--- lang: en title: >- From Low Interest Rates to Interest Rate Repricing: Liquidity Risk, Leverage Constraints, and the Restructuring of Asset-Liability Management in Mature Pension Markets description: >- Taking the 2022 UK liability-driven investment (LDI) liquidity event as its empirical entry point, this paper systematically examines the liquidity ri --- ## Chapter 1 Introduction ### 1.1 Statement of the Problem Since the end of the twentieth century, major advanced economies have been situated for a prolonged period within a macroeconomic environment of declining, and even ultra-low, interest rates. Against this backdrop, mature pension markets — particularly the United Kingdom, the Netherlands, and other markets dominated by defined benefit (hereafter DB) plans — have, in response to the pressure of expanding liability present values as discount rates fell, generally embarked on a path of asset-liability management evolution centered on duration matching, leveraged hedging, and alternative asset allocation. Liability-driven investment (LDI) strategies became mainstream during this period: through interest rate swaps, repurchase agreements, and synthetic duration positions, plans obtained the interest-rate sensitivity required to cover long-duration liabilities using limited notional capital, thereby maintaining book-value solvency balance in a low-rate environment. Existing research has generally narrated this evolution, within a one-directional low-rate environment, as an "advanced practice." The research sequence in which this paper is situated has itself already characterized, in preceding installments, the transmission mechanism from the interest rate environment to asset allocation, the institutional taxonomy of mature markets and the mapping of risk-bearing, as well as the toolbox and governance conditions for pensions' increased allocation to risk assets, overseas assets, and alternative assets during the low-rate period. Relevant analysis by the International Monetary Fund likewise notes that the pension sector has undergone structural change accelerated by prolonged low interest rates (IMF 2025). However, when the macro cycle reversed between 2021 and 2022 and interest rates entered a phase of rapid repricing, the structural exposures accumulated along the aforementioned evolutionary path manifested in concentrated form in an unexpected manner. Between September and October 2022, the UK long-dated government bond (gilt) market experienced a sharp spike in yields amid a budget policy shock; pension plans employing leveraged LDI strategies faced large-scale cash variation margin calls owing to unrealized losses on derivative positions, and were forced to sell gilts to raise cash, which in turn drove gilt prices further down and intensified margin calls further, forming a textbook collateral spiral. The Bank of England (hereafter BoE) launched a temporary targeted gilt-purchase operation on September 28, 2022 to restore market functioning, thereby halting this self-reinforcing downward cycle (BoE 2022). This event delivered a dual shock to the prevailing understanding of pension risk. The first shock was intuitive: it appeared to confirm the simplistic narrative that "rising interest rates harm pensions." The second shock is subtler, and is the one this paper is concerned with: careful fact-checking shows that the funding coverage ratio of UK DB plans in 2022 not only did not deteriorate but improved markedly — the Pension Protection Fund's (hereafter PPF) 7800 Index aggregate surplus rose to a historic peak of £313.8 billion in August 2022, with the funding ratio rising from approximately 105% in August 2021 to approximately 125% in August 2022, and further to approximately 135% by the end of September; by year-end, roughly eighty percent of plans were in surplus on both technical provisions and buyout bases (PPF 2022). In other words, what erupted in the UK pension sector in 2022 was a liquidity crisis, not a solvency crisis. This contrast reveals a fundamental flaw in the prevailing analytical framework: it failed to conceptually distinguish between the solvency path (funding coverage ratio) and the liquidity path (collateral and cash availability) — two essentially distinct lines. This conceptual flaw merits scholarly attention precisely because it is not the incidental oversight of any single commentator, but rather a systemic blind spot of an entire mainstream analytical framework. In a one-directional low-rate environment, solvency and liquidity typically move in the same direction: when rates fall, liabilities expand, assets appreciate, and collateral is abundant, so solvency and liquidity loosen simultaneously; when rates remain low for an extended period, the two likewise tighten together, without abruptly breaking apart. It is precisely this long-run co-movement under a one-directional environment that has habituated analysts to summarizing a plan's overall health along a single dimension — often the funding coverage ratio of solvency. When the macro cycle reverses and interest rates enter a phase of rapid repricing, this co-movement is broken, and the separation between solvency and liquidity is exposed for the first time in extreme form. The analytical value of the 2022 UK episode therefore lies not in its being an isolated extreme case, but in its being a diagnostic window that reveals structural relationships — a rare institutional event through which an entire set of structural relationships, hitherto obscured under a one-directional environment, becomes visible. It should be emphasized that the term "window" here refers only to its diagnostic value in revealing pre-existing structure, not to a natural experiment in the econometric sense from which a causal effect can be estimated; as this paper's methodology (Section 1.3) repeatedly states, this event is, in terms of sample size, essentially equal to one and cannot self-calibrate, and this paper therefore does not, from the outset, read identified causation from this single event. The aim of this paper is precisely to distill the structural relationships revealed by this window into a set of generalizable, falsifiable general judgments, rather than to remain at the level of a review of a specific event. ### 1.2 The Paper's Core Thesis and Its Corrective Positioning This paper assumes a corrective role within the overall research sequence. The preceding three installments of this sequence characterized how pensions increased their allocation to risk assets, overseas assets, and alternative assets during the low-rate period, with a narrative tone inclined toward interpreting alternative and illiquid asset allocation as a more advanced practice. This paper's task is to answer what risks these allocations expose when interest rates rise rapidly and the low-rate environment ends, and to correct the aforementioned one-directional tendency toward an "alternatives-are-advanced" thesis. The substance of this correction does not lie in asserting, in reverse, that "the allocation evolution of the low-rate period was mistaken," but in revealing a more refined thesis: what truly determines pension safety is not the allocation share of risk assets or alternative assets per se, but whether the four factors of return target, liability duration, liquidity needs, and governance capacity are matched with one another. Under this positioning, the paper's core thesis can be stated as follows: interpreting the 2022 UK episode as "rising interest rates harm pensions" is mistaken, but interpreting it as "improved coverage therefore the plans are safe" is a mistake of the same order. The former conflates the dual-sided effect of rising interest rates on liabilities and assets; the latter conflates the solvency dimension with the liquidity dimension. The common root of both is a failure to recognize that a rise in interest rates is a dual-sided event whose net-effect sign is a priori undetermined, and that the funding line and the liquidity line may separate and move in opposite directions during the decumulation (payout) phase. It is worth emphasizing that this paper also maintains a cautious stance toward the existing corrective narrative itself. One popular corrective line of reasoning runs: since what erupted in 2022 was a liquidity crisis rather than a solvency crisis, the solution is to thicken liquidity buffers, constrain leverage, and replicate the "best practices" of "unaffected countries" such as the Netherlands or Denmark. This paper will argue that this line of reasoning, in its final step, falls back into the shallow framing it sought to correct — it treats "improving resilience" as a purely technical problem, while overlooking that improvement itself is an allocation-of-responsibility choice: the setting of any buffer, backstop, or leverage constraint redistributes, at the institutional level, the boundary of "who bears which risk." The deepest-level judgment therefore does not lie in how to thicken buffers, but in recognizing that risk positioning is an institutionally endogenous allocation-of-responsibility choice, thereby understanding why best practices cannot simply be transplanted. ### 1.3 This Paper's Methodological Stance This paper's methodological stance inherits a basic conviction: the core of an economics paper is judgment, not the mechanical stacking of data. That judgment takes priority over data does not mean slighting identification — quite the opposite: a valuable judgment must be falsifiable, non-obvious, and must clearly mark the boundaries of its identification. Accordingly, this paper adheres to the following three principles in handling empirical evidence. First, the principle of slow modeling: identification strategies and pre-registered expectations are established first through adequate theoretical design, and only then is the evidence examined for consistency with those expectations, rather than conducting undisciplined data exploration followed by post hoc rationalization. The identification design and pre-registered expectations of each chapter of this paper are fixed in writing prior to examining any specific evidence, and the non-expected outcomes — that is, which outcomes, were the evidence to point to them, would falsify the judgment and require an honest downgrade — are likewise specified in advance. Second, the principle of identification honesty: this paper strictly distinguishes two categories of evidence. The first category comprises aggregate institutional facts that can be verified through independent sources, including the direction of the funding coverage ratio, the scale of intervention operations, the market share of vehicle structures, the thresholds of regulatory buffers, the values of eligibility thresholds, and the like; such facts are accepted, with in-text parenthetical citations pointing to their sources. The second category comprises causal magnitudes that can only be identified using plan-level micro panel data, including the slope ratio between the two paths, difference-in-differences coefficients for vehicle comparisons, regression discontinuity coefficients at eligibility thresholds, panel coefficients across bailout scopes of application, the pass-through share of contagion networks, and the like; wherever such causal magnitudes cannot be identified in this study owing to data availability constraints, they are uniformly and explicitly designated as "identification design in place, with the causal magnitude awaiting micro data" — no regression coefficient, significance level, or sample size is reported, nor is consistency in aggregate direction ever misconstrued as identified micro-level causation. Third, the principle of load-bearing stratification: the judgments in each chapter of this paper strictly distinguish an identifiable load-bearing layer from a downgraded theoretical or discursive layer. For any proposition that is identifiable, its identification strategy and source of exogenous variation are made explicit; for any proposition that is not identified, it is explicitly labeled as a theoretical-dynamics proposition, an untested empirical hypothesis, or a discursive inference, with its sole exit route from theory to empirics specified. This stratification runs throughout the paper and is the technical implementation of the paper's honesty boundary. It should be further clarified that this paper's repeated practice of actively designating propositions as theoretical or hypothetical does not stem from any deficiency in analytical capacity, but from strict adherence to identification logic. A recurring trap in empirical economic research is the over-interpretation of observed correlation, co-occurrence, or directional consistency as already-identified causation. This trap is especially dangerous in the study of a single event: when the object of study is a unique historical event (such as the 2022 UK LDI crisis), the sample size is, in essence, equal to one, and any inference of the form "in this event, A and B occurred together, therefore A caused B" cannot rule out the possibility that a third factor drove both A and B simultaneously, or that A and B are mutually causal, or that their co-occurrence was purely incidental to this particular circumstance. This paper maintains a high degree of vigilance on this point: for any causal assertion that relies on a single event, this paper requires that it satisfy a self-calibrating condition — namely, that there exists external variation independent of the assertion itself that can serve as a yardstick to test it; if no such yardstick exists, the assertion can only be presented as a theoretical proposition or hypothesis, and must not be passed off as identified causation. This self-calibration requirement is the fundamental criterion by which this paper distinguishes a "defensible judgment" from an "honest downgrade," and is the most concentrated technical expression of this paper's methodological stance. ### 1.4 Structure of the Paper The remainder of this paper is organized as follows. Chapter 2 reviews the literature and characterizes the institutional background, including the mechanism sequence of the 2022 UK LDI liquidity event, as well as the institutional differences in decumulation-phase risk governance across the United Kingdom, the Netherlands, Denmark, and various Continental European jurisdictions. Chapter 3 establishes the theoretical framework, proposing the meta-judgment that "risk positioning is an institutionally endogenous allocation-of-responsibility choice," and formalizing it as a comparative proposition mapping pre-bindable institutional variables onto dimensions of responsibility allocation. Chapters 4 through 8 each develop one of five progressive judgments: Chapter 4 argues that improvement in the funding line does not constitute a safety signal for the liquidity line; Chapter 5 argues that the fact that the Netherlands and Denmark did not experience a crisis cannot be taken to imply institutional superiority, since survivorship bias is a double-edged sword; Chapter 6 argues that the locus of the spiral pathology lies in the mismatch between the moneyness of variation margin and rescuability; Chapter 7 argues that the independent variable of stress testing must be reset from a static base-case endpoint to a path, density, and correlation structure; Chapter 8 argues for the institutional endogeneity of risk positioning. Chapter 9 undertakes a cross-chapter synthesis, clarifying the progressive logic and internal consistency of the five judgments. Chapter 10 is the conclusion, offering policy implications for pension risk governance in China and the limitations of this paper. --- ## Chapter 2 Literature and Institutional Background Review: The 2022 LDI Liquidity Event and Cross-Jurisdictional Decumulation-Phase Governance This chapter proceeds along two lines: first, it traces the strand of literature directly relevant to the judgments advanced in this paper and locates the paper's incremental contribution relative to the existing conventional narrative; second, it characterizes the mechanism sequence of the 2022 UK LDI event and the institutional differences in decumulation (payout) phase governance across jurisdictions, providing the empirical background for the judgments developed in subsequent chapters. ### 2.1 Literature Landscape and This Paper's Positioning The existing literature on risk in the decumulation phase of mature pension markets can broadly be grouped into three strands. The first consists of official and regulatory post-event reviews, exemplified by the Bank of England's record and policy assessment of its 2022 gilt-market intervention, which confirmed the liquidity nature of the crisis, the mechanics of the collateral spiral, and the backstop positioning of the temporary bond-purchase operation (BoE 2022; Breeden 2022). The second strand consists of cross-jurisdictional institutional comparisons; reports from the Dutch central bank and the Danish financial supervisory authority respectively characterize the structural differences in eurozone LDI arrangements and the residual risks that persisted among survivors (DNB 2024; IPE 2023). The third strand consists of macro-level structural narratives: the International Monetary Fund has noted that structural change in the pension sector was accelerated by a prolonged period of low interest rates, and that while the rapid transition to higher rates may be beneficial in the long run, its speed and abruptness triggered liquidity pressure and contagion risk in certain jurisdictions (IMF 2025); long-run observations at the level of allocation accounting appear in global pension asset research (WTW 2026). The conventional narrative that this paper corrects is precisely a default framework formed by the above literature within a one-directional low-interest-rate environment—one that summarizes plan health along a single dimension (often the solvency funding ratio) and tends to understand improvements in decumulation-phase resilience as technical optimizations transportable across jurisdictions. This representative framework of the conventional narrative constitutes the seed against which the judgments of each chapter of this paper are positioned and corrected. This paper's incremental contribution lies not in denying the facts recorded by this literature, but in reorganizing the hierarchy of judgments that these facts support: elevating solvency and liquidity from "two parallel risks" to "at a given point in time, improvement in solvency carries no informational value for liquidity, or even carries the opposite signal"; reconstructing cross-jurisdictional outcomes from "a ranking of superior versus inferior" into "a survivor sample that cannot be used for ranking, plus identifiable structural differences"; and reconstructing resilience improvement from "technical optimization" into "an institutionally endogenous choice of where to place accountability." ### 2.2 The Institutional Logic of Liability-Driven Investment To understand the mechanics of the 2022 UK event, one must first clarify the institutional logic of the LDI strategy. The liabilities of a DB pension plan are, in essence, a series of long-duration defined-benefit commitments whose present value is highly sensitive to the discount rate. Under accounting and regulatory frameworks, liabilities are discounted at market interest rates, so when rates fall, the present value of liabilities expands sharply. To hedge this interest-rate risk, a plan needs to hold assets whose duration matches that of its liabilities. However, if a plan were to match liability duration purely by purchasing long-duration government bonds outright with cash, it would be forced to allocate the vast majority of its assets to low-yielding fixed-income instruments, thereby forgoing the asset returns needed to meet long-term benefit obligations. The LDI strategy arose precisely to resolve this dilemma. Through interest-rate swaps, gilt repo, and synthetic duration positions, a plan can obtain, with a limited amount of nominal capital, the interest-rate sensitivity needed to cover the duration of its entire liability book, while allocating the remainder of its capital to return-seeking assets. This arrangement functioned well in a one-directional low-interest-rate environment: as rates continued to fall, both the liability side and the hedging side expanded in tandem, the plan's solvency balance was maintained, and the leveraged interest-rate sensitivity did not create immediate cash pressure. It was precisely this strong performance in a low-rate environment that led LDI, during its rise, to be widely interpreted as a mature, advanced asset-liability management (ALM) technique. Viewed from the perspective of institutional evolution, the rise of LDI had an inherent logical inevitability. Under a regulatory framework that requires liabilities to be discounted at market interest rates and imposes accounting and solvency constraints on funding ratios, a plan faces a dual pressure: on one hand it must hedge the interest-rate sensitivity of its liabilities to stabilize its funding ratio, and on the other it must obtain sufficient asset returns to close its existing funding gap. Satisfying both objectives simultaneously without leverage would require holding long-duration assets far in excess of the plan's own capital base, which is impractical. Leveraged synthetic duration positions therefore become the only feasible solution—they substitute the notional exposure of derivatives for the holding of physical bonds, thereby achieving both duration matching and return generation with limited capital. This solution is logically self-consistent, but its cost is deferred to the scenario of rising rates: the interest-rate sensitivity magnified by leverage is a gain when rates fall, but is converted into mark-to-market losses and margin pressure when rates rise. In other words, LDI is not an erroneous technique, but rather a technique that is optimal under a specific assumption about the interest-rate environment while being highly dependent on that assumption; when the assumption (that rates remain low or move slowly over the long run) is broken, its embedded fragility surfaces in concentrated form. The concrete form of this fragility is as follows (its concentrated manifestation in the UK in 2022 is documented in BoE 2022). Leveraged synthetic duration positions embed a fragility that lies dormant for a long time in a low-rate environment and suddenly manifests when rates rise: derivative and repo positions must be backed by collateral, and when rate increases produce mark-to-market losses on these positions, the plan must post additional margin. Critically, variation margin for centrally cleared and over-the-counter derivatives typically must be posted in cash, rather than in securities held by the plan. This institutional constraint—that margin "must be posted in cash"—may appear to be a technical detail of clearing rules, but it is in fact the hinge for understanding why the 2022 spiral was self-reinforcing and could not self-correct. If margin could be posted in securities, a plan facing a margin call could simply pledge its gilt holdings directly, without needing to sell into the market to raise cash, and collateral pressure would not transmit into price pressure; but once margin must be posted in cash, the plan is forced to become a seller of gilts, and collateral pressure is thereby converted into price pressure, giving rise to a self-reinforcing downward spiral. Chapter 6 of this paper will distill this hinge into the theoretical proposition of a "topological mismatch between the cash-like nature of margin and central-bank rescuability," and will devote a dedicated analysis to its identification boundary. ### 2.3 The Mechanism Sequence of the 2022 Event The mechanism sequence of the 2022 UK event can be characterized as follows. At the macro level, inflationary pressure and the subsequent fiscal policy shock drove a rapid rise in UK gilt yields; the fiscal policy announcement of late September 2022 (involving approximately £45 billion in unfunded tax cuts) served as the direct trigger, with the 30-year gilt yield jumping by roughly 140 basis points in a single day after the announcement and rising by more than 100 basis points cumulatively over four days (SUERF 2023). At the micro level, the rapid rise in yields produced large mark-to-market losses on LDI plans' synthetic duration positions, triggering cash variation margin calls. Plans sold gilts to raise cash; the sales depressed gilt prices, and falling prices further amplified mark-to-market losses and margin calls, giving rise to a self-reinforcing collateral spiral. The destructiveness of this spiral stemmed not from the ultimate magnitude of the yield move itself, but in large part from the speed of the move: the speed and scale of the yield movement far exceeded the capacity of smaller plans using pooled vehicles to top up collateral—such plans typically had only a one-week, and in some cases a two-week, rebalancing window (Breeden 2022). The compression of speed and timing turned what should have been orderly collateral management into forced selling. At the systemic level, because a large number of LDI plans employed similar leveraged hedging structures, their selling behavior was highly correlated, constituting a crowded trade whose price impact far exceeded what the scale of any individual plan could explain. This systemic-level dynamic is the key to understanding how the 2022 event amplified from liquidity pressure at individual plans into a systemic market failure: an individual plan's sales would, in a normal market, be absorbed by other participants' buying, but when all plans employing similar structures were simultaneously forced to sell by the same interest-rate shock, the market lacked sufficient marginal buyers to absorb the selling pressure, and prices overshot; this overshoot in turn worsened the mark-to-market losses and margin calls of every plan, forming a positive feedback loop across plans. The strength of this cross-plan positive feedback depends on the concentration of plans using leveraged LDI within the gilt market and the homogeneity of these plans' collateral structures—the higher the homogeneity and the greater the concentration, the stronger the positive feedback. It was precisely this systemic-level amplification mechanism that turned what should have been a collateral adjustment self-digested by the market into a spiral that could only be halted by central bank intervention. The BoE launched a temporary, targeted gilt-purchase operation on 28 September 2022, absorbing gilt sales at a daily cap of £5 billion over thirteen working days, with actual purchases of approximately £19.3 billion (comprising £12.1 billion of conventional gilts and £7.2 billion of index-linked gilts), before terminating on 14 October (BoE 2022). This intervention was explicitly positioned as a time-limited backstop tool rather than permanent support, and its time-limited nature carries significant weight for the discussion of commitment credibility in Chapter 8 of this paper. It is worth noting that this systemic-level amplification mechanism itself also provides the intuitive basis for the theoretical network proposition in Chapter 8 of this paper, "the endogenization of central-bank backstops and the contagion topology of the next round"—if a central bank's backstop commitment leads plans to expect that they will be rescued in a crisis, plans may maintain higher leverage and more homogeneous structures, thereby strengthening the cross-plan positive feedback of the next round; but as discussed in Chapter 8, this proposition is classified as a theoretical network proposition, owing to the limited look-through ownership-share data required for identification, and this paper does not treat it as empirically load-bearing. ### 2.4 The Divergence of the Funding Line and the Liquidity Line The most instructive feature of the 2022 event lies in the fact that the funding line and the liquidity line moved in opposite directions during the same period. As noted above, the funding ratio of UK DB plans improved markedly in 2022, its direct driver being precisely that the rise in gilt yields depressed the present value of liabilities discounted at market rates—liabilities fell by more than one-tenth in a single month (PPF 2022). At the same time, it was this very same rise in yields that, through mark-to-market losses on leveraged hedging positions, ignited the liquidity crisis. The improvement in solvency and the exhaustion of liquidity were two opposite-direction consequences of the same interest-rate shock. This divergence phenomenon constitutes the empirical starting point for the entire analysis of this paper. It demonstrates that any framework that conflates solvency and liquidity will produce systematic misjudgments in the decumulation phase: looking only at the funding line, one would mistakenly believe the plan to be safe; looking only at the liquidity line and attributing it to the rise in interest rates per se, one would mistakenly believe that the rise in rates had damaged the pension's solvency. A correct diagnosis requires strictly distinguishing these two lines at both the conceptual and modeling levels, and this is precisely the subject of Chapter 4 of this paper. ### 2.5 Institutional Differences in Decumulation-Phase Governance Across Jurisdictions Part of why the 2022 UK event can offer valuable judgment is that a cross-jurisdictional counterfactual exists. The Netherlands and Denmark, which likewise made extensive use of LDI and interest-rate hedging, did not experience a UK-style collateral spiral in 2022. The existence of this counterfactual makes it possible to "identify the differentiating variables," and also renders "discussing the UK alone" an analytical limitation for lack of a counterfactual. Dutch pension funds experienced large-scale asset sales and margin injections in the first half of 2022: they sold approximately €88 billion in assets (about 4.6% of total pension assets, the largest such volume for any single half-year on record) and injected approximately €82 billion into margin accounts, a considerable portion of which was used to top up margin on interest-rate derivatives; yet, similar to the UK, the funding ratio of Dutch pension funds also improved over the same period (averaging approximately 119% in the first quarter and approximately 122% in the second quarter) (IPE 2022). The reason the Netherlands did not evolve into a UK-style spiral lies in structural differences: euro-denominated LDI rarely or never uses repo to add leverage, part of the margin can be posted using securities rather than cash alone, and it relies on a well-functioning money market (DNB 2024). Danish pension funds also made extensive use of interest-rate derivatives for hedging and operated in a larger, more liquid euro market; but the Danish financial supervisory authority noted after the fact that some pension companies still carried "non-negligible liquidity risk" and needed to forecast margin requirements on an hourly basis—that is, surviving does not equate to being risk-free (IPE 2023). At the level of regulatory governance, jurisdictions have also displayed an institutional divergence in their responses. The UK's Pensions Regulator (hereafter, TPR) issued guidance in April 2023 requiring leveraged LDI to hold, in addition to an operational buffer, a minimum 250-basis-point market-stress resilience buffer, on the assumption that a plan can top up collateral within five working days (TPR 2023). The Irish and Luxembourg regulatory authorities, by contrast, imposed a requirement of resilience to at least a 300-basis-point UK yield shock on sterling-denominated LDI funds, an approach recognized by the European Securities and Markets Authority (ESMA 2022). The BoE, moreover, opened applications in January 2025 for the "Contingent Non-Bank Financial Institution Repo Facility" (hereafter, CNRF), institutionalizing the backstop tool, one of whose eligibility thresholds is holding more than £2 billion in gilts (BoE 2025). These institutional differences are not merely background; they are direct empirical material for Chapter 8's argument that "the accountability boundary diverges with institutional variables." Particular attention should be paid to the differences in the logic of accountability underlying these regulatory responses. The UK's response package is "a minimum plan-level resilience buffer plus a central-bank contingent backstop facility"—on one hand, the 250-basis-point buffer requirement places the responsibility of the first line of defense on the plans themselves, while on the other hand, the CNRF endogenizes the backstop responsibility for tail risk into the central bank's balance sheet. This combination means that the UK has placed the accountability boundary for decumulation-phase risk at a specific position: "the plan's own liquidity resilience plus the central bank's lender-of-last-resort role." By contrast, the response of Ireland and Luxembourg is "a higher fund-level buffer (300 basis points) in place of a jurisdiction-level central-bank backstop"—because these jurisdictions are not the issuer of sterling, their central banks cannot provide lender-of-last-resort support to the sterling gilt market, so their accountability logic must necessarily rest more heavily on the fund's own resilience. This contrast is not accidental, but rather a direct consequence of differences in institutional variables (in particular, the boundary of central-bank authority and the power to issue currency): the issuer of sterling can place part of the accountability boundary on a central-bank backstop, whereas a non-issuer cannot. This observation constitutes the core empirical support for the comparative proposition in Chapter 8 of this paper, "the accountability boundary diverges with institutional variables," and it also foreshadows that chapter's thesis that "best practice cannot simply be transplanted"—Ireland and Luxembourg cannot replicate the UK's "buffer plus backstop" combination precisely because they lack the institutional variable of currency-issuing power that the UK possesses. ### 2.6 Chapter Summary This chapter has reviewed three strands of relevant literature and positioned this paper's incremental contribution, and has then characterized the mechanism sequence of the 2022 UK LDI liquidity event, revealing its most instructive feature—the opposite-direction divergence of the funding line and the liquidity line under the same interest-rate shock. This chapter has further presented the institutional differences in decumulation-phase governance across the Netherlands, Denmark, and other continental European jurisdictions, providing the empirical background for the judgments of subsequent chapters. It should be emphasized that the cross-jurisdictional differences presented in this chapter serve as background and empirical material, rather than as an already-identified causal contrast; elevating these differences to causal identification would require satisfying strict identification conditions, and this honest boundary will be developed in Chapter 5. The core conclusion regarding institutional background is this: the risk landscape of the decumulation phase cannot be characterized along a single dimension; the divergence of solvency and liquidity demands an analytical framework capable of distinguishing the two lines and of further probing their institutional origins. --- ## Chapter 3 Theoretical Framework: Risk Positioning as an Institutionally Endogenous Choice of Accountability ### 3.1 From Technical Correction to Institutional Judgment The theoretical framework of this paper is built upon a chain of progressively deeper questions. The shallowest question is: what kind of crisis erupted in the UK in 2022? The answer is a liquidity crisis rather than a solvency crisis (Chapter 2). The next-level question is: where does the mechanistic lesion of this liquidity crisis lie? The answer points to the collateral spiral welded together by leveraged synthetic duration positions under the constraint of cash variation margin (Chapter 6). A further question is: given that the lesion has been identified, does the path to improvement simply consist of thickening buffers, constraining leverage, and replicating the practices of countries that did not experience a blowup? The theoretical framework of this paper offers a negative, and more profound, answer precisely at this final level. The core of this answer is: improving resilience is not a purely technical problem, but a choice of where to place accountability. Any setting of liquidity buffers, central-bank backstops, or leverage constraints is not "eliminating" risk, but rather redistributing, at the institutional level, the boundary of "who bears what risk, under what circumstances." Thickening a pension's liquidity buffer means shifting a certain risk from the central bank's backstop responsibility back onto the plan itself; establishing a central bank's contingent backstop facility means shifting a certain tail risk from the plan to the public sector's balance sheet. Therefore, risk positioning—that is, where a country places the accountability boundary for decumulation-phase risk—is a choice endogenously determined by that country's institutional structure, rather than a technical parameter that can be optimized independently of institutional background. This "endogenous accountability" perspective stands in sharp contrast to the conventional risk-management analytical framework. The conventional framework tends to understand risk governance as an optimization problem: given the distribution of risk, choose the parameters of buffers, constraints, and backstops so as to minimize some loss function. Under this framework, comparability and transportability exist among the best practices of different countries—if one country observes that another country's buffer setting is superior, it can simply adopt it. However, the "endogenous accountability" perspective reveals an implicit premise of this framework: it assumes that the distribution of risk and the boundary of accountability are exogenously given, independent of institutional structure. This paper's argument precisely negates this premise. The ultimate absorber of risk, the rules for sharing losses, and the availability of backstops are not exogenously given, but are endogenously determined by institutional variables such as bankruptcy-law priority, the ownership structure of clearinghouses, and the boundary of central-bank authority. Therefore, a country's best practice is "optimal" for that country precisely because it was designed for that country's particular accountability boundary; transplanting it to a country with a different accountability boundary renders it no longer optimal, and may even be counterproductive. This shift in perspective is the key step by which this paper moves from technical correction to institutional judgment, and it is this paper's core incremental contribution relative to existing discussions. This meta-judgment constitutes the governing framework for all five groups of judgments in this paper. The technical judgments of Chapters 4 through 7 (the divergence of solvency and liquidity, cross-jurisdictional falsification, the mismatch in the cash-like nature of margin, and the resetting of the measurement scale) are, within this framework, understood as different facets or technical sub-cases of the choice of accountability; Chapter 8, in turn, takes this meta-judgment itself as its direct analytical object. ### 3.2 The Unfalsifiability of Universal Propositions and the Narrowing of Comparative Propositions Characterizing risk positioning as an institutionally endogenous choice of accountability faces a serious methodological risk: this proposition can very easily slide into an unfalsifiable universal form. If the proposition is stated as "any risk positioning is an institutionally endogenous choice of accountability," then it excludes no observation whatsoever—no matter where a country positions its risk, it can always be restated after the fact as "its institutional structure endogenously chose this accountability boundary." When a proposition is true in all possible worlds and no observation can falsify it, it is not an empirical finding but a reformulation at the level of framework. This paper therefore does not adopt the universal form, but instead narrows the meta-judgment into a comparative proposition with observational implications: given a set of ex-ante observable institutional variables X (including bankruptcy-law priority, the ownership structure of clearinghouses, and the boundary of central-bank authority), one can predict that a country will position the accountability boundary for decumulation-phase risk along a particular dimension Y, and that transplanting a practice Y′ that differs from the home country's Y will fail on specific observable indicators. The key to this proposition is that it allows for falsification through successful transplantation: if transplantation succeeds among countries with highly similar institutional variables X, the proposition is supported; if transplantation also succeeds among institutionally disparate countries, the proposition is undermined. It is precisely this condition of "allowing successful transplantation to serve as falsification" that converts an unfalsifiable universal ontological claim into a comparative proposition with observational implications. This narrowing carries far-reaching methodological significance. It means that this paper's highest-level judgment no longer depends on a philosophical assertion that cannot be touched by any evidence, but instead accepts the constraints of empirical testing. Accordingly, the empirical weight of this judgment is preferentially placed on hard identification designs with genuine exogenous variation—the staggered rollout timing of clearing obligations and the discontinuity at eligibility thresholds—rather than on the wording "endogenous accountability" itself (Chapter 8). ### 3.3 The Mapping Mechanism from Institutional Variables to the Dimension of Accountability The mechanism of the comparative proposition can be characterized as a chain running from institutional variables to the dimension of accountability, and then to transplantation failure. Each of the three dimensions of institutional variable X acts on the positioning of the accountability boundary: bankruptcy-law priority determines the order in which different stakeholders are made whole when a plan becomes insolvent, thereby determining who ultimately absorbs the risk; the ownership structure of the clearinghouse (member-owned versus for-profit/commercial) determines the institutional arrangements for margin rules and liquidity transmission; and the boundary of central-bank authority determines which end of the financial system public backstops can act upon. These three dimensions jointly determine where a country places the accountability boundary for decumulation-phase risk—the UK positions it at the liquidity of pension vehicles, while other jurisdictions may position it elsewhere. When a country attempts to transplant a practice Y′ that differs from its own Y (for example, replicating the UK's buffers, backstops, or leverage constraints in a country with different institutional variables X), because the transplanted practice encodes the source country's institutional choice of accountability, it will fail on specific observable indicators (the trigger-point yield, the speed of deleveraging, the share of contagion). The counterintuitive aspect of this mechanism is that it reconstructs "better regulation, thicker buffers, more reliable backstops" as "a redistribution of the accountability boundary"—best practices cannot simply be copied precisely because they encode, in their technical form, the source country's particular institutional choice of accountability, and transplanting them to a country with different institutional variables will cause them to fail on observable indicators. ### 3.4 The Honest Boundary of Layered Load-Bearing The final component of this paper's theoretical framework is a discipline of layered load-bearing that runs throughout the entire paper. This discipline requires that every group of judgments strictly distinguish its identifiable load-bearing layer from its downgraded theoretical or discursive layer. This distinction does not weaken the judgments; rather, it protects their value: a judgment whose identification boundary is clearly marked derives its empirical value precisely from its falsifiability. Specifically, this paper distinguishes evidence into several levels of identifiability. The highest level consists of hard identification designs with genuine exogenous variation, such as difference-in-differences analysis of the staggered rollout of clearing obligations and regression discontinuity at eligibility thresholds; the next-highest level consists of directional verification of aggregated institutional facts, such as the direction of the funding ratio, vehicle shares, and buffer thresholds; below that is a weak existence proposition, such as "if the order of liquidation is forced by collateral availability, then the real cost falls, at least in part, on the assets being liquidated"; the lowest level consists of theoretical-dynamics propositions, hypotheses awaiting empirical testing, and discursive inferences. The honest boundary of this paper is reflected precisely in the explicit labeling of the level to which each proposition belongs, and in the strict adherence to the principle of "not upgrading a lower-level proposition to already-identified causation." ### 3.5 Chapter Summary This chapter has established the theoretical framework of this paper. Its core is a meta-judgment: risk positioning is an institutionally endogenous choice of accountability. To prevent this judgment from sliding into an unfalsifiable universal form, this chapter has narrowed it into a comparative proposition that allows for falsification through successful transplantation—given institutional variables X, one can predict the dimension of accountability Y, and transplanting Y′ will fail on specific observable indicators. This chapter has further characterized the mechanism chain running from institutional variables to the dimension of accountability, and then to transplantation failure, and has established the discipline of layered load-bearing that runs throughout the paper. This framework governs the rest of the paper's chapters: the technical judgments of Chapters 4 through 7 are different facets of the choice of accountability, while Chapter 8 takes the meta-judgment itself as its analytical object. The establishment of this theoretical framework allows this paper, while correcting the conventional narrative, to avoid falling back into the shallow position that "improvement simply means thickening buffers," thereby providing the fundamental premise for a China-specific approach—namely, that it cannot be transplanted and must be endogenously designed. --- ## Chapter 4 An Improvement in the Funding Line Does Not Constitute a Safety Signal for the Liquidity Line ### 4.1 Judgment: Rising Rates as a Two-Sided Event This chapter puts forward the first set of judgments in this paper, and it is also the most direct correction of prevailing understanding. The judgment can be stated as follows: it was wrong to write the 2022 UK episode as "rising rates harmed pensions," but it is an error of the same order to write it as "coverage-ratio improvement, therefore the scheme is safe." The one defensible core judgment of this chapter is this — an improvement in the funding line does not constitute a safety signal for the liquidity line. The theoretical basis of this judgment is that a rise in interest rates is, in essence, a two-sided event. A rate increase simultaneously produces two effects that run in opposite directions: first, it depresses the present value of liabilities discounted at market rates, releasing long-duration liabilities and improving the coverage ratio; second, it depresses the value of collateral and fixed-income assets, triggering margin calls and deleveraging pressure. The sign of the net result of these two effects depends on the degree of duration matching between a scheme's assets and liabilities, and is a priori indeterminate. Therefore, the observed simultaneity of "coverage-ratio improvement and margin calls occurring at the same time" is not a directional causal chain in which "funding improvement triggers liquidity crystallization," but rather a spurious co-movement produced by interest rates as a common driver — the rate move pushed both lines at once, rather than one line causing the other. What requires careful delimitation is the negative-form nature of this judgment. This chapter does not assert the directional causal claim that "funding improvement is a precursor to, or cause of, liquidity crystallization," because such a claim would require additionally identifying, within a compound shock of undetermined sign, "the relative slopes of the two paths and the causal direction of their timing misalignment" — and this cannot be accomplished when interest rates are the common driver. What this chapter asserts is only a negative and weaker proposition: the funding-line-improvement side carries no informational content for the safety of the liquidity line, and may even point in the opposite direction. This is a proposition that negates an inference of safety, not a positive causal discovery. ### 4.2 Mechanism: From a Two-Sided Event to Spurious Co-Movement The mechanistic chain of this judgment can be characterized as follows. A rise in interest rates triggers the simultaneous occurrence of a two-sided event: on one hand, the present value of liabilities falls, releasing long-duration liabilities and improving the coverage ratio; on the other hand, the value of collateral and fixed-income assets falls, triggering margin calls and deleveraging pressure. The sign of the net effect is a function of the degree of asset-liability duration matching, and is a priori indeterminate. As a result, the observed simultaneity of "coverage-ratio improvement and margin calls occurring together" is spurious co-movement driven by interest rates as a common factor, rather than the former causing the latter. Under the pre-registered expectation, in the subsample with a negative duration gap and a high share of cash variation margin, the slope ratio of the two paths would display a timing misalignment in which funding improves even as liquidity crystallizes earlier. The counterintuitive aspect of this mechanism lies not in overturning the news-style narrative, but in overturning the next layer of optimism within the corrective narrative itself. Once existing discussion has already recognized that "what erupted in 2022 was a liquidity crisis rather than a solvency crisis," a natural next inference is that "since solvency has improved, the scheme's fundamentals are safe." The judgment of this chapter precisely negates this next-layer inference: an improvement in the funding line not only carries no information for the liquidity line, but may even show a negative relationship, because the two are driven in opposite directions by the same interest-rate shock. The dividing line between release and spiral rests on the state variable of "what collateral can be posted in" — but because this state variable is itself endogenous (see Section 4.4), the mechanism layer can state this dividing line, while the identification layer must rely on external quasi-random variation. To subject this judgment to the most rigorous scrutiny, the two strongest counterarguments against it must be addressed head-on. The first counterargument targets the causal direction of this judgment. It points out that if this judgment implies the claim that "funding improvement is a precursor to liquidity crystallization," then this is a directional causal claim (temporally prior and causally same-directional); yet the mechanism itself acknowledges that this is a two-sided event — a rate increase simultaneously releases liabilities and marks down assets, with the sign of the net effect determined by the degree of duration matching and a priori indeterminate. A compound shock of undetermined sign cannot, without additionally identifying "the relative slopes of the two paths and the causal direction of their timing misalignment," assert that the "funding-improvement side" is the cause or precursor of the liquidity crisis. The observed simultaneity of "coverage-ratio improvement and margin calls occurring together" is spurious co-movement from a common driver (interest rates pushing both at once); packaging this spurious co-movement, driven by the confounding factor of interest rates, as a causal discovery of a "state-dependent transition function" would commit an identification error. This paper accepts this counterargument, and accordingly strictly confines the judgment to a negative-form, weak assertion: this chapter abandons the directional causal language of "precursor or crystallization," acknowledges that the sign of the net effect is a priori indeterminate, and downgrades the falsifiable proposition to conditional co-occurrence — namely, "in the subsample with a negative duration gap and a high share of cash variation margin, coverage-ratio improvement co-occurs with earlier position closeouts" — bearing its evidentiary weight on the slope ratio of the two paths (rather than on co-directionality). Logical consistency takes priority over narrative tension: rather than asserting an unidentifiable directional causal claim, it is better to hold to an identifiable negative-form proposition. The second counterargument targets the endogeneity of the cash variation margin share as a treatment variable, and the identification status of the distributional-axis proposition. It points out that the one concrete falsifiable claim treats "cash variation margin share" as a quasi-exogenous treatment variable, but whether a scheme posts cash or pledgeable securities is highly collinear with whether it enters a pooled vehicle, its size, and the sophistication of its sponsor — choosing to enter cash variation margin is often tantamount to choosing to enter a pooled vehicle, typically a smaller and less sophisticated scheme, so "a high cash variation margin share causes earlier position closeouts" cannot be separated from "the scheme was inherently fragile, choosing both cash and earlier closeout." Moreover, if "member benefits and sponsor covenant being drained away" is hung on the position of the core judgment and the primary evidentiary weight, this amounts to committing to a distributional finding, while the empirics in fact retreat into the predicament of insufficient covenant granularity, constituting "theory dressed up as already identified." This paper accepts this counterargument and makes a corresponding reassignment of evidentiary weight: the cash variation margin share is downgraded to a mechanism channel, not passed off as exogenous treatment, and is elevated to quasi-exogenous evidentiary weight only where scheme-level quasi-random variation in collateral architecture exists; the distributional axis is moved out of the primary evidentiary weight, narrowing the core judgment to the empirically tractable half concerning the timing misalignment between funding and liquidity, while moving "benefits or covenant being drained away" into the discussion as a speculative inference explicitly flagged as non-empirical. The primary evidentiary weight then shifts from the imagery of the distributional axis to the two genuinely identifiable paths of vehicle-structure comparison and the event study of funding-liquidity timing misalignment. Through direct treatment of these two counterarguments, the judgment of this chapter is refined into a negative-form weak causal proposition with clear identification boundaries, that is falsifiable and non-trivial. ### 4.3 Identification Strategy and Evidence The identification goal of this chapter is to identify the negative-form weak causal proposition of whether "an improvement in the funding line constitutes a safety signal for the liquidity line," explicitly modeling interest rates as the common driver, and bearing evidentiary weight on the slope ratio and timing misalignment between the funding path and the liquidity path, rather than on "co-directionality." The directional causal proposition of same-direction movement and the distributional-axis proposition are not within the identification target, and are instead treated respectively as a theoretical proposition and a discussion-level inference. The identification strategy sets up a primary path and a downgrade exit. The first primary path is the quasi-natural comparison, within the UK, between pooled and segregated liability-driven investment (LDI) vehicles. Pooled LDI (high leverage, shared collateral pools) and segregated LDI (low leverage, independent pools) constitute the cleanest quasi-natural comparison, used to examine the effect of vehicle structure on the timing of liquidity-path crystallization; this comparison source is shared with Chapter 5, and the accounting basis must be kept consistent. The second primary path is an event study of the timing misalignment between the funding path and the liquidity path: using the 2022 gilt shock as the event window, pinning down the timing misalignment between the funding path (coverage ratio) and the liquidity path (margin calls and position closeouts), explicitly modeling interest rates as the common driver, and bearing evidentiary weight on the slope ratio of the two paths. Its identifying assumption is that the transmission lag of the interest-rate shock differs across the two paths, such that the slope ratio can separate out the spurious co-movement. At the level of evidence, the aggregate facts at both endpoints of this chapter have each been independently corroborated, consistent with the two-sided-event structure. The fact that the funding line moved in the direction of improvement has been corroborated: as noted earlier, the PPF 7800 index's aggregate surplus reached a historic peak in August 2022, with the funding ratio rising markedly (PPF 2022). The fact that the liquidity line moved in the direction of crisis has likewise been corroborated: in the same period, forced LDI sell-offs triggered a gilt sell-off spiral, prompting the Bank of England to launch a temporary bond-buying program on September 28 (BoE 2022). The two indeed occurred simultaneously, and the empirical premise for spurious co-movement holds. The vehicle-comparison path is supported by aggregate-level data: pooled LDI accounted for roughly 10% to 15% of the LDI market and was used by smaller DB schemes; segregated LDI was used by larger schemes, hedged roughly 85% of liabilities, and performed well overall owing to mature liquidity management; pooled vehicles, owing to recapitalization frictions, sold approximately 13 percentage points more of their gilt holdings than segregated vehicles, with peak fire-sale discounts reaching 7% to 10% (Breeden 2022). Accordingly, the pre-registered expectation that "pooled vehicles explain the great majority of earlier and larger position closeouts, with segregated vehicles significantly less affected" receives consistent support from the aggregate evidence. Furthermore, the calibrating basis for the two-sided event is consistently manifested across both the UK and Dutch cases. The two-sided structure whereby a rise in interest rates simultaneously depresses the present value of liabilities and the value of asset collateral appears consistently in both the UK case (coverage-ratio improvement accompanied by a liquidity crisis) and the Dutch case (coverage ratio rising to roughly 122% while simultaneously selling €88 billion in assets to meet margin calls) (PPF 2022; IPE 2022), confirming that the "coverage-ratio improvement" side offers no protective information regarding liquidity calls. The spurious co-movement is jointly driven by interest rates, rather than funding improvement triggering liquidity crystallization. The identification design of this chapter follows pre-registration discipline, meaning that expected and unexpected outcomes were fixed in writing before examining any specific evidence, so as to prevent ex post rationalization. With respect to the negative-form core proposition, the pre-registered expectation is: in the subsample with a negative duration gap and a high share of cash variation margin, the ratio of the funding-path slope to the liquidity-path slope shows funding improvement accompanied by earlier liquidity crystallization (timing misalignment). Its unexpected outcome (i.e., the condition for an honest downgrade) is: if the slope ratio of the two paths shows funding improvement accompanied by liquidity relief (true co-directionality, no misalignment), then the judgment that "funding improvement does not constitute a liquidity safety signal" is falsified, and one must fall back to the traditional distinction that "solvency and liquidity are two parallel risks," with this chapter losing its negative-form increment. With respect to the vehicle comparison, the pre-registered expectation is that pooled vehicles explain the great majority of earlier position closeouts, with segregated vehicles significantly later; its unexpected outcome is that if there is no difference in closeout timing between pooled and segregated vehicles, such that vehicle structure does not explain the liquidity path, then the vehicle path is downgraded to a mechanism description, and the primary evidentiary weight retreats solely to the timing-misalignment event study. In addition, there is a falsification condition targeting the common driving role of interest rates: if, after explicitly modeling interest rates, the misalignment between the two paths disappears (i.e., the misalignment is purely artificially produced by the transmission lag of interest rates rather than by structural differences), then the timing misalignment does not count as structural evidence. As for the executed conclusions of this study, the aggregate-direction premises at both endpoints have been corroborated, and the unexpected outcome (true co-directionality) did not occur; however, the micro-estimation of the slope ratio could not be executed owing to limited panel data, and the test of "whether the misalignment disappears after explicit modeling" likewise could not be executed for lack of a panel — these remain open, and nothing is fabricated. This pre-registration discipline allows the judgment of this chapter to gain directional support from aggregate evidence while honestly preserving the identification boundary of its micro-causal claims. ### 4.4 The Honest Boundaries of the Evidence This chapter strictly flags its identification boundaries: any proposition that cannot be identified owing to data-availability constraints is never upgraded to an identified causal claim. First, the delimitation of directional causality. The directional causal claim that "funding improvement is a precursor to, or cause of, liquidity crystallization" has a counterfactual — "if funding had not improved, liquidity would not have crystallized" — that is unobservable because interest rates are the common driver, and hence cannot be identified; this chapter therefore maintains it as a theoretical two-sided-event proposition, not as an identified causal chain. Second, the delimitation of the slope ratio. Estimating the slope ratio of the two paths requires scheme-level panel data and high-frequency or intraday margin coverage-ratio series within the event window; such micro-data are constrained and unavailable, and the numerical value of the slope ratio is neither estimated nor fabricated. The corresponding identification design — using the 2022 gilt shock as the event window, absorbing the common interest-rate shock and time-invariant heterogeneity with year and scheme fixed effects, and focusing on the subsample with a negative duration gap and a high cash share — is already in place, pending execution once scheme-level panel data become available. Third, the delimitation of the cash variation margin share. In the proposition that "a high cash variation margin share causes earlier position closeouts," the treatment variable "cash variation margin share" is an endogenously chosen variable, highly collinear with whether a scheme enters a pooled vehicle, scheme size, and sponsor sophistication. Choosing to enter cash variation margin is often tantamount to choosing to enter a pooled vehicle, typically a smaller and less sophisticated scheme, so "a high cash variation margin share causes earlier position closeouts" cannot be separated from "the scheme was inherently fragile, choosing both cash and earlier closeout." Unless quasi-random variation in scheme-level collateral architecture is obtained (for example, a regulatory or clearing-rule change that forces some schemes to switch to posting cash, grouped by registration timing or a threshold discontinuity via regression discontinuity or difference-in-differences, with the counterfactual being the same scheme posting securities had it not been compelled), this proposition remains a mechanism channel and is not passed off as a quasi-exogenous treatment variable; the difference in collateral architecture between the UK and the Netherlands is not treated as quasi-exogenous, because attributing it in full involves multidimensional differences in leverage, regulation, currency area, and marginal buyers, constituting an omitted-variable problem. Fourth, the delimitation of the distributional axis. The distributional-axis proposition that "member benefits and sponsor covenant were being drained away on the eve [of the crisis]" has not been identified, owing to the undetermined sign of the two-sided event requiring directional decomposition, and owing to its dependence on distributional data at the granularity of sponsor covenants; it is therefore moved out of the core judgment and the primary evidentiary weight, and moved into the discussion as a speculative inference explicitly flagged as non-empirical, to be upgraded to an empirical claim only once covenant-level distributional data are supplemented. ### 4.5 Relationship to Other Judgments and Summary The judgment of this chapter sits at the level of event misdiagnosis within the overall chain of argument. It falsifies, within the UK, the equation "coverage-ratio improvement equals safety" (its empirical premise being the simultaneous occurrence of coverage-ratio improvement and liquidity crisis, see PPF 2022; BoE 2022), and Chapter 5 extends this falsification to a cross-country comparison. The two-sided event of this chapter and the state variable of "what collateral can be posted in" provide the empirical entry point for Chapter 6's "mismatch between margin moneyness and rescuability"; this chapter bears only the negative-form co-occurrence, while Chapter 6 bears the theoretical proposition of moneyness dynamics, and the two do not duplicate the evidentiary weight of the same causal claim. The "slope ratio of the two paths" in this chapter provides the mechanistic motivation for Chapter 7's argument that "path functionals are superior to static endpoint metrics." Furthermore, this chapter supplies the subsequent China proposals with the prerequisite governance premise that "risk monitoring cannot look at the funding coverage ratio alone." The summary of this chapter is as follows. This chapter has argued for the negative-form, weak causal judgment that an improvement in the funding line does not constitute a safety signal for the liquidity line. Its theoretical basis is that a rise in interest rates is a two-sided event whose net-effect sign is a priori indeterminate, and its empirical premise (the simultaneous occurrence of funding improvement and liquidity crisis, and differences in vehicle structure) has been corroborated at the aggregate level. This chapter strictly distinguishes between the identifiable evidentiary paths (vehicle comparison and the timing-misalignment design) and the downgraded theoretical and discussion layers (directional causality, the slope-ratio value, the cash variation margin share, and the distributional axis), maintaining honest boundaries. This judgment corrects the next layer of optimism within the corrective narrative itself, laying the analytical foundation of a strict distinction between funding and liquidity for the chapters that follow. --- ## Chapter 5 The Non-Explosion of Dutch and Danish Schemes Cannot Be Used to Infer Institutional Superiority: Survivorship Bias as a Symmetric Blade ### 5.1 Proposition: The Symmetry of Survivorship Bias This chapter puts forward the second set of propositions of this paper, addressing the methodology of cross-national comparison. The proposition has two layers. The first is a pure falsification layer: the fact that the Netherlands and Denmark did not experience a UK-style spiral in 2022 is not sufficient to prove that their institutional arrangements are superior. The reasoning is that survivorship bias here is a symmetric blade — if the survival of the Dutch and Danish schemes was in fact subcritical luck (that is, this time they simply happened not to hit the tail), then this data provides no information whatsoever about "whether the institutional arrangement is superior." It negates both the inference that "the Dutch and Danish institutional arrangements are superior" and the reverse inference that "the UK is not special, and since nothing exploded this time, the UK is also safe." The second is a layer of identifiable mechanism differences: identifiable load-bearing weight is not placed on the outcome variable of "no explosion this time," but rather on two hard identification paths with exogenous variation — the pseudo-jurisdiction registration comparison between Ireland and Luxembourg, used to falsify the jurisdictional dichotomy of "the UK equals danger, continental Europe equals safety"; and the quasi-natural comparison within the UK between pooled (high-leverage) and segregated (low-leverage) vehicles, used to isolate vehicle structure. The crux of this two-layer proposition is that it demonstrates that "Dutch and Danish survival" — evidence that appears to support one side — in fact provides no information for either direction. Survivorship bias is not evidence supporting one side; it is a blade that fails bidirectionally. True identification lies not in "who did not explode," but in "whether structural differences manifest where exogenous variation exists." ### 5.2 Mechanism: From Symmetric Blade to Identifiable Paths The mechanistic chain of this proposition can be characterized as follows. The non-explosion of the Dutch and Danish schemes in 2022 admits a dichotomy as to its cause: either systemic institutional superiority, or subcritical luck. Because there is no independent, ex ante computable measure of critical distance, these two explanations are indistinguishable. Survivorship bias therefore fails symmetrically: it can be used to infer neither "Dutch and Danish superiority" nor "the UK is not special." The load-bearing weight is thus forced to move off the outcome variable of "non-explosion" and onto two identification paths with exogenous variation — pseudo-jurisdiction registration (schemes registered in Ireland or Luxembourg but exposed to UK gilt dynamics), used to falsify the jurisdictional dichotomy, and the comparison within the UK between pooled and segregated vehicles, used to isolate vehicle structure. The counterintuitive aspect of this mechanism is that it demonstrates that "Dutch and Danish survival" — evidence that appears to support one side — is in fact uninformative for either direction. True identification lies not in "who did not explode," but in "whether structural differences manifest where exogenous variation exists." This position carries constraining implications for the methodology of cross-national comparison as a whole: it requires analysts to abandon reading institutional-quality rankings from the cross-sectional outcomes of a single event-year, and instead to seek the causal manifestation of structural differences within comparisons that have exogenous variation. The strongest counterargument against this proposition arises precisely from the logical rigor of the symmetric blade itself. This counterargument points out: if the survival of the Dutch and Danish schemes is indeed survivorship bias (that is, subcriticality was merely luck this time — the sample simply did not draw the tail) — then this data provides no information whatsoever about "whether the institutional arrangement is superior." Survivorship bias precisely means that the outcome cannot be used to infer institutional quality. The same blade therefore negates both "the Dutch and Danish institutional arrangements are superior" and "the Dutch and Danish institutional arrangements are not superior, and the UK is not special." This chapter therefore cannot have it both ways: it cannot use "survivorship bias" to falsify the opponent's claim of "Dutch and Danish superiority" on one hand, while using that same survival to positively support "the UK mechanism is not special" on the other. Logically, one cannot hold both — one cannot say "this instance of survival carries no information" while simultaneously treating "this instance of survival" as evidence that "the UK is not special." This paper fully accepts the logical constraint of this counterargument, and accordingly confines its claim strictly to agnosticism: the pure falsification layer of this chapter performs falsification only — "the survival of the Dutch and Danish schemes is not sufficient to prove their institutional arrangements superior" — without asserting in the reverse direction that "the UK is not special, and since nothing exploded this time the UK is also safe" (this positive assertion has been judged symmetrically void and deleted). Correspondingly, the identifiable load-bearing weight is moved off the outcome variable of "non-explosion" and onto the two identification paths with exogenous variation. The second strongest counterargument against this proposition arises from the degrees-of-freedom constraint on cross-national decomposition. This counterargument points out: isolating causation from "hedging" to "leverage" requires varying hedging or leverage while holding everything else constant (vehicle structure, regulation, currency area, marginal buyers, clearing arrangements, market depth) — but the three countries differ simultaneously along all of these dimensions, and this is a cross-section of three countries in a single event-year. Difference-in-differences requires parallel trends and multiple periods and units; synthetic control requires a donor pool; with three units and a single treatment period, there are no degrees of freedom to estimate any decomposition. Hence the claim that "Denmark, with high hedging and low leverage, nonetheless survived, showing hedging to be neutral and leverage to be toxic" arbitrarily attributes a single data point's multidimensional differences to "leverage." Furthermore, if one part of the argument claims vehicle structure (pooled funds) is the UK's cause of death, while another part uses Denmark to pin causation on leverage, then "leveraged pooled funds" welds leverage and vehicle together — one cannot simultaneously claim "it is the vehicle, not the strategy or leverage level" and claim "Denmark proves it is leverage"; the two load-bearing lines cancel each other out. This paper accepts this counterargument, downgrading the cross-national comparison to a motivating stylized fact (the reasoning being that what is stuck is not missing data but structural non-identifiability, i.e., degrees of freedom fewer than parameters), downgrading vehicle and leverage to two hypotheses awaiting separation, and using the comparison within the UK between pooled and segregated vehicles to break the collinearity between the two. The cross-national portion bears only the falsification of the jurisdictional dichotomy (the pseudo-jurisdiction registration comparison, for which the evidence is sufficient and defensible), and does not pretend to causal identification of leverage or hedging. Through direct engagement with these two counterarguments, the proposition of this chapter is confined to a methodologically rigorous two-layer statement with a clear load-bearing path. ### 5.3 Identification Strategy and Evidence The identification goal of this chapter is to identify the two identifiable paths of "whether the jurisdictional dichotomy is a false proposition" and "the causal effect of vehicle structure on selling intensity." The decomposition of cross-national leverage and hedging, and the measurement of critical distance, are not within the identification goal, and are treated respectively as a motivating stylized fact and as descriptive phrasing. The first main path of the identification strategy is the pseudo-jurisdiction registration comparison. Schemes registered in Ireland or Luxembourg but exposed to UK gilt dynamics (for example, LDI vehicles registered in Dublin) constitute a comparison for "jurisdiction does not equal risk" — if continentally registered schemes were equally under pressure, then the jurisdictional dichotomy of "the UK equals danger, continental Europe equals safety" is falsified. Its identifying assumption is that registration location is orthogonal to true market exposure, i.e., that registration reflects a choice of tax and regulatory venue rather than a choice of risk. The second main path is the quasi-natural comparison within the UK between pooled and segregated vehicles, testing the effect of vehicle structure on selling intensity, sharing the same comparison source and consistent accounting basis as the main path of Chapter 4. At the evidentiary level, the aggregate anchor for the pseudo-jurisdiction registration comparison receives strong corroboration. Sterling LDI funds resident in Ireland accounted for approximately 30% of net gilt sales by LDI during the crisis period; Ireland is the largest EU registration venue for sterling LDI (accounting for approximately 85% of EU-registered sterling LDI net assets at end-2021), and these funds are mostly managed out of London and merely registered in Dublin or Luxembourg (Central Bank of Ireland 2023). On this basis, the pre-registered direction — that "continentally registered schemes but exposed to UK gilt dynamics were equally under pressure, falsifying the jurisdictional dichotomy" — receives strong corroboration: registration location (a tax and regulatory venue choice) is orthogonal to true market exposure, and the jurisdictional dichotomy of "the UK equals danger, continental Europe equals safety" is confirmed as falsified. The resilience buffer requirements imposed after the crisis by Ireland and Luxembourg on such funds — at least a 300-basis-point UK yield shock — further corroborate the common origin of the exposure (ESMA 2022). The aggregate evidence for the vehicle comparison path is likewise consistently supportive: pooled vehicles sold approximately 13 percentage points more due to recapitalization friction, while segregated vehicles performed well (Breeden 2022). The facts regarding the Netherlands and Denmark are presented as motivating stylized facts. Dutch pension funds sold €88 billion in assets and injected €82 billion in margin in the first half of 2022, yet coverage ratios improved over the same period; the structural difference behind the Netherlands' avoidance of a UK-style spiral lies in the fact that euro LDI rarely or never uses repo to add leverage, permits securities to be posted for part of the margin, and relies on a well-functioning money market (IPE 2022; DNB 2024). Danish pension funds make extensive use of interest-rate derivatives for hedging and operate within a larger, more liquid euro market; Danish regulators subsequently noted that some firms exhibited "non-negligible liquidity risk" and needed to forecast margin requirements on an hourly basis — that is, survival does not equal absence of risk (IPE 2023), consistent with this chapter's position that "subcritical luck and institutional superiority are indistinguishable." The calibration basis of this chapter must be made explicit. This chapter uses pseudo-jurisdiction registration as the ruler for calibrating "whether jurisdiction predicts risk" — schemes registered in Ireland or Luxembourg but truly exposed to UK gilt dynamics constitute a known anchor point where "jurisdictional label and market exposure are decoupled"; if these schemes were equally under pressure (with no systematic difference in net selling and unwinding pressure from UK-registered schemes), then "jurisdiction equals risk" is calibrated as a false signal. The second calibration basis of this chapter is to use the within-UK pooled versus segregated vehicle structure as a quasi-exogenous ruler for calibrating "vehicle versus leverage" — because the two are welded together at the cross-national level, only the internal comparison can provide quasi-exogenous variation in vehicle structure, thereby calibrating the truth or falsity of the cross-national single-point attribution (if the cross-national "pinning it on leverage" is not reproduced in the internal comparison, then the cross-national attribution is false). These two calibration bases together ensure that this chapter does not pass off the multidimensional differences of a single event-year as identification. As for pre-registered expectations, this chapter pre-registers the falsification of the jurisdictional dichotomy as a falsifiable proposition (continentally registered schemes exposed to UK dynamics were equally under pressure), with the unexpected outcome being that if continentally registered schemes were significantly more stable, indicating that jurisdiction genuinely predicts risk, then jurisdictional difference reverts to being a genuine explanatory variable; the vehicle comparison is pre-registered as a falsifiable proposition (pooled vehicles explain the vast majority of selling, segregated vehicles significantly less), with the unexpected outcome being that if the selling intensity of the two shows no difference, then the vehicle-side account is weakened; the cross-national decomposition and critical distance are designated in advance as non-identified. As for the executed conclusions of this study, the direction of the falsification of the jurisdictional dichotomy receives strong corroboration (Ireland-resident funds account for approximately 30% of net selling) (Central Bank of Ireland 2023), and the unexpected outcome did not occur, but the coefficient of the pseudo-jurisdiction comparison regression and the magnitude of the difference-in-differences for the vehicle comparison are both left unestimated and unreported due to limited fund-level micro data; the honest boundary of the symmetric blade is strictly observed — the Dutch and Danish cases serve only as case illustrations, and "non-explosion" is not used to positively support either direction. ### 5.4 The Honest Boundaries of the Evidence This chapter strictly delineates its identification boundaries. First, the honest boundary of the symmetric blade. The Dutch and Danish facts above are presented only as motivating stylized facts. This chapter does not claim that "Dutch and Danish survival disproves that the UK is special" — this positive assertion has been judged symmetrically void and deleted. The outcome variable of Dutch and Danish "non-explosion" bears no identification weight; the identification weight falls on the two identification paths with exogenous variation (the pseudo-jurisdiction 30% net selling figure and the within-UK vehicle comparison). This chapter cannot have it both ways: it cannot use "survivorship bias" to falsify the opponent's claim of "Dutch and Danish superiority" on one hand, while using that same survival to positively support "the UK mechanism is not special" on the other. If anyone uses "nothing exploded this time" to positively support either direction, this violates the honest boundary of this chapter. Second, the delimitation of the cross-national decomposition. Isolating causation from "hedging" to "leverage" requires varying hedging or leverage while holding everything else constant (vehicle structure, regulation, currency area, marginal buyers, clearing arrangements, market depth) — but the three countries differ simultaneously along all of these dimensions, and this is a cross-section of three countries in a single event-year. Difference-in-differences requires parallel trends and multiple periods and units; synthetic control requires a donor pool; with three units and a single treatment period, there are no degrees of freedom to estimate any decomposition. Hence the inference that "Denmark, with high hedging and low leverage, nonetheless survived, showing hedging to be neutral and leverage to be toxic" arbitrarily attributes a single data point's multidimensional differences to "leverage," and is uniformly left unestimated and unreported, remaining a motivating case only. Its upgrade condition is obtaining a cross-national panel (multiple years, multiple countries, including variation in leverage-hedging outcomes in non-explosion years) or constructing an ex ante critical-distance index for out-of-sample prediction. Third, the delimitation of the pseudo-jurisdiction comparison coefficient. Estimating the coefficient of the pseudo-jurisdiction comparison regression (the UK-registration coefficient) requires fund-level cross-sectional micro data on selling intensity, registration location, and control variables; the coefficient is not estimated. The identification design is in place, and the aggregate fact of 30% net selling supports the direction, but does not substitute for identification of the coefficient. The same holds for the magnitude of the causal difference-in-differences effect of the vehicle comparison: the fact that pooled vehicles sold approximately 13 percentage points more is an aggregate stylized fact; converting it into a causal difference-in-differences coefficient with control covariates requires a fund-level panel, and is left unestimated and unreported. Fourth, the collinearity between vehicle and leverage remains unbroken. Because "leveraged pooled funds" welds together "vehicle structure is fatal" and "leverage is fatal," their collinearity remains unbroken, and they are downgraded to two hypotheses awaiting separation. The cross-national portion bears only the falsification of the jurisdictional dichotomy (the pseudo-jurisdiction registration comparison, for which the Irish 30% net-selling evidence is sufficient and defensible), and does not bear the causal identification of leverage. Fifth, the delimitation of critical distance. "Subcriticality and critical distance" have no independent ex ante measure, and belong to reading backward from the outcome (a tautology whereby non-explosion is equated with subcriticality); this is downgraded to descriptive phrasing and does not bear explanatory weight. Unless a critical-distance indicator computable before 2022 (leverage multiplied by collateral moneyness multiplied by market depth) is constructed and subjected to out-of-sample testing, it will remain descriptive only. ### 5.5 Relationship to Other Propositions and Summary The proposition of this chapter sits, in the chain of argumentation, at the level of cross-sectional misuse. Building on Chapter 4's falsification of "coverage ratio equals safety within the UK," it pushes the falsification to the cross-national level — but the cross-national portion is downgraded to a case, with identification falling back on the within-UK pooled versus segregated comparison (sharing the same quasi-natural comparison and consistent accounting basis as Chapter 4). The identifiable output of this chapter (the vehicle comparison and the falsification of the jurisdictional dichotomy) is handed off to Chapter 6, which refines the mechanism down to collateral and margin moneyness; this chapter does not bear the causal burden of moneyness. This chapter's finding that "the Dutch and Danish experience cannot be simply transplanted (survivorship bias plus structural differences)" provides, for the subsequent discussion of a Chinese scheme, the premise that "continental European best practices cannot simply be copied." The summary of this chapter is as follows. This chapter has argued that the non-explosion of the Dutch and Danish schemes cannot be used to infer institutional superiority, and that survivorship bias here is a bidirectionally void symmetric blade. This chapter reconstructs the cross-section of three countries from "a ranking of institutional superiority and inferiority" into "a survivor sample that cannot be used for ranking, plus identifiable structural differences within the UK." The identifiable load-bearing weight falls on the two identification paths of the falsification of the jurisdictional dichotomy (the pseudo-jurisdiction registration comparison, whose aggregate direction receives strong corroboration) and the vehicle comparison, while the cross-national leverage-hedging decomposition and critical distance are downgraded owing to structural non-identifiability or the absence of an ex ante measure. The methodological stance of this chapter — strictly stratifying the defensible (comparisons with exogenous variation) from the indefensible (causal readings of single-event-year cross-sectional outcomes) — is consistent throughout with the honest boundaries maintained across this entire paper. --- ## Chapter 6 The Locus of the Spiral: The Mismatch Between the Moneyness of Margin and Central Bank Rescuability ### 6.1 Judgment: The Moneyness Mismatch as a Theoretical-Dynamics Proposition This chapter advances the third set of judgments of this paper, sharpening the mechanism down to the locus of the spiral. The judgment can be stated as follows: diagnosing the 2022 pathology as "too much leverage or too many alternatives" is a mismatch of instruments; but the core mechanism diagnosis of this chapter has not been identified by the single 2022 rescue episode, and is therefore honestly positioned as a theoretical-dynamics proposition. The content of this theoretical proposition is: the synthetic-duration spiral, welded through central counterparty cash margin into a non-self-healing cash spiral, has its locus in the mismatch between the moneyness of margin (which must be posted in cash) and the topology of central bank rescuability (which operates on the securities side). This proposition can be written as a dynamics inequality and presented as a model proposition. This chapter simultaneously offers a fallback weak version that is defensible as an existence proposition: if the liquidation sequence is forced by collateral availability, the real cost falls at least in part on retained assets (the existence of a stock effect) — but this chapter does not claim the normative superiority of "the scheme should have held on to the overweight." This weak proposition and the theoretical proposition of the moneyness mismatch constitute the two load-bearing layers of this chapter. ### 6.2 Mechanism: From the Synthetic-Duration Spiral to the Moneyness Weld Point The mechanism chain of this judgment can be characterized as follows. A rise in interest rates generates unrealized losses on synthetic-duration positions (LDI derivatives), triggering central counterparty cash-margin calls (which cannot be met with securities); the scheme sells gilts to raise cash, the gilt price falls, and margin calls intensify, forming a spiral. The critical weld point is this: because margin must be posted in cash, the spiral is welded into a non-self-healing cash spiral — even if liquidity on the securities side is ample, it cannot on its own resolve the cash-side margin-call pressure. The locus of the pathology therefore lies in the mismatch between the moneyness of margin (which must be posted in cash) and the topology of rescuability of central bank instruments (which operate on the securities side). The counterintuitive aspect of this mechanism is that it reverses a default proposition of the corrective narrative: that illiquidity equals a liability of the decumulation (payout) phase. That is, "a high share of alternative or illiquid assets" is not the primary cause of the liquidity crisis (this is the existence of a stock effect); the true dynamics lie in the cash-moneyness property on the liability side, not in the allocation ratio on the asset side. This counterintuitive prescription — that the decumulation phase should not rush to correct the overweight, that being unable to sell is in a sense a form of protection, and that the real cost falls on the assets forced into liquidation — is downgraded in this chapter to a conditional: pending acquisition of a portfolio-level holdings panel, no normative directive is issued. The Dutch counterexample precisely corroborates the criticality of this weld point: euro-area LDI can post part of its margin in securities rather than cash alone, and makes minimal use of repo, and hence was not welded into a cash spiral (DNB 2024). This contrast is consistent with the direction of the moneyness-mismatch proposition, suggesting that the pivot determining whether the spiral becomes self-reinforcing is the moneyness of margin — a liability-side attribute — rather than the asset-side allocation structure. The core mechanism diagnosis of this chapter must withstand a sharp counterargument, and this chapter's response to that counterargument is precisely what constitutes the core of its honest boundary. The counterargument points out that a falsifiable version of this judgment requires observing the scenario in which "the rescue acts only on the securities side, while the cash-side spiral continues unabated." But there was only one rescue in 2022 (the BoE's gilt purchases), and it inherently acted on the securities side, or the gilt market. Researchers cannot observe the counterfactual of "what would have happened had the BoE injected directly into the cash side" — there is no control case of a cash-side rescue. Hence the assertion that "a securities-side rescue is ineffective for the cash side" is not an identified comparison, but rather a restatement derived from the definitional scope of the single rescue (purchasing gilts naturally acts on the gilt market). Further, "moneyness-rescuability mismatch" as the locus of pathology is, in substance, a mechanism narrative: it is consistent with the observations, but equally consistent with alternative mechanisms such as "buffers were too thin," "expectations of forced unwinding," or "information spillovers" — a single-point rescue cannot identify among these mechanisms. This paper fully accepts this counterargument, and on this basis explicitly labels the core mechanism diagnosis itself as a theoretical-dynamics proposition, whose empirical identification awaits external scope variation, without claiming that it has already been identified by the 2022 episode. This honest positioning is what distinguishes this chapter from "theory disguised as already identified": this chapter does not package an unidentified mechanism diagnosis as an established causal finding of the 2022 event, but explicitly presents it as a theoretical proposition — one that can be written as a dynamics inequality and that awaits testing on a cross-rescue-episode panel. This chapter must also address a second counterargument directed at its normative prescription. This counterargument points out that the most gripping and counterintuitive prescription of this chapter — that the decumulation phase should not correct the overweight, that being unable to sell is protection, and that the real cost is the stock effect on retained assets — has as its empirical basis a portfolio-level holdings panel that is precisely judged unavailable, constituting "theory disguised as an identified normative version." Worse still, "an overweight that should have been held through was wrongly liquidated" can only be defined by whether a rebound subsequently occurred — a rebound would justify saying "it should have been held," while a continued decline would mean one cannot say "wrongly liquidated" — rendering the proposition unfalsifiably true. This paper accepts this counterargument, downgrades the normative prescription to a conditional proposition, and front-loads its data dependency: "if retained assets systematically underperform following the fire sale (this must be tested with a holdings panel), then the stock effect dominates and holding through is superior"; pending acquisition of the panel, no normative mandate of "the overweight should not have been corrected" is issued, and the load-bearing claim retreats to the weak existence proposition of the stock effect. At the same time, to break the circularity of an ex post definition, this chapter front-loads an ex ante criterion: replacing the ex post label of "wrongly liquidated" with an ex ante computable "should-retain set" (based on liquidity premium, fundamental quality, and correlation with the spiral), using ex ante rules for out-of-sample comparison rather than hindsight. Through its handling of these two counterarguments, the judgments of this chapter are clearly divided into two layers — a defensible weak existence proposition and a proposition explicitly labeled as theoretical concerning the moneyness mismatch — without disguising either layer as an identified causality or an established normative conclusion. ### 6.3 Identification Strategy and Evidence The identification goal of this chapter is to render precise, as a theoretical-dynamics proposition (a dynamics inequality), the mismatch between the moneyness of margin and central bank rescuability, and to design the sole identification exit through which it could be upgraded from theory to empirics (a panel of scope variation across rescue episodes); at the same time, the stock-effect prescription is downgraded to a conditional proposition accompanied by an ex ante criterion. Causal identification of the core mechanism is not claimed at this stage — it is explicitly flagged as awaiting external scope variation. The first main path of the identification strategy is event evidence on the scope of the BoE rescue. The scope of the BoE's 2022 gilt purchases (the securities side) is identifiable — which side a rescue instrument acts on is an observable institutional fact (what is identified is the scope, not the counterfactual ineffectiveness). The second main path is the weak existence proposition of the stock effect: when the liquidation sequence is forced by collateral availability, the real cost falls at least in part on retained assets — defensible as an existence proposition, without claiming normative superiority. The upgrade path is a panel across rescue episodes: if multiple rescue or intervention episodes (across countries, across time, involving central bank instruments of differing scope) are obtained to constitute scope variation, and if a panel of "scope times outcome" shows that securities-side instruments provide no relief to the cash spiral while cash-side instruments do provide relief, then the moneyness mismatch would gain empirical support — this is the sole identification exit by which this chapter could be upgraded from theory to empirics. At the evidentiary level, the aggregate anchor for rescue scope has been verified. The 2022 BoE intervention consisted of gilt purchases (a securities-side instrument); from September 28 to October 14, purchases totaled approximately GBP 19.3 billion, aimed at restoring market functioning and buying resilience time for LDI (BoE 2022). Accordingly, the proposition that which side a rescue instrument acts on is an observable institutional fact holds, and the scope — the securities side — is identifiable. The institutional precondition for the cash-spiral weld point has likewise been verified: margin calls on LDI interest-rate derivatives must be met in cash; to raise cash, gilts are sold, prices fall, and margin calls intensify, constituting the institutional-fact basis for "the synthetic-duration spiral is welded, via central counterparty cash margin, into a cash spiral"; the Dutch counterexample corroborates the criticality of this weld point (DNB 2024). The directional premise of the weak existence proposition of the stock effect has been verified (as an aggregate stylized fact): collectively, owing to recapitalization frictions, gilts were forced into fire sales at discounts of 7% to 10% — the liquidation sequence, forced by collateral availability (which requires cash), meant the cost fell on the assets forced into liquidation (Breeden 2022), supporting the directional claim that "if the liquidation sequence is forced by collateral availability, the real cost falls at least in part on the liquidated assets" (without claiming normative superiority). The calibration basis of this chapter must be made explicit, to explain why the core mechanism diagnosis can only stand as a theoretical proposition rather than an identified causality. This chapter uses the scope of central bank rescue instruments (a known institutional fact) as the yardstick for calibrating the truth or falsity of the "moneyness-rescuability mismatch" — that the 2022 BoE gilt purchases acted on the securities side is a known anchor point, but only when multiple rescue episodes exist, providing contrasting variation between "scope on the cash side" and "scope on the securities side," can this yardstick be used to calibrate the core assertion that "securities-side instruments provide no relief to the cash spiral." A single-point scope is definitional and cannot self-calibrate. This study does not possess such contrasting variation, and the core mechanism diagnosis therefore remains a theoretical proposition. The second calibration basis of this chapter is to use an ex ante computable "should-retain set" (liquidity premium, fundamental quality, correlation with the spiral) as the ex ante yardstick for calibrating the stock-effect prescription — if acting according to the ex ante rule does not predict a superior out-of-sample rebound, then the conditional prescription is calibrated as lacking empirical support, thereby foreclosing the circular self-validation of "ex post wrongful liquidation." Together, these two calibration bases ensure that the theoretical proposition is not disguised as identification by a single-point definitional scope. With respect to pre-registered expectations, this chapter follows the discipline set out below. The upgrade threshold for the moneyness mismatch is written to require scope variation: this chapter does not pre-register that "the moneyness mismatch was already identified in 2022"; its upgrade threshold is that, in a panel of multiple rescue episodes, interventions with "scope on the securities side" provide no relief to the cash spiral while interventions with "scope on the cash side" do provide relief; its unexpected outcome is that if the panel shows securities-side instruments also relieve the cash spiral (e.g., via indirect cash injection through a confidence channel), then the moneyness mismatch is weakened as the locus of pathology, reverting to the alternative mechanism of "buffers or confidence." The existence of the stock effect is pre-registered in falsifiable form: the expectation is that retained assets will systematically underperform the benchmark following the fire-sale window; its unexpected outcome is that if retained assets do not underperform, or even outperform (implying that the assets sold in the fire sale were the low-quality ones), then the stock effect does not dominate, the antecedent of the conditional proposition "holding through is superior" is false, and no normative mandate of "should have held through" may be issued. The should-retain set is accordingly written as an ex ante rule, and this chapter does not pre-register any ex post identification of "wrongful liquidation." As for the execution-stage conclusions of this study, both the scope-variation panel required for the upgrade and the holdings panel are unavailable owing to data limitations; hence neither the causal coefficient of the moneyness mismatch nor the magnitude of underperformance from the stock effect is estimated or fabricated; although the institutional premise that the rescue scope was the securities side has been verified, a single point cannot self-calibrate, and the positioning of the core claim as theoretical is thereby maintained. ### 6.4 The Honest Boundaries of the Evidence This chapter strictly flags its identification boundaries, a point that is especially critical here, because the core mechanism diagnosis of this chapter is itself honestly positioned as a theoretical proposition not identified by the 2022 episode. First, the delimitation of causal identification of the moneyness mismatch. There was only one rescue in 2022, and its scope was definitional (purchasing gilts naturally acts on the gilt market); the counterfactual of "what would have happened had injection been made directly into the cash side" is unobservable, and this assertion is equally consistent with alternative mechanisms such as "buffers too thin," "expectations of forced unwinding," and "information spillovers" — a single-point rescue cannot identify among these mechanisms. The moneyness-rescuability mismatch therefore remains a theoretical-dynamics proposition and inequality, without claiming to have been identified by the 2022 episode. Its upgrade exit is a panel of central bank instruments across rescue episodes, across countries, across time, and across differing scopes (with the coefficient on cash-side scope as the core test statistic); such data are unavailable owing to data limitations, and the coefficient is neither estimated nor fabricated. Second, the delimitation of the ineffectiveness of securities-side rescue. The assertion that "securities-side rescue is ineffective for the cash side" cannot be identified for lack of a counterfactual cash-side rescue, and is therefore downgraded to a theoretical mechanism, not treated as an identified comparison. Third, the delimitation of the stock-effect prescription. The normative prescription that "the decumulation phase should not correct the overweight, being unable to sell is protection, and the real cost is the stock effect" requires a portfolio-level holdings panel to test whether retained assets systematically deteriorate after the fire sale, and its evidentiary basis is limited and thin. It is therefore reformulated as a conditional: if retained assets systematically deteriorate after the fire sale (this must be tested with a holdings panel), then the stock effect dominates and holding through is superior; pending acquisition of the panel, no normative mandate of "the overweight should not have been corrected" is issued. The load-bearing claim retreats to the weak existence proposition of the stock effect, without claiming normative superiority. Fourth, de-circularizing the should-retain set. "An overweight that should have been held through but was wrongly liquidated" can only be defined by whether a rebound subsequently occurred (a rebound justifies saying "it should have been held"; a continued decline means one cannot say "wrongly liquidated"), rendering the proposition unfalsifiably true. To break this circularity, an ex ante criterion must be front-loaded: replacing the ex post label of "wrongful liquidation" with an ex ante computable "should-retain set" (based on liquidity premium, fundamental quality, and correlation with the spiral), using ex ante rules for out-of-sample comparison rather than hindsight. This chapter does not pre-register any ex post identification of "wrongful liquidation." ### 6.5 Relationship to Other Judgments and Summary The judgment of this chapter occupies the layer of mechanism positioning within the argumentative chain. It builds on the vehicle comparison and jurisdictional dichotomy falsification of Chapter 5, sharpening the mechanism down to collateral and the moneyness of margin — but is explicitly flagged as a theoretical-dynamics proposition awaiting external scope variation for identification, and does not reuse the within-UK identification of Chapter 5 to pass itself off as identification of moneyness. This chapter supplies the theoretical path dynamics of the spiral (the cash spiral welded shut), on the basis of which Chapter 7 argues that the independent variable of stress testing must be replaced — from a static base-case endpoint to a path functional; the two chapters share an intuition of path dynamics, but each downgrades a different claim (this chapter downgrades the identification of moneyness; Chapter 7 downgrades the cause of path failure). This chapter's "mismatch between the moneyness of margin and central bank rescuability" directly supplies a core design variable for the subsequent China proposal: if China develops LDI or derivatives hedging, it must front-load the design of a match between the moneyness of margin and the scope of central bank instruments, so as to avoid a similar mismatch (the Dutch contrast of posting part of margin in securities is documented in DNB 2024). In addition, this chapter's decoupling of denominator overweight from liquidity risk runs in the same direction as Chapter 4's proposition that solvency does not equal liquidity — both separate liquidity risk from solvency and allocation ratios. The summary of this chapter is as follows. This chapter has argued that the locus of pathology in the spiral lies in the mismatch between the moneyness of margin and the topology of central bank rescuability. The methodological honesty of this chapter is embodied in the following: the core mechanism diagnosis itself is explicitly positioned as a theoretical-dynamics proposition, rather than an identified causality — because the scope of the single 2022 rescue is definitional and cannot self-calibrate. What is identifiable is the institutional fact of rescue scope and the weak existence proposition of the stock effect; the causal identification of the moneyness mismatch, the ineffectiveness of securities-side rescue, and the normative prescription of holding through the overweight are respectively downgraded to a theoretical proposition, a conditional, and a de-circularized ex ante rule. This chapter provides the mechanism foundation for the path-based redesign of stress testing and for the moneyness design of the China proposal. --- ## Chapter 7 The Independent Variable of Stress Testing: From Static Basis-Point Endpoints to Path, Density, and Correlation Structure ### 7.1 Judgment: A Strict Two-Layer Division This chapter presents the fourth set of judgments in this paper, addressing the methodology of stress testing. The judgment is strictly divided into two layers. The first layer is the methodological layer, which is defensible and belongs to an analytic or axiomatic-level proposition: scenario designs of plus-or-minus 100, 200, or 300 basis points are rigid, and making scenarios more extreme (larger basis-point moves) merely intensifies the same error. Rarity should be defined by a minimum-density region (Mahalanobis distance or highest-density region), not by translation along integer basis points — "more extreme does not mean more basis points" is an analytic, axiomatic-level proposition, and it is the truly robust contribution of this chapter; it only governs how rarity is to be defined. The second layer is the empirical layer, which must be demoted to a hypothesis: "the cause of death in the UK 2022 episode was path rather than magnitude" is a causal proposition about a specific event, and cannot be derived from the first-layer axiom (going from "basis-point endpoints are not a good gauge" to "the cause of death lies in the path" is a category leap, a jump from measurement theory to causal theory, overstepping its proper level) — it requires independent empirical evidence and must therefore be demoted to a hypothesis. The value of this two-layer judgment lies in the following: it flips the negative checklist common in existing discussions — "avoid setting scenarios by subjective intuition" — into an axiomatic foundation for a positive methodological solution: rarity must be defined by a density region, and the prevailing basis-point-endpoint framework has mis-set the gauge (first layer, defensible); while "basis-point endpoints simply cannot capture the cause of death in 2022" is demoted to an empirical hypothesis requiring multi-event panel testing (second layer). ### 7.2 Mechanism: Mis-set Gauges versus the Correct Density-Region Solution; Path Dynamics versus the Cause-of-Death Hypothesis The mechanism of this judgment can be characterized on two layers. The mechanism of the first layer (defensible) is as follows: the prevailing framework defines "extremity" by translation along integer basis points, but under a joint distribution, the probability-density differences among equal basis-point moves can be vast — under correlation structure and fat tails, the same 300 basis points can differ enormously in rarity depending on direction. The correct solution is to define rarity via a minimum-density region (Mahalanobis distance or highest-density region), i.e., "more extreme" should mean "lower density" rather than "larger magnitude." The mechanism of the second layer (demoted to hypothesis) is as follows: mechanistically, path functionals (speed, timing, density) are closer to margin and liquidity spirals than static basis-point endpoints are — a rapid shock compresses the timing of margin calls, thereby triggering a cash spiral (as established in Chapter 6). But its falsifiable version requires a counterfactual path experiment, and 2022 in the real world is a single sample, incapable of identifying whether "path" or "magnitude" was the cause of death. The counterintuitive aspect of this judgment is that it replaces the independent variable of stress testing from "magnitude" with "path, density, and correlation structure" — but what is defensible is the resetting of the methodological gauge (first layer); the attribution of the specific cause of death (second layer) is a hypothesis awaiting testing. The correct counterintuitive posture is to "acknowledge that the gauge was wrong," not to "assert that the cause of death lies in the path." The first and strongest counterargument this judgment must withstand points directly at the risk of overstepping between the first and second layers. This counterargument holds that the chapter's claim that "more extreme does not mean more basis points" is an "axiomatic-level argument." An axiom or definitional proposition (that rarity should be defined by minimum density region rather than integer basis points) does indeed hold, but it only governs how rarity is defined; it cannot yield "path rather than magnitude was the cause of the UK 2022 crash." Going from "basis-point endpoints are not a good gauge of rarity" to "the cause of death lies in the path" is a category leap: the former is a proposition of measurement theory, the latter a causal proposition about a specific event, requiring independent empirical evidence. Lending the certainty of an axiom to a causal conclusion is using an irrefutable definition to endorse a refutable empirical claim. This paper fully accepts this counterargument and accordingly strictly separates the two layers: the first layer (methodology, defensible) has its boundary made explicit as "only governing how rarity is defined," endorsing no cause of death for any specific event; the second layer (empirical, demoted to hypothesis) is separated from the first — "the cause of death in 2022 was path rather than magnitude" must be supported by independent empirical evidence (multi-event panels or intraday data) and cannot be derived from the first-layer axiom. The layers no longer borrow certainty from one another. The second strongest counterargument this judgment must withstand points directly at the sample-size constraint on the second-layer proposition and the causal status of the correlation reversal. This counterargument holds that the falsifiable content of "path dominates magnitude" has two halves: the first half — "same endpoint, different path, different blow-up point" — requires a counterfactual path experiment, but the real-world 2022 episode is a single sample; one can only construct two paths within a model, and a within-model conclusion is a product of the model's specification rather than an identification of reality; the second half — "a complete backtest" — depends on intraday margin-flow data, and is partly demoted to theory. Thus what is highly identifiable is that "the density-region method is mature" (methodology), while what is not identified is that "path was causally more important than magnitude in 2022" (empirical) — treating the availability of a method as the identifiability of a proposition would be an error of identification. Furthermore, the claim that the flip of correlation from negative to positive is the "true culprit" is merely a co-occurrence, in a single event, of correlation reversal and the crash; it cannot rule out the possibility that "the interest-rate shock simultaneously drove both correlation reversal and the crash" (correlation reversal being a common outcome) or that correlation reversal is endogenous to the sell-off (everyone selling stocks and bonds simultaneously drove correlation positive). This paper accepts this counterargument and explicitly states that "the maturity of a method does not equal the testing of a proposition": it separates the availability of the density-region method from the testing of the path-over-magnitude proposition. A feasible empirical test would need to compare, across multiple historical interest-rate shock events (not only 2022), the explanatory power of endpoint magnitude versus path speed for whether a margin spiral is triggered, or to use intraday data to attribute the path in 2022. Until then, path-over-magnitude is presented as a hypothesis, not a tested proposition. The claim that correlation reversal is the "true culprit" is tightened to co-occurrence; "the correlation matrix must be set as a first-class scenario variable" stands independently as a normative recommendation, stated separately from any causal claim. Through the treatment of these two counterarguments, the first-layer contribution of this chapter is robustly preserved, while the second-layer empirical attribution is honestly confined to a hypothesis awaiting testing. ### 7.3 Identification Strategy and Evidence The identification goal of this chapter divides into two layers. The first layer: resetting the rarity gauge of stress testing from translation along integer basis points to a minimum-density region (Mahalanobis distance or highest-density region), and setting the correlation matrix as a first-class scenario variable — this layer is a methodological gauge reset and is defensible. The second layer: designing "path dominates magnitude as the cause of death in 2022" as an empirical hypothesis awaiting testing via multi-event panels or intraday data, without claiming identification at this stage. There are two main paths in the identification strategy. The first main path is minimum-density-region calibration (first layer, highly identifiable): the joint distribution of historical yields and correlations can calibrate the minimum-density region; the method and historical data are both readily available, and this is the defensible core of this chapter. The second main path is setting the correlation matrix as a first-class scenario variable (a normative recommendation, defensible): correlation structure ought to be incorporated in any case, and this stands independently as a normative recommendation, not dependent on causal attribution from a single event. The reserve path is the cause-of-death proposition that path dominates magnitude (only once a multi-event panel or intraday data is obtained): this is an empirical proposition, not yet identified — a feasible empirical test would need to compare, across multiple historical interest-rate shock events (not only 2022), the explanatory power of "endpoint magnitude versus path speed" for whether a margin spiral is triggered (constructing a multi-event panel), or to use intraday data to attribute the path of the 2022 episode. At the evidentiary level, the factual anchors supporting the layered structure have been verified. The fact of correlation reversal is measurable (as fact, not causation): that stock-bond correlation flipped from negative to positive in 2022 (a large-scale simultaneous sell-off of stocks and bonds to meet margin calls) is an observable fact — the fact itself is measurable, but a single event's correlation reversal does not equal causation; this chapter has tightened it to "co-occurring, and theoretically amplifying the spiral." The institutional precondition for path and speed acting as amplifiers has been verified (as a stylized fact): the speed and scale of yield movements far exceeded the capacity of smaller investors in pooled funds to top up margin (typically only a one- or, for a few, two-week rebalancing window), which is precisely the direct mechanism by which pooled vehicles turned into forced sellers; 30-year gilt yields rose more than 100 basis points within four days, and jumped about 140 basis points in a single day after the budget announcement (SUERF 2023; Breeden 2022), supporting the motivation for the second-layer mechanism, but this chapter has clearly flagged this as a hypothesis awaiting testing, not an already-identified cause of death. The methodological premise of the first layer has been verified (as a methodological fact): the minimum-density region is calibrated using the joint distribution of historical yields and correlations for rarity, and the method and data are mature and available — this is an analytic, axiomatic-level contribution, only governing how rarity is defined, and does not depend on single-event causation. The calibration basis of this chapter must be made explicit. At the first layer, this chapter uses the joint distribution of historical yields and correlations as a known yardstick to calibrate the authenticity of the "rarity gauge" — the minimum-density region uses the historical joint distribution to test whether "equal basis-point moves in different directions or correlation regimes have significantly different densities"; if the historical joint distribution is approximately isotropic (equal basis points roughly equal to equal density), then the density region adds no increment relative to basis-point translation, and the first-layer contribution is calibrated as a technical refinement rather than a gauge correction. At the second layer, this chapter uses a panel of multiple historical interest-rate shock events as an out-of-sample yardstick to calibrate the cause-of-death hypothesis — only by comparing the explanatory power of magnitude versus path across multiple events can the authenticity of a single-point attribution for 2022 be calibrated (a single sample cannot self-calibrate; a path constructed within a model is a product of its specification). These two calibration bases strictly separate "method maturity" from "proposition tested," precluding the certainty of an axiom from being lent to an empirical claim. As to pre-registered expectations: the first layer of this chapter is pre-registered on a weak-falsification grid (equal basis-point moves showing significantly different densities in different directions), with the unexpected outcome being that if the historical joint distribution is approximately isotropic, the first layer shrinks to a technical refinement; the second layer is written as a hypothesis requiring a sample size greater than one (across a panel of multiple historical shock events, the explanatory power of path speed for triggering a spiral is significantly higher than that of endpoint magnitude), with the unexpected outcome being that if the panel shows magnitude's explanatory power is not lower than path speed's, the hypothesis is falsified and reverts to "magnitude remains the primary gauge, path a second-order correction"; correlation reversal is written down as co-occurrence, not causation. As to the execution conclusions of this study: because the historical joint distribution was not actually computed, the density differences of the first layer are reported with direction known but specific values not compiled; the coefficients of the second layer's multi-event panel are neither estimated nor compiled owing to data limitations; and the lead-lag test for correlation reversal was likewise not executed owing to the unavailability of intraday data. ### 7.4 The Honest Boundaries of the Evidence This chapter strictly flags its identification boundaries, the core of which is to strictly separate "method maturity" from "proposition tested." First, the delimitation of the path cause-of-death proposition. The falsifiable version of "path dominates magnitude as the cause of death in 2022" requires either a counterfactual path experiment or intraday margin-flow data; the real-world 2022 episode is a single sample, and the two paths constructed within a model are a product of the model's specification, not an identification of reality. It is therefore maintained as an empirical hypothesis. The coefficients of a multi-event panel regression (comparing the explanatory power of the path-speed coefficient and the endpoint-magnitude coefficient for triggering a spiral) are uniformly neither estimated nor compiled; the identification design (a panel of multiple historical shock events or intraday path attribution) is in place, but the data are unavailable owing to limitations. Second, the delimitation of correlation reversal as "true culprit." Treating the flip of correlation from negative to positive as the "true culprit" treats a single event's correlation as causation, and cannot rule out the interest-rate shock simultaneously driving both correlation reversal and the crash (correlation reversal being a common outcome), or correlation reversal being endogenous to the sell-off (everyone selling stocks and bonds simultaneously driving correlation positive). It is therefore tightened to co-occurrence, not treated as causal identification; data for an intraday lead-lag test are unavailable. "The correlation matrix must be set as a first-class scenario variable" stands independently as a normative recommendation, stated separately from any causal claim. Third, the quantitative delimitation of the first layer's anisotropy test. Although the method for testing density differences of equal basis-point moves in different directions is available, this study did not actually compute it using historical joint-distribution data; the first layer is therefore reported as methodologically defensible with direction known, but specific density-difference values are not compiled. Correspondingly, one unexpected outcome of the first layer is: if the historical joint distribution is approximately isotropic (equal basis points roughly equal to equal density), then the density region adds no increment relative to basis-point translation, and the first-layer contribution shrinks to a technical refinement rather than a gauge correction — this determination can only be made through actual computation, which this study leaves open and unresolved. It must be specially emphasized that there is a boundary discipline between the layers: the second layer is not endorsed by the first-layer axiom. An axiom or definitional proposition (that rarity should be defined by minimum density region rather than integer basis points) does indeed hold, but it only governs how rarity is defined, and cannot yield "path rather than magnitude was the cause of the UK 2022 crash." Going from "basis-point endpoints are not a good gauge of rarity" to "the cause of death lies in the path" is a category leap: the former is a proposition of measurement theory, the latter a causal proposition about a specific event, requiring independent empirical evidence. Lending the certainty of an axiom to a causal conclusion is using an irrefutable definition to endorse a refutable empirical claim — this chapter strictly refuses this overstepping. ### 7.5 Relationship to Other Judgments and Summary The judgment of this chapter occupies the level of measurement correction within the paper's chain of argument. It builds on the spiral path dynamics of Chapter 6 (the cash spiral welded shut), on that basis arguing that the independent variable of stress testing must be replaced with path functionals — but while this chapter builds on Chapter 6's path intuition, the two are demoted along different dimensions: Chapter 6 demotes the identification of moneyness, while this chapter demotes the attribution of the cause of death via path; this chapter does not borrow Chapter 6's dynamics to endorse the second-layer causal claim. This chapter's methodological gauge reset provides Chapter 8 with a technical interface for "the design of stress testing itself is also a choice of accountability" — how rarity is defined is a matter of who defines tolerable risk. This chapter's "defining rarity via a density region" provides, for the subsequent China proposal, a methodologically correct solution for the design of stress-testing frameworks (avoiding the setting of basis-point scenarios by subjective intuition), and is a directly transplantable methodological layer (as distinct from the institutional layer emphasized in Chapter 8 as non-transplantable). The summary of this chapter is as follows. This chapter has argued that the independent variable of stress testing must be reset from static basis-point endpoints to path, density, and correlation structure. The chapter is strictly divided into two layers: the first-layer methodological gauge reset is an analytic, axiomatic-level contribution, defensible and not to be demoted; the second-layer proposition that "path dominates magnitude as the cause of death in 2022" is a causal proposition about a specific event, and, being unidentifiable from a single sample, is demoted to an empirical hypothesis awaiting testing via multi-event panels or intraday data, while the "true culprit" claim about correlation reversal is tightened to co-occurrence. The core honest boundary of this chapter is its refusal to lend the certainty of the first-layer axiom to the second layer's empirical claim, i.e., a strict distinction between "method maturity" and "proposition tested." This layering is consistent throughout with the philosophy running through this entire paper of "strictly separating what is defensible from what is not." --- ## Chapter 8 Risk Localization as an Institutionally Endogenous Allocation of Blame ### 8.1 Judgment: A Comparative Proposition under Pre-Bound Mapping This chapter presents the fifth set of judgments of this paper, which is also the meta-judgment of the entire piece, and simultaneously the direct correct answer to the tendency to "copy best practices wholesale." The judgment can be stated as follows: we no longer assert that "risk localization is always an institutionally endogenous allocation of blame" (this universal proposition excludes no observation and is true in all possible worlds; it is a framework-level restatement rather than an empirical finding), but instead adopt a comparative proposition that pre-binds institutional variables to blame-allocation dimensions. The content of this comparative proposition is as follows: given institutional variable X (bankruptcy-law priority, clearinghouse ownership, central-bank mandate boundaries), one can predict that a given country will localize risk in dimension Y, and that transplanting a practice Y′ that differs from the country's own Y will fail on specific observable indicators; the key lies in allowing successful transplantation to falsify the proposition — if transplantation succeeds between countries with highly similar institutional variable X, the proposition is supported; if transplantation also succeeds between institutionally disparate countries, the proposition is undermined. This is the highest-dimensional judgment of the entire paper. It flips the whole chapter from "how to improve resilience" to "improvement itself is the political economy of blame allocation," but it collapses the highest-dimensional judgment from an unfalsifiable universal ontological claim into a comparative proposition that pre-binds the mapping and allows falsification via successful transplantation. This is also the only direct correct answer to "copying best practices wholesale," and it provides the judgment increment supplying the premise of "non-transplantable, must be endogenously designed" for the China proposal that follows. ### 8.2 Mechanism: From Institutional Variables to Blame-Allocation Dimensions to Transplant Failure The mechanism chain of this judgment has already been expounded in the theoretical framework of Chapter 3; here it is refined in conjunction with the identification strategy. Institutional variable X (bankruptcy-law priority, clearinghouse ownership, central-bank mandate boundaries) determines that a given country locates the blame boundary for decumulation (payout) phase risk in dimension Y (the UK locates it in the liquidity of pension vehicles; other countries may locate it elsewhere). Transplanting a practice Y′ that differs from the country's own Y (copying the UK's buffers, backstops, or leverage constraints wholesale into a country with a different institutional variable X) will fail on specific observable indicators (trigger-point yields, deleveraging speed, contagion share). But if transplantation succeeds between countries with highly similar institutional variable X, the proposition is supported; if transplantation also succeeds between institutionally disparate countries, the proposition is undermined — this design of "allowing successful transplantation to falsify" is the key feature distinguishing this judgment from a universal ontological claim. The counterintuitive aspect of this mechanism is that it reconstructs "better regulation, thicker buffers, more reliable backstops" as "redistribution of the blame boundary" — buffers do not eliminate risk but shift it from one party bearing blame to another (a shift rather than an elimination, but one that first requires a distinguishable criterion to bear the weight). Best practices cannot be copied wholesale precisely because they encode the source country's institutional blame-allocation choice X; transplanting them into a country with a different institutional variable X will fail on observable indicators. The mechanism by which the three institutional variables act on the blame-allocation dimension can be characterized in finer detail. First, bankruptcy-law priority. It determines the order in which different stakeholders (beneficiaries, sponsors, guarantee institutions, the public sector) are repaid when a scheme becomes insolvent or a sponsor defaults, thereby determining, at the most fundamental level, who ultimately absorbs decumulation-phase risk. A bankruptcy law that places beneficiaries in a priority repayment position, versus one that places a guarantee institution or the public sector in a backstop position, leads to entirely different blame boundaries and, in turn, entirely different incentives for risk governance. Second, the ownership structure of clearinghouses. Member-owned clearinghouses and commercialized clearinghouses follow different logics in setting margin rules, in the transmission of liquidity, and in loss-sharing during a crisis; whether a clearinghouse requires cash margin, whether it accepts securities as collateral, and how its margin model adjusts with volatility all directly shape whether a spiral will become self-reinforcing (as discussed in Chapter 6). Third, central-bank mandate boundaries. These determine which end of the financial system public backstops can act upon, which institutions can receive support, and the term and conditions of that support. As discussed in Chapter 2, the issuer of sterling can place part of the blame boundary on central-bank backstops, whereas a non-issuer cannot — this difference is precisely a direct consequence of the institutional variable of central-bank mandate boundaries. These three institutional variables jointly determine where a country places the blame boundary for decumulation-phase risk, and also determine on which observable indicators best practices originating from that country will fail when transplanted. The first and strongest counterargument that this judgment must withstand goes directly to the unfalsifiability of the universal proposition. This counterargument points out: the proposition form "risk localization is always an institutionally endogenous allocation of blame" is a universal proposition that excludes no observation — no matter where a country locates risk, it can always be restated after the fact as "its institutional structure endogenously chose this blame boundary." When a proposition is true in all possible worlds and no observation can falsify it, it is not an empirical finding but a framework-level restatement. The falsifiability point offered — "best practices fail after transplantation" — is extremely weak (transplant failure has many alternative explanations: implementation drift, timing, unrelated shocks, insufficient dosage), and cases of successful transplantation are not addressed as to whether they would falsify it. This paper fully accepts this counterargument and, accordingly, narrows the proposition into a comparative proposition with observational implications — given institutional variable X, one can predict blame-allocation dimension Y, and transplanting Y′ will fail on specific observable indicators, with the key being that successful transplantation is allowed to falsify it (if transplantation succeeds between countries with high X-similarity, the proposition is supported; if transplantation also succeeds between institutionally disparate countries, the proposition is undermined). Importantly, this paper carries this narrowing through to the highest-dimensional level of statement: the paper's one-sentence thesis and the chapter title's main clause are both downgraded synchronously with the proposition itself, each carrying its own falsifiability boundary, rather than relying on a subordinate clause to retroactively qualify it — this is to prevent the downgraded universal claim from quietly reviving at the highest load-bearing position. The truly load-bearing paths for identification (the comparative proposition, conditional slopes, clearing topology) have all been correctly narrowed, with none bearing identification in universal form. The second and strongest counterargument that this judgment must withstand goes directly to the bidirectional self-confirmation of buffer displacement, the counterfactual attribution of sunset backstop commitments (sunset here refers to an explicit, pre-announced expiration commitment by the central bank to exit at a set date; hereafter referred to as "sunset commitment"), and the data dependency of backstop-endogenization topology. This counterargument points out: that "buffers only displace, never eliminate" risk confirms the proposition under either outcome is a hallmark of unfalsifiability; that "the sunset commitment succeeded only because this particular expiration happened to occur in a non-fragile state, not because the commitment itself deterred" is a counterfactual attribution that treats an unobservable counterfactual as an established fact in order to deny deterrence; that "central-bank backstops endogenize the next round's contagion topology" is a strong structural proposition about future or counterfactual topology, but the pension and money-market-fund look-through shares needed for identification are admittedly stuck for lack of data, and absent such shares, "same-entity short-circuiting causing a contagion conduit" is a theoretical-network inference that cannot be distinguished from "the two markets merely happen to share participants but contagion was never realized." This paper accepts this counterargument and makes three corresponding adjustments: first, buffer displacement is downgraded to a conditional-slope hypothesis, giving "displacement versus elimination" a distinguishable observational criterion (the slope and curvature of trigger-point yield as a function of the buffer), with the commitment that if trigger probability approaches zero or exhibits a structural break as the buffer increases, "non-elimination" is refuted; second, the sunset commitment is split into two layers — the identifiable layer is the pricing evidence of the dynamic inconsistency of the announcement, while the unidentifiable layer ("it would have failed if fragile; success was mere luck") is flagged as a model-inferred counterfactual, replaced with the conditional proposition that "commitment credibility depends on the fragility state at expiration," instead of an outright denial of deterrence; third, backstop-endogenization topology is downgraded to a theoretical-network proposition, retreating to a currently measurable existence proposition (pensions being simultaneously a seller of government bonds and a supplier of repo cash). Through the treatment of these two counterarguments, the meta-judgment of this chapter is narrowed into a comparative proposition that allows falsification via successful transplantation and whose load-bearing path rests on hard identification design, rather than an unfalsifiable universal ontological claim. ### 8.3 Identification Strategy and Evidence The identification target of this chapter is to identify the comparative proposition that pre-binds institutional variables to blame-allocation dimensions (allowing falsification via successful transplantation), with empirical weight preferentially placed on hard identification paths (the quasi-natural experiment of the phased rollout of clearing mandates, and the regression discontinuity at eligibility thresholds). Propositions such as the universal ontological claim, "buffers never eliminate risk," and "backstop endogenization of the next round's topology" are not identification targets and are respectively narrowed to a comparative proposition, a conditional-slope hypothesis, and a theoretical-network proposition. The first hard identification path of the identification strategy is the phased rollout of clearing mandates as a quasi-natural experiment. The phased rollout of clearing mandates (such as mandatory central clearing) is a genuinely exogenous variation and offers the hardest identification — using difference-in-differences or event studies organized around rollout timing or batches, with the treatment group being instruments or institutions already brought under mandatory clearing and the control group being those not yet included. Its identifying assumption is that the timing of the phases is determined exogenously by the regulatory calendar, independent of any individual institution's contemporaneous risk. This path is elevated to the path bearing priority empirical weight in this chapter, to compensate for the downgrading of other paths. The second hard identification path is regression discontinuity at eligibility thresholds: the discontinuity at eligibility or qualification thresholds constitutes a regression discontinuity design — the discontinuous jump in blame localization or risk-bearing between institutions on either side of the threshold identifies the causal effect of the institutional variable on localization. In addition, the third path is the comparative proposition of institutional variables to blame-allocation dimensions (which requires pre-binding X to Y and allowing falsification via successful transplantation, rather than a universal ontological claim), and the backup path is the pricing evidence from the exit commitment of sunset backstop instruments (high-frequency price responses in the announcement window can be identified on the ground, but counterfactual attribution cannot be identified). At the evidentiary level, the aggregated anchors of institutional facts have been verified. The institutional facts of eligibility thresholds have been verified: the BoE's CNRF opened for applications in January 2025, limited to insurers, DB pension schemes, and LDI funds, requiring eligibility thresholds such as holding more than £2 billion in gilts, activated only when the BoE judges that market dysfunction threatens financial stability, exchanging gilts for cash (BoE 2025). On this basis, the institutional premise of "the discontinuity at eligibility or qualification thresholds constitutes a regression discontinuity, with a jump in blame localization between institutions on either side of the threshold" (the existence of a genuine £2 billion threshold) has been verified. The empirical premise of buffer displacement has been verified: the TPR required in April 2023 that leveraged LDI funds hold a minimum of 250 basis points of market-stress resilience buffer in addition to operational buffers (assuming replenishment capacity within five days) (TPR 2023), while Ireland and Luxembourg required at least 300 basis points of UK-yield resilience for sterling LDI funds (ESMA 2022); on this basis, the existence and direction of rightward shift of "buffers set a trigger point that shifts rightward to a higher yield" has been verified. The institutional premise of sunset-commitment pricing evidence has been verified: the BoE's 2022 gilt purchases had a strict sunset (explicitly framed as a backstop, "not permanent," to buy time for LDI), terminating on 14 October and beginning resale on 29 November (BoE 2022); on this basis, the institutional premise of "the dynamic inconsistency of the sunset-commitment announcement can be priced via deleveraging speed" (the commitment is indeed time-limited and carries an inherent credibility gap) has been verified. The existence of institutional variable X has been verified: the very existence of the CNRF is itself an expansion of the central-bank mandate boundary X, consistent with the UK's Y of "locating the blame boundary in the liquidity of pension vehicles plus central-bank backstops"; the cross-jurisdictional difference in X (Ireland and Luxembourg imposing a 300-basis-point buffer rather than the UK's combination of 250 basis points plus the CNRF) corroborates that "the blame-allocation dimension diverges with institutional variable X" (ESMA 2022; BoE 2025). The calibration basis of this chapter must be made explicit. This chapter uses cases of successful transplantation as a bilateral yardstick to calibrate the truth or falsity of the mapping from institutional variables to blame-allocation dimensions — only by allowing "successful transplantation" as the anchor of falsification (if transplantation succeeds between countries with high X-similarity, the proposition is supported; if transplantation also succeeds between institutionally disparate countries, the proposition is undermined) can the pre-bound mapping calibrate the universal ontological claim into a comparative proposition with observational implications (the universal version lacks this yardstick and is true in all worlds, hence uncalibratable). The second calibration basis of this chapter is to use the timing of the phased rollout of clearing mandates as a known exogenous anchor — the timing of the phases is determined by the regulatory schedule, predates the 2022 crisis, and is independent of any individual institution's contemporaneous risk, making it a genuinely exogenous yardstick, against which "whether institutional variables move the blame boundary" is calibrated (the regression discontinuity at eligibility thresholds belongs to the same category of hard identification anchor). These two calibration bases together ensure that the meta-judgment does not pass itself off as identification in universal form or through ex post restatement. With respect to pre-registered expectations, this chapter observes the following discipline. The comparative proposition is pre-registered in a falsifiable bilateral format: the pre-bound mapping predicts that transplantation of Y succeeds between countries with high X-similarity and that transplantation of Y′ fails on specific indicators between institutionally disparate countries; its unexpected outcome would be that if both "transplantation fails between countries with high X-similarity" and "transplantation also succeeds between institutionally disparate countries" occur simultaneously, the mapping is falsified and the proposition undermined, reverting to "risk localization is determined by technology rather than institutions" — this is precisely the falsifiability boundary carried by the chapter title itself. The phased rollout of clearing is pre-registered in a falsifiable hard-identification format: it is expected that instruments brought under mandatory clearing will show a significant jump in blame localization or liquidity-bearing relative to the control group; its unexpected outcome would be that if there is no difference before and after the phased rollout and the clearing mandate does not move the blame boundary, the causal path from institutional variable to localization is weakened. Buffer displacement is written as a distinguishable format: this chapter does not pre-register the universal negation "buffers never eliminate risk," but only tests the slope and curvature of trigger-point yield as a function of the buffer, with the commitment that if trigger probability approaches zero or exhibits a structural break as the buffer increases, "non-elimination" is refuted. The sunset commitment is written as a conditional proposition, and backstop endogenization is written as an existence proposition. As for the execution conclusions of this study, the bilateral falsification of the comparative proposition, the difference-in-differences coefficient of the phased clearing rollout, and the regression discontinuity coefficient of eligibility thresholds are all unobtainable due to limited cross-country comparable outcome datasets and institution-level panels; they are uniformly neither estimated nor fabricated. Although the institutional premise of the existence of differences in X has been verified, bilateral falsification is not asserted as identified, and the falsifiability boundary carried by the chapter title is thereby maintained. ### 8.4 The Honest Boundaries of the Evidence This chapter strictly demarcates its identification boundaries, the core of which is to rigorously distinguish the existence of institutional facts from the identification of causal coefficients. First, the delineation of the difference-in-differences coefficient for the phased rollout of clearing. The identification design is the hardest (the timing of the phases is exogenous), but the difference-in-differences coefficient for "blame localization or liquidity-bearing interacted with inclusion in mandatory clearing and its aftermath" requires an institution-by-time panel, and such data are limited and unobtainable; the coefficient is uniformly neither estimated nor fabricated. The institutional fact of the phased rollout exists (such as the batch-wise onboarding of the CNRF), but this does not substitute for the causal coefficient. Second, the delineation of the regression discontinuity coefficient for eligibility thresholds. The £2 billion gilt threshold genuinely exists, but estimating the regression discontinuity of "a jump in blame localization" between institutions on either side of the threshold requires institution-level running-variable data, which is neither estimated nor fabricated. Third, the delineation of the bilateral falsification test for the comparative proposition. The bilateral falsification test of the comparative proposition requires a cross-country comparable-outcome dataset of "successful transplantation among countries with high X-similarity versus failed transplantation among institutionally disparate countries," and such data are limited and unobtainable; at the aggregate level, the existence of differences in X can be confirmed (the UK's institutional combination differs from that of Ireland and Luxembourg), but a systematic comparison of "success or failure of transplantation" is not asserted as identified. Fourth, the delineation of the slope and curvature of buffer displacement. That "buffers only displace, never eliminate" risk confirms the proposition under either outcome — if the buffer is thickened and no failure occurs, one says "the trigger point was shifted to a higher yield, it just didn't reach that level this time"; if the buffer is thickened and failure still occurs, one says "the buffer only displaced, it never eliminated" — both outcomes confirming the claim is a hallmark of unfalsifiability. To break this circularity, "displacement versus elimination" must be given a distinguishable observational criterion: using trigger-point yield as a function of the buffer and testing its slope and curvature (a linear rightward shift indicates displacement; the existence of a buffer threshold at which trigger probability collapses structurally indicates elimination), which can be distinguished using scenario data. This chapter commits: if trigger probability is observed to approach zero or exhibit a structural break as the buffer increases, "non-elimination" is refuted. The criterion of slope and curvature of trigger-point yield as a function of the buffer requires scenario data, which are limited and unobtainable; it is therefore maintained as a conditional-slope hypothesis, without asserting the universal negation of "never eliminated," nor empirically establishing "structural collapse." The buffer facts (250 basis points and 300 basis points) support only the existence and rightward shift of the trigger point, not the universal negation of "never eliminated" (TPR 2023; ESMA 2022). Fifth, the delineation of "sunset commitment succeeded only through luck" and backstop-endogenization topology. That "the sunset commitment succeeded only because this particular expiration happened to occur in a non-fragile state, not because of commitment deterrence" is a counterfactual attribution — the fact that the expiration was not fragile precisely renders "commitment deterrence versus luck" inseparable (fragile-expiration performance was never observed), treating an unobservable counterfactual as an established fact in order to deny deterrence is over-speculation about a counterfactual that never occurred. Hence the sunset commitment is split into two layers: the identifiable layer (retained) is the pricing evidence of the dynamic inconsistency of the sunset-commitment announcement (the high-frequency relationship between deleveraging speed and the credibility gap); the unidentifiable layer (downgraded to discussion) is "it would have failed if fragile, success was mere luck," flagged as a model-inferred counterfactual, replaced with the conditional proposition that "commitment credibility depends on the fragility state at expiration," instead of an outright denial of deterrence. "Central-bank backstops endogenizing the next round's contagion topology" is a strong structural proposition about future or counterfactual topology, but the pension and money-market-fund look-through shares required for identification are limited and unobtainable, so it is downgraded to a theoretical-network proposition, retreating to a currently measurable existence proposition — that pensions are simultaneously marginal participants in both the government-bond market and the repo market (verifiable if the shares can be partially disclosed). This existence proposition has verified support: Dutch pension funds simultaneously sold both equities and bonds while replenishing margin, and UK pension funds were simultaneously sellers of gilts (PPF 2022; IPE 2022), but the shares and topology are not estimated. ### 8.5 Relationship to Other Judgments and Summary The judgment of this chapter sits at the highest institutionally endogenous level in the argumentative chain, serving as the capstone of the entire paper's progressive chain. If the corrections of Chapters 4 through 7 were written as merely "thicken the buffer, copy best practices," they would fall back into shallowness; this chapter elevates the corrections to the level of an institutional-endogeneity judgment — Chapter 7's "stress-test design is itself a question of who defines rarity" is a technical sub-case of this chapter's "blame allocation." This chapter elevates the technical corrections of the preceding four chapters to a meta-judgment of political economy, but narrows it into a falsifiable comparative proposition (without swallowing up the identifiable legs of the preceding four chapters). This chapter is the only judgment that directly provides the premise of "non-transplantable, must be endogenously designed" for the China proposal that follows: the correct answer to "copying best practices wholesale" is that best practices encode the source country's blame-allocation choice X, and since China's X differs, endogenous design rather than wholesale copying is required. This chapter's "the comparative proposition is not universal" complements rather than contradicts Chapter 5's "cross-country decomposition is unidentifiable, with the Netherlands and Denmark as cases" — Chapter 5 cautiously flags that cross-country decomposition is unidentifiable, whereas this chapter elevates cross-country differences into a falsifiable comparative proposition, using hard identification paths (the phased rollout of clearing and eligibility thresholds) rather than the three-country decomposition on which Chapter 5 was stuck. The summary of this chapter is as follows. This chapter has argued that risk localization, in the sense of a mapping that pre-binds institutional variables to blame-allocation dimensions, is an institutionally endogenous allocation of blame. This chapter flips the entire chapter from "how to improve resilience" to "improvement itself is the political economy of blame allocation," but by narrowing it into a comparative proposition that allows falsification via successful transplantation, it avoids the unfalsifiability of a universal ontological claim. Empirical weight is preferentially placed on the institutional premises of the two hard identification paths of phased clearing rollout and eligibility thresholds (verified), while the causal coefficients, the bilateral falsification of the comparative proposition, the slope of buffer displacement, the counterfactual attribution of the sunset commitment, and the topological shares of backstop endogenization are each correspondingly downgraded. This chapter provides the fundamental non-transplantable premise for the China proposal, and is the final destination of the entire paper's corrective logic. --- ## Chapter 9 Cross-Chapter Synthesis: The Progressive Logic and Internal Consistency of the Five Sets of Judgments ### 9.1 The Structure of the Progressive Chain The five sets of judgments in this paper are not parallel but constitute a progressive chain that deepens layer by layer. The structure of this chain can be understood as a stepwise descent through "levels of misdiagnosis." Chapter 4 operates at the level of event misdiagnosis. It corrects the most superficial misjudgment of the nature of the 2022 event—the conflation of a solvency crisis with a liquidity crisis. Its core contribution is establishing a strict distinction between the funding line and the liquidity line, and pointing out that an improvement in the funding line does not constitute a safety signal for the liquidity line. The judgment at this level is negative and relatively weak, but it clears the conceptual ground for all subsequent analysis: without this distinction, later discussions of mechanism, measurement, and institutions would all be built on a conflation. Chapter 5 operates at the level of cross-sectional misuse. It corrects the methodological error of reading institutional superiority or inferiority from cross-country cross-sectional outcomes. After Chapter 4 established that "improved coverage does not equal safety," a natural follow-up question arises: "then are the Dutch and Danish systems, which did not blow up, institutionally superior?" Chapter 5 points out that this question itself presupposes an untenable inference—survivorship bias is a symmetric knife, and the non-occurrence of a blowup in a single event-year provides no information about institutional superiority or inferiority. This level shifts the identifiable load-bearing weight from "who did not blow up" to "the manifestation of structural differences where exogenous variation exists." Chapter 6 operates at the level of mechanism identification. Having cleared away, in the preceding two chapters, the nature of the event and the methodology of comparison respectively, Chapter 6 asks where the mechanistic lesion of the spiral lies. It pinpoints the lesion to the mismatch between the moneyness of margin and the topology of bailout-ability, but honestly locates this core mechanistic diagnosis as a theoretical-dynamics proposition—because the scope of the single 2022 rescue is definitional and cannot be self-calibrating. This level provides the pivot for understanding the spiral's self-reinforcement and lays the dynamical foundation for the measurement reset of the next level. Chapter 7 operates at the level of measurement correction. After Chapter 6 established the path dynamics of the spiral, Chapter 7 asks: given that the spiral is driven by path dynamics, should the independent variable of stress testing be switched from a static basis-point endpoint to a path functional? Its answer is strictly stratified into two layers—the methodological recalibration (defining rarity via the density domain) is defensible, while the empirical attribution that "the path was the cause of death in 2022" must be downgraded to a hypothesis. This level converts the mechanistic insight of the preceding chapters into a methodology for stress-test design. Chapter 8 operates at the level of institutional endogeneity. This is where the progressive chain closes. After the preceding four chapters complete the technical correction from event to mechanism to measurement, Chapter 8 raises a meta-judgment: all these improvements (buffers, backstops, leverage constraints, stress-test design) are themselves choices about the allocation of accountability, and risk location is institutionally endogenous. It elevates the technical correction to political economy, but by narrowing it to a falsifiable comparative proposition, it avoids sliding into an unfalsifiable universal ontology. ### 9.2 Verification of Internal Consistency The five sets of judgments are not only progressive but also mutually consistent, with no conflicts in load-bearing weight. Their consistency is verified one by one below. First, Chapter 4's negative weak-causal claim and Chapter 8's endogenous-accountability comparative proposition do not conflict. Chapter 4 operates at the level of event misdiagnosis (the conflation of solvency and liquidity), while Chapter 8 operates at the level of institutional endogeneity (the institutional determination of accountability boundaries); the two represent different levels of progression rather than competing claims at the same level. Chapter 4 identifies the factual-level distinction that "solvency does not equal liquidity," while Chapter 8 identifies the institutional origin behind this distinction; the two are complementary. Second, Chapter 5's "cross-country non-identifiability" and Chapter 8's "comparative proposition that permits falsification via successful transplantation" are complementary rather than contradictory. Chapter 5 cautiously notes that the cross-country decomposition of leverage hedging is not identifiable within a single event-year (three units, a single treatment period, no degrees of freedom), and treats the Netherlands and Denmark only as motivating cases; Chapter 8, by contrast, elevates cross-country differences into a falsifiable comparative proposition, but its empirical weight rests on hard identification designs (difference-in-differences across phased clearing implementation and regression discontinuity around eligibility thresholds), not on the three-country decomposition on which Chapter 5 was stuck. What Chapter 8 undertakes is precisely the falsifiable structure that Chapter 5 cautiously flagged; the division of labor between the two is clear. Third, Chapter 6's "decoupling of denominator overallocation from liquidity risk" and Chapter 4's "solvency does not equal liquidity" point in the same direction. Both separate liquidity risk from solvency and from the allocation ratio of assets: Chapter 4 separates it along the solvency dimension, Chapter 6 along the asset-allocation-ratio dimension, jointly pointing to the conclusion that "liquidity risk has dynamics independent of both solvency and allocation." Fourth, Chapters 6 and 7 share the intuition of path dynamics but each downgrade a different assertion. Chapter 6 provides the theoretical path dynamics of the spiral (the cash spiral being welded shut), downgrading the identification of moneyness; Chapter 7, building on this, argues for a path-based reset of the stress-test independent variable, downgrading the attribution of the path as cause of death. Chapter 7 does not borrow Chapter 6's dynamics to endorse its own second-layer causal assertion; the downgrading boundary between the two is clear. Fifth, Chapter 7's methodological stratification is consistent with the honest-boundary philosophy running through the entire paper. Chapter 7 strictly stratifies what is defensible (the methodological recalibration of scale) from what is not (empirical causal attribution), refusing to lend the certainty of an axiom to an empirical assertion—this stance is itself a concentrated embodiment of the paper's overall discipline of stratified load-bearing. ### 9.3 The Systematicity of the Honest Boundary The honest boundary running through the five chapters constitutes a systematic feature of this paper's methodology, worth summarizing at the synthesis level. For every set of judgments, this paper clearly distinguishes its identifiable load-bearing path from its downgraded theoretical or discursive layer. The identifiable load-bearing paths include: the pooled and segregated quasi-natural comparisons shared by Chapters 4 and 5, and the difference-in-differences across phased clearing implementation and the regression discontinuity around eligibility thresholds in Chapter 8. These designs are complete at the design level, with clear sources of exogenous variation. By contrast, the causal magnitudes that can only be identified with scheme-level micro panel data—the slope ratio in Chapter 4, the difference-in-differences magnitude in Chapter 5, the causal coefficient of moneyness mismatch in Chapter 6, the path-as-cause-of-death claim in Chapter 7, and the difference-in-differences and regression-discontinuity coefficients for accountability location in Chapter 8—are uniformly maintained at a downgraded status because of micro-data constraints. Their upgrade pathways (variation in scheme-level collateral architecture, panels across multiple rescue scopes, panels across multiple historical shock events, institution-level accountability panels) are each spelled out, and all remain unexecuted due to data unavailability rather than any design flaw. Particularly worth emphasizing are two boundaries of "a single point cannot self-calibrate." The scope of the single 2022 rescue in Chapter 6 is definitional (purchasing government bonds naturally acts on the government-bond market), and cannot be used to self-calibrate the core assertion that "securities-based instruments provide no relief for the cash spiral"; the single 2022 sample in Chapter 7 cannot be used to self-calibrate the cause-of-death attribution that "path dominates magnitude." These two boundaries are strictly maintained, with no single point masquerading as identification. The honest boundary of this paper is also reflected in the bidirectionality of the downgrading direction. The pre-registered "unexpected result" for each set of judgments (i.e., the result that, if the evidence points toward it, would falsify the judgment and require an honest downgrade) is preserved in every case, and no downgraded proposition has been quietly upgraded to "identified" merely because the aggregate evidence "appears supportive." Consistency in aggregate direction is reported only as "directional consistency" or "premise holds," which is not equivalent to micro-level causal identification. Moreover, the contrary or boundary evidence retrieved during verification (such as the Danish regulator's note that some firms still carry "non-negligible liquidity risk," or the fact that segregated mandates, despite hedging 85% of liabilities, nonetheless performed well) is consistent with the honest boundaries of each chapter and overturns none of the core judgments—rather, it supports Chapter 5's conclusion that "survival does not equal absence of risk, a symmetric knife" and Chapter 4's conclusion that "vehicle structure, rather than hedge ratio alone, drove earlier position closures." ### 9.4 The Generalizability of the Methodological Contribution The technical content of the five sets of judgments in this paper each addresses a specific problem of the pension decumulation phase, but the methodological stance running through them has generalizability beyond this specific context, worth distilling at the synthesis level. The first generalizable methodological contribution is the strict delineation of "the boundary of causal identification in single-event studies." The 2022 pension crisis, and more broadly any unique historical event (a financial crisis, a policy shock, an institutional transformation), faces the fundamental methodological constraint of a sample size of one. The criterion this paper proposes and applies—"a single point cannot self-calibrate," meaning that a causal assertion resting on a single event can be identified only when an external variation independent of that assertion is available as a yardstick, and otherwise can only be presented as a theoretical proposition or hypothesis—provides an operable standard for handling this class of research. The value of this criterion lies in the fact that it neither abandons analysis merely because the sample size is one (which would render a great many important historical events unstudiable) nor relaxes the rigor of identification merely because analysis is necessary (which would reduce analysis to post hoc rationalization). Instead, it draws a clear line between the two: what is defensible is judgment calibrated against an external yardstick, and what must be downgraded is single-point attribution without a yardstick. The second generalizable methodological contribution is the strict distinction between "measurement propositions and causal propositions." The stratification in Chapter 7 of this paper—the methodological recalibration of scale (how rarity is defined) is defensible, while the cause-of-death attribution for a specific event must be downgraded to a hypothesis—reveals an overreach that is common yet often overlooked in applied econometrics and risk management: inferring from "a given measurement scale is unreasonable" that "the conclusion measured using that scale is causally wrong." These two classes of propositions belong to different logical categories, and the certainty of the former cannot be lent to the latter. This paper's adherence to this distinction offers a paradigm for preventing such overreach in stress testing, risk measurement, and quantitative methodology more broadly. The third generalizable methodological contribution is the transformation of "a universal ontological proposition into a falsifiable comparative proposition." Chapter 8 of this paper narrows "risk location is an institutionally endogenous choice of accountability" from an unfalsifiable universal ontology into a comparative proposition that permits falsification via successful transplantation, demonstrating how a framing judgment with deep intuitive appeal but no observational implications can be transformed into a refutable proposition subject to empirical testing. The technical core of this transformation—binding in advance the mapping from independent variable to dependent variable, and explicitly specifying which observations would falsify that mapping—has methodological demonstration value for political economy, comparative institutional analysis, and the construction of social-science theory more broadly. Together, these three methodological contributions constitute an implicit claim of this paper: the value of an economic judgment lies not in the strength or certainty of its conclusion, but in the clarity of its identification boundary and its refutability. A weaker judgment that clearly marks its identification boundary and specifies its falsification conditions carries greater academic value than a stronger assertion that marks no boundary and admits no falsification. The consistent application of this claim throughout this paper is precisely its methodological increment relative to the existing discussion. ### 9.5 Chapter Summary This chapter has conducted a cross-chapter synthesis of the five sets of judgments. The five sets of judgments constitute a progressive chain running from event misdiagnosis, through cross-sectional misuse, mechanism identification, and measurement correction, to institutional endogeneity, with the level of misdiagnosis descending step by step and each level clearing the ground for the next. The five sets of judgments are mutually consistent, with no conflicts in load-bearing weight: the event level and the institutional level are progressive and complementary; cross-country non-identifiability and the falsifiable comparative proposition have a clear division of labor; the decoupling of solvency from liquidity and the decoupling of the denominator from liquidity point in the same direction; and the mechanism level and the measurement level share an intuition while each downgrading a different assertion. The honest boundary running throughout the paper—stratified load-bearing, non-self-calibration of single points, bidirectionality of the downgrading direction—constitutes a systematic feature of this paper's methodology. This synthesis shows that the value of this paper lies not in offering a set of settled causal conclusions, but in offering a set of judgments that are clearly bounded in identification, refutable, and non-obvious, together with an honest marking of the path from judgment to identification. --- ## Chapter 10 Conclusion: Implications for Pension Risk Governance in China and the Limitations of This Paper ### 10.1 Review of the Core Conclusions Taking the 2022 UK LDI liquidity event as its empirical entry point, this paper has proposed and argued for five progressive sets of judgments concerning liquidity risk, leverage constraints, and the restructuring of asset-liability management in mature pension markets during the low-interest-rate decumulation phase. The core conclusions of these five sets of judgments can be summarized as follows. First, the risk landscape of the decumulation phase cannot be characterized along a single dimension. A rise in interest rates is a two-sided event whose net-effect sign is a priori undetermined; the funding line and the liquidity line may separate in opposite directions, and an improvement in the funding line does not constitute a safety signal for the liquidity line. Taking the UK in 2022 as an example, the significant improvement in funding coverage (PPF 2022) and the emergency gilt purchases triggered by the collateral spiral (BoE 2022) were driven in opposite directions by the same interest-rate shock. Any framework that conflates solvency with liquidity will produce systematic misjudgments during the decumulation phase. Second, the value of cross-country comparison does not lie in reading a ranking of institutional superiority or inferiority from the outcomes of a single event-year. Survivorship bias is a symmetric knife; the fact that the Netherlands and Denmark did not blow up cannot be used to infer institutional superiority. What is identifiable is comparison with exogenous variation—the jurisdictional dichotomy falsification test provided by the fact that Irish-resident sterling LDI accounted for roughly 30% of net selling during the crisis period (Central Bank of Ireland 2023), and the structural difference of pooled vehicles selling roughly 13 percentage points more (Breeden 2022)—rather than the outcome variable of "who did not blow up"; the Danish regulator's follow-up on residual liquidity risk (IPE 2023) further confirms that survival does not equal absence of risk. Third, the mechanistic lesion of the spiral lies in the mismatch between the moneyness of margin and the topology of central-bank bailout-ability, rather than in the level of leverage or the share of alternative assets per se. The structural difference in the Netherlands—posting part of margin in securities and relying less on repo (DNB 2024)—precisely counter-confirms the criticality of this weld point. This diagnosis holds as a theoretical-dynamics proposition; its causal identification awaits variation in scope across rescue episodes. Fourth, the independent variable of stress testing must be reset from a static basis-point endpoint to path, density, and correlation structure. The methodological recalibration of scale—defining rarity via the density domain—is defensible, while the empirical attribution that "the path was the specific cause of death" (whose institutional premise is that yields rose more than 100 basis points over four days, a speed far exceeding the capacity to post additional margin, see SUERF 2023) must be downgraded to an untested hypothesis. Fifth, risk location is an institutionally endogenous choice of accountability. This meta-judgment is narrowed into a comparative proposition that permits falsification via successful transplantation—given that institutional variables predict the dimension of accountability, transplanting a different practice will fail on observable indicators. The difference between the UK's "250-basis-point buffer plus contingency backstop facility" (TPR 2023; BoE 2025) and Ireland's and Luxembourg's "300-basis-point buffer" (ESMA 2022) is precisely a manifestation of the accountability boundary bifurcating with institutional variables. Improving resilience is not a purely technical problem but a redistribution of the accountability boundary. ### 10.2 Policy Implications for Pension Risk Governance in China The five sets of judgments above provide a set of implications for the ex ante design of China's pension risk governance framework, grounded in the fundamental premise that such a framework "cannot simply be transplanted and must be designed endogenously to the institution." This premise itself derives from the meta-judgment of Chapter 8: best practices encode the accountability choices of the source country's institutions, and since China's institutional variables differ from those of the UK and the Netherlands, endogenous design rather than direct copying is required. Under this premise, the policy implications of this paper can be specified as follows. First, the risk-monitoring framework cannot look only at funding coverage. The judgment of Chapter 4 shows that an improvement in the funding line does not constitute a safety signal for the liquidity line. In designing its risk-monitoring framework, China's pension system must, in addition to the solvency path (funding coverage ratio), independently establish a monitoring dimension for the liquidity path (collateral and cash availability), so as to avoid misjudging scheme safety during an interest-rate repricing phase by looking only at funding coverage. This implication is directly constrained to the design of account-level risk monitoring, grounded in Chapter 4's conclusion regarding the two-sided event and the separation of the two lines. Second, if China develops LDI or derivative hedging instruments, it must design in advance a match between the moneyness of margin and the scope of central-bank instruments. The judgment of Chapter 6 shows that the lesion of the spiral lies in the mismatch between the moneyness of margin (which must be posted in cash) and the topology of central-bank bailout-ability (which acts on the securities side). If China develops leveraged duration-hedging instruments in the future, it must, at the outset of institutional design, consider in advance: whether margin can be posted in securities rather than cash alone, and whether the central bank's emergency facilities can act on the cash side. The Dutch comparison shows that allowing part of margin to be posted in securities and reducing reliance on repo leverage is a structural condition for avoiding the welding-shut of the cash spiral. This implication is constrained to the level of instrument design, grounded in Chapter 6's mechanistic conclusion regarding the moneyness mismatch. Third, the stress-testing framework should adopt the density domain to define rarity, avoiding the subjective, intuition-based setting of basis-point scenarios. The judgment of Chapter 7 shows that rarity should be defined by the minimum density domain (Mahalanobis distance or highest-density region) rather than by a shift of an integer number of basis points, and that the correlation matrix should be set as a first-class scenario variable. This is one of the few implications in this paper that can be directly transplanted at the methodological level (as distinct from institutional-level implications, which cannot be directly transplanted): in designing its pension stress-testing framework, China can directly adopt the density-domain method for defining rarity and incorporate correlation structure into scenario design, thereby avoiding the miscalibration of existing frameworks that equate "more extreme" with "a larger number of basis points." This implication is constrained to the methodological design of stress testing, grounded in the first-layer methodological conclusion of Chapter 7. It should be noted that the second-layer empirical attribution of Chapter 7—that "the path was the cause of death"—remains an untested hypothesis; therefore, China's stress-test design may adopt the methodological scale of the density domain, but should not make strong normative arrangements based on the as-yet-unidentified path-as-cause-of-death proposition. Fourth, the institutional design of risk governance must recognize its essential nature as a choice of accountability. The meta-judgment of Chapter 8 shows that any setting of buffers, backstops, and leverage constraints is a redistribution of the accountability boundary. In designing its pension risk-governance framework, China must clearly recognize that: setting a liquidity buffer shifts a certain risk from the public backstop back onto the scheme itself; establishing a central-bank contingency backstop shifts a certain tail risk from the scheme to the public sector. This recognition requires policy designers to clarify, at the institutional level, the accountability boundary of "who bears what risk under what state of the world," rather than treating risk governance as a technical parameter that can be optimized independent of institutional context. This implication is constrained to the top-level design of the governance framework, grounded in the meta-judgment of Chapter 8 regarding the endogeneity of accountability. There is an internal hierarchical structure among the four implications above that must be clarified. The first three implications (the two-line distinction in risk monitoring, the matching of margin moneyness, and the density-domain scale for stress testing) are concrete, actionable design recommendations, corresponding respectively to the judgments of Chapters 4, 6, and 7; the fourth implication (the recognition of the essential nature of the accountability choice), by contrast, is an overarching meta-principle, corresponding to the meta-judgment of Chapter 8, which provides the fundamental positioning for the first three recommendations. The significance of this hierarchical structure is that the first three recommendations are not, in themselves, "technical modules" that can be transplanted independent of institutional context; rather, they must be designed endogenously under the constraint of the fourth meta-principle, in conjunction with China's own institutional variables. For example, exactly how the matching of margin moneyness (the second recommendation) should be implemented depends on China's clearing-house ownership structure and the boundaries of the central bank's mandate (institutional variable X)—if China's central bank is able to provide cash-side emergency support to the relevant market, then the accountability boundary can be partly placed on the public backstop; if not, then the accountability boundary must fall more heavily on the scheme-level cash buffer. Therefore, the policy implications of this paper are not a checklist to be ticked off item by item, but a framework that must be designed endogenously under the guidance of the meta-principle, in conjunction with the country's own institutional variables. This is precisely the fundamental difference between this paper's approach and the approach of "directly copying best practices," and it is the ultimate policy-level landing point of this paper's corrective logic. Furthermore, the policy implications of this paper also imply a judgment about governance capacity that must be made explicit. Chapter 1 of this paper already points out that what truly determines pension safety is not the allocation ratio of risk assets or alternative assets per se, but whether the four factors of return target, liability duration, liquidity needs, and governance capacity are matched with one another. The first three factors are technical dimensions that can be quantified and designed; the fourth (governance capacity) is an institutional dimension—it determines whether a country can effectively identify its own institutional variables, whether it can endogenously design its risk-governance framework under the guidance of the meta-principle, and whether it can coordinate the response of the scheme level and the public sector in a crisis moment. Therefore, the deepest implication of this paper for pension risk governance in China is not any single concrete technical recommendation, but the building of governance capacity itself: only when governance capacity is sufficient to support the accurate identification of institutional variables and the clear-eyed design of accountability boundaries can the foregoing concrete recommendations truly take effect. Absent this governance capacity, any technical module copied from elsewhere will fail within China's own institutional environment—this is precisely the extension of Chapter 8's thesis that "best practices cannot be transplanted" into the dimension of governance capacity. Taken together, the common feature of the policy implications above is that each recommendation is bound to a specific account level or design level, a specific constraint dimension, and the specific conclusion of the corresponding judgment in this paper. This binding itself embodies the methodological stance of this paper: policy recommendations should not be generalized assertions detached from an identification boundary, but should be concrete arrangements with a clear judgmental basis and a clearly specified scope of application. What is especially important is that all these implications are subordinate to one fundamental premise—China's approach cannot simply copy the best practices of mature markets, because those practices encode the institutional accountability choices of the source country; China must, on the basis of identifying its own institutional variables, endogenously design its risk-governance framework. ### 10.3 Data Availability and the Boundaries of This Research This paper has several limitations that must be honestly flagged; these limitations precisely define the identification boundary of this paper's judgments and the direction for future research. To give the reader a systematic grasp of this paper's identification boundary, this section consolidates the downgrading statements scattered across the chapters into a standardized discussion of research boundaries. First, micro-level causal magnitudes are, in general, unidentified. Owing to data-availability constraints, this paper was unable to obtain scheme-level micro panel data. Consequently, the core causal magnitudes in each chapter of this paper—the slope ratio between the solvency and liquidity paths, the difference-in-differences coefficient for the vehicle comparison, the causal coefficient of the moneyness mismatch, the comparison of explanatory power between path and magnitude, and the difference-in-differences and regression-discontinuity coefficients for accountability location—all remain downgraded and unestimated. What this paper can provide is a directional verification of aggregate institutional facts and the completion of identification designs, not the identification of micro-level causation. This limitation is not a design flaw but a data constraint; each chapter specifies the sole exit route from downgraded status to identified status, pending the corresponding micro data. Second, several core mechanistic propositions are positioned as theoretical propositions rather than as identified causation. The moneyness-and-bailout-ability mismatch in Chapter 6 is positioned as a theoretical-dynamics proposition because the definitional nature of the scope of the single 2022 rescue precludes self-calibration; the path-as-cause-of-death claim in Chapter 7 is positioned as an untested empirical hypothesis because 2022 is a single sample and cannot support identification; the comparative proposition in Chapter 8 is not asserted as identified, owing to the absence of a cross-nationally comparable dataset of transplantation outcomes. The empirical identification of these propositions awaits, respectively, a panel across multiple rescue scopes, a panel across multiple historical shock events, and a dataset of cross-country transplantation outcomes—all beyond the data conditions of this paper. Third, the cross-country decomposition is structurally unidentifiable. The cross-country leverage-hedging decomposition in Chapter 5, given a cross-section of three countries in a single event-year with no degrees of freedom to estimate any decomposition, is structurally unidentifiable, and hence the Netherlands and Denmark serve only as motivating cases. Its upgrade requires either a cross-country panel (spanning multiple years and countries, including variation from years without a blowup) or the construction of an ex ante critical-distance index for out-of-sample prediction—neither of which is satisfiable under the conditions of this paper. Fourth, this paper's judgments take the UK event as their principal empirical entry point, and extrapolation to other institutional environments must be undertaken with caution. Although this paper's meta-judgment (the endogenous-accountability comparative proposition) is itself designed to address the problem of extrapolation, the two-sided falsification test of this proposition has not yet been carried out; hence the empirical strength of this paper's thesis that "best practices cannot be transplanted" is constrained by the fact that the comparative proposition has not yet been identified. This paper strictly maintains its honest boundary on this point: this thesis holds as a comparative proposition with observational implications that permits falsification via successful transplantation, but its two-sided empirical falsification test awaits cross-nationally comparable data. ### 10.4 Concluding Remarks The fundamental methodological belief of this paper is that the core of an economics paper is judgment, not the mechanical accumulation of data; and that the value of a judgment lies precisely in its refutability and in the honest marking of its identification boundary. Accordingly, this paper has not padded its argument with fabricated causal coefficients or exaggerated identification claims, but instead constitutes its contribution through a set of judgments that are clearly bounded in identification, refutable, and non-obvious, together with an explicit marking of the level of identifiability at which each proposition stands. What this paper corrects is not only the naive misjudgment that "rising interest rates harm pensions," but also the next, shallower layer of the corrective narrative itself—"improved coverage therefore means safety," "thickening the buffer suffices," and "copying best practices suffices." This multi-layered correction ultimately converges on an overarching judgment: risk governance in the decumulation phase is, in essence, an institutionally endogenous choice of accountability, and its improvement cannot be optimized as a purely technical problem detached from institutional context. For the ex ante design of China's pension risk-governance framework, the core implication of this judgment is that what truly matters is not the allocation ratio of risk assets or alternative assets, but whether the four factors of return target, liability duration, liquidity needs, and governance capacity are matched with one another, and whether a country has, on the basis of identifying its own institutional variables, endogenously designed its risk-governance framework. ## Appendix Tables This paper consolidates the comparative materials scattered across the preceding chapters into the following three numbered exhibits, so that readers may systematically grasp the paper's mechanism comparisons, identification strategies, and data sources; all three tables are summarized from the body text and introduce no new claims. Table 1 Core Mechanism Comparison: The One-Directional Low-Rate Environment versus the Rate-Repricing Decumulation (Payout) Phase | Dimension | One-Directional Low-Rate Environment | Rate-Repricing Decumulation (Payout) Phase | |---|---|---| | Funding line (funding ratio) | Liabilities inflate, assets appreciate; the funding ratio is under pressure but changes slowly | Present value of liabilities falls; the funding ratio improves markedly (PPF 2022) | | Liquidity line (collateral and cash) | Collateral is ample, no immediate cash pressure | Cash variation margin calls surge; collateral spiral (BoE 2022) | | Relationship between the two lines | Move in the same direction; can be summarized along a single dimension | Diverge in opposite directions; a single dimension necessarily misjudges the situation | | Leveraged synthetic duration positions | Amplify returns; fragility lies dormant | Unrealized losses trigger margin calls; fragility becomes concentrated and manifest | | Monetary character of margin | The cash-payment constraint does not constitute pressure | The cash-payment constraint welds together into a self-reinforcing cash spiral | | Mainstream narrative | LDI as advanced practice | Must shift from "alternative-advancement thesis" to "matching determines safety" | Table 2 Competing Hypotheses and Identification Strategies for the Five Judgments | Judgment | Conventional Narrative (Corrected) | This Paper's Defensible Identification Path | Portions Downgraded to Theory or Hypothesis | |---|---|---|---| | Judgment One (Chapter 4) | Improved funding ratio therefore the scheme is safe | Event study of vehicle comparison and funding-liquidity timing misalignment | Directed causality, slope-ratio values, cash-margin share, distribution axis | | Judgment Two (Chapter 5) | The Netherlands and Denmark did not blow up, therefore their institutions are superior | Pseudo-jurisdiction-of-registration comparison falsifying the jurisdictional dichotomy; internal vehicle comparison | Cross-border leverage-hedging decomposition, critical distance | | Judgment Three (Chapter 6) | The cause lies in excessive leverage or excessive alternatives | Institutional facts of the rescue's scope of application; weak proposition on the existence of a stock effect | Causality of monetary mismatch, ineffectiveness of securities-side rescue, prescriptive norm of holding through overallocation | | Judgment Four (Chapter 7) | A more extreme scenario simply means a larger basis-point move | Defining rarity in the density domain (methodological, defensible) | Attribution of cause of death to path rather than magnitude, causal status of correlation reversal | | Judgment Five (Chapter 8) | Simply replicate the best practice of countries that did not blow up | Difference-in-differences across phased clearing stages; institutional preconditions for regression discontinuity at the eligibility threshold | Bilateral falsification of the comparative proposition, slope of buffer translation, counterfactual of time-bound commitment, topology of backstop endogenization | Table 3 Data Sources and Availability | Institutional Fact | Key Figures | Source Institution (Year) | Judgment Served | |---|---|---|---| | UK DB PPF 7800 total surplus | GBP 313.8 billion (historical peak, August 2022); funding ratio approximately 105% to approximately 135% | PPF (2022); Professional Pensions (2022) | Chapter 4 | | BoE 2022 gilt intervention | Launched 28 September, actual purchases of approximately GBP 19.3 billion, terminated 14 October, explicit time limit | BoE (2022) | Chapters 4, 6, 7, 8 | | Speed of the 30-year gilt yield shock | Jumped approximately 140bp in a single day after the budget, rose more than 100bp over four days | SUERF (2023) | Chapter 7 | | Structure of UK vehicles | Pooled vehicles account for approximately 10% to 15%; pooled vehicles sold approximately 13pp more, fire-sale discount of 7% to 10% | Breeden (2022) | Chapters 4, 5, 6 | | Irish-resident GBP LDI | Accounted for approximately 30% of net selling during the crisis; accounted for approximately 85% of EU-registered GBP LDI | Central Bank of Ireland (2023) | Chapters 5, 8 | | Dutch pension funds, first half of 2022 | Sold approximately EUR 88 billion, injected approximately EUR 82 billion; funding ratio approximately 119% to approximately 122% | IPE (2022) | Chapters 5, 7 | | Structural differences between the Netherlands and Denmark | Euro-area LDI uses repo less, can post securities as margin | DNB (2024) | Chapters 5, 6 | | Danish FSA follow-up | Some firms have a "non-negligible liquidity risk" | IPE (2023) | Chapter 5 | | TPR resilience buffer | Guidance issued April 2023; minimum 250bp, assuming a 5-day replenishment | TPR (2023) | Chapter 8 | | Ireland/Luxembourg buffer | At least 300bp of UK-yield resilience (endorsed by ESMA) | ESMA (2022) | Chapter 8 | | BoE CNRF eligibility threshold | Opened January 2025; must hold more than GBP 2 billion in gilts | BoE (2025) | Chapter 8 | | Structural change in the pension sector | Accelerated by prolonged low rates; rapid rate hikes triggered liquidity pressure and contagion | IMF (2025) | Chapters 1, 2 | --- ## References (Works Cited) Bank of England. "Bank of England Announces Gilt Market Operation." *Bank of England News*, 28 Sept. 2022, www.bankofengland.co.uk/news/2022/september/bank-of-england-announces-gilt-market-operation. Bank of England. "The Bank of England Opens New Contingent Non-Bank Financial Institution Repo Facility for Applications." *Bank of England News*, 28 Jan. 2025, www.bankofengland.co.uk/news/2025/january/boe-open-new-contingent-non-bank-lending-facility-for-applications. Bank of England. "Financial Stability Buy/Sell Tools: A Gilt Market Case Study." *Bank of England Quarterly Bulletin*, 2023, www.bankofengland.co.uk/quarterly-bulletin/2023/2023/financial-stability-buy-sell-tools-a-gilt-market-case-study. Breeden, Sarah. "Risks from Leverage: How Did a Small Corner of the Pensions Industry Threaten Financial Stability?" Speech at ISDA and AIMA, Bank of England, 7 Nov. 2022, www.bankofengland.co.uk/speech/2022/november/sarah-breeden-speech-at-isda-aimi-boe-on-nbfi-and-leverage. Central Bank of Ireland. "Irish-Resident LDI Funds and the 2022 Gilt Market Crisis." *Financial Stability Notes*, Central Bank of Ireland, 2023, www.centralbank.ie/docs/default-source/publications/financial-stability-notes/irish-resident-ldi-funds-and-the-2022-gilt-market-crisis.pdf. De Nederlandsche Bank. "Limited Liquidity Risks in Euro LDI Funds." *DNB Supervision News*, De Nederlandsche Bank, 2024, www.dnb.nl/en/sector-news/supervision-2024/limited-liquidity-risks-in-euro-ldi-funds. European Securities and Markets Authority. "ESMA Agrees Investment Restrictions for GBP LDI Funds in Ireland and Luxembourg." *ESMA News*, 2022, www.esma.europa.eu/press-news/esma-news/esma-agrees-investment-restrictions-gbp-ldi-funds-ireland-and-luxembourg. International Monetary Fund. *Pension Funds and Financial Stability*. Global Financial Stability Note GFSNEA2025001, International Monetary Fund, 2025. Investment and Pensions Europe. "Danish FSA Says Pension Fund Liquidity Risk Needs Close Follow-Up." *IPE News*, 2023, www.ipe.com/news/danish-fsa-says-pension-fund-liquidity-risk-needs-close-follow-up/10063809.article. Investment and Pensions Europe. "Dutch Pension Funds Sell €88bn in Assets in Rebalancing Exercise." *IPE News*, 2022, www.ipe.com/dutch-pension-funds-sell-88bn-in-assets-in-rebalancing-exercise/10061980.article. Pension Protection Fund. "The PPF 7800 Index: September 2022." *Pension Protection Fund*, 2022, www.ppf.co.uk/ppf-7800-index. 亦见 Professional Pensions. Professional Pensions. "DB Funding: September 2022." *Professional Pensions*, 2022, www.professionalpensions.com/news/4057337/db-funding-september-2022. SUERF — The European Money and Finance Forum. "Putting Out the NBFIre: Lessons from the UK's Liability-Driven Investment Crisis." *SUERF Policy Notes and Briefs*, 2023, www.suerf.org/publications/suerf-policy-notes-and-briefs/putting-out-the-nbfire-lessons-from-the-uks-liability-driven-investment-crisis. The Pensions Regulator. "New LDI Guidance Published by TPR to Ensure Schemes Minimise Risk." *TPR Press Releases*, 24 Apr. 2023, www.thepensionsregulator.gov.uk/en/media-hub/press-releases/2023-press-releases/new-ldi-guidance-published-by-tpr--to-ensure-schemes-minimise-risk. Willis Towers Watson, Thinking Ahead Institute. *Global Pension Assets Study 2026*. Thinking Ahead Institute, Willis Towers Watson, 2026.