HMRC Discovery Assessments under TMA 1970 s.29: Case-Law Depth for Landlords

The four-bracket time-limit framework under TMA 1970 sections 34, 36 and 36A is the headline structure for HMRC discovery assessments against landlords: 4 years ordinary, 6 years careless, 20 years deliberate, 12 years offshore innocent error. Our companion page on the four-bracket framework walks the headline structure with worked examples. This page is the case-law-depth companion. Where the headline framework decides which window is open, the case-law decides whether HMRC can actually walk through it.

Three pieces of case-law architecture do most of the operational work in landlord discovery cases: the post-Tooth abandonment of the staleness doctrine, the post-Sanderson competent-officer test under s.29(5), and the pre-FA 2008 architecture on brought-forward rental losses. This page walks each in operational language and maps five landlord scenarios onto the framework.

The staleness doctrine is dead, and that matters

Before May 2021 the working operational defence to an out-of-time discovery assessment ran in two layers. Layer one was the s.29(5) competent-officer test: HMRC must show the officer could not reasonably have been aware of the under-assessment at enquiry-window closure. Layer two was the staleness gloss: even where layer one was on HMRC's side, the courts had built a freestanding doctrine that a long delay between the officer's discovery and the actual s.29 assessment could invalidate the assessment. The Upper Tribunal decisions in Charlton v HMRC [2013] UKUT 770 (TCC), Pattullo v HMRC [2016] UKUT 270 (TCC) and Beagles v HMRC [2018] UKUT 380 (TCC) had progressively refined this layer-two doctrine. By 2020 it was being argued in most defended landlord discovery cases as a fall-back to layer one.

HMRC v Tooth [2021] UKSC 17 (judgment 14 May 2021) ended the gloss. The Supreme Court held that the statutory language of section 29(1) does not support a freestanding staleness doctrine. Discovery is a threshold met when the officer comes to the relevant view. Once made, the discovery does not lapse and does not need to be rediscovered. Subsequent delay before the s.29 assessment is issued does not unmake it. The Upper Tribunal authorities that had built the staleness layer were overtaken to that extent.

The operational consequence for landlords is layered. The fall-back delay-based defence is dead. The threshold defence and the competent-officer defence both survive. The real shift is in what arguments to run and in what order: where staleness used to be the second argument after s.29(5), the only remaining second argument is the procedural delay lever under s.28A(4) (covered below), which works on a quite different mechanism (closing the enquiry, not invalidating the assessment).

The s.29(5) competent-officer test: post-Sanderson operational practice

Section 29(5) is the non-behaviour unlock condition. Where HMRC seeks to make a s.29 assessment after the s.9A enquiry window has closed and is not invoking the s.29(4) behaviour-based unlock, s.29(5) requires that an officer could not have been reasonably expected, on the basis of information then available, to be aware of the under-assessment before the enquiry-window closing point.

The leading Court of Appeal authorities are Langham v Veltema [2004] EWCA Civ 193 and HMRC v Sanderson [2016] EWCA Civ 19. Veltema established the basic framing: "information made available" includes anything filed with HMRC or made available to HMRC by the taxpayer at or before the relevant time. Sanderson tightened the test on HMRC's side by clarifying two points. First, the hypothetical officer is one of ordinary diligence rather than a specialist; the officer is not assumed to have specialist knowledge of every line in every manual. Second, the information bar is reasonably high in operational terms; HMRC is not assumed omniscient.

For landlord cases the post-Sanderson operational picture has three working strands.

• What was on the return itself. The white-space disclosure box, the rental-income figures, the schedule of properties (where attached), the foreign-income page (where used). If the matter giving rise to the under-assessment was not disclosed in or alongside the return, s.29(5) generally falls in HMRC's favour.
• What was on related returns. Partnership returns, joint-property elections, Form 17 declarations, CGT returns. A landlord whose joint-property income was correctly disclosed on the Form 17 declaration but mis-allocated on the SA return has an arguable s.29(5) position if HMRC had the Form 17.
• What HMRC actually received from third parties. Connect-system data matches, Land Registry feeds, Common Reporting Standard exchanges from overseas tax authorities, third-party reports from lettings platforms. Post-Sanderson the question is what an officer of ordinary diligence would have done with that data, not what HMRC's specialist Connect analysts would have spotted.

The narrower HMRC's actual information position at the enquiry-window closing date, the stronger the s.29(5) defence. Where the matter was effectively disclosed in or alongside the return, s.29(5) often fails for HMRC; where the matter required HMRC to obtain or analyse third-party data after the closure point, s.29(5) often succeeds for HMRC.

The post-Tooth deliberate-behaviour reset

Tooth did not just kill the staleness doctrine. It also reset the deliberate-behaviour test that controls (a) the s.29(4) behaviour unlock for cases where the enquiry window has closed, (b) the 20-year extension under s.36(1A), and (c) the Schedule 24 FA 2007 deliberate penalty bands (70 per cent and 100 per cent maxima).

The Supreme Court held that deliberate, in this context, means subjective knowing falsity: the taxpayer (or a person acting on their behalf) must have known the figure being submitted was wrong at the time of submission. Negligent submissions, careless submissions, reckless submissions and submissions made under "wilful blindness" do not qualify. The pre-Tooth HMRC practice of characterising "blind eye" cases as deliberate is overtaken.

The downstream effects in landlord cases are material. First, the 20-year bracket under s.36(1A) is harder for HMRC to invoke than it was. Many cases HMRC characterises as deliberate on the opening enquiry letter fall back to careless on a rigorous Tooth analysis. Second, Schedule 24's 70 per cent and 100 per cent maxima become harder to defend in tribunal where the underlying behaviour, properly analysed, was careless rather than knowingly false. Our Schedule 24 penalty bands page walks the band-by-band figures.

The point applies symmetrically to the agent: s.36(1A) reads "by the person or another person acting on the person's behalf". An accountant who knowingly submitted a false figure on a landlord client's return imputes deliberate behaviour to the landlord for the 20-year window even if the landlord was unaware. The defensive analysis on the landlord side is to demonstrate full and accurate disclosure to the agent and an arm's-length supervisory relationship.

The brought-forward losses trap: pre-FA 2008 architecture

Most landlord discovery cases play out within the four-bracket framework on the year the loss of tax arose. The case that breaks the pattern is the landlord with substantial accumulated rental losses brought forward from earlier years. ITTOIA 2005 s.118 allows property losses to be carried forward and set against future rental income from the same property business indefinitely. A landlord who absorbed losses in 2005/06 to 2007/08 and is still utilising those losses in 2026/27 has a live discovery-reach question that the headline four-bracket framework does not directly answer.

The historic pre-FA 2008 architecture treated a brought-forward loss claim as a relief computation rather than an assessment. The argument ran that, even where HMRC could in principle make a discovery assessment on the loss-utilisation year (the year in which the loss is used to reduce current rental tax), it could not reach back to the loss-origination year (the year in which the loss arose) once that year's ordinary time limit had closed. The reach into the origination year depended on whether one of the behaviour-driven extensions under s.36(1) or s.36(1A) was available on the origination year's facts, not on the utilisation year's.

FA 2008 Sch 39 was part of the package that consolidated and tightened the discovery framework. Among its effects, it brought loss-carry-forward claims for post-amendment loss years more firmly into the discovery architecture, so that the pre-FA 2008 distinction between origination-year and utilisation-year reach lost most of its practical bite for new loss years. The pre-FA 2008 architecture, however, still governs loss-origination years that arose before the amendments came into force. For landlords with long-running accumulated rental losses, the question is loss-year by loss-year: did the loss arise before or after the FA 2008 commencement, and if before, is the origination year still within s.29 reach on its own facts?

The practical consequence: if the loss-origination year is no longer within s.29 reach (because its ordinary time limit has closed and no behaviour-driven extension applies on that year's facts), HMRC cannot rewrite a closed-year loss it has no power to reopen, even where it is validly enquiring into the loss-utilisation year. A landlord who carried forward a £40,000 loss from 2006/07 and is utilising £10,000 of it in 2026/27 has an out-of-time defence on the £10,000 reduction if the 2006/07 year is now beyond s.29 reach on its own facts.

The s.28A(4) closure-notice direction: the surviving delay lever

Staleness is dead. The procedural lever that survives is TMA 1970 s.28A(4). Where HMRC has opened an enquiry under s.9A and is sitting on it without progressing to a closure notice, the taxpayer may apply to the First-tier Tribunal for a direction requiring HMRC to issue a closure notice within a specified period. The tribunal must grant the direction unless HMRC have reasonable grounds for not issuing it (s.28A(6)).

This is not a substitute for staleness. Staleness invalidated assessments; s.28A(4) closes enquiries. But for the landlord stuck in a long-running open enquiry where HMRC is repeatedly issuing further information requests without converging on a position, s.28A(4) is the operational tool to force the enquiry to a head. Most s.28A(4) applications are not contested by HMRC: once an application is on file the case officer typically issues the closure notice rather than defend the delay. Where the application is contested, the tribunal applies the s.28A(6) reasonable-grounds test, which weights heavily against HMRC where the information already received is sufficient to issue a closure notice on the merits.

Operationally, the s.28A(4) lever is usable from around 9 to 12 months of open enquiry where HMRC has not progressed substantively. Earlier than that is often premature; later than 12 months is often appropriate. Our companion page on the s.28A closure-notice mechanics walks the procedural detail.

Five worked landlord scenarios mapped onto the framework

Scenario 1: ordinary four-year on-shore, no behaviour, s.29(5) on a knife edge. A higher-rate landlord with two UK buy-to-lets transposed figures from a lettings-agent statement into a 2022/23 self-assessment return, declaring £14,000 of rental income that should have been £18,000. No carelessness on a Sanderson analysis: the figures were transcribed from the agent statement; the discrepancy was not visible without cross-checking against the bank account. HMRC opens a discovery enquiry in November 2026 having received CRS-flagged Land Registry data and run a Connect-system match. The 2022/23 enquiry window closed 31 January 2025; the assessment is within the four-year limit under s.34 to 5 April 2027 but outside the s.9A window. HMRC must therefore rely on s.29(4) or s.29(5). Behaviour analysis fails (no carelessness on the agent-statement transposition); HMRC pivots to s.29(5). The s.29(5) analysis turns on whether HMRC could reasonably have spotted the discrepancy from the agent statement filed with the return. If the agent statement was attached, s.29(5) likely fails for HMRC; if not, it succeeds.

Scenario 2: careless six-year, Sch 24 careless behaviour established. A landlord with three buy-to-lets misclassified £30,000 of capital improvements (a re-wired kitchen, a new bathroom, a replacement boiler) as deductible repairs across 2020/21 and 2021/22. HMRC opens an enquiry in October 2026. Behaviour: careless on a Sch 24 para 3(1) analysis (failure to take reasonable care; competent professional advice on the capital-vs-revenue line was not sought). Time limit: six years under s.36(1) from end of 2020/21 (5 April 2021) = expires 5 April 2027. HMRC is within the bracket. The defensive analysis is band-down within Sch 24 (careless unprompted disclosure for the 0 per cent floor under para 10) and possible suspension under para 14.

Scenario 3: deliberate twenty-year, post-Tooth deliberate test in play. A landlord knowingly omitted £40,000 of cash rental income per year over 2018/19 to 2022/23 (five years, £200,000 of undisclosed income). HMRC discovers via informant in October 2026. HMRC characterises behaviour as deliberate on the opening enquiry letter and asserts the s.36(1A) 20-year window. On a Tooth analysis the deliberate test is met (the landlord knew the figures being submitted were false; cash rental was deliberately routed outside the declared bank account). The 20-year window applies. Exposure on the income-tax side runs to the full five years, with Sch 24 deliberate-band penalties at 70 per cent of potential lost revenue (unprompted floor 20 per cent; prompted 35 per cent). Where the cash routing involved active concealment (third-party bank account, false invoicing), the band steps up to deliberate-and-concealed (100 per cent maximum; 30 per cent / 50 per cent floors).

Scenario 4: offshore twelve-year innocent error under s.36A. A landlord with a Spanish holiday let took written advice from a UK adviser in 2014 that the income was exempt under the UK-Spain double-tax treaty (the advice was wrong; the property rental income is fully UK-taxable subject to foreign tax credit). The landlord disclosed the Spanish income on Spanish returns but not on UK self-assessment returns from 2014/15 to 2024/25. HMRC discovers via CRS data feed in October 2026. Behaviour analysis: not careless (full disclosure to a competent adviser; reasonable reliance on advice); not deliberate (no knowing falsity). HMRC's reach is via s.36A subsection (2): 12 years from the end of each relevant tax year. The 2014/15 year has 12-year reach to 5 April 2027 (open by a margin); 2015/16 has reach to 5 April 2028; and so on through 2024/25 to 5 April 2037. Eleven separate years are within s.36A's reach. Penalties on the offshore innocent-error limb under Sch 24 are typically zero per cent (no behaviour band engages on a Hanson-line reliance-on-advice analysis), but the FA 2017 Failure-to-Correct framework under Sch 18 imposes a standalone offshore-correction obligation with its own minimum penalty for failures to disclose by 30 September 2018 (covered separately in our offshore-disclosure coverage).

Scenario 5: brought-forward losses pre-FA 2008 trap. A landlord with a long-running BTL portfolio absorbed £80,000 of accumulated rental losses in 2005/06 to 2007/08 (pre-FA 2008 amendments) from a period of high finance-cost interest and adverse tenancy losses. The losses have been carried forward under ITTOIA 2005 s.118 against subsequent rental profits. In 2026/27 the landlord utilises £15,000 of the accumulated loss against current rental income. HMRC opens an enquiry in 2028/29 challenging the original 2005/06 computations. The s.29 reach into the 2005/06 origination year depends on (a) whether the loss claim was made on a return then within s.29 discovery scope (pre-FA 2008 architecture: the claim was a relief computation, not an assessment, so s.29 reach was narrower than the post-FA 2008 framework) and (b) whether the behaviour analysis for 2005/06 meets the s.36(1) careless or s.36(1A) deliberate unlock on that year's facts. If neither, the 2005/06 year is closed for s.29 purposes and HMRC cannot rewrite the loss. The 2026/27 utilisation challenge then falls away: HMRC cannot reduce a brought-forward loss it has no power to reopen.

Records to defend an out-of-time discovery

The post-Tooth landscape puts most of the weight on the s.29(5) competent-officer analysis when HMRC is not relying on the s.29(4) behaviour unlock. That analysis is evidence-driven. Records that show what was on the return, what was attached to it, what white-space disclosures were made and what HMRC subsequently received from third parties are the load-bearing evidence in a s.29(5) defence.

TMA 1970 s.12B requires self-assessment records to be kept for five years after 31 January following the tax year. A 2026/27 record must be kept until 31 January 2033. For corporation tax CA 2006 s.388 sets a six-year floor from the end of the accounting period. The s.12B(5) record-keeping penalty is capped at £3,000 per year of failure; the bigger exposure is on the substantive Sch 24 / Sch 41 penalties on consequential tax loss, not on s.12B itself.

Property Tax Partners' standing recommendation is to keep records for seven years (one year above the s.12B floor) for landlord clients. Where a discovery assessment is made outside the ordinary four-year window, the seven-year buffer routinely makes the difference between a tractable s.29(5) defence and a stretched-evidence position. The records to keep include: the return itself in filed form; any schedules attached; the white-space disclosure box; the lettings-agent statements; bank statements showing rental receipts; mortgage statements; repair invoices; tenancy agreements; deposit-protection records; HMRC correspondence; and (for offshore matters) foreign-tax-credit certificates and overseas-territory tax returns.

The operational playbook for landlords on a discovery assessment

When a s.29 discovery assessment arrives the response framework is:

1. Within the first 7 days. Identify the year of assessment and the time-limit bracket HMRC is relying on. The bracket should be discernible from the assessment letter; if not, request clarification within the appeal window.
2. Within the first 14 days. Test the time-limit position on the headline four-bracket framework. An assessment outside the relevant window is invalid and can be vacated on appeal. Our four-bracket framework page walks the year-by-year arithmetic.
3. Within the first 21 days. Test the s.29(4) / (5) unlock. Where HMRC is using s.29(5), what was the information available to HMRC at the enquiry-window closing date? Where HMRC is using s.29(4), what is the evidence for careless or deliberate conduct? On the deliberate end, the Tooth subjective-knowing-falsity test is the controlling threshold.
4. If brought-forward losses are in play. Run the loss-by-loss origination analysis. Pre-FA 2008 origination years often fall outside s.29 reach on their own facts and the utilisation-year challenge then falls away.
5. Within the 30-day appeal window. Lodge a notice of appeal under TMA 1970 s.31A. ADR engagement does not pause the 30-day clock. Late appeals require a reasonable-excuse application under the Martland three-stage framework.
6. If the wider enquiry is open and HMRC is stalling. Consider a s.28A(4) closure-notice direction application after around 9 to 12 months of open enquiry. This is the procedural lever that survives the staleness-doctrine death.

The time-limit defence under s.34 / s.36 / s.36A is procedural and decisive. An assessment outside the relevant bracket falls regardless of the underlying merits. The s.29(5) competent-officer defence operates inside the bracket and turns on evidence. The brought-forward losses trap operates on the year-of-origination side and turns on the pre-/post-FA 2008 architecture. Getting all three layers right is the highest-value first analytic step in any discovery-assessment response.

Penalty interaction once a discovery is confirmed

The s.29 / s.34 / s.36 / s.36A framework determines whether HMRC can assess at all. The Schedule 24 FA 2007 framework then determines what penalty attaches to any assessment that does get made. The two are independent: a 6-year careless discovery assessment under s.36(1) attracts a Sch 24 careless penalty; a 20-year deliberate assessment under s.36(1A) attracts a Sch 24 deliberate penalty; a 12-year offshore innocent-error assessment under s.36A can attract no penalty at all on a Hanson-line reliance-on-advice analysis. Our Schedule 24 band matrix page walks the bands and the offshore Category 1 / 2 / 3 uplifts.

For voluntary-disclosure routes that operate within the same penalty architecture (Let Property Campaign for residential rental, Worldwide Disclosure Facility for offshore, Digital Disclosure Service for general UK), our coverage of the LPC three-step process sets out the operational mechanics.

If HMRC has issued a discovery assessment against you and you are working through which time-limit bracket applies, what s.29 unlock condition HMRC is relying on, or whether the brought-forward losses trap shrinks HMRC's reach into the loss-origination year, the form at the foot of the page is the route to a structured first-pass assessment.