Full explanation
Overview
Three weeks after year end, someone finds it: last year's inventory was overstated by two hundred thousand pounds. The accounts are already printed. Now the question isn't whether to fix it — it's whether last year's numbers get rewritten, or this year's just absorb the hit.
IAS eight governs how you choose accounting policies, and what happens when a policy, an estimate, or a past number turns out to be wrong. IAS ten deals with the gap between your year end and the day the accounts are signed — and which events in that window you have to fold back in.
Who does what
IAS eight: the rules for selecting accounting policies, and the three different ways you correct course — a change in policy, a change in estimate, or a prior-period error, each treated completely differently. IAS ten: events after the reporting period — the ones you adjust for, and the ones you only disclose.
Retrospective vs prospective
Two words do all the work in IAS eight, so let's pin them down. RETROSPECTIVE means you rebuild the past: you restate prior-period comparatives, and adjust the opening balance of equity for the earliest period presented, as if the new treatment had always been in place. PROSPECTIVE means you leave the past exactly as reported and apply the new treatment only from the date of change onward, in the current and future periods. Policy changes and error corrections are retrospective. Estimate changes are prospective. That's the whole fault line.
Policy, estimate, or error
Everything in IAS eight turns on one classification, and people get it wrong constantly. A change in accounting POLICY is a change in the basis of recognition or measurement — for example, switching cost formula, or moving from the cost model to the revaluation model. A change in ESTIMATE is a revision of a judgement about an uncertain amount — a useful life, a provision, an expected credit loss — as new information arrives. And an ERROR is a mistake: something misstated using information that WAS available and should have been used. Policy and error look backward. Estimate looks forward. That single distinction decides everything that follows.
A set hierarchy
Choosing a policy in the first place. If a standard specifically applies to a transaction, you apply that standard — no discretion. If none does, management uses judgement, working down a set hierarchy: first, the requirements in other IFRS standards dealing with similar issues; then the definitions and recognition criteria in the Conceptual Framework; and only then may you look to the most recent pronouncements of other standard-setters, other accounting literature, and accepted industry practice — but never in conflict with the first two levels. Policies must be applied consistently to similar transactions unless a standard permits categorising them differently.
Definition of an accounting estimate
A real amendment sharpened this line in twenty twenty-three. IAS eight now formally DEFINES an accounting estimate as a monetary amount subject to measurement uncertainty. The practical effect: developing an estimate involves selecting an estimation or valuation technique AND the inputs it uses. A change to either — a new technique, a revised input — is a change in ESTIMATE, accounted for prospectively, NOT a change in policy. Unless, of course, it's correcting a past mistake, in which case it's an error. The amendment exists precisely because entities were mislabelling estimate changes as policy changes to force retrospective treatment.
Get the label right and the accounting almost falls out on its own. Get it wrong, and you've either rewritten history you shouldn't have, or buried something you should have surfaced.
Retrospective — as if always applied
Change in accounting policy: RETROSPECTIVE application. You account for it as if the new policy had always been applied. Meridian switches its cost formula and the effect on periods before last year is a four hundred thousand pound reduction in inventory. You restate the opening balance of retained earnings for the earliest period presented by four hundred thousand pounds, restate last year's comparative figures on the new basis, and only the current year uses the new policy from the start with no catch-up in profit or loss. Nothing about this touches THIS year's profit — it all lands in equity and the comparatives.
A third balance sheet
One consequence people forget: when a retrospective policy change — or an error correction — has a material effect on the opening position of the earliest comparative period, you must present a THIRD balance sheet. Two years of profit or loss, but three balance sheets: current year end, prior year end, AND the beginning of the prior year, restated. It's the only way a reader can see the restatement rolling through.
Estimate vs policy
Estimate versus policy, on one set of numbers, so you can see the difference. Meridian's pre-tax profit last year was reported at one million pounds. Now apply a two hundred thousand pound downward adjustment. As a POLICY change: last year's reported profit is restated to eight hundred thousand, the comparative in this year's accounts shows eight hundred thousand, and opening retained earnings drops. As an ESTIMATE change: last year stays at one million exactly as printed, and the full two hundred thousand hits THIS year's profit instead. Same adjustment, same amount — landing in a completely different year, on a completely different line.
When a new IFRS forces it
Not every policy change is voluntary. When the IASB issues a new or amended standard — a recent example being the twenty twenty-three narrowing of the deferred-tax initial recognition exemption in IAS twelve — and it changes how you must recognise or measure something, that IS a change in accounting policy. But you follow the new standard's own TRANSITIONAL PROVISIONS, which may require full retrospective restatement, a modified approach, or a cumulative catch-up in opening equity with no comparative restatement at all. IAS eight's general retrospective rule only governs a VOLUNTARY policy change, or a mandatory one where the new standard is silent on transition.
Prospective — the past stands
Change in accounting estimate: PROSPECTIVE only. Meridian bought a machine for one million two hundred thousand pounds, depreciating it straight-line over twelve years — one hundred thousand a year. After four years, carrying amount eight hundred thousand pounds, engineers reassess the remaining life at just four more years, not eight. You do NOT restate the past. From this year forward, you spread the eight hundred thousand over four years — two hundred thousand a year. The prior years stay exactly as reported. The change is recognised in profit or loss in the period of the change, and future periods, only.
Prior column: moves, or doesn't
This is where the statement of changes in equity earns its place. For a policy change or an error, the prior-period column is restated line by line, and the opening retained earnings figure carries a visible adjustment with a note reference. For a change in estimate, the statement of changes in equity is untouched — no prior column moves, because nothing about the past was wrong. If you ever see a restated opening equity balance for what was described as an estimate change, something is mislabelled.
Which brings us to the one nobody wants to be responsible for — a genuine error in numbers that have already been published.
Retrospective restatement
Prior-period error: RETROSPECTIVE restatement, the same shape as a policy change. That two-hundred-thousand-pound inventory overstatement from the opener — discovered after issue. You correct it by restating the comparative amounts for the prior period in which the error occurred, or, if it's older, restating the opening balances of assets, liabilities and equity for the earliest period presented. The correction never runs through current-year profit or loss. And crucially, this is only an error if the information needed to get it right was available when those statements were authorised. If it genuinely wasn't, it's a change in estimate instead — forward-looking, no restatement.
Impracticability
The escape hatch: impracticability. If it is genuinely impracticable to determine the period-specific effect of an error or a policy change — the data no longer exists, or the estimates needed can't be made objectively with hindsight — you restate from the earliest date that IS practicable, and disclose that fact. It's a narrow exception, not a convenience. The bar is 'cannot after making every reasonable effort,' not 'would be difficult.'
Three different packages
Disclosure, and it differs sharply by type. Policy change: the nature of the change, the reason the new policy gives more reliable and relevant information, and the amount of the adjustment for each financial statement line item affected and for earnings per share, for the current and each prior period presented. Estimate change: the nature and amount, or a statement that quantifying the future effect is impracticable. Error: the nature of the error, and the amount of the correction for each line item and for earnings per share, for each prior period presented, plus the amount at the beginning of the earliest period.
That covers what's inside the reporting period. IAS ten is about the window that opens the moment the period closes.
Up to authorised for issue
IAS ten covers events between the end of the reporting period and the date the financial statements are AUTHORISED FOR ISSUE — and the standard requires you to disclose that date and who gave the authorisation, because everything hinges on it. Two categories. ADJUSTING events provide evidence of conditions that already existed at the reporting date — you update the numbers. NON-ADJUSTING events reflect conditions that arose only after — you leave the numbers alone and, if material, disclose.
The condition existed at year end
Adjusting events — the condition existed at year end, you just learned more. A court case settling after year end that confirms the entity already had a present obligation. Evidence that an asset was impaired at the reporting date — a customer going bankrupt after year end, confirming a receivable was already uncollectible. The sale of inventory after year end giving evidence of its net realisable value at the reporting date. The discovery of fraud or error showing the statements were wrong. In every case, you adjust.
The condition arose after
Non-adjusting events — the condition arose after the reporting date, so the numbers stand. A dividend declared after year end is the classic one: NO liability is recognised at the reporting date, even if it relates to that year's profit — it's disclosed in the notes only. Also non-adjusting: a major business combination after year end, a sharp decline in the fair value of investments after year end, a restructuring announced after year end, a major loss from a fire after year end. Disclose the nature and an estimate of the financial effect, or say why an estimate can't be made.
Going concern gone
One event overrides the whole framework. If, after the reporting period, management determines it intends to liquidate the entity or cease trading, or has no realistic alternative but to do so, the financial statements must NOT be prepared on a going concern basis. This is treated as fundamentally adjusting — not a note, a wholesale change in the measurement basis of every asset and liability — because going concern underpins the entire set of accounts.
Tight, non-negotiable
IAS ten disclosure, tight but non-negotiable. The date the financial statements were authorised for issue and who gave that authorisation — and, if the entity's owners have the power to amend the statements after issue, that fact. For each material non-adjusting event: its nature, and an estimate of its financial effect, or a statement that such an estimate cannot be made. And if events after the period give information about conditions at the reporting date, you update the related disclosures too, not just the numbers.
The presentation side
IFRS eighteen, effective January twenty twenty-seven, reshapes the presentation side. It renames IAS eight itself to Basis of Preparation of Financial Statements and relocates the general presentation requirements. For restatements specifically: the requirement to present that third balance sheet on a material retrospective change carries straight through, and IFRS eighteen's new defined subtotals and category structure mean a restated comparative income statement must be rebuilt on the new structure too — so a policy change adopted around transition touches both the numbers and the face of the statements.
Three things matter
From an investor's chair, three things matter here. The authorisation date tells you how much post-year-end information the accounts could have captured — a long gap means more should have been reflected. The dividend trap: a declared dividend sits in the notes, not the balance sheet, so a naive payout-ratio calculation off the face of the statements can mislead. And restatements break comparability: when last year's error is corrected retrospectively, the comparative you're reading this year is NOT the number that was published last year — always check for a restatement note before trending anything.
Three FAQs
Three questions that come up every time. First — is correcting a rounding or classification slip an error? Only if it's material or if immaterial slips were made intentionally to achieve a presentation; genuinely trivial reclassifications are just fixed. Second — if I change an estimate AND it turns out last year's was actually wrong, which is it? If the information was available last year, it's an error, restate; if it only emerged now, it's an estimate, go forward. Third — does a change in estimate ever hit equity directly? No. It runs through profit or loss, or through the carrying amount of an asset or liability — never a direct adjustment to opening retained earnings. That's the tell for a mislabelled policy change.
Put all three corrections and a subsequent event on one company's numbers.
Four events, four treatments
Meridian, one year, four moves. A cost-formula change: retrospective, opening retained earnings down four hundred thousand pounds, comparatives restated, a third balance sheet presented. A useful-life revision on the machine: prospective, depreciation rises from one hundred thousand to two hundred thousand pounds a year, no restatement. A two-hundred-thousand-pound prior-year inventory error: retrospective, comparatives restated, current profit untouched. And a two-hundred-and-fifty-thousand-pound dividend declared three weeks after year end: non-adjusting, disclosed in the notes, no liability on the balance sheet. Four events, four different treatments, one rulebook.