way every year and written down where a reader can find it. Think of a household that has decided it will always value the gold in the locker at what it paid, never at today's rate. Valuing at cost is a method, chosen once and applied to every item in the locker. Nobody had to choose it; a different household could reasonably choose the other way. Once the same rule runs through every year, this year's number and last year's number mean the same thing. Anjani Stationers measures its paper and ink stock at weighted average costA way of pricing stock in which all the units bought are pooled and given one average cost per unit, so a later purchase at a different price changes the average rather than being tracked separately., so the Rs 22,00,000 of closing stock is priced by pooling every reel bought during the year and taking an average. Weighted average cost is the method, and the method does not change.

Who chooses it? The people running the business choose it, from the set of methods the rules permit, and then they have to say which one they chose. The declaration lives in the notes to the accountsThe pages that sit behind the statements themselves, explaining what the headline numbers are made of and which methods were used to arrive at them., usually as the first note. Almost nobody reads the first note, and that is a shame: it is the shortest note in any set of accounts and the one that says the most. The first note is where a reader finds out that a business writes off its machines over ten years rather than fifteen, or that it provides against overdue customer dues at half rather than at nothing.

What is an accounting estimate, and how is it different from a policy?

A policy states how to measure, and almost never states the number. An accounting estimate is the judgement that fills in the one number the chosen method still needs, and it is always a judgement about something nobody can yet know. The policy says the van's cost is spread evenly over the years it is used. Fine. Over how many years? Nobody in the world knows. Somebody sat down, looked at how far the van would run and what similar vans last, and decided six. The decision to say six is the estimate, and the moment it is made a precise-looking figure of Rs 2,00,000 a year appears in the accounts as though it had come off an invoice.

Four such judgements sit inside one small set of accounts. Look at what actually happened at Anjani Stationers in year one. The van, bought for Rs 12,00,000, was given a useful lifeThe number of years a business expects to get use out of something it has bought, decided in advance so that the cost can be spread across those years. of six years. The printing machine, bought for Rs 30,00,000, was given ten. The Sunrise Public School group had Rs 6,00,000 overdue and Meera Rao judged that about half of it would never arrive, so Rs 3,00,000 was charged against profit. And the closing stock of Rs 22,00,000 rests on a count and on an average that assumes every reel is still worth what it cost. Four judgements, four numbers, and not one of them is checkable by adding anything up.

Put the two side by side. A policy could have been different in principle, and the business had a genuine choice between permitted methods. An estimate could not have been different in principle: there is only one true answer, and nobody will know it until later. The van will last however long the van lasts. Six years was not a preference; it was a forecast.

What is an accounting error, and how is it different from an estimate that turned out wrong?

An accounting error is a fact that existed at the time and was recorded wrongly, or was not recorded at all, so the figure published was wrong on the day it was published. No judgement is involved. Nobody weighed anything. A number was simply not right. An error might be a transposed figure, a payment posted to the wrong account, a delivery counted twice, or, as at Anjani Stationers, a supplier's invoice that came in, got filed and never reached the books.

The event is small and completely ordinary. A paper supplier delivered reels during year one and sent an invoice for Rs 4,00,000, dated inside year one. The paper was used. The notebooks printed on it were sold, and the sale of those notebooks sits in year one's revenue. The invoice, though, was put in the wrong file and never entered in the purchase records. The invoice surfaces during year two, when the supplier's statement is reconciled. Nobody hid it. Nobody decided anything. Filing mistakes of this kind happen in every business that has ever run on paper, and the result of this one is that year one's accounts reported the sales of those notebooks without the cost of the paper they were printed on.

Readers slip at exactly this point, so now hold the missing invoice against an estimate that turned out wrong. Suppose the van turns out to last four years rather than six. Was the six-year estimate an error? It was not. On what was known when year one was prepared, six years was a reasonable judgement about a van whose running had not yet been observed. Nothing in year one's accounts was wrong when they were signed. The missing invoice is different in kind: the invoice existed, dated, in the correct year, and simply was not there. One is a judgement that later information improved. The other is a fact that was always available and got lost.

Which single question sorts a change into the right box?

Three tests are not needed. One test, asked in one direction, places the change: was the earlier figure wrong on the day it was prepared, using what was known on that day? If the earlier figure was defensible on what was known at the time, it was never wrong, so nothing gets rewritten; if it was not defensible on what was known at the time, the published figure was wrong and has to be corrected. Everything else follows from that one answer.

Two smaller questions sit either side of it. If nothing was wrong, the next question is whether the method itself changed or only a judgement inside it: a changed method is a policy change, a revised judgement is an estimate change. And the test is applied without hindsightKnowledge that only became available later. Judging a past decision by what is known now rather than by what was known then.. The test asks what was known then, not what is known now. Asking what was known then is the single discipline that stops every refined judgement from becoming a retrospective accusation.

How is a change in an estimate handled, and from when?

During year two the van turns out to be doing far more running than anybody expected, and the life is re-estimated from six years to four. Work through what that does. The van cost Rs 12,00,000. Two years of charge had already been taken before year one began, and year one took a third, so Rs 6,00,000 has been charged and Rs 6,00,000 is still carried. Under the original six-year view, three years of life remained and the plan was Rs 2,00,000 a year for three more years. Under the four-year view, three years of life have already gone and exactly one remains, so the whole remaining Rs 6,00,000 is charged in that one year.

A change in an estimate takes effect from the year of the change onward, and it never reaches back into a year that has already been reported. Year one keeps its Rs 2,00,000 of van depreciation and its Rs 38,00,000 of profit, both exactly as first published. Year two carries Rs 6,00,000 instead of Rs 2,00,000, Rs 4,00,000 more than the original plan. By the end of year two the van is fully written down, so years three and four of trading carry nothing for it at all. Notice what has not changed: the total charge over the van's whole life is still Rs 12,00,000, its full cost. Only the timing moved.

Why forward rather than backward? Because on the day year one was prepared, six years was an honest reading of what was known, and no new fact makes an old judgement retrospectively dishonest. If every refined estimate reopened the years before it, then no published set of accounts would ever be final, a lender could never rely on a signed statement, and the carrying amountThe value at which something appears in the accounts today: usually what it cost, less the part of that cost already charged against profits. of everything a business has would be permanently provisional. Judgements get better with time. Getting better is a feature of judgement, not a fault in the accounts that used the earlier judgement.

How is an error corrected, and why does it reach backwards?

The missing Rs 4,00,000 invoice is handled in the opposite direction, and for a reason that follows directly from the test. Year one was wrong when it was signed. The paper was bought in year one, used in year one, and turned into notebooks sold in year one. So the cost belongs to year one, and the only honest fix is to go back into the year that was already published and put it where it always belonged.

Correcting a prior periodA financial year, or part of one, that has already been closed and reported. Anything belonging to it has to be dealt with in that year rather than in the year now running. error means restating the year that was reported, so year one's profit falls from Rs 38,00,000 to Rs 34,00,000 and year two carries none of the cost at all. The paper was never year two's expense. Not for one day. When year two's accounts are published, year one appears beside them as the comparative figuresLast year's numbers, printed beside this year's in the same statement so a reader can see the movement between the two., and those comparative figures are shown restated with a note saying what was corrected and by how much. A reader who saw the original Rs 38,00,000 is told plainly that the correct figure was Rs 34,00,000.

Restating is uncomfortable, and the discomfort is exactly why the mistake described below gets made. Restating a published year means telling people the number they were given was wrong. Booking the invoice as a year two cost, closing the file and moving on is much pleasanter. Every reason for doing that is a reason about the person doing the books, and none of them is a reason about which year used the paper.

What actually separates an estimate from an error, side by side?

The two events set out above are almost the same size and are handled in exactly opposite directions, so they sit well side by side. The van revision moves Rs 6,00,000 of charge, of which Rs 4,00,000 is extra compared with the original plan. The missing invoice moves Rs 4,00,000. If size decided the treatment, they would be treated the same way. Size decides nothing. The estimate revision travels forward into the year now running and the years after it. The error travels backward into the year already closed. The only thing that determines the direction is whether the earlier figure was wrong at the time.

The third and fourth rows are mirror images. In the estimate case, year one is untouched and year two takes the hit. In the error case, year one takes the hit and year two is untouched. Two events, similar amounts, discovered in the same year by the same person, and the profit that moves is a different year's profit in each case. Two rows carry the whole of the distinction.

Read the default state first, as static text, before pressing anything. With the van re-estimate selected at Rs 4,00,000, year one stands at Rs 38,00,000 and year two at Rs 34,00,000. Now press the third button. The bars do not move. Year one still stands at Rs 38,00,000 and year two at Rs 34,00,000, even though the event is a completely different kind of thing and the honest answer is year one Rs 34,00,000 and year two Rs 38,00,000. An error pushed into the wrong year draws exactly the same picture as an honest change in judgement, and the identical picture is why the two are confused and why the reveal button exists.

How much of Anjani Stationers' Rs 38,00,000 profit is judgement rather than arithmetic?

Now put a number on the whole idea. Year one's profit before tax of Rs 38,00,000 is built from gross profit of Rs 1,14,00,000, less salaries of Rs 54,00,000, rent of Rs 12,00,000, insurance of Rs 2,00,000, depreciation of Rs 5,00,000 and the provision against the Sunrise Public School group of Rs 3,00,000. Ask of each line: is there a document behind it? Salaries have payslips. Rent has a lease and a bank payment. Insurance has a policy and a receipt. Depreciation and the provision have nothing of the kind.

Depreciation of Rs 5,00,000 and the provision of Rs 3,00,000 together account for Rs 8,00,000 of the Rs 38,00,000 reported: 21 per cent of the profit rests on judgement rather than on any invoice. There is no supplier who sends a bill for depreciation. There is no receipt for a provision. Both are somebody at a desk deciding how long a machine will run and how much of an overdue amount will never arrive, and then writing the answer into a statement that looks, to any casual reader, entirely arithmetical.

To see how much room judgement leaves, look at a mirror business. Bharati Notebooks Private Limited, an invented printer in the same trade, bills the same Rs 2,40,00,000, collects the same cash and holds the same assets. Bharati Notebooks estimates its printing machine's life at fifteen years rather than ten, so it charges Rs 2,00,000 of depreciation rather than Rs 3,00,000, and it raises no provision against its own overdue school. Its profit comes out at Rs 42,00,000 against Anjani Stationers' Rs 38,00,000, 10.5 per cent higher on identical trade and identical cash in the bank. Neither business has broken a rule, and neither has made an error. Two sets of honest judgements produced two different profits from the same year of work.

How do a lender, an analyst and a household actually use this?

A lender uses it as a comparability check before it uses anything else. When a bank looks at Anjani Stationers' two years side by side to decide whether to renew a working line, the first question is not what the profit was; it is whether the two figures were prepared the same way. So the credit officer turns to the first note, reads the methods, and looks for one of three things: a method that changed, an estimate that was revised, and a prior year restated. Each one changes what the trend means. A profit that fell because a judgement got tighter is a very different animal from a profit that fell because the schools stopped buying.

An analyst uses it to strip out the part of a movement that is not trade. Take the Rs 4,00,000 of extra van depreciation in year two. Nothing about the printing business changed; a view about a vehicle changed. So an analyst reading the two years would note the revision, keep it in the profit because it is a real cost of using a real van, and separately record that Rs 4,00,000 of the fall is a timing effect that will not repeat once the van is fully written down. Both facts matter, and neither is visible without reading the notes.

A household meets exactly the same three ideas without ever naming them. Deciding always to value the household gold at what it cost is a policy, guessing that the scooter has four more years in it is an estimate, and forgetting to write down the plumber's bill is an error, and the third one is the only one that means last month's household budget was wrong. The household is applying the same test as the credit officer, in a smaller currency. When the plumber's bill turns up two months late, the honest household says last month cost more than it thought, and does not pretend the plumbing happened this month.

Anjani Stationers Private Limited, Bharati Notebooks Private Limited, Anjani Kulkarni, Meera Rao and the Sunrise Public School group are invented.
Educational material. Not advice on any investment, tax, budget or market position.