The Profession That Welcomed It

The matching adjustment

When a company hands its pension scheme to an insurer — a bulk purchase annuity, or "buyout" — the insurer receives a premium and assumes an obligation to pay pensions for fifty years or more. The price of that obligation depends on the rate at which future pensions are discounted, and here UK regulation contains a remarkable concession.

An ordinary investor values a safe liability using a safe rate. An annuity writer is permitted to do something better: under the matching adjustment (MA), it may discount its pension liabilities at the yield of the assets it happens to hold against them. Buy higher-yielding assets, and the liability itself shrinks. The justification is not absurd. Pensioners cannot surrender their annuities; there can be no run on an annuity book. An insurer holding a bond to maturity bears only the risk of default, not the risk of market prices wobbling in between. So, the argument runs, the portion of a bond's spread that merely compensates for illiquidity — for locking money away — is genuinely earned by an investor who never needs the money back early, and may fairly be banked.

The catch is the word "portion." Some of any credit spread is compensation for default risk, which the insurer retains in full. The regulation therefore prescribes a deduction — the fundamental spread (FS) — representing retained credit risk. Only the remainder flows into the discount rate:

Liability discount rate = risk-free rate + MA, where MA = asset spread − FS.

And since the asset spread is itself measured over the risk-free rate, the two formulations collapse into the same thing:

risk-free + (spread − FS) = asset yield − FS

— which is the concession in its plainest form: the insurer discounts its pension promises at the yield of its own portfolio, less the prescribed deduction, while everyone else must discount at the risk-free rate alone.

Both ingredients deserve a closer look, because neither is quite what an outsider would assume.

The asset spread is the excess of the portfolio's yield over the regulator's risk-free curve. For a traded corporate bond this is observable: the market sets the price, the price implies the yield, the yield implies the spread. But for the assets that increasingly dominate annuity books — restructured lifetime mortgages, regeneration debt, private placements — there is no market price. The yield is whatever the insurer's own valuation model says it is, on an asset the insurer may itself have originated and rated. The first input to the formula, in other words, is to a significant degree self-declared.

The fundamental spread is a prescribed deduction, published by the regulator in tables by credit rating, sector and maturity, built from two components: the expected cost of defaults (long-run historical default probabilities and recovery rates for each rating class) and the expected cost of downgrades (the loss from having to sell a downgraded asset and replace it to keep the portfolio matched). Both components are derived from decades of rating-agency data on publicly traded bonds, averaged across cycles. A floor applies — broadly 30 per cent of the long-term average spread for non-financial issuers, 35 per cent for financials — and in benign markets the floor is frequently the binding number, meaning the deduction is anchored to a thirty-year average rather than to anything happening in the world today. Nothing in the calculation looks at the specific asset; two instruments with the same rating and term receive the same FS, whether one is a Tesco bond with a forty-year trading history and the other a tranche of equity release mortgages rated by the insurer holding it.

The downgrade component deserves a moment, because it is where the current formula quietly gestures at the risk the reform wanted to price properly. It is computed from a rating transition matrix — the agencies' long-run average probabilities that a bond moves from one rating to each other rating in a year, default included. For each possible downgrade, the regulator estimates the loss an insurer would take from selling the now-cheaper asset and buying back a higher-rated, lower-yielding one to keep the portfolio matched; those losses are weighted by their historical probabilities and spread across the asset's life. It is, in effect, a reserve for the cost of being forced to rotate out of deteriorating credit. But notice three things. It is built on the same averaged, public-bond data — so it is blind to self-rated illiquid assets that have no transition history at all. It moves with the long-run matrix, not with markets — so, like the default component, it barely stirs in a crisis, exactly when downgrades cluster. And it sits in quiet tension with the matching adjustment's own founding claim: the MA earns its illiquidity premium precisely because the insurer never has to sell — yet the downgrade charge reserves only modestly for selling, on the assumption forced rotation is rare. The firm cannot fully have it both ways. The honest reading is that the downgrade component is a static, average-based proxy standing in the place where the supervisor wanted a dynamic, uncertainty-sensitive charge — which is the credit risk premium the reform proposed and the law omitted. The partial overlap was convenient: because the formula already gestured at downgrade risk, the industry could argue a fuller credit risk premium would be double-counting — a weaker charge cited to ward off a stronger one.

Everything in this story turns on that subtraction. If the FS is calibrated correctly, the MA is a sensible recognition of genuine economics. If the FS is too small, then part of what the insurer books as risk-free "illiquidity premium" is actually unrecognised default risk — and because the MA reduces liabilities immediately, the error is not a slow leak but a day-one capital injection. Every gilt swapped for a higher-spread asset manufactures surplus on the spot, surplus that funds the pricing of the next buyout. The FS is not one parameter among many. It is the valve through which the whole bulk annuity market is pressurised.

Notice what the construction implies. The FS is blind to the actual assets, and it is almost completely static: in March 2020, when credit spreads doubled in a fortnight of genuine panic, the published deduction barely moved — meaning the entire blow-out was classified, by construction, as illiquidity premium, and annuity balance sheets strengthened in the teeth of a credit crisis.

Why those properties matter becomes clear when you look at what the assets actually are.

What sits behind the pensions

The day before a buyout, a mature pension scheme typically holds 70–80 per cent gilts and gilt-like instruments. The day after, the same pension cashflows are backed by a portfolio the insurer has rebuilt for yield — because yield, via the MA, is capital.

A representative large writer's book splits roughly as follows: a fifth to a quarter in sovereigns and cash, much of it pledged as collateral for the enormous swap overlay that manufactures the inflation-linkage pension benefits require; a third or so in public investment-grade corporate bonds; and a quarter to a third in what the industry calls "direct investments" or illiquids — lifetime mortgages restructured into notes, infrastructure and project-finance debt, social housing loans, commercial property lending, urban regeneration finance. Add the structured slice, and on a look-through basis something approaching 40 per cent of the assets behind a modern annuity book is private or structured credit — assets with no market price, no trading history, and in many cases no external rating. They are valued by models, and rated, frequently, by the insurers holding them.

Now recall the FS's two properties. The deduction meant to capture the default risk in this portfolio is calibrated on public bond data that describes none of these assets, and it does not respond when credit conditions deteriorate. The higher the spread an insurer can source — or assign — within a rating bucket, the larger the MA, the larger the day-one surplus. As the PRA's own analysis put it, the structure positively incentivises the hunt for the highest-spread assets each rating notch will tolerate. The drift of annuity portfolios from gilts toward self-rated private credit is not a corporate fashion. It is the rulebook, doing exactly what it rewards.

Releasing the capital twice

The MA is the first release valve. There is a second.

A growing share of new buyout business — on some recent deals, 20 to 40 per cent — is passed straight through to offshore reinsurers under funded reinsurance: the UK insurer cedes both the liabilities and the premium, and holds in their place a collateralised IOU. The collateral pools backing those treaties, typically domiciled in Bermuda and often managed by reinsurers affiliated with private equity houses, run noticeably hotter than the books their UK cedants retain — roughly twice the weight in structured credit and private lending — at capital requirements lighter than the PRA's. The pensioner's security now depends on a chain: a UK insurer's claim, on an offshore vehicle, holding privately originated assets, valued by the vehicle's affiliates, supervised elsewhere. The Bank of England was sufficiently alarmed to issue a dedicated supervisory statement on funded reinsurance in 2024; the business has continued to grow.

Layer the two mechanisms and the elegance is undeniable. The MA converts asset spread into immediate capital onshore; funded reinsurance exports the residual liability to a jurisdiction where the same trade can be run again, harder. At each stage, capital is "released" — which is to say, the buffer between the pension promise and the credit cycle is thinned, lawfully, with the proceeds recycled into winning the next deal.

Originating your own yield

There is a final refinement, and it is the one that should give even a sympathetic reader pause: the largest writers increasingly manufacture the assets themselves.

Legal & General made the model explicit strategy. Its capital arm sponsored and took equity in regeneration schemes, build-to-rent housing, later-living developments, science parks; the projects issued debt; and the group's annuity fund bought the debt, structured to be MA-eligible. The same loop runs through equity release: the group originates lifetime mortgages at retail, restructures them into notes, and the annuity book holds the notes. The branding — "self-manufactured assets," inclusive capitalism — was cheerful and unembarrassed, and the underlying logic is not foolish: fifty-year pension liabilities are the natural funder of fifty-year illiquid projects, and origination capacity is scarce.

But observe what the arrangement does to the one number that matters. The "spread" on internally originated debt is not set by any market; it is set within the house — by a group that is simultaneously the project's sponsor, equity holder, originator, valuer and creditor. That spread, minus a fundamental spread calibrated on public bonds that resemble these assets not at all, becomes matching adjustment, becomes day-one capital, becomes pricing power. The insurer is, in a precise and structural sense, marking its own homework, and the regulation converts the marks into money.

None of this is hidden, and none of it is illegal. Which is why the location of the safeguard matters so much. With assets unpriced by markets, valuations produced by interested models, and liabilities that cannot run and therefore never force a reckoning, the only line of defence left is the calibration of the fundamental spread — and the institutions charged with getting it right.

What the machine is worth

How much do these mechanisms actually matter? Enough that without them the product could not exist at its current price. The arithmetic is worth walking through, because each layer can be priced. Take an illustrative £1 billion buyout of pensions in payment, with liabilities of roughly twelve years' duration, and build the balance sheet up from a clean baseline — an insurer holding gilts and discounting at the risk-free rate, the way the rest of the financial world must value a promise.

Layer one: the MA on ordinary credit. Rotate the gilts into public investment-grade bonds at, say, 110 basis points of spread, deduct a fundamental spread of around 50, and the MA of ~60 basis points lifts the discount rate across a twelve-year duration: liabilities fall by roughly 7 per cent. About £70 million of surplus, created on day one, by reclassification.

Layer two: the private assets. Shift a third of the portfolio into illiquids at 200–250 basis points — spreads the insurer's own models declare, on assets its own desks originate — and the blended portfolio spread rises to perhaps 160 basis points while the fundamental spread, anchored to rating buckets and its long-term floor, barely moves. The MA roughly doubles. Liability reduction: ~12 per cent, or another £50–60 million attributable purely to the illiquid allocation.

Layer three: the reform dividend. The 65 per cent risk margin cut releases perhaps another £20 million on a book of this size.

Layer four: funded reinsurance. Cede a fifth of the book to Bermuda and the capital that would have backed it — held against assets hotter than anything the PRA would bless onshore — is partly returned through the reinsurer's pricing. Call it £10–15 million, and rising with the cession rate.

Stack the layers: in the region of £150 million of day-one value per £1 billion of pensions, conjured by valuation and venue rather than by any cashflow yet received — against typical disclosed new-business margins in this market of well under half that. Run the stack backwards and the conclusion is stark: priced without the machine, every buyout in Britain would be a heavily loss-making transaction. The mechanisms are not an enhancement to the business model. They are the business model — which is why the industry fought for the fundamental spread as if its existence depended on it. It did.

The aggregates confirm the scale. The capital base of the entire bulk annuity sector — the own funds standing behind every transferred pension in the country — is in the region of £80 billion; the day-one value manufactured by the stack above, compounded across a £40–50 billion-a-year market, is a material fraction of it. Funded reinsurance exposure alone has reached some £40 billion and is growing quickly — on the regulator's own projection it could climb toward £110 billion over the coming decade — quickly enough that in the spring of 2026 the PRA moved to tighten its capital treatment. Its 2025 stress test put numbers on the danger: the recapture of a single largest funded-re counterparty, some £12 billion of liabilities, would cut solvency coverage by around ten percentage points and drain roughly £3 billion of surplus — and that on today's relatively small exposures. The scenario, of course, was designed by the same authorities whose fundamental spread is the quantity in dispute.

The capital that insures itself

Pause on where that £150 million goes first, because it reveals the architecture's strangest property. Writing the deal creates a capital requirement — the SCR, a 1-in-200 stress dominated by credit risk on the very assets just bought — of perhaps £90 million. The own funds that cover it are the £150 million the matching adjustment just created. No shareholder injected anything; no coupon has been received. The buffer held against the assets failing is manufactured from the assumption that the assets will not fail. Without the MA, the same deal would show a £40 million-plus capital shortfall, and the shareholder would have to fund it with real money — which is what writing annuities used to require. With it, every deal arrives carrying its own capital, which is how a sector holding roughly £80 billion absorbs £50 billion of new promises a year without ever pausing to raise any.

The circle then closes a second time, inside the capital calculation itself. The SCR's credit stress asks what happens if spreads blow out — and answers, under the rules, that the MA rises as spreads widen, lifting the discount rate and absorbing most of the hit. The same assumption is thus applied twice: once to create the capital, once to shrink the requirement it must cover. And in the scenario that actually matters — defaults, not spread noise — the logic inverts fatally: the asset loses its value at the same moment the capitalised future spread on it, sitting in own funds, is revealed never to have been coming. The capital evaporates precisely when the risk it guards against materialises. A bank's capital is paid-in money and realised earnings — things that exist whether or not the loan book performs. An annuity writer's MA-derived capital is a bet on the book performing, relabelled as the protection against it not performing. A landlord who counts thirty years of future rent as equity, then pledges that equity as his reserve against the tenants defaulting, has the same balance sheet.

The defence, once more, is the partition: the FS keeps the default budget inside the liabilities, so what reaches own funds is — if the line is drawn correctly — only spread that never depended on credit performance. Every road in this system leads to the same valve.

Where the windfall goes

Surplus conjured on day one does not sit idle. It flows to three destinations, and each completes a different part of the circuit.

Some of it is competed away into the price. The buyout market is brutal enough — Woods's point again — that the MA benefit is largely bid through to the customer in cheaper premiums, which is why schemes have at times bought out for close to, or even less than, the value of their gilt portfolios. A corporate sponsor exiting its pension obligations is thus part-funded by prudence that, if the fundamental spread is miscalibrated, was never really there. The sponsor banks the saving and departs; the discount it enjoyed becomes risk the pensioner unknowingly retains.

Some of it funds the next deal. The day-one surplus on transaction N becomes the pricing capacity for transaction N+1 — which is how a sector with roughly £80 billion of capital absorbs £40–50 billion of new liabilities every year without pausing for breath. New business, on this arithmetic, can be close to self-financing the moment it completes. Remove the machine and the market would not merely reprice; it would largely stop.

And some of it leaves the building. Solvency surplus — inflated by spread that will not actually be earned for decades — is the wellspring of "capital generation," and capital generation is what the big annuity writers pay out. The largest of them have built their entire equity story on the dividend: a billion pounds and more a year, drawn substantially from annuity surplus, distributed in cash to shareholders whose executives are themselves remunerated on the very capital-generation metrics the matching adjustment manufactures. The people who set the spreads on self-originated assets are paid on the surplus those spreads create; the loop has no exterior.

Sit with the asymmetry, because it is the moral centre of the whole arrangement. The matching adjustment recognises fifty years of future credit spread today, and a portion of that recognition is paid out, this year, in cash. If the fundamental spread proves too thin a decade hence, the error will be discovered by whoever is still standing — remaining policyholders, the compensation scheme, ultimately the state. The dividends of the good years cannot be recalled. Gains are distributed in real time; losses are discovered in arrears and socialised. A machine with that payout profile does not need malice to be dangerous. It only needs time.

The warning, and the decision

To understand why the Treasury chose as it did, remember what it needed. Post-Brexit Britain had promised itself a dividend: levelling up, green energy, new housing, rebuilt infrastructure — an investment programme of a scale the public finances, strained by the pandemic and then by the gilt crisis, could not begin to fund. The government needed someone else's balance sheet, and the only pools of long-term sterling capital large enough were the very pension promises migrating onto insurers' books. The insurance industry grasped the position perfectly and made the offer explicit: loosen the capital rules, and we will be your infrastructure investors — the Association of British Insurers attaching a number to it, over £100 billion of productive investment in the decade ahead. The reform of Solvency II thus stopped being a technical exercise in prudential calibration and became something far harder for a supervisor to win: the funding mechanism for a government's economic story, dressed as the repatriation of rules from Brussels.

The Prudential Regulation Authority understood all of this and said so — and it is essential to be exact about what it said. The PRA's quarrel was never with the matching adjustment itself, which it regarded as economically sound for genuine hold-to-maturity investors; its quarrel was with the deduction. Through 2021 and 2022, as the review gathered pace, the PRA supported most of the reform package — including a deep cut to the risk margin, a buffer its own chief executive had long criticised — but argued that the fundamental spread had to be repaired in exchange: made sensitive to market conditions, granular by asset, and explicit about the credit risk premium it currently ignores. Sam Woods, the deputy governor running the PRA, identified the engine driving the problem in terms no campaigner could improve on: the buyout market is so competitive that even well-intentioned firms are compelled to chase the most MA-productive assets available or lose the business to rivals who will.

The PRA did not stop at diagnosis; it designed the cure, in full. The existing fundamental spread is built from two parts — expected losses (historic default rates by rating) and the cost of downgrades — anchored to a thirty-year average and floored:

FS (current) = expected loss + cost of downgrade

The PRA's proposed replacement had three:

FS (proposed) = expected loss + credit risk premium + valuation uncertainty

Each new component answered a specific failure. The credit risk premium — calibrated, in the PRA's quantitative study, at no less than around 35 per cent of the credit spread, and linked partly to current market spreads rather than a multi-decade average — was compensation for the uncertainty of losses, the fact that defaults arrive in correlated waves rather than at their tidy historical mean. The valuation uncertainty allowance was aimed squarely at assets with no market price — the equity release notes and regeneration debt the insurers value with their own models — charging explicitly for the fact that nobody truly knows what such an asset is worth until the day everyone needs to. And greater granularity addressed the absurdity that two assets of the same rating but very different spreads receive the same deduction, which is precisely what rewards the hunt for the highest yield in each rating bucket.

It is worth grounding this in a name, because the abstraction hides the stakes. Consider the kind of risk that Thames Water came to represent: a regulated utility, investment-grade rated, its debt long treated as a safe "gilts-plus-a-margin" holding precisely the sort of asset an annuity book loads up on — until the market's faith in the regulatory settlement cracked and the debt repriced sharply downward. Which component would have helped? Not expected loss: that is anchored to the credit rating, and the rating saw nothing wrong until late, because it was grading faith in a regulator rather than a balance sheet. The credit risk premium would have: it withholds a third of the spread up front precisely against the possibility of exactly this kind of broad, correlated repricing of a whole sector once treated as quasi-safe — and as utility spreads widened, a market-linked premium would have forced recognition during the deterioration rather than after it. And for the many Thames-shaped assets that annuity books hold in private, unlisted form — water and energy-network debt, social-infrastructure loans with the same regulatory-promise credit story and no market price — the valuation uncertainty allowance was the charge for not being able to see the loss coming. The reform, in short, was built to price and to surface precisely the risk that the visible utility crises of the 2020s exemplified. What it could never do is see the danger earlier than the rating agencies, since all three components still lean on the ratings — which is itself the deepest limitation, and a separate essay.

The industry resisted further repair, making one respectable technical argument: a market-sensitive FS would make hold-to-maturity balance sheets swing with credit spreads, importing the very procyclicality the MA exists to remove. In November 2022, the Treasury chose. The risk margin was cut by 65 per cent. MA eligibility was widened to assets with merely "highly predictable" cashflows — language built for exactly the infrastructure and housing debt the government wanted funded. And the fundamental spread — the valve — was retained on its existing methodology, essentially untouched:

FS (enacted) = expected loss + cost of downgrade

No credit risk premium, no valuation uncertainty allowance, no market sensitivity. Of the three-part deduction the supervisor had designed, calibrated and tested, precisely none of the new components survived into law. The complete, ready-to-implement answer to "how much of this spread is really risk" sits to this day in the Bank's archive, while the regime runs on the two-part formula its own designer called inadequate.

The Bank of England's response sits in the parliamentary record: implementing the government's package, it estimated, would raise the annual probability of a life insurer failure from 0.5 to 0.6 per cent — a relative increase of around 20 per cent. The Treasury never rebutted the estimate. It noted a lack of consensus and proceeded. The supervisor's objection was not answered; it was overridden — a thing elected governments are entitled to do, though this one never stood up and said so in those terms.

Parliament then completed the arrangement. The Financial Services and Markets Act 2023 handed the PRA a new statutory secondary objective: to facilitate the international competitiveness and growth of the UK economy. The supervisor that had just been overruled for weighing safety too heavily was formally instructed, by law, to weigh growth in all its future judgements. The defeat was not merely administered; it was codified.

The empty chair

Which brings us, finally, to the profession.

The actuaries are the people inside this machine. They set the spreads on the self-originated assets, calibrate the models that value them, compute the matching adjustment, certify the reserves, and sign the regulatory returns. No group in the country was better placed to tell the public whether the supervisor's warning was sound. And their chartered body — the Institute and Faculty of Actuaries, bound by royal charter to the public interest — did respond, on the very day of the Treasury's decision. It welcomed the package, supported the wider MA eligibility, and on the fundamental spread declared itself encouraged, explaining that it shared the Treasury's concerns about the volatility the PRA's remedy would have introduced.

One statement, and it tells you everything about where the institutional weight settled. The argument the body chose to amplify was the industry's; the analysis it declined to engage was the supervisor's; and the conflict it declined to mention was its own — a membership employed, almost to a person, by the firms whose capital position the fundamental spread determines. No conspiracy is needed to explain this, only Upton Sinclair's old observation about the difficulty of understanding something one's salary depends on not understanding, operating at the scale of an institution.

The few who said otherwise prove the rule by their reception. The academics who argued for years that the matching adjustment itself manufactures capital from unearned spread were received as cranks rather than engaged as colleagues — and it is worth being precise that the PRA never went that far: the supervisor's case was the narrower and more moderate one, that the mechanism is sound but its deduction is too small. The profession declined to engage even with that. A former chief economist of the Bank of England asked publicly, more than once, where the actuarial profession stood on all this; the question was not answered so much as outlasted. Within the profession, dissent has channels, and the channels function as digestion: working parties in, sessional papers out, position unchanged.

Nor is the failure mode novel. After Equitable Life collapsed under guarantees its actuaries had valued and blessed, the Penrose Report criticised the profession's role and the Morris Review of 2005 delivered the verdict: insular, under-challenged, too close to the firms it served — findings damning enough that actuarial standard-setting was taken away and handed to external oversight. Seventeen years later, offered the cleanest possible opportunity to show that something had changed, the profession's body reproduced the pattern at twenty times the scale.

There is even a mechanism now ensuring the silence persists. Among the consolations the PRA salvaged from its defeat is an attestation regime: a named senior individual at each insurer must certify annually, on personal liability, that the fundamental spread suffices for the assets actually held. Whatever its prudential merits, notice its side effect: every actuary's private doubt about the number is now captured in a confidential signature to the regulator. The collective professional voice that might one day say, in public, we who do this work believe the valve is set wrong has been disassembled into sealed compliance artefacts, one signatory at a time. A government confident in its calibration would not need the signatures. The hedge is the confession.

And consider what the signature means for the assets where it matters most. For a tranche of debt on a project the insurer's own group sponsors, the chain runs entirely within one institution: the group's desk set the spread, its credit function assigned the rating, its models produced the valuation, and the resulting matching adjustment became its capital, its dividend and its bonus pool. The attestation — the regime's only asset-specific check on that chain — is then performed by a senior employee of the same group, certifying that the spread his colleagues priced contains no risk beyond what the deduction covers. The personal liability is real; the Senior Managers regime makes a careless signature career-ending, and the first attestation rounds did extract voluntary spread additions that the old regime would never have produced. But the structure should be named for what it is: the audit problem with the auditor in-house; the rating-agency problem with issuer and agency under one roof. The PRA, having been refused the formula that would have given an outside answer, publicly ruled out delivering it through the attestation either. So for the assets the supervisor worried about most, the entire distance between the current regime and pure self-certification is one individual's signature and the regulator's appetite to challenge it. The loop has no exterior — including, now, its own verification. The regime does not check whether the number is right; it checks whether the right person signed it — and that person works for the company whose profit depends on the number being flattering.

We know how this ends, because we have run the experiment before — three times, and written a report each time. Auditors were once entangled with the firms they audited, until the conflict produced enough wreckage that independence was made law. Credit-rating agencies stamped structured mortgage debt AAA while the issuers paid their fees, and the catastrophe that followed in 2008 is the textbook case of an interested party's risk estimate believed until it wasn't — and note that the agency, for all its conflict, was at least a separate company, a weaker entanglement than an insurer rating debt it originated itself. Banks used their own internal models to decide how much capital their own assets required, conjured the requirement downward, and were eventually disciplined by regulators imposing floors on what the models could claim. Each reform arrived the same way: after the failure, never before, written in the language of "this must never happen again." What is singular about the matching adjustment is the timing. This is the first occasion on which we have watched the structure assemble in advance, had the supervisor identify the flaw, hand the government the precise instrument to correct it — a formula-level charge owned by the regulator rather than the firm — and put the instrument back in the drawer. The scandal, if the word applies, is not that the homework is marked in-house. It is that we have read every previous report on the dangers of marking your own homework, and chose, this time, with the remedy already drafted, to do it anyway.

If the fundamental spread proves adequate, none of this will be remembered — and fairness requires saying that it might. The sector's record is unblemished: no UK annuity writer failed in 2008, in 2020, or in the gilt crisis of 2022, when it was the pension schemes, not the insurers, that needed rescuing from their own collateral calls. The hold-to-maturity logic is genuinely strong, and the counterfactual was no idyll — leaving the promises scattered across thousands of underfunded schemes, backed by the fifty-year covenant of whatever happens to a high-street retailer, was its own slow-motion failure. The consolidation brought capital, expertise and supervision to liabilities that previously had none of the three.

But note what that defence cannot do. The unblemished record has been earned entirely in an era of falling rates and central banks standing behind credit markets; the regime has never met its test. And the architecture is built so that no test arrives early. Pensioners are protected by the Financial Services Compensation Scheme at 100 per cent, without cap — so the people bearing the tail risk have neither the means nor the motive to police it, and the ordinary discipline of nervous customers is switched off by design. There is no market price for a third of the assets, no surrender pressure, no maturity wall. If the valve is set wrong, the error does not surface next quarter; it compounds silently, capitalised into day-one profits and recycled into the pricing of the next deal, for a decade or more — until a genuine credit cycle marks the self-originated books to reality. Privatised spread in the good years; a socialised tail, via the FSCS's levies and ultimately the state, in the bad one. Errors in this system are invisible until they are enormous, and that is a property of the architecture, not of anyone's behaviour.

If that day comes, the inquiry will ask its ritual question: where were the experts? The answer is already filed, dated 17 November 2022. They were present. They had read the supervisor's warning, and they understood the machine better than anyone alive.

They welcomed it.

C.J. Marsden writes on political economy at ir35andmore.com. A companion essay, "The Overruled Objection," examines the Treasury's decision itself.

Sources for verification: HM Treasury, Review of Solvency II: Consultation — Response, 17 November 2022 (risk margin cut 65%; FS methodology retained); Bank of England letter to the Treasury Committee (failure probability 0.5% → 0.6%, ~20% relative increase); Sam Woods, Solvency II: Striking the balance, July 2022; PRA DP2/22 (credit risk premium; incentives toward high-spread assets); PRA SS5/24 (funded reinsurance); PRA PS10/24 (MA attestation); FSMA 2023 (secondary competitiveness and growth objective); ABI statement, 17 November 2022 (£100bn pledge); IFoA media statement, 17 November 2022; Penrose Report (2004); Morris Review, Final Report (2005); PRA, Life Insurance Stress Test 2025 (single largest funded-re counterparty recapture ≈ £12bn liabilities, ~10pp coverage reduction, ~£3bn surplus); PRA CP8/26, April 2026 (funded reinsurance exposure ~£40bn, projected toward £110bn; CDA tightening, ~2–4% to ~10%); aggregate BPA own funds ~£80bn per PRA estimates.