Covenant Risk and the Cost of Finding Out Late: Risk Without a Test Date

Consider one clause in one head-lease. At acquisition the buyer’s counsel reads it, notes the unexpired term in the report on title, and moves on; the term is long enough for the hold, and the price has been agreed on that basis. Six years later the same clause is read again, by the lender to the prospective purchaser, whose credit policy sets a minimum residual term after the loan matures. The clause has not changed. The term is six years shorter and the lender’s minimum has not moved. The asset lends at a lower advance rate, or not at all, and the exit price absorbs the difference. The same words were read twice, for two different parties, and the second reading cost the equity what the first could have saved.

Every risk in a deal carries two dates: the date on which it is priced, and the date on which somebody with a stake in the answer reads the document and draws the conclusion. The cost of the risk to the equity is set less by its size than by the gap between those dates, because a risk discovered before pricing is a negotiation and a risk discovered after pricing is a loss.

Credit has spent a century closing that gap by contract. A loan agreement names a metric, a threshold and a date on which the metric is tested, and it puts the test in the hands of a counterparty with an incentive to run it. Roberts and Sufi found that more than 90% of long-term private credit agreements are renegotiated before their stated maturity, mostly early in the life of the contract and rarely because of distress. Credit expects to learn things mid-life and writes the occasions into the paper. It carries a second scheduled date as well, maturity: the Mortgage Bankers Association puts USD 875 billion of US commercial and multifamily mortgage balances, 17% of the USD 5.0 trillion outstanding, at maturity during 2026.

Real estate equity leaves the gap open. The title class of the asset, the terms of the head-lease and the concentration of lease expiries are all knowable on the day of acquisition and all priced on it, and none is tested again on any date fixed by contract. They are tested by events: a sale, a refinancing, an expiry, an objection from a neighbour. The claim of this paper is that who looks, and when, does more to set the cost of structural risk than anything in the risk itself. An earlier note dealt with where a memo’s numbers come from; this one deals with the date on which anyone reads them again.

The Smart Bricks Research Lab is working through that second date, drawing on the credit-contract literature, the Bayes lending surveys, MSCI’s lease events series and the published record of cases in which a document was read late. What follows is the covenant as a discovery mechanism, three real estate risks that lack one, and a way to give them one.

Risk	Knowable at entry	Who runs the test	When it is found	What is still curable
ICR or LTV covenant	Knowable at entryYes, from the facility agreement	Who runs the testThe lender, on a contractual test date	When it is foundQuarterly or at valuation, with a cure period	What is still curableEquity cure, waiver, re-margin, prepayment
Head-lease unexpired term	Knowable at entryYes, from the title and head-lease	Who runs the testNobody, until the buyer's lender	When it is foundAt exit or refinance, when the residual term fails a lending test	What is still curableExtension or regear, if opened years earlier
Lease-expiry concentration	Knowable at entryYes, from the abstracts	Who runs the testNobody, until the void	When it is foundAt expiry, when roughly half of UK leases fell vacant in 2018	What is still curableRenewal, if opened before the s.25 window closes
Restrictive covenant on use	Knowable at entryYes, from the register	Who runs the testNobody, until the objector	When it is foundAt change of use or exit, by injunction rather than damages	What is still curableRelease or s.84 application, before any breach
Escalating ground rent	Knowable at entryYes, from the lease	Who runs the testNobody, until the buyer's valuer	When it is foundAt sale or mortgage, when the unit will not lend	What is still curableDeed of variation, before marketing

A covenant fixes the date on which someone must look. The structural facts of a property have no such date, so they are read by whoever is pricing the asset next, and priced against the seller.

Source: Smart Bricks Research Lab, drawing on Landlord and Tenant Act 1954 s.25; MSCI UK Lease Events Review (Nov 2019); CMA leasehold investigation (2019-2022); Alexander Devine v Housing Solutions [2020] UKSC 45

A financial covenant has four parts, and they are worth separating because real estate borrows the first two and forgets the other two. There is a metric: interest cover, loan-to-value, debt yield. There is a threshold. There is a test date, usually quarterly, on which the borrower delivers a compliance certificate and the lender checks it. And there is a cure right: a period, and a set of permitted responses, an equity injection, a prepayment, a waiver negotiated for a fee. Around these four sits a fifth element that appears nowhere in the document: a counterparty who wants the test run, because its capital is at stake.

Read this way, the covenant is a piece of information architecture before it is a constraint. Its purpose is to convert a slow deterioration into a dated event that both sides can see and act on. The proof is the frequency of violation and the mildness of its consequences. Nini, Smith and Sufi, working from SEC filings between 1996 and 2008, found that between 10% and 20% of US non-financial firms report being in violation of a financial covenant in any given year. That is a population being renegotiated rather than a population in default: after a violation, firms cut acquisitions, capital expenditure, leverage and payouts, and their chief executives are more likely to be replaced. The test date is where creditor control is exercised in fact, and far more often through changed terms than through enforcement.

Roberts and Sufi’s renegotiation evidence points the same way: in their sample of a thousand private credit agreements, renegotiation altered the amount, the maturity and the pricing, and the contingencies written into the original contract shaped the bargaining that followed. The covenant did not stop the borrower deteriorating. It fixed the date on which deterioration became a conversation with terms attached, while the borrower still had room to move.

UK commercial real estate lending shows the machinery running under load. The Bayes Business School lending report for year-end 2023 recorded 60% of lenders reporting covenant breaches across their books and 61% reporting defaults; the interest-cover distribution had deteriorated visibly over two years, and defaulted loans averaged 3.9% of the sample. These figures exist because the loans were tested on dates and the results were collected. A portfolio of properties produces no equivalent series, because nothing in an equity hold requires the equivalent test.

The facility itself is narrow about what it tests. The market-standard form conventionally carries two ratio tests, one on value and one on interest cover, and most facilities in practice test no more than three. Nothing in the facility tests the unexpired term of the head-lease, the concentration of expiries in year three of the loan, or whether a restrictive covenant on the title conflicts with the business plan. The lender reads those once, at origination, in its own counsel’s report, after which they join everything else in the deal that has a price and no date.

• 10% to 20%

  US non-financial firms reporting a financial covenant violation in any given year, 1996-2008

  Nini, Smith and Sufi, RFS 2012

• Over 90%

  Long-term private credit agreements renegotiated before stated maturity, rarely in distress

  Roberts and Sufi, JFE 2009

• 51%

  UK CRE loans with ICR above 2x at year-end 2023, down from 66% a year earlier

  Bayes CRE Lending Report YE 2023

• 22%

  UK CRE loans below 1.4x ICR at year-end 2023, from 14% in 2021

  Bayes CRE Lending Report YE 2023

• USD 875bn

  US commercial and multifamily mortgages maturing in 2026, 17% of the USD 5.0tn outstanding

  MBA, February 2026

Violation is common, renegotiation is the norm, and the calendar is public. Credit does not avoid deterioration; it dates it.

Source: Nini, Smith and Sufi (2012); Roberts and Sufi (2009); Bayes Business School (Lux, 2023); Mortgage Bankers Association (9 Feb 2026)

Before the contrast is pushed further it needs a complication, because the covenant’s discovery machinery is only as current as its inputs. Loan-to-value is tested against a valuation. Interest cover is tested against a rent roll. Both inputs lag the world they describe.

The Bayes year-end 2023 report is direct about the consequence. It notes that many covenant breaches had been recorded over the preceding six to twelve months, and that with valuations being delayed, lenders had so far taken no further action. Property values fell 5.9% on average over 2023 by the report’s measure, and the report observes that loan-to-value ratios are always slow to change. The mechanism is plain. A loan-to-value covenant is tested when a valuation lands; the lender commissions the valuation; and a lender that would rather not crystallise a breach can let the commissioning date drift. The test is scheduled and the input is not, so discovery happens on the calendar after the calendar has been quietly moved.

Interest cover has the same weakness in a different place. The compliance certificate reports passing rent as at the quarter end. A break notice served three weeks after the quarter end reaches the next certificate, and a tenant in arrears reaches it only when the arrears meet whatever the facility defines as a default. The rent roll behind the certificate is kept current by people, and carries whatever they have got round to entering. So the covenant test, even when it fires on time, fires on a picture taken some weeks or months earlier.

This matters for the argument in two ways. First, the credit side is no model of perfection against which real estate equity is being measured; it is a scheduled process with known input lag, and the point of comparison is the schedule rather than the accuracy of what the schedule reads. Second, it sets a limit the paper returns to at the end. A test date discovers what its inputs report, on the date they report it, and a firm that gives its structural risks a test date has bought a schedule rather than foresight.

Take the three structural risks in turn. Each was fully knowable at entry, each was static through the hold, and each was priced when a third party read the relevant document at a market event.

Title class. In England, title class is a mass fact rather than an edge case. The housing ministry counts 4.83 million leasehold dwellings, 19% of the stock and 38% of the stock in London, and its factsheet on remaining terms finds 412,000 households in homes with 80 years or less to run. Those are residential figures, cited here only for scale, but the 80-year line they describe is the clearest example anywhere of a structural threshold that arrives on a fixed calendar. Nothing about the lease changes on the day the term crosses the line except that the buyer’s valuer and the buyer’s lender now read it differently, and the unit becomes harder to mortgage and to sell. The date of the crossing is known to the day at acquisition. Nothing in the underwriting standard requires it to be written down as a date on which anybody must act, the way a covenant test is written down.

Restrictive covenants on the register behave the same way with a sharper edge. In Alexander Devine Children’s Cancer Trust v Housing Solutions, a developer built 13 of 23 affordable homes on land burdened by a covenant that forbade building, knowing of the covenant and choosing to proceed, and then applied to have it modified. The Supreme Court refused in November 2020, holding that it was not in the public interest for a party to secure modification by relying on a state of affairs its own deliberate breach had created. The covenant had sat on the register throughout. Once it was read for the beneficiary rather than for the developer, the remedy was an injunction and a question about demolition, which no sum of money could settle. A risk of this kind cannot be cured by paying once the breach exists, which is why the date on which it is read matters more here than for almost any other risk in the deal.

Head-lease terms. The doubling ground rent is the cleanest documented case of a clause priced at entry as harmless and discovered at sale as disqualifying. The Competition and Markets Authority opened its investigation in June 2019 into leases whose ground rent doubled every ten to fifteen years. The clause had been in each lease from the day it was signed. Its effect, in the authority’s words, was that owners could often struggle to sell or mortgage their homes, because lenders reading the clause years later declined to lend against it. By August 2022 the authority reported that over 20,000 leaseholders had been helped by the investigation, through commitments by freeholders to remove the terms. That number counts people who had a clause read for them late, at the moment they tried to sell, by a counterparty pricing against them. Commercial head-leases carry the same species of term, alienation restrictions, forfeiture provisions, reviews geared to an index the business plan never modelled, read once by counsel at entry and next by the buyer’s lender at exit.

Expiry and break concentration. A weighted average unexpired lease term is the figure most memos use to describe income risk, and it is an average, so it has no cliff. MSCI’s lease events series shows what the cliff looks like when it arrives. In 2018, of the leases in the UK index that reached expiry, 29% were renewed by the tenant and 56% fell vacant for a quarter or more, the highest vacancy rate at expiry since the series began in 1998 and the lowest renewal rate since 2007; weighted by rent the outcome was worse. Where a break option arose, tenants exercised it in roughly one case in five, and 26% of the contracted income exposed to a break was lost, in line with the long-run average. An expiry, on those numbers, is close to a coin toss on the income behind it, and the coin is tossed on a date written into the lease years before the current owner read it.

The mechanics around the date are as binary as any covenant test. In Friends Life v Siemens Hearing Instruments, the Court of Appeal in 2014 held a tenant’s break notice void because it omitted the statement, required by the lease, that it was given under section 24(2) of the Landlord and Tenant Act 1954. The tenant had served on time, in writing, with an unmistakable intention to break, and one missing sentence kept it bound to the lease, because a unilateral option must be exercised exactly as the contract says. That is the character of a covenant test: a fixed date, a fixed form, a binary outcome. The difference is that a lender’s facility agreement requires the covenant test to be scheduled and tested; no equivalent requirement exists for a landlord’s own break options, so whether one is diarised depends on the individual asset manager’s practice rather than on a standard.

In each of these cases the document was complete on the day the price was agreed. What moved was the reader, and the party the reader was pricing for.

• 56%

  UK leases expiring in 2018 that fell vacant for a quarter or more (61% weighted by rent), highest since 1998

  MSCI UK Lease Events Review, Nov 2019

• 29%

  Expiring leases renewed in 2018 (23% by rent), lowest since 2007

  MSCI UK Lease Events Review, Nov 2019

• 26%

  Contracted income lost where a break arose in 2018; 79% of breaks not exercised by rent

  MSCI UK Lease Events Review, Nov 2019

• 412,000

  English households in leasehold homes with 80 years or less remaining, 2023-24

  MHCLG, May 2026

• Over 20,000

  Leaseholders freed from escalating ground rent terms that left homes hard to sell or mortgage

  CMA, August 2022

• 6 to 12 months

  Statutory window for a landlord's section 25 notice; a defective break notice is void

  LTA 1954 s.25(2); Friends Life v Siemens [2014] EWCA Civ 382

None of these facts moved during the hold. Each was priced when a third party read the document at a market event.

Source: MSCI in association with BNP Paribas Real Estate (Nov 2019); MHCLG leasehold statistics (2025, 2026); Competition and Markets Authority (2021, 2022); legislation.gov.uk; The National Archives Find Case Law

In credit the counterparty finds the problem on a schedule. In real estate the next buyer finds it in the price.

Smart Bricks Research Lab

Why do these risks go unscheduled in firms that schedule almost everything else? The answer is structural rather than negligent. The reporting frameworks an institutional investor answers to are built around periodic performance and periodic valuation. INREV’s guidelines, the NCREIF and MSCI indices and ILPA’s reporting templates each ask what the asset is worth and what it returned over the period. None asks how far ahead the firm has looked, or when it last looked.

Valuation is where structural risk is supposed to be captured, and the valuation standard is candid about how loosely it captures it. The RICS Red Book requires a valuer to reflect the strength of the tenant covenant in the figure, and it prescribes no method for scoring that strength and no cycle on which the view must be refreshed. The valuer reads the tenancy schedule on the valuation date and forms an opinion. Between valuation dates the opinion stands, and nothing in the report records how far ahead the valuer looked at the expiry profile or when the head-lease was last read against the business plan.

Nor does any framework measure the thing this paper is about. No published standard says how many months before a lease event a firm must have logged it, none reconciles the voids a firm forecast to the voids it went on to experience, and none records how long a structural exposure or a covenant breach went unexamined before someone acted. A firm can report against every mainstream framework and never produce the number that matters here: the months between the date a fact became knowable and the date somebody looked.

The industry knows how to publish a calendar; it publishes one for outsiders and not for itself. Listed REITs in the United States disclose lease expiration schedules, showing the rent that expires in each coming year, and European listed companies report a vacancy rate under EPRA’s guidance. An expiration schedule is dated and a weighted average lease term is not, and the same average can describe a smooth ladder or a portfolio with most of its rent expiring in a single year. On the MSCI outcome figures above, the second portfolio carries something close to an even chance of losing that rent to vacancy when the year arrives, and the average says nothing about it. The schedule the market requires of a listed landlord, dated and disaggregated, is not a schedule any framework requires a private fund to run for itself.

Borrow the covenant’s four parts and apply them. Each structural risk becomes a scheduled test with a metric, a written threshold, a test date, a cure window and a named owner, and each test is computed from the documents rather than from a summary of them.

Head-lease headroom. The metric is the unexpired term, less the planned hold, less the minimum residual term a purchaser’s lender will require at exit. The threshold is the firm’s own, written down at acquisition, with the 80-year residential line kept in view as a reminder that thresholds in this market are real and arrive on a known day. The test runs on every valuation date and whenever a title or head-lease document changes. The cure is an extension, a regear or a deed of variation, and the test must run early because each of those takes quarters to negotiate, against a counterparty who knows exactly what the calendar is doing to the equity’s position.

Expiry concentration. The metric is the share of contracted rent reaching expiry or an exercisable break within any rolling twenty-four months, with breaks weighted by the observed exercise rate. MSCI’s 2018 outcomes, roughly one lease in five broken and about a quarter of the exposed rent lost, are a defensible opening weight, reviewed each year. The threshold is written. The test date is monthly, because the metric is computed from the lease abstracts and the abstracts change. The cure window is fixed by statute: a landlord’s section 25 notice under the 1954 Act must be served not more than twelve nor less than six months before the termination date it specifies. A renewal conversation that has not begun twelve months out has already let the statute shape the outcome, and a leasing effort that has not been pointed at the coming void has an appointment with the empty-rates clock, which allows three months without business rates on most empty property, and a further three months on industrial premises, before the full liability falls due.

Title class and use. The metric is every restrictive covenant, easement and rent-escalation clause on the title, read against the intended use in the business plan and against saleability and mortgageability at exit. The threshold is zero unresolved conflicts. The test runs at entry and again on any change of use or capital plan, because the plan is what turns a dormant covenant into a live one. The cure is a release negotiated with the beneficiary, indemnity insurance, or an application under section 84 of the Law of Property Act 1925, and Alexander Devine is the reminder that an application made after the breach may be refused because of it.

Around the three tests sit the design rules credit already uses. Each test has one named owner. The tolerance is written before the test is switched on, because a test without a tolerance fires constantly and is soon ignored. Each test recomputes when a document lands: a new lease, an amendment, a valuation, a revised plan. And exceptions reach the investment committee on the same cadence as compliance certificates reach a lender, with the evidence attached. None of this asks an analyst to remember a date. It asks that the documents be read by something that recomputes when they change, which is work suited to a system and poorly suited to a spreadsheet kept by hand.

1. 1

   Head-lease headroom

   Unexpired term minus planned hold minus the buyer's lender's minimum residual term • Threshold written at acquisition • Tested at every valuation date and on any title or head-lease change • Cure: extension, regear or variation, opened years out • Owner: asset manager

2. 2

   Expiry concentration

   Share of contracted rent expiring or exercisable in any rolling 24 months, break-weighted at the observed exercise rate • Threshold written at acquisition • Tested monthly from the abstracts • Cure: renewal opened before the s.25 window; leasing pipeline pointed at the void • Owner: leasing lead

3. 3

   Title class and use

   Every restrictive covenant, easement and rent-escalation clause read against intended use, saleability and mortgageability • Zero unresolved conflicts with the business plan • Tested at entry and on any change of use or capex plan • Cure: release, indemnity, s.84 application, before any works • Owner: legal

4. 4

   Exception reporting

   Each failed test reported to the committee with its evidence, on the same cadence as covenant compliance certificates • Written tolerance per test • Every cycle • Cure: decision recorded, or tolerance revised with reason • Owner: CIO

Borrowed from the covenant: a metric, a threshold, a test date, a cure window and an owner. Each is computed from the documents and recomputed when a document changes.

Source: Smart Bricks Research Lab

The cost of scheduling structural risk is mostly a calendar and an owner. The three schedules can be built for an existing portfolio from the lease abstracts, the head-leases and the reports on title, all of which are already in the data room, and the arithmetic in each is trivial; the build is measured in analyst-days rather than months. The standing cost is the person who clears the exceptions each cycle and the annual review of the tolerances. A firm that cannot name the person who owns the exception list has not yet paid for the test.

The limits deserve equal candour. A test date finds risk it cannot remove. A head-lease with too little term for the next buyer’s lender has, at exit, one cure, which is price; the test moves the discovery earlier so that the price is paid deliberately at entry, or the extension is opened while there is time to negotiate it, but the test itself removes nothing. The same holds for a covenant on the register: the test surfaces the conflict before the works begin, and the beneficiary still has to be found and paid or persuaded.

The test also inherits input lag, exactly as the loan-to-value test does. The Bayes finding that breaches were recorded while delayed valuations deferred any action is the limit case: a scheduled test with stale inputs discovers late on schedule. A structural test computed from abstracts that lag the leases has the same defect, and the only remedy is recomputation on document change.

The outcomes behind the thresholds are cyclical. MSCI describes break exercise rates as highly cyclical, driven by the gap between in-place and market rents, and its series shows the rent-weighted share of income lost at a break ranging from 23% in 2017 to 42% in 2011. A weight taken from 2018 is a base rate from one year in one market, and a firm that treats it as a forecast has misread the test. Scheduled discovery reports what is knowable today about a date in the future, how much rent reaches an expiry, how much term remains, what the register says; it does not say what the tenant will decide or what a lender will require in six years. What it does is move the reading of the document from the party who profits from a late reading to the party who pays for it.

Credit’s calendar is public. The Mortgage Bankers Association publishes the maturity wall each year, and reported it shrinking from USD 957 billion in 2025 to USD 875 billion in 2026; the market can see the dates coming and price them. A portfolio’s structural calendar could be equally visible to the firm that owns it, built from the documents it already holds. The only discovery that costs the equity nothing is the one made before the price is set, and a test written into the acquisition memo is the closest thing to it available afterwards. That is the calendar’s original argument, and the entry price is where it begins.