Financial Restatement: When Past Numbers Are Corrected

A restatement corrects a prior period because that period carried a material error, so the comparatives are presented again as if the error had never happened and the correction is described in a note. A restatement is best learned through what it is not. A revised estimate is not an error, and a change of accounting treatment restates comparatives without anything having been wrong. Only one of the three admits a mistake.

Consider something ordinary. A household totals up what a wedding cost: the caterer, the hall, the cards, and a figure for the tailoring that nobody could pin down that week because three bills had not come in. Two months later the bills arrive and the tailoring figure was out by a fifth. Two quite different reasons could have produced that. Either those bills were sitting on a shelf the whole time and nobody looked, or they genuinely had not been raised yet. The number moves by the same amount under both. What the figure says about the household is not remotely the same in the two cases, and that gap is the whole subject.

The reader arrives holding the statements and the notes, the annual report and the parts it is built from, and the idea that an auditor works to an amount they have judged material rather than to perfection. Audit materiality covers that idea in its own right. The new event is narrow, and it gets misread more often than almost anything else in a set of accounts: a figure that was already published, already audited and already read is presented again with a different number on it. Three quite separate things can produce that moment. Only one of them is a confession.

What are the three ways a past figure gets touched again, and which one admits a mistake?

The three arrive looking identical in the accounts and mean entirely different things. Separate them before anything else. The first is the correction of a prior period errorA mistake in figures already published, made by getting the treatment wrong, doing the arithmetic wrong, leaving something out, or misusing information that was on hand at the time.. Something was wrong. A figure was misstated, the misstatement was material, and the accounts as published did not say what they should have said. The second is a change of accounting treatment, where a different permitted method is adopted and applied by retrospective applicationPresenting earlier periods as though the new method had always been in use, so that the years shown side by side are drawn on the same basis.. The earlier years are then drawn on the new basis. Nothing was wrong. The old method was permitted when it was used and the new one is permitted now. The third is a change in estimateA revision of a judged figure, such as the share of a bill that will never be collected, or the number of years a machine will keep working. New information has arrived since the last judgement was made.. Nothing was wrong either. New information turned up, a judged figure is revised, and the revision is carried forward from the date it was made.

Now watch which periods each one touches. That is where the confusion is manufactured. The first two both reach backwards. Under both, the comparativesThe earlier year printed beside the current year in a set of accounts. The two can then be read against each other. are presented again with different figures in them, and under both, the balance the business started the year with is adjusted. The third reaches nowhere at all. Not one figure printed in any earlier year moves, not one comparative is touched, and the revised estimate simply appears in the current year and the years after it.

Two of the three restate a comparative and only one of the three says anything was wrong. A reader looking at a restated comparative column has not yet seen an admission of error. They have seen that a figure moved. The sentence that tells them which of the three moved it is in the note, and until they have read that sentence they are holding a fact with no meaning attached. The distinction is not a subtle one that only matters to specialists. The distinction decides whether the correct response is a raised eyebrow or a shrug, and readers get it wrong in both directions every reporting season.

What counts as an error, and what test separates it from a revised estimate?

An error is narrower than the word sounds in ordinary speech. Four things produce one. A requirement can be misapplied, so a transaction was put through on a basis that was never available for it. The arithmetic can simply be wrong. A wrong formula in one column of a working file does it, and that is far more ordinary than anyone likes to admit. Something can be left out, so a liability that existed was never recorded at all. And information can be misused. This fourth cause is by a distance the most argued about.

Misuse of information carries the whole weight, so read it slowly. The information in question has to be information that was available when the accounts were approved. Nobody can misuse information that did not yet exist. The test for an error is therefore what was knowable at the time rather than what is known now. Take the household and its tailoring bills again. If those bills were on the shelf in March and nobody opened them, the figure written in March was wrong when it was written, and it was wrong using material sitting in the same room. If the bills genuinely had not been raised until May, then the March figure was the best figure available in March, and the May number is not a correction of it. The May number is a later and better figure produced by later and better information.

So the dividing line is a moment in time, not a size and not a direction. Everything sitting on the near side of the day the accounts were approved was, in principle, knowable at the timeCapable of being found out by whoever prepared the accounts, using material that already existed on the day those accounts were approved. Not the same thing as actually known., and getting it wrong is a prior period error. Everything that arrived after that day is new, and using it now revises an estimate. The test never asks whether the figure turned out to be right. A provision covering bills that may never be collected can be set at one amount, and those bills can go on to behave nothing like it. The original amount is still not an error. An estimate is a judgement made on what was available, not a forecast that was supposed to come true.

How does a correction actually appear in the accounts a reader is holding?

Three things appear, and a reader who knows to look for all three will never mistake one mechanism for another. The first is the comparative column itself, presented again with corrected figures in every line the error touched and labelled as restated. The second appears only sometimes. Where the effect of the correction reaches back before the earliest year printed, a further balance sheet is presented at the beginning of that earliest year. The reader can then see the position the corrected figures started from. Three balance sheets where two were expected look strange on first encounter, and they are not a sign of anything worse than an error whose effect began early.

The third is the note, and the note is where the whole thing actually lives. The note names what went wrong, gives the effect on each line affected, gives the effect on earnings per share, and adjusts the opening retained earningsThe accumulated profit a business carries into a year from all the years before it. Correcting an earlier year changes what was carried in, so this balance is adjusted. that were carried into the earliest period shown. The columns tell a reader only that a correction happened. The note is the only place that says what it was, how big it was and which of the three mechanisms produced it. A reader who registers the restatement and skips the note has collected the fact and left the content behind. Feeling informed on nothing is worse than not noticing at all.

What happens if a year one inventory count was wrong by Rs 2,50,000?

Anjani Stationers Private Limited, an invented notebook maker, has restated nothing. Its statutory audit ended in an unqualified opinion, with nothing qualified and no matter emphasised, and no period of its has ever been corrected. The restatement traced below is therefore a counterfactual, built so the mechanism can be followed on figures already familiar.

Suppose the goods in the warehouse had been counted wrongly at the end of year one, and the year one closing inventory had gone into the accounts Rs 2,50,000 higher than the goods actually on the floor. Suppose further that nobody noticed until the year two audit, when the count was reperformed and the difference came out. Follow the arithmetic and it moves in a very particular pattern. Cost of materials consumedOpening stock plus what was bought during the year, less the stock still on hand at the end. Whatever is missing from the closing figure has, by arithmetic, been treated as used up. is worked out by taking what was there at the start, adding what was bought, and deducting what is still there at the end. Deducting too much at the end charges too little to the year, so year one cost of materials consumed was understated by Rs 2,50,000 and year one profit before tax was overstated by exactly the same Rs 2,50,000. On the balance sheet, year one inventory and year one equity were each Rs 2,50,000 too high. On 4,00,000 shares, the year one earnings per share was overstated by Rs 0.625 a share.

Is that big enough to matter? On this engagement the auditor judged Rs 1,90,000 to be material to the accounts as a whole, set performance materiality at Rs 1,42,500 and treated anything below Rs 9,500 as too trivial to accumulate. The auditor arrived at those three amounts for this one engagement. Not one of them is a rule anybody must follow, a benchmark anybody publishes, or a fixed share of any figure. Against the first of them, Rs 2,50,000 is 1.32 times the amount judged material. The counterfactual error therefore sits comfortably past the point where it could simply be left alone. Size is part of the question and never the whole of it. A misstatement below the amount the auditor judged material is still a misstatement and is still corrected. It just does not send anyone back to redraw a published year.

Read the bottom half of that table carefully. The reason those year two figures do not move is more precise than it first looks. A comparative is not somehow sealed off. The error was found during the year two audit, before the year two accounts were drawn. Those accounts were prepared using the corrected opening inventory from the start. Had the miscount survived one more year, the year two accounts would have been built on the wrong opening figure. Year two cost of materials consumed would have been overstated by Rs 2,50,000 and year two profit understated by the same amount. The balance sheet would have quietly corrected itself by the year two year end, and two profit figures would have stayed wrong in opposite directions. An inventory error washes out of the balance sheet in a year and does not wash out of the profit line at all. Finding it late is therefore worse than finding it early, even though the closing balance eventually comes right on its own. The figures above are worked before any tax effect. The tax charge and the tax balance of a corrected year depend on matters covered separately.

Now run the same line the other way. The contrast is the point. Suppose nothing had been miscounted at all, and instead the provision covering bills that may never be collected had simply been revised because a school group that had always paid on time went quiet. A school group going quiet after the year one accounts were approved is new information, and new information revises an estimate. Watch what moves. The year one comparative: nothing. The retained earnings carried into year two: nothing. Earnings per share for year one: nothing. No further balance sheet, no restated column, no note describing an error. There is no error to describe. The revision appears in year two and in the years after it, and the accounts already published stay exactly as they were printed. The same rupee amount can either rewrite a published year or touch nothing at all, and which of the two happens is decided entirely by when the information arrived rather than by how large the number is.

What does a restatement say about the people who prepared the accounts?

A restatement says that something was wrong. That single fact is the entire content of it, and the temptation to read more is enormous. A correction of a prior period error establishes that a published figure did not say what it should have said. The correction does not establish that anybody was dishonest. Carelessness is not established either. Nor is unfitness for the job, and no part of the disclosure carries the material that would support any of those three conclusions.

Look at what actually produces errors. The ordinary causes are far duller than the imagined ones. A genuinely difficult judgement can be resolved one way when it is made and differently on later reflection, with entirely reasonable people on both sides of it and a written file behind each. A system can change, so a ledger is moved or a mapping is rebuilt and one account lands in the wrong place for a year while everything around it works perfectly. A complex transaction can be read carefully and understood wrongly. The word complex means exactly that, rather than serving as an excuse for it. And a requirement can be applied incorrectly by people doing their honest best with a rule that is hard to apply. This last cause describes a large share of everything that has ever gone wrong in a set of accounts anywhere.

Both halves hold at once, and the honest position needs both. Restatements are uncommon, and they are serious when they happen, and neither fact is a reason to wave one away. Treating a restatement as evidence of bad faith is unfair to the people involved and, far more often than not, simply wrong. The disclosure that a figure was corrected contains nothing whatsoever about why it was wrong or who let it be. A reader who makes that jump has not found something. Such a reader has decided something, using material that could not settle it, about people who cannot answer back in the document. Vaidehi Rao, who signs these accounts off as controller, holds reasons behind every judgement in them that never appear in the accounts themselves. The same holds for whoever prepares any set of accounts anywhere. Unwritten reasons are a fact about what accounts are rather than a fact about her.

How should a reader treat restated figures?

Four steps, and they take a few minutes rather than an afternoon. The first is to use the restated figures. They are the corrected ones, and the business itself has withdrawn the originals. The second is to read the note. Only the note names which of the three mechanisms produced the change, what caused it and how large it was. The third is to ask whether the affected area is one that recurs. A warehouse count happens every single year, so a counting error says something about a process that will run again. A one-off transaction misread once cannot be misread again in the same way. The fourth is to check whether the audit opinion on the restated period changed. Any such change is a fact to be looked up rather than a judgement to be formed.

Then hold the conclusion that almost everybody gets backwards. A restatement makes the past more reliable rather than less. The figure now in front of the reader is the corrected one, and the people who put the wrong one there have taken it off the table. The instinct runs the other way so strongly that it is worth saying twice. Before the correction, a reader was working with a wrong number and did not know it. After the correction, that reader is working with a right number and knows exactly which one moved, by how much, and why. The accounts got better. Reliability did not go down. The reader's comfort did, and the two are not the same thing at all.

Think of it the way a household thinks about a bill it queried. The moment the shopkeeper checks the ledger and says the total was wrong by two hundred rupees, the household knows more than it did an hour earlier, not less. Nobody concludes that the shop cannot count. The household concludes that this particular total is now right and that the ledger got looked at. No other total in the book can claim as much. The published accounts of a business work the same way, and the only difference is the number of people watching.

Panels do not travel, so the readings this one produces are written out below in plain text. At the default, nothing has happened, no period moves and nothing is disclosed, matching the actual position of Anjani Stationers. Set the switch to something was wrong and drag the slider to Rs 2,50,000 and the panel restates the year one comparative, adjusts the retained earnings carried into year two, leaves every year two figure alone, and puts Rs 0.625 a share against year one earnings per share. Drag the same slider down to Rs 90,000 instead and the panel stops restating anything. Rs 90,000 sits below the Rs 1,90,000 this auditor judged material, and a misstatement of that size is corrected without sending anyone back to redraw a published year. Switch to a different permitted treatment at any size and the comparative still moves while the stamp turns green. Nothing was wrong. Switch to new information has arrived, and no setting of the slider, at any amount up to Rs 5,00,000, will make a single earlier period move. The panel does nothing sharper. The reach switch changes only whether a further balance sheet appears. A revised estimate never reaches back, so under that setting the switch does nothing at all.

Who uses this, and what do they do with it?

Three people meet a restated column in the same week. Watch how little their three jobs have in common.

A lender is checking whether a covenant was ever actually met. Covenants are written against reported figures, so a restated year is a genuine practical problem rather than an abstract one: a ratio that cleared its threshold on the original numbers may not clear it on the corrected ones. The lender therefore rebuilds the test on the restated figures. The far more useful second step is to ask what the covenant should be measured against in future. The question is a conversation with the business rather than a conclusion about it. An analyst does something else entirely. An analyst has a model with the earlier year hard-coded into it, and the restatement means the growth rates, the margins and the per share figures in that model are all computed off a superseded base. Rebuilding it is dull, unavoidable work, and the analyst who skips it will report a growth rate that is arithmetically wrong in a direction nobody can see.

And Vaidehi Rao, as finance controller, uses it in the opposite direction from both of them. The person preparing the accounts reads the note they are about to write and asks whether a reader with no context could tell, from that paragraph alone, which of the three mechanisms this was. If the answer is no, the paragraph gets rewritten before it goes anywhere near a printer. That is the whole of good disclosure in one habit. A note that says the comparatives have been restated and stops has told the reader nothing and has made it near certain that half of them will assume the worst. A note that says what happened, why, how much and which mechanism produced it leaves nothing for anybody to invent. Such a note costs three extra sentences and is the cheapest reputational protection available to any business that ever has to correct anything.

Notice that all three uses are practical and none of them is a verdict. The lender rebuilds a test. The analyst rebuilds a model. The controller rewrites a paragraph. Nobody in that list needs to form a view about anybody's character in order to do their job. The ones who try tend to do the job worse, because the time goes into the theory rather than into the arithmetic that actually changed.

What can a reader never establish from a restatement?

Three questions sit permanently outside what the disclosure can answer, and saying so plainly is what separates a reader who is useful from one who is dangerous.

The first is whether the error was avoidable. The disclosure shows what went wrong, how large it was and which lines it touched. The disclosure does not show whether a different process, a different system or a different pair of eyes would have caught it. None of that is in the document, and no amount of staring will put it there. The second is whether anybody was at fault. The disclosure establishes that a figure was wrong. No person is named, no decision is described, and nothing in the disclosure could distinguish a difficult judgement made honestly from any other cause. The third is whether the same area holds another error. A correction shows that this one was found and put right. The correction says nothing whatsoever about what the next count, the next mapping or the next complex transaction will produce.

All three questions are outside what the disclosure carries. A reader who answers any of them from a restatement note has stopped reading and started deciding. Here the cost is paid by people who cannot reply, and that matters more than almost anywhere else in a set of accounts. A supplier who treats a restated comparative as a sign that something is being concealed will shorten the credit it gives to a business whose only offence was to put a figure right in public. Putting a figure right in public is exactly the behaviour everybody claims to want. Punishing the correction is how corrections become fewer, and fewer corrections is a strictly worse world for every reader in it.

Anjani Stationers Private Limited, Sunrise Public School, Chitra Binding Works and Vaidehi Rao are invented.
Educational material. Not advice on any investment, tax, budget or market position.