HMRC v Tooth [2021] UKSC 17

The leading modern authority on discovery assessments. Tooth decides two things that matter every day in practice: a tax return must be read as a whole, including its white space; and a “deliberate inaccuracy” means a statement the taxpayer knew was wrong and intended HMRC to rely on. It also (obiter, but decisively) buries the staleness doctrine that practitioners had relied on since 2013.

The Facts

Mr Raymond Tooth is a solicitor. In 2008–09 he participated in a marketed tax planning arrangement (the “Romangate” scheme) intended to generate an employment-related loss. The commercial purpose of the arrangement, from his perspective, was to carry that loss back under the loss relief provisions and set it against his income for the earlier tax year, 2007–08.

The difficulty was mechanical rather than conceptual. The tax return software used by his advisers did not contain a field that permitted an employment loss carried back from a later year to be entered on the employment pages of the 2007–08 return. The software would either reject the entry or produce an invalid submission. The advisers therefore entered the loss figure in the partnership pages of the return, a box that the software would accept, and then completed two separate “white space” narrative boxes explaining, in terms, exactly what had been done and why: that the figure represented an employment-related loss for 2008–09 being carried back to 2007–08, and that the entry had been made in the partnership pages because the software permitted no alternative.

Nothing was concealed. The return, read from beginning to end, told HMRC precisely what the taxpayer was claiming, on what basis, and why the figure appeared in an unexpected place.

HMRC’s Response

In August 2009 HMRC wrote to Mr Tooth stating that it was opening an enquiry into the claim. The enquiry was purportedly opened under Schedule 1A to the Taxes Management Act 1970, which governs claims made outside a return. The claim in this case had been made in a return, so the correct statutory vehicle was an enquiry into the return itself under s9A TMA 1970. The Schedule 1A notice was therefore legally ineffective. HMRC had, in substance, done nothing that stopped the clock.

The enquiry window under s9A closed. The 2007–08 return became final. Then, on 24 October 2014, more than five years after HMRC first engaged with the entry, Officer Wilson of HMRC issued a discovery assessment under s29(1) TMA 1970 in the sum of approximately £475,000. Because the ordinary four-year and extended six-year windows had long expired, HMRC could only sustain the assessment by relying on the twenty-year window in s36(1A)(a) TMA 1970, which requires the loss of tax to have been “brought about deliberately” by the taxpayer.

Procedural History

The case produced four different analyses on its way to the Supreme Court, which is itself instructive: the same facts were capable of supporting materially different legal characterisations at every level.

The pattern matters for practitioners. Mr Tooth won at every level, but the reasons changed each time. By the time the case reached the Supreme Court, HMRC had lost the deliberate inaccuracy point twice and the taxpayer had lost the staleness point nowhere. The Supreme Court reversed the lower courts on staleness while still dismissing HMRC’s appeal, which is exactly why the staleness reasoning is obiter.

The Issues Before the Supreme Court

1. Inaccuracy: Did the 2007–08 return contain an “inaccuracy” at all, given the white space explanation?
2. Deliberateness: If it did, was that inaccuracy “deliberate” within the meaning of s118(7) TMA 1970, so as to open the twenty-year window under s36(1A)(a)?
3. Discovery: Did Officer Wilson make a “discovery” within s29(1) TMA 1970 in October 2014?
4. Staleness: If a discovery had been made earlier by another officer, could the 2014 assessment be defeated on the ground that the discovery had become “stale”?

The Statutory Framework

Three provisions do the work.

• s29(1) TMA 1970: the gateway. An officer must “discover” that an amount which ought to have been assessed has not been assessed, or that an assessment has become insufficient, or that excessive relief has been given.
• s36(1A)(a) TMA 1970: the twenty-year window, available where a loss of income tax or CGT has been “brought about deliberately” by the taxpayer or a person acting on their behalf.
• s118(7) TMA 1970: the definition provision. A loss of tax or situation is brought about deliberately by a person where it “arises as a result of a deliberate inaccuracy in a document given to HMRC by or on behalf of that person”.

The whole appeal turned on the grammar of s118(7). Does “deliberate inaccuracy” describe (a) an inaccuracy that was itself intended, a statement the maker knew to be wrong and intended should mislead, or (b) an entry that was deliberately made and which turns out to be inaccurate? HMRC argued for the second reading, which would have swept in every conscious act of completing a return that later proves wrong.

The Authorities Considered

Cenlon Finance Co Ltd v Ellwood [1962] AC 782

The House of Lords held that “discovery” is a word of wide meaning: it covers any case in which an officer comes to the conclusion that there has been an insufficiency, whether that conclusion arises from new facts, from a change of view on known facts, or from the correction of an error of law. Viscount Simonds observed that the word should not be narrowly construed. Cenlon is the foundation on which the Supreme Court built its rejection of staleness: if a discovery is simply the officer’s subjective conclusion, there is no textual basis for saying that conclusion decays over time.

HMRC v Charlton [2013] STC 866 (UT)

This is the decision that generated the staleness doctrine. At paragraph [37] the Upper Tribunal (Warren J and Judge Berner) said that a discovery must have “the quality of newness” and suggested that a discovery could lose that quality with the passage of time. Practitioners built a substantial body of argument on that sentence. The Supreme Court in Tooth expressly disapproved “the latter part of para 37” of Charlton. The “newness” requirement survives, it goes to whether the particular officer has genuinely come to a new conclusion, but the temporal decay element does not.

Pattullo v HMRC [2016] STC 2043

The Upper Tribunal (Lord Glennie) had applied and developed Charlton, accepting that a discovery could become stale where HMRC delayed unreasonably in acting on it. This line of authority was likewise displaced by Tooth.

Langham v Veltema [2004] EWCA Civ 193

Not directly in issue, but structurally important. Veltema established that the s29(5) “hypothetical officer” protection operates only on the information the taxpayer actually supplied in or with the return, not on everything HMRC could have found out. The Supreme Court in Tooth pointed to s29(5) as one of the real statutory protections available to a taxpayer facing a delayed assessment. The protection that staleness had been doing duty for.

Auxilium Project Management Ltd v HMRC [2016] UKFTT 249 (TC)

The FTT’s well-known formulation is that a deliberate inaccuracy occurs when a taxpayer knowingly provides HMRC with a document containing an error, with the intention that HMRC should rely on it as accurate. Auxilium is an FTT decision and therefore not binding, but the Supreme Court’s reasoning in Tooth is consistent with it, and the two are now routinely cited together at first instance.

The Ratio Decidendi

1. A tax return must be construed as a whole

The Court held that in deciding whether a document given to HMRC contains an inaccuracy, the document must be read in its entirety. Boxes on a return form are not free-standing statements. Where a figure is entered in a particular box and a narrative entry elsewhere in the same return explains what the figure represents and why it appears there, the two must be read together. On that approach, Mr Tooth’s return contained no inaccuracy at all: read as a whole it stated, correctly, that he was claiming to carry back an employment loss.

Lord Briggs and Lord Sales rejected the submission that HMRC’s computer systems read returns box-by-box and that the law should follow the machine. The construction of a legal document does not depend on the limitations of the reader’s software.

2. “Deliberate inaccuracy” requires an intention to mislead

The second and principal holding is a point of statutory grammar with very large practical consequences. In s118(7) the adjective “deliberate” qualifies the noun “inaccuracy”, not the act of making the statement. What is required is a statement that was deliberately inaccurate: one which the maker knew to be inaccurate and intended should be relied on as accurate. A statement that is deliberately made but merely turns out to be inaccurate does not satisfy the test, however careless or even reckless the maker may have been about its accuracy.

Even if the return had contained an inaccuracy, therefore, there was no deliberate inaccuracy. Mr Tooth had not intended to mislead anybody; he had done the opposite, drawing attention to what he was doing. The twenty-year window in s36(1A)(a) was accordingly unavailable and the October 2014 assessment was out of time.

The Obiter Dicta, and Why They Are Nonetheless Decisive

1. There is no concept of “staleness”

The Court held that there is “no place for the idea that a discovery which qualifies as such should cease to do so by the passage of time”. The statutory scheme sets out a carefully calibrated set of time limits in ss34 and 36 TMA 1970, four, six and twenty years. Parliament having specified those limits, there is no warrant for the courts to superimpose an additional, undefined requirement that HMRC act promptly after making a discovery. The relevant passage in Charlton was disapproved.

This overturns the practitioner orthodoxy that prevailed between 2013 and 2021. Any advice, precedent or template grounds of appeal drafted before May 2021 that runs staleness as a free-standing ground should be treated as unsafe.

2. “Discovery” is subjective and officer-specific

The question is whether the particular officer who decides to make the assessment has subjectively come to the conclusion that there has been an insufficiency. Officer Wilson had done so in October 2014. That was a discovery, notwithstanding that other officers had looked at the same material years earlier.

3. There is no doctrine of collective HMRC knowledge

Knowledge held by one officer is not attributed to another for the purposes of s29(1). This cuts both ways: it defeats the argument that an earlier officer’s awareness spoils a later officer’s discovery, but it also means HMRC cannot aggregate departmental knowledge to defeat a s29(5) argument.

4. Where the real protections lie

The Court identified the protections that remain available to a taxpayer facing a long-delayed assessment: the statutory time limits themselves; the conditions in s29(4) (careless or deliberate conduct), and s29(5) (the hypothetical officer test); and public law principles, enforceable by judicial review, including the doctrine of conspicuous unfairness and abuse of power. These are now the correct places to look.

How Tooth Has Been Applied Since

Paul Harrison v HMRC [2023] UKUT 38 (TCC). The taxpayer argued that the staleness passages in Tooth were obiter and therefore not binding, so that the pre-Tooth line of authority survived. The Upper Tribunal rejected that. It held that the Supreme Court’s treatment of staleness was clearly given in order to provide general guidance on an important question, was exhaustive, was reached on full argument, and dealt expressly with the competing arguments. There is accordingly no concept of staleness in the direct tax discovery regime. HMRC’s Enquiry Manual at EM3260 reflects the same position.

On the deliberate limb, Tooth is now the starting point in every behaviour-category dispute, whether the issue arises under s36(1A) TMA 1970, Schedule 24 FA 2007 (inaccuracy penalties), or Schedule 41 FA 2008. The subjective, intention-to-mislead standard has proved durable and is routinely applied alongside the Auxilium formulation. Later decisions have had to grapple with whether “blind-eye” knowledge, suspecting an inaccuracy and deliberately declining to check, is enough; the better view is that it can be, because the taxpayer in that position does intend HMRC to rely on a statement they suspect is wrong.

Practitioner Application

What to argue after Tooth

• Read the return as a whole and say so early. If there is a white space entry, a covering letter, a computation or a schedule that explains the position, put it in front of HMRC at the earliest opportunity and frame the argument as one of construction: the document, properly read, is not inaccurate.
• Attack the twenty-year window on state of mind, not on quantum. HMRC frequently reasons backwards from the size of the understatement to a finding of deliberateness. Tooth forecloses that. Ask HMRC to identify the evidence of intention to mislead, and press for it in writing before the review or appeal.
• Do not plead staleness. Plead the statutory time limits, s29(5), and (where HMRC’s delay is genuinely egregious) consider whether the facts support a public law challenge instead.
• Test the discovery on subjective grounds. Because discovery is officer-specific, the question is what this officer concluded and when. Request the officer’s notes. A discovery that was never actually made by the assessing officer is still a valid line of attack.
• Check the enquiry vehicle. Tooth is a reminder that HMRC does sometimes use the wrong statutory power. An enquiry opened under the wrong provision is a nullity and does not extend any time limit.

Where Tooth does not help

• It does not assist against the six-year window, which rests on carelessness: a failure to take reasonable care, requiring no intent at all.
• It does not prevent HMRC from assessing late; it only requires HMRC to stay within the statutory limits.
• It does not protect a taxpayer whose return contains a bare inaccurate figure with no explanatory material. The white space is doing a great deal of work in this case.

Frequently Asked Questions