Differences in Asset Allocation Between Defined Benefit and Defined Contribution Pensions Under Low Interest Rates: A Comparison Based on Asset-Liability Management and Risk-Bearing Mechanisms

Literature and Institutional Background Review

Existing discussions of DB/DC allocation differences under low interest rates generally proceed along two traditional lines of inquiry. The first is the tradition of institutional comparison, which takes as its point of departure the nominal definition that "DB has the employer bear the risk, DC has the individual bear the risk," and from there treats DB's liability-driven investment (LDI) and duration matching, and DC's long-term equity and life-cycle allocation, as natural corollaries of the risk attribution of the two institutional types. This traditional narrative correctly captures the difference between the two institutional types in nominal risk attribution, and broadly accords with observable allocation patterns, but it treats institutional type as an exogenously given classification variable, without asking why the institutional form itself is as it is, or why it shifts under low-rate pressure. The second is the tradition of put-option pricing in financial theory, whose classic conclusion holds that once the writer of the residual guarantee put is fixed, low funding combined with maximal risk-asset weighting is the option holder's dominant strategy, a conclusion that holds regardless of whether the risk-bearing party is risk-averse (Sharpe 1976; Treynor 1977). This paper's option-position perspective is precisely a continuation of this theoretical tradition, extended from the context of corporate finance to the principal-agent chain of pension institutions.

At the level of institutional facts, this paper's background review is anchored in a number of publicly verifiable aggregate facts. In the case of the United Kingdom, following the establishment of accounting standards that revalue DB liabilities as an on-balance-sheet item, together with the regulatory requirement that wind-up be funded to buyout level, private-sector DB schemes have undergone marked de-risking and a wave of closures; pension funds' ownership share of the UK equity market fell from about 32.4% in the early 1990s to about 1.5% in recent years, while their ownership share of the bond market rose from about 28% to about 63% (British Progress 2024; ii.co.uk 2023). In the case of the Netherlands, the Future of Pensions Act (Wet toekomst pensioenen) established a wholesale transition toward new defined-contribution-type arrangements, while retaining a collective buffer through an explicitly parameterized solidarity reserve (DNB 2023; NautaDutilh 2023). In the case of Australia, the mandatory Superannuation Guarantee has raised statutory contribution rates in annual increments on a nationally uniform legislative timetable, bringing an increasing share of workers into DC-type account systems (ATO 2024). These institutional facts together form the empirical foundation for this paper's five axes of analysis, but this paper's position is that describing what an institution "is" does not amount to explaining "why" its allocation is as it is — the latter is the direction this paper seeks to advance. Full sources for the relevant institutional facts are given in the Works Cited at the end; the body of this paper marks the source of every claim resting on a verifiable fact using MLA in-text parenthetical citations.

Chapter 1 Introduction

1.1 Statement of the Problem

Since the early twenty-first century, major developed economies have experienced a sustained period of low, and even ultra-low, interest rates. For pension systems, falling interest rates exert pressure from two sides simultaneously: on one hand, the present value of long-term liabilities expands as the discount rate falls, widening the funding gap of DB schemes; on the other hand, the expected returns of traditional fixed-income assets are compressed, forcing pension entities of all types to re-choose between "raising expected return" and "controlling risk exposure." Alongside this pressure, a widely observed structural fact is the continuing shrinkage of DB schemes in the private sector and the rising share of DC schemes. This trend takes on a similar direction but strikingly different paths across jurisdictions: the UK's path is dominated by the closure wave and de-risking of private-sector DB; the Netherlands is marked by the wholesale transition of collective contracts toward new defined-contribution-type arrangements; Australia, through annual increments in mandatory contribution rates, brings an increasing share of workers into DC-type account systems. On the surface, these jurisdictions share the same direction of "moving from DB toward DC"; however, if the analysis stops at this direction itself, it misses a more fundamental question: why the same interest-rate shock is filtered into signals of different nature and different visibility across different jurisdictions, and why it triggers consequences that diverge, and even run in opposite directions.

The standard explanation for this fact usually rests at the level of the institutional definition that "DB has the employer bear the risk, DC has the individual bear the risk," and proceeds naturally to the empirical generalization that "DB tends to respond to low rates with liability-driven investment and duration matching, while DC tends to raise expected returns through long-term equity and life-cycle allocation." This generalization is not entirely wrong; it correctly captures the nominal difference in risk attribution between the two institutional types, and broadly accords with observable allocation patterns. But this paper holds that, as an answer to the focal judgment, this line of explanation suffers from two fundamental deficiencies. First, it treats institutional type as an exogenously given classification variable, without asking why the institutional form itself is as it is, or why it shifts under low-rate pressure; in other words, it takes the institution as a premise of analysis rather than an object of analysis. Second, it treats "the rising share of DC" as the endpoint of analysis rather than a phenomenon to be explained — risk does not vanish as DB shrinks; it merely shifts from the balance sheets of employers and sponsors to individuals (longevity risk, investment risk, sequence risk), and even to more concealed risk bearers (future taxpayers, absent generations, collective buffers in a vacuum of property rights). Interrogating the direction, mechanism, and consequences of this shift is the point of departure that sets this paper apart from any textbook-style comparison.

It should be clarified that what this paper criticizes is not the correctness of the proposition "DB bears risk, DC transfers risk," but rather the posture of treating it as the endpoint of analysis. A pension textbook may correctly state this proposition at the outset and then move on to a technical comparison of the two institutional types; this paper's position is that it is precisely after this proposition has been stated that the real economic question begins — why the institution stops risk here rather than there, whether allocation pressure in turn reshapes the form of the institution, and who ultimately bears, in what silence, the risk that has been shifted away. These questions cannot be answered by the nominal definition of institutional type; they require opening up the institution to examine its internal principal-agent chain, property-rights structure, measurement representation, and the larger institutional configuration in which it is embedded. This paper's five axes of analysis open up the institution from precisely these five directions: Chapter 4 examines the option stopping point along the principal-agent chain; Chapter 5 examines the property-rights structure and continuous spectrum; Chapter 6 examines measurement representation; Chapter 7 examines the embedded systemic configuration; Chapter 8 examines the contractual nature of asset-side inflows. Each axis attempts to answer the mechanistic question of "how institutions lead to allocation," rather than resting on the nominal description of "what institutions are."

1.2 Focal Judgment and Literature Gap

This paper's focal judgment rests on a pair of bidirectional causal relations. Direction A is that institutions determine the boundary of allocation: DB's liability rigidity constrains it to conduct asset-liability management (ALM), while DC's risk transfer allows it to carry long-term equity. Direction B is that allocation pressure in turn reshapes institutions: under low interest rates, DB's funding gap becomes unbearable, forcing closure waves, the individualization of collective contracts, and the emergence of hybrid forms — the institution is thus an endogenous outcome of allocation pressure, not merely its antecedent. These two directions are not mutually exclusive but operate simultaneously: the institution constrains allocation at a given point in time, while allocation pressure rewrites the institution over a longer horizon. The difficulty of the focal judgment lies precisely in the fact that these two directions are intertwined in real data — an episode of institutional transition is often both the result of accumulated past allocation pressure and the precondition for future allocation behavior.

The existing literature has gaps in both directions. In Direction A, a large body of comparative research uses dummy variables for institutional type to explain cross-country allocation differences, but institutional type is highly co-variant with the depth of the capital market in which it is embedded, the regulatory framework, plan size, and other factors; it is often unclear whether what the dummy variable absorbs is the institutional mechanism itself or these confounders — a regression that uses a "DB dummy" to explain a tendency toward duration matching may merely be restating the tautology that "mature, large plans have longer liabilities," rather than revealing the institutional mechanism itself. In Direction B, institutional transitions (such as the individualization of collective contracts) are almost always endogenous to the allocation pressure of the period; event studies find it difficult to separate the mechanism whereby "pressure forces out institutional change" from the mechanism whereby "allocation follows institutional change," and so causal claims about origins mostly lack a clean identification basis; treating the political process by which one jurisdiction completes a transition ahead of others as an exogenous instrument appears to offer identification, but in fact dresses an actually unidentified causality in the language of identification — because that political process is itself a product of allocation pressure.

This paper does not attempt to paper over this difficulty with the language of identification; instead, it explicitly separates the identifiable-consequence side of Direction A from the not-cleanly-identifiable-cause side of Direction B, so that every claim carries its own evidentiary status. This separation is not an avoidance of Direction B but an honest positioning of it: Direction B is the soul of this paper's focal judgment (who bears the risk that has been shifted away, how allocation pressure reshapes institutions), but under current identification techniques it cannot obtain clean causal evidence; this paper therefore presents it through political-economy theory and case studies, and explicitly labels its evidentiary status as theoretical and case-based, rather than identified. Honestly leaving the unidentifiable cause side at the theoretical level is precisely intended to prevent it from contaminating the identifiable-consequence side of Direction A — the latter being the part on which this paper can stand in a falsifiable way.

1.3 Core Judgment and Contributions

The overarching judgment this paper advances and argues is: benefit contingency (the degree of risk sharing) is a continuous institutional state variable, of which DB/DC are degenerate corner points on this continuous spectrum; "de-risking" as movement along the spectrum is an intergenerational redistribution of wealth toward generations currently present, rather than the mere elimination of risk. Under this overarching judgment, this paper unfolds five mutually orthogonal axes of analysis, characterizing respectively: the risk stopping point and option hardness (Chapter 4), the continuous spectrum and the ontology of property rights (Chapter 5), the measurement development layer (Chapter 6), system-level institutional complementarity (Chapter 7), and the contractual nature of asset-side inflows (Chapter 8).

The paper's contributions accordingly fall into three layers. First, at the theoretical layer: it re-parameterizes the pension institution as a pair of continuous state variables, providing a single, unified metric for institutional taxonomy for subsequent research, with the DB/DC dichotomy restated as a special case under this parameterization. The value of this re-parameterization lies not in its falsifiability (it is a reselection of the unit of analysis, not a hypothesis testable against data), but in its organizing power and downstream reusability — it allows cross-country differences, hybrid forms, and institutional transitions all to be organized and compared within the same coordinate system. Second, at the judgment layer: it places the political-economy inquiry of "who bears the risk that has been shifted away" at the center of the analysis, pointing out that "de-risking," under conditions of a property-rights vacuum, is in substance a distributional decision as to whose account the value of the buffer is written into. This judgment advances pension analysis from the technical-prudential level to the distributional-political level: de-risking is no longer merely a prudential question about how to reduce the volatility of the funding gap, but a political question about how the value of the collective buffer is distributed across generations. Third, at the methodological layer: it designs, for each of the five axes, an identification strategy that distinguishes Direction A from Direction B, and establishes falsifiable thresholds prior to any data. This methodological self-awareness — designing before implementing, judgment preceding data, honestly labeling identification status — is itself part of this paper's contribution: it demonstrates a research posture that, in a domain of identification difficulty, neither abandons the sharpness of judgment nor uses the language of identification to paper over gaps in evidence. It should be specifically noted that, owing to data availability constraints, the empirical conclusions reported in this paper are limited to publicly verifiable institutional aggregate facts; the micro-level causal identification on which the core propositions of the five axes depend, since the fund-level panel and experimental data required are unavailable under the conditions of this study, are uniformly and honestly reported as "design in place, evidence pending," with no claim of identified causality made. This honest boundary runs throughout the paper and is a direct expression of this paper's methodological stance that judgment precedes data.

1.4 Structure of the Paper

The remainder of this paper is organized as follows. Chapter 2 outlines the comparative background of DB/DC institutions under low interest rates and the main forms of transition, illustrating with a three-country path how institutions lead to allocation. Chapter 3 establishes the theoretical framework of the continuous spectrum, defining the two state variables of degree of risk sharing and distributional core, and explaining the Direction A/Direction B identification distinction and the discipline of honest boundaries adopted throughout this paper. Chapters 4 through 8 unfold the five axes of analysis in turn, each chapter following the unified structure of "judgment — mechanism — identification strategy — honest labeling of evidentiary status": Chapter 4 discusses how put hardness drives the sign of allocation; Chapter 5 discusses the ontology of the continuous spectrum and buffer property rights; Chapter 6 discusses the marginal effect of the measurement development layer; Chapter 7 discusses system-level institutional complementarity as a lens on the direction of fragility; Chapter 8 discusses how the contractual nature of asset-side inflows determines illiquid-allocation capacity. Each of these five chapters corresponds to one established analytical thread, and marks, at every load-bearing point of the argument, the judgment it serves and the verifiable facts it cites. Chapter 9 conducts a cross-axis synthesis, explaining how the five axes jointly characterize the same institutional phenomenon at different margins, and how they are unified by the overarching judgment at both the ontological and normative levels. Chapter 10 summarizes the policy implications and research limitations, and points to directions for future research.

Chapter 2 Institutional Background and Forms of Transition

2.1 The Asymmetric Impact of Low Interest Rates on the Two Institutional Types

The impact of falling interest rates on DB and DC is asymmetric in its mechanism. For DB, low interest rates directly amplify the present value of liabilities through the discount-rate channel: the more mature the plan and the longer the liability duration, the more severely the same decline in the discount rate amplifies the present value of liabilities, mechanically raising the contribution obligation. This amplification is mechanical and predictable — it does not depend on the sponsor's intentions or market sentiment, but is a direct consequence of discounting mathematics; for a mature plan with a liability duration of twenty years, a one-percentage-point decline in the discount rate is enough to raise the present value of liabilities by about twenty percent, a shock of a magnitude far exceeding the concurrent volatility on the asset side, and constituting the principal source of pressure faced by DB plans under low interest rates. For DC, there is no concentrated liability present-value line item; low interest rates mainly affect the terminal-value distribution of individual accounts by depressing expected asset returns, with the consequences dispersed to the participant level, and lacking any unified on-balance-sheet item to aggregate them into a gap that fluctuates day to day. This structural difference means that the same interest-rate shock is, on the DB side, aggregated into a highly visible gap that fluctuates day to day and enters the field of view of governance, while on the DC side it is dispersed into the slow drift of the terminal-value distributions of countless individual accounts, neither fluctuating day to day nor entering any centralized field of decision-making. This asymmetry is the common background for understanding the mechanisms in subsequent chapters: the same interest-rate shock, entering the two institutional systems, is filtered into signals of different nature and different visibility, thereby triggering allocation responses in different directions. Chapter 6 will single out this mechanism of "signals being filtered differently" as the measurement development layer for in-depth analysis.

2.2 Institutional Differences and Allocation Facts Across Major Jurisdictions

In the United Kingdom, following the establishment of accounting standards that revalue DB liabilities as an on-balance-sheet item, together with the regulatory requirement that wind-up be funded to buyout level, private-sector DB schemes have undergone marked de-risking and a wave of closures. Pension funds' ownership share of the UK equity market fell from about 32.4% in the early 1990s to about 1.5% in recent years, while over the same period their ownership share of the bond market rose from about 28% to about 63% (British Progress 2024). A distinction of accounting basis must be made here: the 32.4%→1.5% figure is pension funds' ownership share of UK equities, not the equity allocation ratio within any single fund; within a fund's own portfolio, equity allocation has fallen from about 60%–70% to a portfolio dominated by liability-matching bonds (ii.co.uk 2023). These two bases must not be conflated, and this paper strictly distinguishes them in subsequent citations.[^1]

In the Netherlands, institutional reform has taken a different path. The Future of Pensions Act took effect on 1 July 2023, with all plans required to comply by 1 January 2028, and accumulated entitlements to be transferred to the new DC-type contract through a collective value transfer (invaren) (NautaDutilh 2023). The new Dutch system has not moved toward pure individualization but has introduced an explicitly parameterized solidarity reserve: the regulatory basis stipulates that the maximum size of the solidarity reserve is 15% of the pension fund's total retained assets (including the reserve itself); if funded by contributions, the annual injection from contributions per participant may not exceed 10% of total contributions, the injection from excess returns may not exceed 10% of that year's positive collective excess return, and the 15% ceiling must be met by 1 January 2037 or earlier (DNB 2023). This explicitly parameterized reserve arrangement provides direct institutional evidence for this paper's Chapter 3 proposition that "the degree of risk sharing is a continuously adjustable state variable."

In Australia, the mandatory Superannuation Guarantee has established statutory employer contributions since 1 July 1992, with the contribution rate raised in annual increments on a nationally uniform legislative timetable, from an initial 3% (with separate tiers for small and large employers) to 9% in the 2002–03 fiscal year, followed by a freeze at 9.5% between 2014 and 2021, reaching 12% as of 1 July 2025 (ATO 2024). This legislative timetable of annual rate increases is imposed on all funds simultaneously, with its timing determined by the federal legislative process rather than by any fund's contemporaneous allocation pressure, constituting a near-exogenous dosage staircase on which the identification design of Chapter 8 relies.

2.3 From Institution to Allocation: A Mechanistic Reading of the Three-Country Paths

This paper's position is that cross-country institutional facts alone do not constitute an explanation of cross-country allocation differences — describing what each country's institution "is" does not amount to explaining "why" each country's allocation is as it is. This section therefore treats the three main jurisdictions with a focus on how their institutions lead to their allocation, rather than on what their institutions are.

In the United Kingdom, the effect of institution on allocation is realized through a clear mechanism chain. Accounting standards revalue DB liabilities as an on-balance-sheet item, turning the liability into a line item that fluctuates day to day and enters governance's field of view; the regulatory requirement that wind-up be funded to buyout level converts the widening of the gap into a hard claim against the sponsor; the two together give the sponsor a strong incentive to compress the variance of the gap, and its allocation response is a large-scale shift from equities to bonds. The fall in pension funds' ownership share of the UK equity market from about 32.4% to about 1.5%, and the rise in their ownership share of bonds from about 28% to about 63% (British Progress 2024), is precisely the aggregate consequence of this mechanism chain — but it should be noted that this consequence is jointly driven by two channels, development (the increased visibility from putting liabilities on the balance sheet) and genuine constraint (wind-up requirements and rating reclassification), and cannot be attributed to either channel alone (Chapter 6 will discuss this confounding in detail).

In the Netherlands, the effect of institution on allocation is realized through a different mechanism. Rather than moving toward pure individualization, the Netherlands has explicitly parameterized the degree of risk sharing as an adjustable reserve dial: the ceiling and drawdown rules of the solidarity reserve are written into legislation, allowing a collective buffer to be retained above individual accounts, thereby institutionally underpinning the capacity to absorb interest-rate shocks. This explicitly parameterized reserve arrangement places the new Dutch system in the middle of the continuous spectrum rather than at a corner point, and its allocation behavior is accordingly neither pure DB liability matching nor pure DC individualized aggressive allocation, but an intermediate form moderated by the thickness of the reserve.

In Australia, the effect of institution on allocation is realized through the contractual nature of mandatory inflows. The mandatory Superannuation Guarantee, by raising the statutory contribution rate in annual increments on a nationally uniform legislative timetable, constructs a predictable, uninterruptible sequence of mandatory inflows; on the asset side, this inflow sequence constitutes a "synthetic long-duration asset." Per the mechanistic expectation of Chapter 8, this contractual nature should, in direction, support a relatively higher allocation to illiquid and private assets — this directional expectation is consistent with the prevailing observations regarding the allocation forms of large Australian funds, but the causal effect involved remains, under the conditions of this study, at the stage of "identification design in place, evidence pending"; this paper does not make a claim of established causality here (Chapter 8 will discuss this mechanism and its identification design in detail). The comparison of the three-country paths shows that the same direction of "moving from DB toward DC," via different institutional mechanisms, leads to allocation consequences that diverge, and even run in opposite directions — precisely what cannot be captured by taking discrete institutional types as the starting point of analysis.

2.4 Hybrid and Transitional Forms the Dichotomy Cannot Accommodate

It is worth emphasizing that the dichotomy of pure DB and pure DC struggles to accommodate the large number of hybrid and transitional forms that exist in reality. The Dutch collective defined-contribution arrangement, the UK's collective defined contribution (hereafter CDC) schemes, and public-sector "last-man-standing" arrangements all fall between the two poles. Taking UK CDC as an example: the Royal Mail Collective Pension Plan is the first and, to date, only CDC scheme authorized by the Pensions Regulator, authorized in April 2023 and launched on 7 October 2024, with contributions structured as 6% from the employee plus 13.6% from the employer, totaling 19.6% (House of Commons Library 2024). The distinctiveness of CDC lies in the fact that it neither leaves investment and longevity risk entirely with the employer (as in DB) nor transfers them entirely to the individual (as in pure DC), but instead disperses risk across the collective membership and across generations through collective risk-sharing and target-benefit adjustment mechanisms — precisely a typical point in the middle of the continuous spectrum. The existence of these forms in itself constitutes an empirical falsification of the dichotomous framework: a dichotomy already empirically falsified by a large number of hybrid forms, if still taken as the basic unit of analysis, will systematically overlook the distributional arrangements that these forms carry. It is for precisely this reason that, wherever a form does not fit the dichotomy, this paper lists it separately and explains its significance for the bidirectional causal judgment, rather than forcing it into the category of pure DB or pure DC.

2.5 Summary

This chapter has outlined the asymmetric impact of low interest rates on DB and DC, the institutional differences and allocation facts of the major jurisdictions, a mechanistic reading of the three-country path from institution to allocation, and the hybrid and transitional forms that the dichotomous framework struggles to accommodate. The comparison of the three-country paths shows in particular that the same direction of "moving from DB toward DC," via different institutional mechanisms — the UK's accounting recognition and wind-up regulation, the Netherlands's explicit parameterization of reserves, and Australia's contractual nature of mandatory inflows — leads to allocation consequences that diverge, and even run in opposite directions. These background facts together point to a judgment: taking discrete institutional types as the starting point of analysis can neither explain the different consequences of the same shock across different jurisdictions, nor accommodate the large number of intermediate forms, still less answer the mechanistic question of how institutions lead to allocation. Describing what an institution "is" does not amount to explaining "why" its allocation is as it is, and it is the latter that this paper seeks to advance. The next chapter will build on this to establish the theoretical framework of the continuous spectrum, providing a common theoretical coordinate system for the five axes of analysis.

Chapter 3 Theoretical Framework: The Institutional Continuum and the Identification Distinction

3.1 Two State Variables: Degree of Risk-Sharing and the Distribution Kernel

This paper proposes to characterize the morphology of pension institutions using a pair of continuous state variables. The first variable is the degree of risk-sharing, denoted s, taking values on the closed interval [0,1], measuring the share of risk borne across generations or by the sponsor. s equal to 0 corresponds to pure individual bearing, i.e., the textbook DC corner; s equal to 1 corresponds to full intergenerational or sponsor bearing, i.e., the textbook DB corner. Real-world institutions almost never sit at the corners: the Netherlands' solidarity reserve, the UK's CDC, and public-sector "last-person-bears-it" arrangements are all points within the continuum. This observation is not a minor taxonomic patch but a substantive challenge to the binary framework — if the vast majority of real-world institutions fall within the continuum, then categories defined by the corners are no longer the natural unit for describing reality, but rather two special cases extracted from the continuum. The second variable is the distribution kernel, denoted φ — the rule that allocates the gains and losses of the collective buffer across generations, between active and retired members, and among members with different degrees of organization, including parameters such as the buffer ceiling, the smoothing window, and intergenerational weights. The value of φ is not an actuarially given technical constant but a political target set through negotiation among social partners: how high the buffer ceiling is set, how long the smoothing window is stretched, and whether intergenerational weights favor active or retired members — the setting of each parameter is an expression of a distributive stance, not a technical calibration.

Under this parameterization, the DB/DC dichotomy is restated as degenerate corners of the (s, φ) continuum. It must be made clear that the value of this re-parameterization proposition lies in its organizing power and downstream reusability, not in its falsifiability — it is a framework-level restatement that provides a single common coordinate for institutional taxonomy for this paper and subsequent research, and does not itself bear the burden of empirical identification. This point must be stated honestly, lest readers mistake this paper as asserting an empirical finding testable against data: restating institutions as a continuum is not a falsifiable hypothesis but a re-selection of the unit of analysis, whose justification lies in whether it better organizes cross-country variation, better accommodates hybrid forms, and better serves downstream judgment. The Netherlands' new system explicitly writes the ceiling and drawdown rules of the solidarity reserve into legislation (see Chapter 2, Section 2.2), which is precisely evidence that s, as a continuously adjustable dial, is made explicit in real-world institutions (DNB 2023): when a law decomposes "the share of intergenerational bearing" into a set of adjustable numerical parameters and specifies their ceilings one by one, the degree of risk-sharing as a continuous variable ceases to be a theoretical fiction and becomes an actual dial in the hands of legislators.

This continuum framework also provides a shared coordinate system for understanding the subsequent five chapters. Chapter 4's put hardness is an option-theoretic characterization of the "risk stopping point" on the continuum; Chapter 5's degree of risk-sharing and distribution kernel are the ontological definition of the continuum itself; Chapter 6's measurement-visibility layer explains how movements on the continuum are filtered into visible signals; Chapter 7's systemic complementarity embeds the continuum within a larger institutional configuration; and Chapter 8's mandate strength is the projection of the continuum onto the inflow side of the asset end. The five axes share the premise of a "continuum," and on each axis the DB/DC dichotomy degenerates into the corner of the corresponding continuous variable — this sharing is not a coincidence but the consistent application of this paper's position that institutions should be understood as continuous state variables.

3.2 De-risking as Intergenerational Wealth Redistribution

Within the continuum framework, "de-risking" is reunderstood as a political displacement along the s continuum toward 0, rather than the elimination of risk. The key to this reunderstanding lies in the question of the property-rights ownership of the value of the collective buffer. The collective buffer is driven up during periods of low interest rates and high valuations, but its ownership often exists in a vacuum: it is neither a vested entitlement of active members, nor a defined benefit of retired members, nor an asset the sponsor may freely withdraw. This property-rights vacuum is not an oversight in institutional design but an intrinsic feature of collective pension arrangements — the very reason the collective buffer can perform its function of intergenerational smoothing is precisely that it is not monopolized by any single generation; but precisely because it is not monopolized, its ownership becomes an unsettled, negotiable question at every point of institutional change. Precisely because property rights remain undetermined, every movement along s is simultaneously a redivision of wealth between generations and between active and retired members — whether the value of the buffer is credited to the accounts of members present at the time (especially those close to retirement with high political organization) or retained as an intergenerational smoothing reserve carries opposite welfare signs. When an instance of "de-risking" credits the value of the buffer to the accounts of members present at the time, it is a one-time gain for those present members, and simultaneously an implicit expropriation of generations not yet entered or absent; calling this process "de-risking" obscures its distributive substance. This assertion is independent of any comparison between institutional types; it is a normative insight that this paper can defend, and its empirical vehicle — to whom the value of the buffer flows during a transition — exists and is observable.

The power of this normative insight lies in reframing "de-risking," which is typically framed as a technical-prudential issue, as a distributive-political issue. In standard prudential discourse, de-risking is responsible fiduciary conduct: reducing the volatility of the funding gap, protecting the certainty of benefits, and reducing pressure on the sponsor's balance sheet. These statements are not false, but they systematically omit a question — to whom the value of the buffer flows when it is redistributed in the process of de-risking. This paper does not argue that de-risking is always improper, nor that the value of the buffer is always credited to accounts of members present at the time; what this paper argues is that any de-risking decision is simultaneously a distributive decision, and this distributive dimension should not be obscured by prudential discourse. Recognizing this is the key step in advancing pension institution analysis from the technical level to the political economy level. It is also this insight that directly connects Chapter 5's property-rights ontology with the focal judgment of this paper as a whole — who bears the risk that has been transferred out: the value of the buffer within the property-rights vacuum is precisely a concrete form of that transferred risk, and the distribution kernel φ, which determines to whom it flows, is precisely the object of the political-economic contest.

3.3 The Identification Distinction Between Direction A and Direction B

Throughout this paper, a pervasive identification distinction is adopted, one that is the methodological linchpin of this paper. Direction A refers to the causal effect of institutional morphology on allocation outcomes (institution → allocation): given that the institutional morphology has already been exogenously determined, its consequences for allocation are examined. Direction B refers to the reverse reshaping of institutional morphology by allocation pressure (allocation pressure → institution): examining how current-period allocation and political pressure reciprocally determine the morphology and displacement of institutions. These two directions correspond to the two halves of the focal judgment stated in Chapter 1 — Direction A is "institutions determine the boundaries of allocation," and Direction B is "allocation pressure reshapes institutions." The two operate simultaneously and are intertwined in reality, but they occupy fundamentally different positions with respect to the feasibility of identification.

This distinction is methodologically decisive. Direction A has a credible identification path: when variation in institutional morphology arises from sources exogenous to a given plan's current-period allocation pressure (such as the staggered adoption of legislation across different jurisdictions or industries), this exogenous variation can be used to identify the causal effect of institutions on allocation; in this case, the arrow points from institution to allocation, and the exogenous variation pins down the direction of the arrow, allowing identification to hold. Direction B, by contrast, generally lacks clean identification: institutional transitions are endogenous to current-period allocation pressure, and any estimate pressed onto a transition event struggles to separate the two directions — a transition may be both the result of accumulated past allocation pressure and a precondition for future allocation behavior; event studies cannot separate "pressure forcing out institutional change" from "allocation following institutional change."

A trap that must be especially guarded against is treating the political causes of institutional change as an exogenous instrument. On the surface, the fact that one jurisdiction completes a piece of legislation before another appears exogenous in timing and thus usable as an instrument; but in reality, this timing is precisely the product of allocation pressure and political bargaining. Using it as an instrument to identify "institution → allocation" in fact still identifies the downstream consequence of the common antecedent — allocation pressure — rather than the independent effect of institutions on allocation; more seriously, if this is claimed to identify Direction B (allocation pressure → institution), then identification-bearing language is being used to dress up a causal relationship that is fundamentally unidentified — because the political process itself is precisely a manifestation of allocation pressure, not an exogenous shock independent of it. Therefore, this paper's treatment of Direction B is a consistent, honest downgrade — presenting it as political-economic theory and case illustration, without using identification-bearing language, and in particular avoiding treating the political causes of institutional change (who lobbied, why a given jurisdiction legislated first) as an exogenous instrument. This self-restraint is not conservatism but the baseline of identification honesty: it is better to honestly leave Direction B at the theoretical level than to use identification-bearing language to mask the fact that it is unidentified.

3.4 The Discipline of Honest Boundaries

From the distinction above derives this paper's discipline of honest boundaries, which constrains all empirical statements throughout the text. First, wherever subsequent chapters label something "micro-data unavailable / directional," it must not be upgraded in the text to "identified / causally confirmed." This targets a common temptation: when the identification design is in place, an exogenous source of variation genuinely exists, and directional evidence is also supportive, quietly rewriting "design in place" as "identified," and "directionally consistent" as "causally confirmed." This paper explicitly rejects this slippage — a design being in place does not equal identification being achieved, and directional consistency does not equal causal confirmation. Second, the falsifiability thresholds pre-set in each chapter must be preserved in full in the main text and must not be loosened after the fact in pursuit of a stronger conclusion; the value of a falsification threshold established prior to the data lies precisely in the fact that it does not change according to the researcher's ex post preference, and therefore this paper fully lists these thresholds in the evidentiary-status section of every chapter, including those that could overturn the chapter's judgment once the data appear. Third, any cited, publicly verifiable institutional aggregate fact is marked with its source via an MLA in-text parenthetical citation together with its evidentiary status; whenever the data required are unavailable under the conditions of this research, it is explicitly marked that the identification design is in place but the evidence remains to be obtained. This discipline is not a modest posture toward the strength of conclusions but an intrinsic requirement of this paper's methodological stance: the core of pension research is judgment, and data serve to test judgment, not the reverse; the value of a piece of research lies first in whether its judgment is sharp and falsifiable, and only then in whether its data are already in place. Reversing this order — obtaining the data first and then seeking a judgment — is precisely what this paper seeks to avoid.

3.5 Summary

This chapter has established a continuum framework with the degree of risk-sharing and the distribution kernel as state variables, restated the DB/DC dichotomy as degenerate corners, reunderstood "de-risking" as intergenerational wealth redistribution under a property-rights vacuum, and established the Direction A/Direction B identification distinction and the discipline of honest boundaries that run throughout the paper. This framework provides the subsequent five chapters with a shared theoretical coordinate system and shared methodological constraints: theoretically, the five chapters share the ontological premise of the continuum; methodologically, the five chapters share the Direction A/Direction B identification distinction and the honest boundary that "a design being in place does not equal identification being achieved." Within this framework, the following five chapters will unfold specific mechanism analyses and identification designs along five mutually orthogonal margins — option hardness on the liability side, the continuum and buffer property rights, measurement visibility, systemic complementarity, and the contractual nature of inflows on the asset side — each chapter following the unified structure of "judgment — mechanism — identification strategy — honest labeling of evidentiary status."

Chapter 4 The Risk Stopping Point and the Hardness of the Writer of the Residual Guarantee Option

4.1 Judgment

The core identifiable claim advanced in this chapter (confined to Direction A) is: the sign of the configuration—DB de-risking and DC aggressiveness—under low interest rates is causally driven by the hardness of the option position of the writer of the residual guarantee put, and this hardness can be orthogonal to, or even the reverse of, the nominal risk preference of the bearing party. The single proposition this chapter claims to have cleanly identified is this: exogenous institutional differences that cut across put hardness (such as the staggered adoption, across jurisdictions, of legislation authorizing trustees to consider non-shareholder interests) causally move the equity share and duration of the configuration along Direction A, and the sign exhibits a counterintuitive feature—the harder the guarantee, the more de-risking; the softer the guarantee, or the more it is skewed toward the tail, the more risk-taking.

The sharpness of this claim lies in its shift of the driver of the configuration's sign away from "who bears the risk" and toward "how hard the option position is at the stopping point." Conventional analysis expects that who bears the risk determines the configuration tendency—DB, with the employer bearing risk, is conservative; DC, with the individual bearing risk, is aggressive. This chapter's claim, instead, is that the sign of the configuration can be orthogonal to, or even the reverse of, the nominal bearing party: a DB plan nominally underwritten by the employer will de-risk if its put is written by a party subject to hard recourse (a hard put); a DC plan that has nominally transferred the risk may, if its stopping point is softened by legal safe harbor (a soft put), see its default configuration pushed toward a higher equity center of gravity instead. This counterintuitive sign—the harder the guarantee, the more de-risking; the softer the guarantee, the more risk-taking—is precisely the incremental judgment by which this chapter departs from the textbook narrative of "the risk-bearer is conservative, the risk-transferee is aggressive."

Elements that must honestly be downgraded to theoretical propositions, not identified by this chapter, include the following. First, "the nature of the stopping-point option—that is, who truly writes the stopping point of the residual put—is itself the political-economic cause": this is the mechanism theory explaining "who ends up bearing the risk that was ostensibly transferred," and it is this paper's contribution to the focal point of judgment; but its causal side (how the institution comes to stop the put at this particular point, and whether configuration pressure reshapes that stopping point in reverse) belongs to Direction B, and the chapter offers no clean identification there, only theory and case illustration. This downgrade does not blunt the chapter's identificatory sharpness: the chapter's identifiable core claim stops at the consequence side (given put hardness, the configuration sign follows), while the deeper causal inquiry into "why the put is politically stopped at this point" is presented as normative argument within political economy—the ultimate writer of the soft put is the silent taxpayer and the absent generation, and this identification is itself an answer to "who ends up bearing the risk that was transferred," albeit a theoretical answer rather than an econometrically identified one. Second, "the intergenerational option held by the living generation over the absent generation" under collective defined-contribution arrangements is downgraded to a mechanism hypothesis; the net present value of its intergenerational transfer must be characterized by an overlapping-generations model, the evidence leans toward inference, and no identifiable claim is made.

4.2 Mechanism: The Endogenous Stopping Point on the Principal–Agent Chain

The mechanism of this chapter can be stated as a single causal chain. Low interest rates raise the present value of DB liabilities and widen the funding gap; residual risk is pushed along the chain "sponsor → trustee → beneficiary → taxpayer → absent generation" toward different stopping points; the hardness of the put writer at the stopping point determines the sign of the configuration. The key concept here is the "stopping point": residual risk does not remain with the nominal bearer designated by the institutional type, but travels along the principal–agent chain until it meets a party that can (or is forced to) write the residual guarantee put. Where this stopping point lies depends on the legal and contractual structure of each link in the chain, not on the institution's nominal label.

In the hard-put case (strong employer-recourse contracts, contractual trigger ladders, equity-credit rules that cause rating reclassification as debt), the writer of the put faces genuine funded-level recourse and has an incentive to compress gap variance, so the higher the funded ratio, the more de-risking occurs, producing de-risking in the mean segment. The intuition behind this logic is: when the residual put is written by a party subject to hard recourse, that party bears the entire downside of a widening gap without enjoying the symmetric upside of a narrowing gap (a fully funded plan does not return surplus to the employer), so its dominant strategy is to compress gap variance—that is, to de-risk. A contractual trigger ladder (e.g., a gap reaching a given threshold triggers additional contributions or collateral requirements) concretizes this hard recourse into a set of discrete penalty thresholds; the closer to the threshold, the stronger the pressure to de-risk. In the soft-put or tail-put case (weak employer, "last person standing," public backstop, silent taxpayer, the intergenerational pool under collective defined contribution), the put holder has an upside option motive and tends to retain or even add risk, flipping to risk-taking in the extreme distress segment. The intuition behind this flip is: when the ultimate writer of the residual put is a party that will not actively seek recourse (future taxpayers, an absent generation, or a pool backstopped by a public guarantor), the current decision-maker in effect holds a call option—if the aggressive configuration succeeds the plan is saved, and if it fails the loss falls on the silent writer; this asymmetric payoff structure makes risk-adding, rather than de-risking, the dominant strategy in the extreme distress segment, which is precisely the behavioral manifestation of the so-called "pension put."

On the DC side, legal choice (the safe-harbor liability shield) moves the stopping point from the individual back to the institution, pushing up the equity center of gravity of the default glide path. The key point here is that a "soft put arising from legal choice" already exists on the DC side itself, which means DC is itself an object of put analysis rather than a clean "put-free" control—a recognition that directly determines the adjustment to the identification strategy that follows. Specifically, when the law provides a process-based liability safe harbor for default investment choices (i.e., a trustee is shielded from outcome liability as long as the prescribed process was followed), the party that sets the default glide path in effect obtains a soft put: it can push the equity center of gravity higher in pursuit of higher expected returns without bearing personal liability for adverse outcomes, because that liability has been absorbed by the safe harbor. The existence of this soft put means that the DC default configuration is not determined purely by participant risk preference, but is likewise subject to the option structure at the stopping point—except that DC's stopping point is shaped by legal choice rather than by contractual recourse.

The a priori theory of option pricing provides the theoretical foundation for the orthogonality described above: the sign of the configuration is a function of the difference in put hardness between the two sides, and can be orthogonal to the nominal risk preference of the bearing party. Under the classic result of put option pricing, once the put writer is fixed, low funding plus maximal risky-asset weighting is the dominant strategy for the option holder—a result independent of "whether the employer is risk-averse" (Sharpe 1976; Treynor 1977). This orthogonality is the theoretical pivot of this chapter's identifiable claim: it implies that the sign of the configuration can be driven by put hardness rather than by risk preference, so that empirically, put hardness should still significantly predict the direction of the configuration after controlling for risk preference—this constitutes a falsifiable test, and is also the key point by which this chapter departs from the trivial explanation that "different parties simply have different risk preferences."

4.3 Identification Strategy

The identification design of this chapter distinguishes a primary strategy, alternative strategies, and a downgrade exit. The primary identification source (Direction A, cleanest on the DB side) is the staggered adoption, across jurisdictions, of relevant legislation as an exogenous negative shock to put hardness. Beginning in the mid-1980s, a total of 35 U.S. states adopted, in staggered fashion, constituency statutes authorizing directors to consider non-shareholder interests (Atif, Nadarajah, and Richardson 2023); published literature uses this staggered adoption as a quasi-natural experiment and finds that adoption lowered the cost of debt and loan spreads, with the mechanism explicitly stated as mitigating the conflict between residual claimants (shareholders) and fixed claimants (creditors) (Gao, Li, and Ma 2021). This legal rebalancing of "residual claim versus fixed claim" is precisely the exogenous source of variation this chapter requires: authorizing trustees to consider non-shareholder interests weakens shareholders' hard recourse over the pension put, softening the put. The identification logic is: the timing of adoption across states is determined by each state's legislative agenda, not by the current funded status of any specific plan, so the timing of adoption is exogenous relative to the configuration pressure of any individual plan; the difference-in-differences in configuration between the treatment and control groups before and after adoption identifies the causal effect on configuration of an exogenous decline in put hardness. On this basis, a cross-jurisdictional difference-in-differences design can be constructed, with pre-adoption parallel trends serving as a test of the identifying assumption—if the treatment group already exhibits an abnormal configuration trend prior to adoption, exogeneity is called into question and must be downgraded accordingly.

Alternative identification sources include: using reform of the pension insurance premium formula as a quasi-experiment for internalizing put pricing, and, within public plans, using elected versus appointed boards and proximity to elections as a mediating identification. The logic of the former is: when pension insurance premiums shift from a flat rate to a risk-linked rate, the soft put originally borne by the insuring institution (and, behind it, the collective of premium payers) is explicitly priced, thereby weakening the upside option motive of the put holder; the theoretical expectation is tail de-risking—that is, internalizing premium pricing converts a soft put that originally encouraged risk-taking into a priced, and therefore risk-suppressing, position. The logic of the latter deserves particular explanation: in public DB plans, the ultimate writer of the residual put is the future taxpayer, a textbook soft put; the theoretical expectation is that the closer to an election, the stronger the incentive for an elected board to conceal the gap by raising the discount rate and increasing equity and alternative allocations (because making the gap explicit carries a political cost for incumbents), while an appointed board, not directly facing electoral pressure, should be decoupled from this motive. This contrast of "elected versus appointed, moderated by electoral proximity" provides an internal, falsifiable identification design for the behavior of the soft put—if the increase in discount rate and in equity/alternative allocation co-move within the same governance body and intensify with electoral exposure, while appointed trustees remain decoupled (serving as a placebo), the behavioral mechanism of the soft put is supported; otherwise it is weakened.

On the DC side, given that the clean control of "DC has no put" has been contaminated by this chapter's own theory, the identification strategy is modified accordingly: the approach abandons using "DC does not de-risk" as a control for "the absence of a hard put," and instead uses a continuous measure of put hardness constructed uniformly across DB/DC and independent of the configuration outcome (safe-harbor strength plus whether the default fund carries an implicit public backstop guarantee) to test the sign proposition within DC itself; the primary identification still relies on an event study around the exogenous timing of safe-harbor legislation and case law, with the measure used only to characterize treatment intensity and not to bear the identification load. The methodological point of this modification is that a placebo contaminated by theory ("DC has no put") should not continue to serve as a pillar of identification—logical consistency should take priority over convenience of identification; using a continuous measure constructed independently of the configuration outcome, and testing the sign within DC rather than across DB/DC, circumvents the circularity, because the DC-internal distinction between hard put and soft put (strong safe harbor plus no implicit backstop, versus weak safe harbor plus an implicit backstop) does not depend on a cross-institutional label contrast. In addition, this chapter sets a downgrade exit for this modification: if the continuous measure of put hardness cannot be cleanly constructed (for instance, if the "implicit safety net" cannot be coded orthogonally to configuration willingness), then identification on the DC side is explicitly marked as "not identified, design pending," and the DC side serves only as a qualitative argument that "a legal-choice soft put exists," not as a clean placebo; in that case the identifiable core claim retreats to the DB-side stakeholder-legislation difference-in-differences alone, which does not depend on the DC measure and stands independently. This downgrade exit is established prior to the data, ensuring that even if construction of the DC-side measure fails, the chapter's core identifiable claim (on the DB side) does not thereby lose its footing.

4.4 Robustness Counterarguments and Responses

Counterargument One (Identification Circularity): The sign identification of "the harder the guarantee, the more de-risking" depends on a clean measure of put hardness; but this chapter maintains that DC also has a legal-choice soft put, so the originally clean control of "DC has no put" is contaminated by its own theory, and identification falls into circularity—the measure used to determine hardness and the configuration sign being explained may be measuring the same thing. In response, this chapter's treatment is as follows: it accepts the counterargument, abandons using "DC has no put" as a clean control, and instead uses a measure constructed independently of the configuration outcome to test within DC; primary identification instead relies on the staggered adoption of exogenous institutional differences, so that the direction of the arrow from "hardness → configuration" is pinned down by the exogenous timing of institutional change; and a downgrade exit is established—if the continuous measure of put hardness cannot be cleanly constructed, identification on the DC side is explicitly marked as unidentified, and the identifiable core claim retreats to the DB-side stakeholder-legislation difference-in-differences alone (which does not depend on the DC measure and stands independently).

Counterargument Two (Overreach into the Causal Side and Confounding with Scale): First, what this chapter really wants to say is that "the nature of the stopping-point option is itself the political-economic cause," but this belongs to Direction B; even if the staggered timing of stakeholder legislation is exogenous, what it identifies is only the configuration consequence given an already-fixed institutional form (Direction A), and it cannot identify the more ambitious causal-side claim that configuration pressure reshapes the put stopping point in reverse; treating the political cause of legislative adoption as identification of the causal side would fall into the trap of dressing up unidentified causation in the language of identification. Second, the counterintuitive sign of "the harder the guarantee, the more de-risking" might merely reflect a confound of scale and maturity—plans with hard puts tend to be larger and more mature, and de-risking could be explained by duration-matching needs rather than by the option position. In response, this chapter's treatment has two parts. On overreach into the causal side, the chapter strictly maintains the directional distinction: the causal side is uniformly downgraded to theory and case illustration, the core claim asserts only the consequence side, and the political cause of legislative adoption is never treated as an exogenous instrument for the causal side; "why the put is politically stopped at this point" and "whether configuration pressure reshapes the stopping point in reverse" are explicitly marked as normative arguments within political economy (the ultimate writer of the soft put being the silent taxpayer and the absent generation), with no clean identification anywhere in the chapter. Only if some future study finds exogenous variation "independent of contemporaneous configuration pressure, with the arrow running unambiguously from configuration pressure to the put stopping point" can the causal side be elevated from case illustration to identification; until then, this chapter refuses to describe the causal side in identificatory language. On the scale confound, this chapter explicitly controls for the log of asset scale, plan maturity, and funded ratio in the identification specification, recasting the core test as "put hardness still significantly predicts the direction of de-risking after controlling for duration-matching needs," and uses tail reversal (risk-taking in the extreme distress segment) as a falsifiable feature that a pure duration story cannot produce—a plan driven purely by duration-matching needs would not flip to risk-taking in the extreme distress segment, so if tail reversal appears, it is evidence unique to the option mechanism rather than to duration matching; conversely, if the sign does not reverse in the tail and the magnitude of de-risking is fully absorbed by scale and duration, the option mechanism is falsified and the explanation reverts to duration matching. This treatment converts the sharpest alternative explanation (duration matching) into a competing hypothesis that can be adjudicated by data, rather than evading it through wording.

On the boundary with other axes of analysis, this chapter draws an explicit division of labor with Chapter 5: this chapter addresses "with whom the risk stops, and how put hardness determines the sign of the configuration"—a perspective of principal–agent stopping points and option positions; Chapter 5 addresses "the degree of risk-sharing itself as a continuous dependent variable, and to whom the property rights of the buffer belong"—a perspective of ontology and property rights. The intergenerational option under collective defined contribution belongs to this chapter (as a mechanism hypothesis); the continuum and buffer property rights belong to Chapter 5. This division of labor ensures that the two chapters characterize two orthogonal facets of the same institutional form, rather than overlapping accounts.

4.5 Evidentiary Status and Pre-Registration Comparison

The verifiable layer of this chapter has been confirmed. The staggered adoption of stakeholder legislation (35 U.S. states adopting in staggered fashion beginning in the mid-1980s, with the mechanism being a rebalancing of the conflict between residual claimants and fixed claimants; published literature using this staggered adoption as a quasi-natural experiment finds that adoption lowered the cost of debt and loan spreads) has been independently verified as a publicly verifiable fact (Atif, Nadarajah, and Richardson 2023; Gao, Li, and Ma 2021); the theoretical lineage of put option pricing (that once the put writer is fixed, low funding plus maximal risky-asset weighting is the dominant strategy for the option holder) holds directionally (Sharpe 1976; Treynor 1977); the factual foundation of UK DB de-risking following the balance-sheet recognition of liabilities under accounting standards (liability recognition on the balance sheet, plus the requirement since June 2003 that wind-up be funded to buyout level) serves as directional background, jointly exerting hard-recourse pressure on the sponsor, but this fact is collinear with the visible leg and the real-constraint leg, and is not attributed to put hardness alone (British Progress 2024). These verifiable facts confirm that the exogenous source of variation required by this chapter's load-bearing leg does indeed exist in reality, and also confirm that the theoretical foundation of the option mechanism holds up; but they are aggregate institutional facts and do not by themselves constitute identification of the causal effect of put hardness.

The micro-identification layer of this chapter, however, could not be executed, owing to data availability constraints. A fund-level panel of put hardness (self-constructed coding of contractual trigger ladders, safe-harbor strength, and implicit public backstops) is not obtainable under the conditions of this study—contractual trigger clauses are largely private (requiring regulatory data or an actuarial consulting sample), safe-harbor case-law strength requires self-constructed coding with the relevant litigation numbering only in the single digits, and cross-country comparability is weak—so the primary difference-in-differences equation cannot be estimated, and this paper reports no regression coefficients, significance levels, or sample sizes. This honest gap must be explicitly flagged: the core identifiable claim of this chapter (that put hardness drives the sign of the configuration) remains, under current conditions, at the stage of "identification design in place, evidence pending," not "already proven." What can be confirmed is that the identification design (staggered-adoption difference-in-differences, pre-adoption parallel-trends test, and controls for the confounds of scale, maturity, and duration matching) is in place, and that the exogenous source of variation required by the load-bearing leg (staggered adoption of stakeholder legislation) does indeed exist in reality and has been independently verified (Atif, Nadarajah, and Richardson 2023); its execution must await the commissioning of micro-level data.

Correspondingly, the pre-specified falsifiability thresholds are fully preserved, and this preservation is critical—it means that when micro-level data eventually become available, the chapter's judgment will still be held to the tests set at the outset, unrelaxed by the data. Specifically: the primary sign expectation is that the treatment group's equity share rises relatively following adoption of stakeholder legislation (the put turning softer leads to risk-taking, i.e., the reverse verification of "harder implies more de-risking"); tail reversal (de-risking in the mean segment, risk-taking in the extreme distress segment) is the key falsifiable feature—if the tail does not reverse, the option mechanism is falsified; the coefficient on put hardness must remain significant after controlling for risk preference, and the institutional dummy variable's coefficient should collapse once hardness is added; if the magnitude of de-risking is fully absorbed by scale and duration and the tail does not reverse, the option mechanism is judged to have lost to the duration-matching explanation; if the continuous measure on the DC side fails to be constructed, it is marked as unidentified and does not count as counter-evidence against the DB-side main conclusion. The causal side (Direction B) presets no identificatory expectation whatsoever; that the writer of the soft put is the silent taxpayer and the absent generation is presented as normative argument, not as an identified claim. The full comparison between this chapter's pre-registered thresholds and its evidentiary status has been itemized in this section; comparable comparisons for the other analytical axes appear in the respective evidentiary-status-and-pre-registration-comparison sections of Chapters 5 through 8; the cross-axis comparison of core mechanisms and primary identification sources is given separately in Table 1 at the end of the text (a mechanism-comparison table, which does not carry an item-by-item summary of pre-registration thresholds).

4.6 Summary

This chapter resets the "given risk attribution" implied by institutional type as an endogenous stopping point pushed out along the principal–agent chain by low interest rates, and demonstrates that it is option hardness (rather than nominal attribution) that governs the sign of the configuration. This is an identifiable Direction A increment that directly answers the consequence side of the focal judgment question, "who ends up bearing the risk that was transferred"—the ultimate writer of the soft put is the silent taxpayer and the absent generation, and the sign of the configuration is determined by the option hardness at the stopping point, which can be orthogonal to, or even the reverse of, the nominal bearing party. This chapter's identifiable core claim stops at the consequence side (given put hardness, the configuration sign follows); its causal side (why the put is politically stopped at this point) honestly remains at the level of theory and case illustration, without being dressed in identificatory language; owing to data availability constraints, the micro-causal coefficients on the consequence side likewise remain at the stage of "identification design in place, evidence pending." The option-position perspective of this chapter and the property-rights ontology perspective of the next chapter operate on two orthogonal facets of the same institutional phenomenon: the former characterizes the risk stopping point and the sign of the configuration, the latter characterizes the continuum itself and the buffer property rights.

Chapter 5 The Continuous Spectrum, Buffer Property Rights, and Intergenerational Redistribution

5.1 Judgment

Building on the framework of Chapter 3, this chapter further argues that the risk-sharing degree s and the distribution kernel φ are the theoretical first-order variables of pension institutions, while DB/DC are degenerate corner cases; "de-risking" along the s spectrum is intergenerational wealth redistribution to the currently present generation, not risk reduction. The judgment has three layers, each honestly labeled as to its standing. The first layer is a framework proposition: DB/DC are degenerate corners of the (s, φ) spectrum; this proposition serves as the single source of truth for this paper's taxonomy, carries no empirical identification burden, and its value lies in organizing power and downstream reusability. The second layer is a normative insight: movement along the s spectrum is not necessarily welfare-reducing risk reduction, but is often a one-off wealth redistribution to the currently present generation; this claim is independent of any collapse test and is independently falsifiable. The third layer concerns the identifiable consequence-side mechanism of Direction A: once s is fixed, its causal consequences for allocation behavior are identifiable — funds with thicker reserves exhibit lower allocation elasticity in response to interest-rate shocks; sponsors whose bearing capacity is closer to the threshold are more likely to move along the spectrum toward 0.

5.2 Mechanism

The existence and location of the spectrum (theoretical skeleton, not causal identification). Low interest rates raise the present value of liabilities, and the marginal cost to the sponsor of maintaining high s rises sharply, prompting the sponsor to use its bargaining power to push s toward 0. This "de-risking" is not the elimination of risk but a political displacement along the (s, φ) spectrum, with the direction of displacement determined by the bearing-capacity threshold.

Property-rights vacuum and redistribution (the core of the normative insight). The property-rights vacuum of the collective buffer is the core of the mechanism. The buffer is inflated during periods of low interest rates and high valuations, but its ownership is not clearly defined — it is neither the vested entitlement of active members, nor the defined benefit of retired members, nor an extractable asset of the sponsor. Precisely because property rights are undetermined, every movement along s constitutes a re-slicing of wealth between generations and between active and retired members. Wholesale collective conversions provide an observable vehicle for observing this redistribution: at the point of such conversions, the value of the buffer is rewritten into accounts, and its destination (credited to active members' accounts, retained across generations, or used to lower the employer's future contributions) can be coded.

Self-destructive endogenous loop (case study, Direction B, not carrying causal burden). A sponsor's use of bargaining power to push s toward 0 raises, under accounting rules, the present value of liabilities and widens the funding gap, which imposes further de-risking pressure, until the plan closes. This is a self-reinforcing dynamic: de-risking is supposed to relieve pressure, but under mark-to-market accounting, de-risking (shifting toward long-duration liability-matching bonds) instead raises valuation exposure to liability-sensitive assets, or otherwise widens the accounting gap, thereby triggering pressure for further de-risking, cycling in this way until the plan closes. This is the internal DB dynamics of the endogenous movement of s; this chapter explicitly labels it as the theory and case study of Direction B: conversion and closure events are endogenous to the current-period allocation pressure, and event studies cannot separate "pressure forcing institutional change" from "allocation following institutional change," so this chapter does not attempt causal identification here, offering only a case presentation.

Why the dichotomy persists (meta-level theoretical cap, not carrying empirical burden). A natural question is: given that the DB/DC dichotomy has been empirically falsified by a great many hybrid forms, why does it nonetheless persist tenaciously in discourse and practice? This chapter's answer is: the DB/DC dichotomy persists as a depoliticizing technology. It conceals the distributive struggle (the value of φ — who gets the buffer, how intergenerational weights are set) within the label's implicit parameters, presenting what is essentially a political negotiation over "who bears the risk, who gets the buffer" as an apparently technically neutral institutional choice. When a struggle over intergenerational distribution is packaged as a technical decision of "choosing DB or DC," the distributive substance of the struggle is removed from the negotiating table — this is precisely the function of the dichotomy as a depoliticizing technology. This explains the persistence of the dichotomy: it serves to conceal φ from open political negotiation. This chapter adopts this judgment as a theoretical cap at the level of institutional reproduction, carrying no empirical burden.

5.3 Identification Strategy and Three Separable Empirical Objects

This chapter designs three mutually independent empirical objects that do not share identification.

Model A (normative insight: the distributive consequences of de-risking). The goal is to test whether "de-risking along the s spectrum is equivalent to wealth transfer to the currently present generation," and whether the direction and magnitude are systematically determined by observable features of the distribution kernel φ. The identification strategy is a cross-section of conversion events plus heterogeneity by φ features; this chapter does not claim clean exogeneity, explicitly labeling this as descriptive and heterogeneity characterization, with falsifiability coming from the directional constraint of the φ→direction mapping rather than from an exogenous instrument. The purpose of this design is to build the challenge of "treating a case as a regularity" (see Counter-argument 2 below) directly into the design — not settling for the single-point observation that "one particular conversion did in fact credit the buffer to active members' accounts," but requiring that the direction of buffer flow vary systematically with observable features of φ, thereby upgrading the normative insight from an isolated case to a falsifiable directional proposition. The data are a sample of wholesale collective conversions, with each event coded for the change in the value of active members' accounts in the conversion year, the destination of the buffer's value (three categories: credited to active members' accounts, retained across generations, or used to lower the employer's future contributions), and features of φ (the proportion of representation of governance body members, whether design authority rests with the employer or with joint trustees, and a proxy for the degree of political organization of members near retirement). In terms of specification, the change in the value of active members' accounts is regressed on governance representation, the locus of design authority, and control variables, supplemented by a multinomial regression of buffer-flow destination on φ features.

Model B (reserve absorption elasticity, Direction A, identifiable). The goal is to test the elasticity of allocation to interest-rate shocks once s is fixed — the thicker the solidarity reserve, the more blunted the equity and duration allocation response to interest-rate shocks. The intuition is: a fund with a thick reserve can use the reserve to absorb the short-term volatility caused by interest-rate shocks without having to immediately adjust its allocation; a pure individual DC scheme with no reserve lacks this buffer, and its allocation must bear the shock directly. The identification strategy is Direction A (institution→allocation): cross-fund differences in reserve thickness plus the time series of interest-rate shocks, with reserve thickness driven by legislated or contractual buffer ceilings, exogenous to any single fund's current-period allocation preferences; the quasi-exogenous movement of s is identified only at points where legislative timing is independent of the fund's own allocation pressure, using legislative staggering as a discontinuity (such as legislative changes to regulatory buffer ceilings, or mandatory timing of industry-wide collective agreement renewals). It must be made clear here that this legislative-staggering identification targets Direction A (legislation renders the institutional form of s exogenous, and its allocation consequences are tested), not Direction B — the legislative timing renders the institutional parameter of reserve thickness exogenous, with the arrow running from institution to allocation, which is precisely the definition of Direction A. The specification regresses the allocation response to interest-rate shocks on the interaction of reserve thickness and the interest-rate shock, controlling for fund fixed effects and time fixed effects; the interaction term is expected to show that funds with thicker reserves have a significantly smaller response magnitude than pure individual DC schemes with no reserve. The explicit parameterization of the Dutch solidarity reserve (not exceeding 15%, drawdown not exceeding 10%) provides institutional empirical evidence that s can be explicitly parameterized by legislation (DNB 2023); however, it should be noted that the Dutch solidarity reserve only took effect in 2023 and conversion is to be completed by 2028, so the early-period panel window for reserves and allocation is inherently short, and even if micro-data become available later, the power of Model B's early estimates will remain limited.

Model C (bearing-capacity screening, Direction A, identifiable). The goal is to test the effect of sponsor bearing capacity on the choice of direction of movement along the spectrum (maintaining DB, converting to CDC, hybridization, or closure). The identification strategy is Direction A: using the mechanical sensitivity of the contribution rate to the discount rate (determined by plan maturity and duration, exogenous to management's intentions) as an instrument for bearing-capacity pressure — for a highly mature plan, a decline in the discount rate mechanically amplifies the contribution obligation, and this amplification is unrelated to whether the sponsor subjectively wishes to de-risk, and can therefore serve as an exogenous proxy for bearing-capacity pressure. The data are a sponsor-level panel (contribution-to-payroll ratio, changes in credit rating, whether it is an unfunded public plan) paired with form-conversion events. The specification is a multinomial logit predicting the probability of form conversion using the bearing-capacity proxy, legislative availability, and industry fixed effects. This model carries a directionally explicit falsifiable sub-hypothesis: unfunded public plans, lacking balance-sheet constraints and thus lacking the mechanical pressure to be forced along the spectrum, should undergo significantly fewer form conversions; if the data show no systematic difference in conversion rates between public and private plans, the bearing-capacity mechanism is called into doubt. The Royal Mail's conversion to CDC can serve as a threshold anchor for bearing capacity approaching its limit triggering form conversion (though the specific contribution figures must be used with the correction noted below).

5.4 Robustness Counter-arguments and Their Treatment

Counter-argument 1 (the first-order status of s is a circular proposition): The most natural test of the first-order status of s is "after controlling for s, the explanatory power of the DB/DC label dummy variable collapses"; but any proxy for s necessarily includes the dimension of benefit contingency, and benefit contingency is precisely the defining distinction of DB/DC, so that after controlling for s, the collapse of the label's coefficient is bound to occur under any data. The problem with this test is that it is true regardless of how the world actually is — because the variance of the label has already been absorbed by s, what it proves is that "s and the label are measuring the same thing," not that "s is more first-order"; a test that is true regardless of how the world is, is not a testable implication but a tautology. This chapter accepts this counter-argument, does not resist it forcibly, and does not defend the collapse test: it removes the collapse test from the list of testable implications, downgrades the empirical first-order status of s to a theoretical proposition (a framework-level reparameterization, whose value lies in organizing power and downstream reusability), and shifts the empirical pillars entirely onto three non-circular legs — the normative insight (Model A, a welfare judgment that does not depend on "s being more first-order" and requires only observing who receives the buffer flow at the time of conversion), reserve absorption elasticity (Model B, Direction A, the allocation consequences once s is fixed), and bearing-capacity screening (Model C, Direction A, the effect of bearing capacity on the direction of movement along the spectrum). At the same time, this chapter leaves an exit that could reverse the conclusion for this counter-argument: if a measure of s independent of benefit contingency can be constructed (built purely from the buffer's property-rights structure and bargaining structure, such as the labor-management composition of the governance body, the statutory value of the buffer ceiling, and reserve drawdown rules, orthogonal to the label's definition), then the collapse test would be de-circularized, and the first-order status of s could be upgraded from a theoretical proposition to an empirical assertion; this chapter's default assumption is that such an independent measure is difficult to construct, and so the first-order status of s remains a theoretical proposition. This default is an honest default, not a pessimistic one — it faithfully reflects the fact that, under current identification technology, the first-order status of s cannot be cleanly tested, and converts this fact into an explicit upgrade path that future research can fill in.

Counter-argument 2 (treating a case as a regularity): Even if "de-risking equals wealth transfer" holds in individual conversions, it may simply be an incidental product of the particular conversion's design, rather than a general regularity of movement along the s spectrum — with a different φ (design authority resting with trustees rather than the employer, or more balanced representation), the same "de-risking" could very well retain the buffer as a cross-generational reserve, or even lower the employer's future contributions. If "the direction and magnitude of wealth transfer" is entirely determined by φ, and φ varies from case to case, then "de-risking equals transfer to the currently present generation" is not a normative regularity but merely "this time φ happened to be so." This chapter takes this challenge seriously and does not defend against it — it does not deny the observation that the buffer flowed to active members' accounts at the time of conversion, but it does deny the validity of extrapolating from that observation to a normative proposition. Accordingly, this chapter frames the modeling of the normative insight as a dose-response and heterogeneity design of "observable features of φ → direction and magnitude of wealth transfer" (i.e., Model A), rather than as a single-sign assertion of "de-risking → wealth transfer"; and it writes into the pre-registration a falsifiable downgrade threshold: if the direction of transfer indeed varies systematically with features of φ (the higher the representation, the larger the magnitude credited to active members' accounts; where design authority rests with the employer, contributions to the employer are more likely to be reduced), then the normative insight is upgraded to the stronger and falsifiable proposition that "the structure of φ determines the distributive consequences of de-risking"; if the direction of transfer bears no systematic relationship to features of φ, then the normative insight is downgraded to "cases exist where de-risking equals transfer, but the direction cannot be generalized." This bidirectional upgrade/downgrade threshold is written in advance, not an explanation tailored to the data after the fact, but a falsifiable structure established prior to the data that can be adjudicated by the data in either direction.

5.5 Evidentiary Status and Pre-registration Comparison

The verifiable layer of this chapter has been confirmed. The explicit continuous parameters of the Dutch solidarity reserve (not exceeding 15% of total assets, contribution injections not exceeding 10% of total contributions, and injections of excess returns not exceeding 10% of the current year's positive collective excess return, with the 15% ceiling to be met by 2037) have been verified as publicly verifiable facts (DNB 2023); the Future of Pensions Act took effect on July 1, 2023, all plans must comply by January 1, 2028, and accumulated entitlements must be transferred into new DC contracts via collective value transfer, which has been verified as a publicly verifiable fact (NautaDutilh 2023). Together, these two facts constitute the institutional evidence for this chapter's first- and third-layer judgments: the former directly confirms that the risk-sharing degree is explicitly parameterized as an adjustable dial in a real-world institution (the institutional vehicle for the framework proposition), while the latter provides an observable conversion-event vehicle for the normative insight (the value of the buffer being rewritten into accounts at the point of collective value transfer). The logical falsifiability of the normative insight — that "de-risking moving along the s spectrum is equivalent to a one-off wealth transfer to the currently present generation" — holds, and its empirical vehicle (who receives the buffer's value flow at conversion) exists and is observable — but it must be noted that logical falsifiability holding and the empirical vehicle existing do not amount to this normative proposition having been empirically confirmed; confirmation must await the micro-coding of Model A, which is not obtainable under the conditions of this study.

An honest numerical correction must be made here. There exists a widely circulated candidate figure for the normative insight's principal anchor number, namely that in a certain Danish industry-wide plan's wholesale conversion, "the buffer increased by 25% and was credited to active members' accounts"; independent verification has failed to confirm this specific value of "25%"; the closest verifiable fact obtainable is that the plan converted its entire with-profits scheme into a unit-linked scheme, with a one-off payment of approximately an amount equivalent to 20% of members' savings (not 25%, and phrased as a one-off payment rather than a buffer increase). This paper therefore downgrades this principal anchor figure: the direction (wholesale conversion indeed resulting in a one-off distribution to active members, of a magnitude reaching double digits as a percentage) holds directionally, but the specific value of "25%" cannot be verified by this paper, and the main text notes it as "approximately 20% (pending verification against the original actuarial source)" (Pensions in Denmark 2024). Likewise, this paper makes a correction to the bearing-capacity threshold anchor: Royal Mail's actual contribution rate is 6% from employees plus 13.6% from the employer, totaling 19.6%; the previously circulated trajectory of "contributions 17%→50%" — the "50%" therein is in fact the benefit-efficiency ratio of CDC relative to DC (benefit improvement of up to 50% per pound contributed), not a contribution rate, so this specific figure is withdrawn, retaining only the directional proposition that "bearing capacity approaching its threshold triggers form conversion," a direction supported by the fact of CDC authorization and the differential closure rates between public and private sectors (House of Commons Library 2024).

The micro-identification layer of this chapter could not be executed owing to data availability constraints. The event-by-event coding database of wholesale conversions required for Model A, the fund-level allocation panel and interest-rate-shock series required for Model B, and the sponsor-level panel and form-conversion events required for Model C are all unobtainable under the conditions of this study; the corresponding regressions and multinomial logit cannot be estimated, and this paper reports no coefficients. Furthermore, the Dutch solidarity reserve took effect in 2023, with conversion to be completed by 2028, so the early-period panel window for reserves and allocation is inherently short; even if micro-data become available later, the power of Model B's early estimates will remain limited, and the confidence intervals may be wide. The pre-registered thresholds are fully retained: Model A's principal hypothesis is that the higher the governance representation, the larger the jump in the value of active members' accounts in the conversion year, and its downgrade threshold (if no systematic relationship exists, downgrade to an isolated case) is retained; Model B expects that reserve thickness blunts the interest-rate response; Model C expects that bearing capacity significantly predicts form conversion, and that public plans without balance-sheet constraints convert significantly less often. The first-order status of s (the framework proposition) is not tested in any model.

5.6 Summary

This chapter has argued for the status of the risk-sharing degree and the distribution kernel as theoretical first-order variables, reconceiving "de-risking" as intergenerational wealth redistribution under a property-rights vacuum, and honestly confining the empirical first-order status of s to a theoretical proposition — because its most natural collapse test necessarily holds under any data, and is therefore a tautology. This chapter's empirical pillars shift onto three non-circular legs: the normative insight (the distributive consequences of de-risking varying systematically with features of the distribution kernel), reserve absorption elasticity (Direction A), and bearing-capacity screening (Direction A); all three legs remain at the design stage owing to the unavailability of micro-data, but each of their pre-registered falsifiable thresholds is fully retained. This chapter's property-rights ontology perspective and Chapter 4's option-position perspective jointly depict two orthogonal cross-sections of the same institutional form: the latter asks "where does the risk come to rest, and how does the hardness of the put determine the sign of allocation," while the former asks "who owns s itself as a continuous dependent variable, who owns the buffer." At this point, Chapters 4 and 5 have depicted this institutional phenomenon from the two directions of the liability side and ontology; the next chapter turns to econometric visualization of this layer situated between institution and behavior.

Chapter 6 The Measurement-Development Layer: Discount Benchmarks, Recognition Tiers, and Statistical Moments

6.1 Judgment

This chapter proposes that, beneath institutional differences and above behavior, there exists an independent measurement-development layer, composed of three decomposable dials: the discount benchmark (DB uses the AA corporate bond curve; certain DC gap-disclosure conventions use the 10-year Treasury rate plus an associated actuarial conversion), the recognition tier (whether the liability is booked on the balance sheet, or enters only individual footnotes and disclosures), and the order of the statistical moment displayed (DB displays the second-order variance of funded status, DC displays the first-order level of account balances). This layer can be orthogonal to risk preferences: the same interest-rate shock is filtered into different visible signals via DB liability mark-to-market versus DC account disclosure, and this asymmetry in measurement development shapes allocation responses at the margin.

This chapter maintains two honesty boundaries throughout, which delineate the precise scope of its claims. First, development is a mechanism standing alongside "recognition changing real constraints," not the sole exclusive mechanism. This is a deliberate narrowing of a stronger, more tempting claim ("allocation bifurcation was caused by development rather than real constraints"). Recognition, in events such as accounting standards mandating on-balance-sheet liability recognition, does two things at once: it raises visibility (the development leg), and it changes real stakes (the real-constraint leg: on-balance-sheet recognition triggers debt covenants, rating reclassification as debt, restricted distributable reserves, and changes in regulatory capital). When rating agencies reclassify a gap as debt, and de-risking follows to defend the rating tier, this is explicitly a real-constraint channel, and it no longer serves as supporting evidence for "development independent of preference." Both legs can be orthogonal to risk preference, but on recognition events they are perfectly collinear, and the chapter's current design cannot separate them. Acknowledging this collinearity is the first cornerstone of the chapter's honesty boundary — it means the chapter abandons any attempt to use historical recognition events to prove that "development is independent of real constraints." Second, the chapter's core assertion claims only that "the development layer can move allocation at the margin"; the development-driven historical account of bifurcation is downgraded to a hypothesis — the DC display experiment proves only forward-looking marginal elasticity (behavior changes at the margin when development is imposed), and cannot establish that development historically caused DB/DC allocation bifurcation (such as whether the death of UK DB was driven by development). Historical causation is explicitly labeled a hypothesis, not dressed up in identification language.

Within this honesty boundary, the identifiable core of the chapter's judgment is "the existence of the development layer plus marginal movability" — a forward-looking display-randomization experiment can cleanly demonstrate that development is movable, and an event study of rating-methodology revisions can cleanly demonstrate the effect of the real-constraint leg. The theoretical and normative extension of the chapter's judgment, however, ties the authority to set development rules back to political economy: whoever defines liabilities, and whoever chooses the discount benchmark and recognition tier, determines allocation to a considerable degree. This extension matters because it reveals that measurement development is not a neutral technical arrangement — the choice of discount benchmark, the setting of the recognition tier, and the stipulation of which order of moment is displayed are each rules that can be contested through political struggle, and control over these rules constitutes an indirect but real influence over allocation. This political-economy extension is presented as theory, not identified by this chapter, but it connects the chapter's measurement perspective to the paper's overall focus of judgment — "who bears the risk that has been transferred out": those who set the development rules are also, to a considerable degree, the shapers of risk redistribution.

6.2 Mechanism

The development leg (three-layer filtering, each orthogonal to preference). The same interest-rate shock enters two sets of accounts and is filtered layer by layer into different visible signals. The first layer is benchmark filtering: DB liabilities are discounted using the AA corporate bond curve, while DC gaps (if displayed) are discounted using the Treasury rate plus an actuarial conversion; the magnitude of the development gap is approximately equal to the negative change in the credit spread. When spreads widen, the "observed deterioration in liabilities" is asynchronous across the two sets of accounts, forming a benchmark wedge in allocation response — the same substantively identical liability is displayed as a different figure under the two discounting conventions, and this difference, which fluctuates with the credit spread, is itself an observable development wedge. The second layer is recognition-and-attention filtering: DB liabilities are placed on the balance sheet, procyclical and fluctuating daily, entering the field of view of governance actors and producing de-risking pressure; DC gaps enter only individual footnotes and disclosures, participant attention is weak, and allocation remains unmoved. The key distinction here is between "confirmation" (recognition, where the liability enters the body of the balance sheet) and "disclosure" (where information enters only footnotes or individual statements) — the same information, displayed with different positioning and different degrees of mandate, enters decision-making view to different degrees, and thus triggers different behavioral responses. The third layer is statistical-moment filtering: DB displays the second-order moment (the variance of funded status, gap volatility), DC displays the first-order moment (the level of account balances); actors who see variance have high hedging demand, actors who see levels have low hedging demand, and the order of moment displayed thus endogenously shapes hedging behavior — a participant shown only the account balance (the level) lacks the informational trigger to hedge the variance of their terminal outcome, whereas a trustee shown the variance of funded status is directly cued to a hedging need. The core claim of the development leg is: part of allocation bifurcation is not a difference in preferences, but a difference in visible signals — it is not that the two types of actors want different things, but that the two institutional arrangements let them see different signals.

The real-constraint leg (the mechanism is "recognition changes real stakes," not "visibility"). Mark-to-market accounting revalues long-term liabilities as a daily-fluctuating line item, and recognition, once accounting standards make it mandatory, changes real stakes: on-balance-sheet recognition triggers debt covenants, rating agencies reclassify the gap as debt, gap volatility becomes debt volatility, and defending the rating tier bears directly on actual financing costs incurred, hence de-risking follows. This is not "seeing it, and so reacting," but "seeing that it triggers a real penalty, and so reacting" — the distinction between the two is the crux of the chapter's honesty boundary: in the development leg, the response originates from visibility itself (seeing variance directly produces hedging demand); in the real-constraint leg, the response originates from the real penalty triggered by visibility (the covenant and rating consequences triggered by on-balance-sheet recognition). The initiator of this leg is the corporate finance department (defending the rating, defending covenants), and the mechanism is real incentives, hence it belongs to the real-constraint leg. Acknowledging the existence of this leg and explicitly attributing rating reclassification to it is a deliberate narrowing of this chapter relative to a stronger claim ("allocation bifurcation is driven purely by development") — the chapter does not claim development is the sole mechanism, only that development is an independent mechanism standing alongside real constraints, and that its independence is demonstrated by a forward-looking experiment (rather than a historical event).

The two legs are collinear. In historical events such as accounting standards mandating on-balance-sheet liability recognition, which simultaneously impose both development and real constraints, the two legs are perfectly collinear: strict recognition simultaneously raises visibility and triggers covenant, rating, and capital constraints. Sponsor fixed effects can absorb time-invariant preferences, but cannot absorb the time-varying shock of "recognition changing real stakes at the moment of treatment," so any estimate resting on such a break can only prove "recognition stringency → de-risking (the joint force of both legs)," not "development is independent of real constraints."

6.3 Identification Strategy

This chapter designs two identifiable legs and one historical item explicitly labeled as inseparable. The core strategy of this design is: since the development leg and the real-constraint leg are perfectly collinear and inseparable on historical recognition events, rather than forcing separation on historical events, the two mechanisms are placed into two different, each individually clean, designs for separate identification — the development leg is placed in a forward-looking display-randomization experiment (holding real constraints fixed, varying only the visible signal), and the real-constraint leg is placed in an event study of rating-methodology revisions (holding the visibility dimension fixed, varying only real financing costs). The historical item honestly acknowledges its inseparability, and serves only as a descriptive fact of the joint force of both legs, unless an independent source of variation can be found that orthogonalizes the two dimensions. This strategy of "separate placement rather than forced separation" is the methodological core of the chapter's response to the collinearity difficulty.

Leg A (the development leg, forward-looking margin, clean). Designed as a DC display-randomization field experiment: DC participants are randomized on how their retirement savings information is displayed — the order of moment displayed (first-order balance only, versus accompanied by a second-order band of retirement monthly income variance or a worst-case scenario), the presentation of the discount benchmark (different conversion conventions), and the display position and degree of mandate (footnote versus front screen, i.e., an experimental analog of confirmation versus disclosure). The width of the variance band serves as a continuous dose, enabling dose-response identification — if hedging behavior increases monotonically with the width of the variance band, this not only proves that development is movable but also characterizes its marginal elasticity. Outcome variables are target-date fund glide-path acceleration, annuity or delayed-claiming inquiry rates, additional-contribution rates, and (where available) actual reallocation. This design is clean precisely because real constraints (legal status, contracts, account structure) remain entirely unchanged within treatment, and only the visible signal is changed, so any behavioral difference is attributable to development itself, thereby cutting off the counterargument that "development is merely a renaming of real constraints." This is the core of the chapter's identifiable core: the independence of the development layer is not proven by historical events (on which the two legs are collinear), but by this forward-looking marginal experiment — under conditions where real constraints are held fixed by experimental design, if a change in the visible signal can move allocation, this is direct evidence of development as an independent mechanism. As an observational supplement, the elasticity of benchmark filtering can be estimated using the legislative effective-date break of mandatory retirement-income conversion disclosure combined with credit-spread jumps.

Leg B (the real-constraint leg, direction A, clean). Designed as an event study of rating-methodology revisions: the dates on which rating agencies revise their methodology for treating pension gaps serve as a quasi-natural experiment, proximity to a rating tier serves as continuous treatment intensity, and discontinuous rebalancing by covered DB plans in the revision window serves as the effect. Uncovered private DB plans plus contemporaneous DC plans serve as a placebo; if they jump synchronously, the effect is not rating-driven. This leg is clean and belongs to the real-constraint leg precisely because rating reclassification changes real financing cost (the rating tier corresponds directly to actual financing cost) rather than visibility; sponsor fixed effects absorb time-invariant preferences, and the methodology revision provides a time-varying exogenous shock. The significance of separating this leg from Leg A is: Leg A proves the development leg (visibility can move allocation), Leg B proves the real-constraint leg (rating reclassification can move allocation), and the two legs separately identify the two mechanisms, thereby allowing the chapter's explicitly labeled collinear legs to stand separately on different bodies of evidence by design — Leg A holds real constraints fixed and varies only visibility in a forward-looking experiment, Leg B holds the visibility dimension fixed and varies only real constraints (rating methodology) in an event study.

Leg C (separating the two historical legs; downgraded to a hypothesis if this proves unattainable). This requires finding an independent source of variation that "changes visibility but not real constraints (or vice versa)," such as a regulatory change that alters only disclosure format without altering the on-balance-sheet recognition obligation, or a two-by-two cross-jurisdictional matrix of "strict accounting recognition but no rating coverage" versus "lenient accounting but rating coverage," attempting to orthogonalize the visibility dimension from the real-constraint dimension. This is the most difficult leg of the chapter, and the one most likely to fail. The chapter therefore sets an honesty gate: if this orthogonal variation cannot be constructed, then the break at which accounting standards mandate on-balance-sheet recognition serves only as a descriptive fact of the joint force of both legs, not attributed to either leg alone; the development leg's contribution to historical bifurcation is permanently labeled a hypothesis, and the main text does not write causal sentences about it. This gate is set prior to the data, prohibiting the after-the-fact upgrading of a hypothesis to a causal claim merely because a stronger conclusion is desired.

6.4 Robustness Counterarguments and Responses

Counterargument One (development is a redundant mediator): So-called "development" is not an independent mechanism at all; DB de-risking occurs because on-balance-sheet recognition of liabilities triggers real debt covenants, rating reclassification, and regulatory capital constraints — all of this is real constraint; "visibility" is a redundant mediator, and removing it, real constraints alone can fully explain the massive allocation shift following mandatory accounting-standard recognition. The chapter accepts the core of this counterargument and narrows accordingly: on historical recognition events, the two legs are perfectly collinear, and the chapter acknowledges it cannot prove "development is independent of real constraints," and has already explicitly attributed rating reclassification to the real-constraint leg. But the independence of the development layer is not proven by historical events, but by the forward-looking marginal experiment — the DC display experiment, under conditions where real constraints are entirely unchanged, changes only the visible signal; if hedging behavior increases with the width of the variance band, this proves that visibility itself can move allocation at the margin, which is a residual the counterargument cannot explain. The counterargument correctly redirects the core assertion from "development caused historical bifurcation" to "development can move allocation at the margin," and the chapter accepts this narrowing.

Counterargument Two (a mismatch of level and timing): Even if a forward-looking experiment proves that development can move DC behavior at the margin, this is irrelevant to the core narrative — the core narrative is the historical grand bifurcation of DB and DC (DB moving toward bonds, DC remaining unmoved). A contemporary display experiment conducted on DC participants cannot, in terms of timing, establish that historical bifurcation was driven by development (the development environment, participants, and products of that era were all different); in terms of external validity, the marginal participants moved in the experiment (those sensitive to display) need not be the actors who historically drove aggregate allocation — aggregate DB de-risking was decided by sponsors and trustees under accounting and regulatory pressure, not by the attention elasticity of DC individuals. Using forward-looking, DC-side, marginal evidence to support a historical, DB-side, aggregate claim is an overreach across levels and across time. The chapter does not confront this counterargument head-on, but instead directly accepts it as a discipline of scope, and writes this discipline into the pre-registration, so that the main text cannot quietly upgrade marginal evidence into a historical claim: the core assertion is explicitly labeled "movable at the margin," historical causation is explicitly labeled a "hypothesis," and the DC-side experimental evidence is pre-declared not to extrapolate to DB-side aggregate history. The residual tension of the counterargument (what actually drove DB-side aggregate history) is acknowledged as an open question this chapter does not resolve, deferred to the causation layer as a theoretical presentation, with the accounting-standard break serving as a descriptive fact of the joint force of both legs, not as single-leg causation. The substance of this response is: the chapter does not pretend to have solved a problem it cannot solve, but draws the boundary clearly and honestly labels the open question as open.

6.5 Evidentiary Status and Pre-Registration Comparison

The chapter's verifiable layer has been confirmed, but requires one important convention correction, which is essential to the correct use of the chapter's descriptive break. Verification confirms that the widely cited figures "equities 32.4% → approximately 1.5%, bonds 28% → 63%" exist, but their convention is pension funds' share of ownership of the UK equity market (in 1992, pension funds held 32.4% of UK equities, today approximately 1.5%), not the equity allocation ratio within a single fund; the shift in within-fund equity allocation is sourced as moving from approximately 60%–70% toward liability-matching bonds (British Progress 2024; ii.co.uk 2023). This distinction must be written into the main text: if 32.4% and 1.5% are directly taken as "within-fund equity share," this is a convention error — 32.4% is a market ownership share, not a within-fund allocation ratio. For the purposes of this chapter's argument, the directional fact that "UK DB indeed moved substantially toward bonds following the recognition event" holds and may be used as a descriptive break; but this break is the result of the joint force of the development leg and the real-constraint leg, and cannot be attributed to a single leg. In addition, the spike in gilt and swap rates triggered by the September 2022 fiscal statement and the ensuing LDI margin-call spiral, and the record-setting risk transfer driven by the subsequent improvement in buyout funding levels (buy-in and buyout transactions reached a historic high of approximately £49.1 billion in 2023, involving more than 250 deals, an increase of approximately 25% over the prior year), provide a directional evidentiary foundation for "recognition stringency plus real constraint → de-risking (joint force of both legs)," not single-leg causation (Pensions Age 2024).

The chapter's micro-identification layer could not be executed due to data availability constraints. The recordkeeper participant panel or survey experiment sample required for Leg A, the rating-methodology-revision date library and tier panel required for Leg B, and the hand-coded cross-jurisdictional recognition-stringency data required for Leg C are all unavailable under this study's conditions; the corresponding randomization-experiment regressions and event-study regressions cannot be executed, and this paper does not report dose-response coefficients or event-study coefficients. It must be particularly noted that Leg A is the sole clean evidence for the chapter's identifiable core (marginal movability of the development layer); without obtaining it, this cannot be verified, so the chapter's core judgment, under this study's conditions, remains at "design in place, evidence pending." The pre-registration thresholds are fully preserved: Leg A's primary hypothesis is that the variance-band-width dose coefficient is positive and monotonic; if not significant or non-monotonic, the marginal movability of the development layer is falsified; Leg A's separation tests allow some dials to be null; Leg B's primary hypothesis is that the tier-proximity effect is positive and the placebo coefficient is near zero; the relationship between the two legs is not expected to be separable at the historical break, and if orthogonal variation is unattainable, the core assertion remains at "marginal movability plus joint-force descriptive," with historical causation written as a hypothesis; Leg A's DC-side marginal evidence is pre-declared not to extrapolate to DB-side aggregate history.

6.6 Summary

This chapter has argued that between institutions and behavior there exists a measurement-development layer composed of the discount benchmark, the recognition tier, and the order of the statistical moment, and has honestly narrowed its core assertion to "can move allocation at the margin," downgrading the development-based account of historical bifurcation to a hypothesis. This narrowing is a deliberate concession relative to a stronger claim: the chapter does not claim that development is the sole or historical cause of allocation bifurcation, only that development is an independent mechanism standing alongside real constraints, and that its independence is demonstrated by a forward-looking display-randomization experiment (rather than a historical recognition event). The collinearity of the development leg and the real-constraint leg on historical recognition events is explicitly registered as this chapter's identification boundary, and the question of what actually drove historical bifurcation is honestly labeled an open question. This chapter's visibility perspective, together with Chapter 4's put-hardness perspective and Chapter 8's inflow-nature perspective, act on different margins in explaining "why DC is allocated the way it is"; the three coexist at different margins rather than reducing to one another — this coexistence is yet another instance of this paper's methodological stance that "institutional phenomena arise from multiple mechanisms superimposed at different margins."

Chapter 7 System-Level Institutional Complementarity: A Theoretical Lens on the Direction of Vulnerability

7.1 Judgment

This chapter extends the causal arrow one layer outward: pension systems are not isolated institutions but are embedded in a larger institutional configuration comprising varieties of capitalism, the financial system, and the skills regime; the same interest-rate shock may have opposite consequences under complementary versus mismatched configurations, so that the direction of vulnerability (amplification versus absorption, rather than its level) is a system-level property of the configuration, rather than a property inherent to the pension-system type itself. Furthermore, the DB/DC distribution itself may be an endogenous outcome that co-evolves with the financial system, rather than an exogenous classificatory variable.

The significance of this perspective lies in how it reframes the common question of "which system is more fragile." The common framing presupposes that vulnerability is an inherent property of the institutional type — whether DB is more fragile, or DC is more fragile. This chapter's answer is: the direction of vulnerability is not an inherent property of either institutional type, but a system-level property of the larger configuration in which that institution is embedded. The same DB system, when embedded in a highly complementary configuration (market-led finance plus LDI leverage), will amplify an interest-rate shock into a systemic event; when embedded in a low-complementarity configuration (bank-led finance plus public reserves), the same shock will instead be absorbed as a localized loss. Hence the direction of vulnerability cannot be read off from the institutional type; it can only be read off from the systemic configuration in which the institution is embedded — and this reframing bears directly on the downstream judgment of whether "importing mature-market experience" can be done wholesale: if one country's pension system is transplanted into an environment where the complementary configuration has not co-evolved into place, its direction of vulnerability may be the opposite of that in its native environment.

This chapter must maintain an honest positioning throughout: the entire axis is presented as a theoretical and descriptive framework, explicitly marked as not having achieved identification, and is not offered as an identified causal claim. This positioning is not a gesture of modesty on the chapter's part, but a candid acknowledgment of identification difficulties — at the design stage this chapter attempted to use legal origin as an instrumental variable for financial marketization, and demographic shocks as a quasi-exogenous source of funding pressure, but both legs failed under critical scrutiny (see the section on identification strategy), and so the entire axis is honestly downgraded to a theoretical lens. What can be sustained is that the lens of institutional complementarity can organize and describe the cross-national pattern of "why the same shock has opposite consequence directions in the UK, the Netherlands, and Japan," and can provide a structural rationale for "mature-market experience cannot simply be transplanted" (an unco-evolved complementary configuration is the structural root cause of transplant failure). What cannot be claimed is that this chapter offers an identified causal effect — both the directional bifurcation of "high complementarity amplifies, low complementarity absorbs" and the claim that "the distribution is a co-evolved outcome" are hypotheses awaiting pre-registration, requiring an independent complementarity measure constructed prior to the crisis and free of LDI exposure, plus a design that excludes common demographic causes, before either can be elevated to causal status; until then, both are presented uniformly as a theoretical lens, descriptive association, and hypotheses awaiting pre-registration, without identificational language. Correspondingly, the gate this chapter supplies for the downstream judgment of "cannot be transplanted" does not rest its weight on the weakest evidence — the skills-mirroring channel — but stands on the comparatively stronger descriptive characterization of amplification/absorption and the directional pattern of co-evolution, and this gate is explicitly marked as supplying a theoretical lens rather than an identified causal threshold.

7.2 Mechanism

The theoretical skeleton of this chapter decomposes the net effect of an interest-rate shock on a given country's pension system into a structural equation containing a complementarity-amplification term: the net effect is approximately equal to the direct liability revaluation plus the direct asset-price effect, plus or minus a complementarity-amplification term (this term equals a directional coefficient multiplied by the complementarity between the pension regime and the financial system, multiplied by the interest-rate shock). It must be stated immediately that writing down this structural equation does not mean estimating its coefficients and claiming causality — quite the opposite: the purpose of this structural equation is to organize theoretical intuition, and its causal application is locked by a pre-registered gate (see the section on identification strategy). Writing the structural equation without estimating its coefficients is part of this chapter's honest positioning: the structural equation helps clarify the theoretical logic of "how complementarity moderates the direction of the shock," but turning it into an estimated causal equation requires first satisfying the gate condition. Three sub-mechanisms follow.

Amplification and absorption (the primary anchor of the theoretical lens, plus case illustration). Highly complementary configurations (liberal market economies, market-led finance, LDI leveraged hedging) form a pro-complementarity spiral: mark-to-market liabilities trigger LDI hedging plus leverage, rising interest rates trigger margin calls, long-duration government bonds are sold off, yields rise further, triggering further margin calls. This spiral is "pro-complementary" precisely because each link is triggered by the institutional logic of the preceding link: mark-to-market accounting requires hedging, hedging is implemented via leverage, leverage triggers margin calls when rates rise, margin calls force asset sales, asset sales push up yields and trigger a new round of margin calls; it is precisely the mutual meshing of these institutional links that amplifies an interest-rate shock — one that could otherwise have been absorbed locally — into a systemic event. The UK's 2022 gilt and LDI spiral serves as the illustrative case. Configurations with low complementarity or mismatched buffers (coordinated market economies, bank-led finance, public reserves) instead absorb the shock: the shock is internalized by the banking system, public buffers, or fiscal mechanisms, without being amplified through a market spiral, with Japan's public reserve and bank-led structure serving as the absorption case — in such configurations, the meshing between the pension system and the financial system is looser, lacking a transmission chain that amplifies the shock stage by stage, so the shock remains local. The mechanism should be read not as "which institutional type is more fragile" but as "the same directional shock is amplified into a systemic event by a complementary structure, and absorbed into a localized loss by a mismatched or buffering structure"; the direction of vulnerability is determined by the system-level property of the configuration, not by any inherent fragility of the pension-system type.

Co-evolution (hypothesis). The pension asset pool and financial market depth are mutually reinforcing sources of growth for each other: market depth supplies investable assets and liquidity, which drives pension funding, and the resulting inflow of pension assets in turn deepens the market further, giving rise to multiple equilibria (a market-led equilibrium and a bank-led equilibrium); the DB/DC distribution may therefore be an endogenous outcome co-evolving with the financial system, rather than an exogenously given classificatory variable. The implication of this hypothesis is profound — if the institutional distribution is itself the result of co-evolution with the financial system, then explaining differences in configuration by institutional type carries an endogeneity risk: the institution and the configuration may both be joint products of a deeper factor, namely the depth of the financial system. This is a hypothesis awaiting pre-registration, not an identified direction, because its most natural test (a direction that reverses with the horizon in a panel VAR) is observationally equivalent to the alternative explanation of "common demographic drivers," and this common cause has not been excluded.

Skills mirroring (exploratory, the weakest evidence). DB can be viewed as the financial form, on the pension side, of the incentive-compatibility constraint for specific-skills investment: long-term employment commitments and long-term benefit commitments mutually collateralize each other — employers exchange long-term benefit commitments for employees' investment in specific skills, and employees exchange specific-skills investment for the stability of long-term employment, each supporting the other (see Estévez-Abe, Iversen, and Soskice 2001). From this perspective, freezing or closing DB under low interest rates loosens the incentive compatibility of long-term employment, thereby suppressing investment in specific skills, so that part of the cost of de-risking falls on the specific-skills regime rather than solely within the pension system itself. This item is explicitly marked as the weakest of the three sub-mechanisms in terms of evidence, offered as an exploratory supplement — its identification is constrained by a triple data gap: pension-side and skills-side data reside in different databases requiring manual matching, firm-level "specific-skills investment" lacks a clean indicator, and aligning DB-freeze events with skills data must be done manually. Accordingly, this chapter does not place the load-bearing weight of the downstream "cannot be transplanted" gate on this weakest sub-mechanism.

7.3 Identification strategy: abandoning causal instrumental variables in favor of theory and description plus a pre-registered gate

This chapter explicitly abandons, and does not revive, two causal identification legs that were once considered: using legal origin as an instrumental variable for financial marketization (because legal origin passes through dozens of non-"financial marketization" channels — the labor market, the welfare state, the skills regime, corporate governance, and so forth — the exclusion restriction fails, and what is identified is an entire bundle of institutions rather than a single channel), and using demographic shocks as a quasi-exogenous source of funding (because demographic structure simultaneously drives up savings, market deepening, and funded pension assets, the common cause is not excluded, and it is downgraded to a descriptive covariate).

This chapter instead adopts a theoretical and descriptive strategy: a cross-national panel provides descriptive characterization (the correlational pattern between complementarity proxies and shock direction, explicitly marked as non-causal); the UK's 2022 event serves as a case study (not a quasi-natural experiment for causal estimation); a panel VAR provides exploratory directional characterization (explicitly marked as not having excluded the common demographic cause). To turn "whether causal language may be used" into a gate that can be overturned by data, this chapter pre-registers a causalization gate: only if two conditions are jointly satisfied — the existence of a complementarity measure constructed prior to the crisis and free of LDI exposure (constructed orthogonally to subsequent spiral outcomes), and a pre-registered mapping from complementarity to direction with a falsifiable threshold (the data must admit the possibility of counterexamples such as "high complementarity yet no spiral" or "low complementarity yet a spiral") — may this chapter elevate "high complementarity amplifies" to a causal assertion; otherwise it remains permanently at the level of a theoretical and descriptive lens.

The construction principle for this independent complementarity measure warrants specific elaboration, since it is the first condition of the gate and also the primary shortfall in this chapter's causalization. This measure must use only pre-crisis (e.g., a static cross-section from several periods before the shock) variables of financial-structure and pension-structure complementarity (such as the degree of market dominance, the depth of institutional investors, the share of pension asset inflows back into the market, and the degree of corporate-governance coordination), and must explicitly exclude any LDI exposure, leverage, or ex-post spiral outcome variables — otherwise it would fall into the circularity of using an outcome to predict an outcome (using LDI leverage to predict the LDI spiral is equivalent to using a proximate cause of the outcome to predict the outcome itself). Once constructed, the measure must also be ex ante orthogonal to the outcome variable of "whether a subsequent spiral occurred" (the construction must not let the measure observe the outcome), and must preserve a falsifiable threshold (there must exist observations of high complementarity without a spiral, or low complementarity with a spiral). This chapter's position is: such a complementarity measure has not yet been shown to be constructible, and therefore, until it is constructed and passes the orthogonality and falsifiability thresholds, this chapter does not elevate "high complementarity amplifies" to causal status. This chapter's default landing point is that the gate is not passed, so the main text by default presents descriptive association and hypotheses awaiting pre-registration — what is needed to elevate this to causal status is such a measure that satisfies the gate, not an adjustment of this chapter's wording.

Accompanying this is a design that excludes the common demographic cause. For the co-evolution hypothesis, demographic structure (population aging) simultaneously drives up savings, market deepening, and funded pension assets, and is thus a common driver of both market depth and funded assets; consequently, "the direction between pension assets and market depth reverses with the horizon" is observationally equivalent to "common demographic driving," and the common cause has not been excluded. To free this hypothesis from this trap of observational equivalence, a design excluding the common demographic driver must be devised — for instance, testing whether the direction reversal persists after controlling for demographic structure, or seeking variation in market depth that does not pass through the demographic channel. Until such a common-cause-excluding design is achieved, co-evolution can only remain a hypothesis. Furthermore, given the rarity of extreme events (see Counter-argument Two), one alternative path is to identify the amplification term within a structural model, rather than estimating an amplification coefficient from a rare cross-section; but this path is model identification rather than data identification, and the credibility of its conclusions depends on the plausibility of the model specification, which must be explicitly marked as model identification and subjected to corresponding robustness scrutiny.

7.4 Robustness counter-arguments and treatment

Counter-argument One ("downgrading to a lens" is mere rhetoric). Once the structural equation for the complementarity-amplification term is written down, the most natural next step is to estimate its coefficients and claim causality, so "downgrading to a lens" is merely a matter of wording. This chapter accepts the force of this counter-argument, and locks it down rigidly with a design-level gate rather than relying on self-restraint: turning "whether causal language may be used" into the aforementioned pre-registered gate, fixed in place before any data is brought to bear; the default is that the gate is not passed. The direct response to the charge of "mere rhetoric" is this: the distinction lies not in wording but in falsification structure — the lens version predicts no single-country outcome and only organizes cross-national patterns, and hence cannot be falsified by any single country; the causal version must pre-register a falsifiable threshold for the mapping from complementarity to direction. This chapter currently remains at the former, precisely because the falsifiable threshold for the latter has not yet been constructed.

Counter-argument Two (overfitting to a single event). The UK gilt and LDI spiral is an event with a sample size of one, and using a single case to support "high complementarity amplifies" — precisely the point the lens most wants to explain — is overfitting to a single event rather than a cross-national regularity. This chapter treats this as the weakest link in the entire axis, offers no defense, and only honestly delimits its scope, in three parts. First, no single case is used to support causality: the UK spiral serves in this chapter only as an illustrative case, not as causal evidence for "high complementarity amplifies"; the absence of causal evidence has already been locked into the causalization gate as requiring "an independent complementarity measure constructed prior to the crisis and free of LDI exposure, plus a falsifiable threshold." Second, the "directional bifurcation" is downgraded to the level its statistical power can support: cross-national descriptive characterization (after controlling for the DB share and the duration gap, the probabilistic pattern of pension-related market dysfunction occurring in market-led financial countries during periods of rapidly rising interest rates), explicitly marked as descriptive association, whose small sample size supports only pattern description rather than causal estimation; this level does not fail merely because the sample size is one, since it never claims causality in the first place. Third, the statistical-power shortfall is written into the design constraints — the history of observable events combining extreme complementarity, a large shock, and market dysfunction is inherently sparse (the UK in 2022 is the primary case; the Netherlands and Japan did not trigger such events in the same period), and cross-sectional inferential power is inherently weak; one possible alternative is to identify the amplification term within a structural model rather than estimating it from a rare cross-section, but this is model identification rather than data identification and must be explicitly marked as such; if the comparable events cannot be expanded to a number sufficient to support cross-sectional inference, nor can a structural model substitute for this, the directional bifurcation in this chapter remains permanently at the level of description and hypothesis. This is not a failure but an honest boundary of the system-level lens: the value of the lens lies in organizing cross-national differences and supplying downstream structural rationale, not in forcibly extracting a causal coefficient from a rare event — attempting to extract a causal coefficient from an event with a sample size of one is itself a misunderstanding of the nature of this lens.

7.5 Evidentiary status and pre-registration cross-check

The verifiable layer of this chapter only upgrades the factual foundation of descriptive association and case study, and never upgrades to a causal assertion. The UK's 2022 fiscal statement triggered a sharp rise in yields, LDI margin calls, and a spiral of long-duration government bond sell-offs, which triggered emergency central bank intervention; this stands in contrast with the Netherlands and Japan, which did not trigger such events in the same period, and serves as an illustrative case of pro-complementarity among a liberal market economy, market-led finance, and LDI (not a quasi-natural experiment for causal estimation) (Pensions Age 2024). The Netherlands' collective transition serves as a rare live sample of transition within a coordinated market economy (not treated as causal identification) (NautaDutilh 2023). The cross-national descriptive characterization of each country's DB/DC shares and asset structure takes the publicly authoritative statistics of the Organisation for Economic Co-operation and Development (OECD) as the standard data source, but the specific share figures must remain consistent with the single source of truth for the taxonomy across the entire volume, and this chapter does not adopt separate figures so as not to break cross-chapter consistency — this self-restraint stems from the requirement of a single source of truth for the institutional taxonomy: quantitative facts shared across chapters must follow a unified standard, and no chapter may source its own figures separately, or else the same fact would appear with different values across chapters, damaging overall consistency. The gate this chapter supplies for the downstream "cannot be transplanted" judgment must be fixed, prior to downstream consumption, as a theoretical lens rather than an identified causal threshold, so as to prevent downstream work from mistakenly citing the lens as identified causality; and the load-bearing weight of the gate rests on the comparatively stronger descriptive characterization of amplification/absorption and the directional pattern of co-evolution, rather than on the weakest evidence — the skills-mirroring channel.

This chapter's causal identification is explicitly marked at the outset as not having been achieved in its directional claim, and this text does not attempt to restore it — this differs from earlier chapters: in earlier chapters the identification design was in place and only went unexecuted owing to data unavailability, whereas in this chapter identification was, in principle, unachievable owing to the failure to exclude the common cause and the failure of the exclusion restriction. Here it must be honestly stated that two identification legs have already been abandoned and are not to be revived: first, using legal origin (common law versus civil law) as an instrumental variable for financial marketization, because legal origin passes through dozens of non-"financial marketization" channels — the labor market, the welfare state, the skills regime, corporate governance, and so forth — the exclusion restriction fails catastrophically, and what is identified is an entire bundle of institutions rather than a single channel (see the related lineage in La Porta et al. 1998); second, using demographic shocks as a quasi-exogenous source of funding, because demographic structure (population aging) simultaneously drives up savings, market deepening, and funded pension assets, the common cause is not excluded, and it is downgraded to a descriptive covariate. The status of the three specifications is as follows: the descriptive-association specification requires a pre-crisis, LDI-exposure-free independent complementarity coding and a cross-national event panel, neither of which is available, and the number of extreme events is fundamentally too small, so no causal coefficient is reported and no "effect" language is used; the co-evolution hypothesis specification requires a design excluding the common demographic cause, which is unavailable, so it remains at the level of hypothesis; the skills-mirroring specification requires matching across three databases and a firm-level indicator of specific-skills investment, neither of which is available, so it is treated as exploratory and is not placed in the downstream gate's load-bearing position. The causalization gate defaults to not being passed, so the main text of this chapter remains at the level of theoretical lens, descriptive association, and hypotheses awaiting pre-registration. The pre-registration thresholds are fully preserved in their entirety: the falsification threshold for the descriptive association must exist and be declared in advance, and post hoc redefinition of complementarity to revive it is prohibited; the survival condition for the co-evolution hypothesis is that the direction reversal still holds after controlling for demographic structure, failing which it is judged to be an illusion of common demographic driving and is falsified; the survival threshold for skills mirroring is that even if the sign matches expectations, it serves only as exploratory weak evidence; until an independent complementarity measure satisfying the gate emerges, all statements in the main text remain at the level of theoretical lens.

7.6 Summary

This chapter situates the pension system within the larger configuration of varieties of capitalism, the financial system, and the skills regime, argues that the direction of vulnerability is a system-level property of the configuration, and honestly positions the entire axis as a theoretical and descriptive lens, explicitly marking that causal identification has not been achieved. This positioning differs from earlier chapters: in earlier chapters the identification design was in place and only went unexecuted owing to data unavailability, whereas in this chapter identification was, in principle, unachievable owing to the exclusion-restriction failure of the legal-origin instrument and the failure to exclude the common demographic cause; the entire axis is therefore honestly downgraded to a theoretical lens, and its causalization must await the emergence of an independent complementarity measure satisfying the gate. What this chapter supplies for the downstream judgment of "cannot be transplanted" is a structural rationale (an unco-evolved complementary configuration is the root cause of transplant failure), not an identified causal threshold; this supply rests its load-bearing weight on the descriptive characterization of amplification/absorption and the directional pattern of co-evolution, rather than on the weakest evidence — the skills-mirroring channel — and placing the downstream gate's load-bearing weight on the comparatively stronger evidence rather than the weakest evidence is one way in which this chapter acts responsibly toward downstream research.

Chapter 8 The Contractual Nature of Asset-Side Inflows and the Capacity for Long-Term Illiquid Allocation

8.1 Judgment

This chapter shifts the explanatory focus of allocation capacity from liability-side duration to the contractual nature of the asset-side contribution flow. This shift itself constitutes the chapter's incremental judgment relative to conventional analysis: conventional explanations attribute DB's capacity to hold long-duration and illiquid assets to its long liability duration, whereas this chapter argues that what truly determines the capacity for long-term illiquid allocation is not the duration anchor on the liability side, but the contractual nature of the inflow on the asset side. The core judgment is: allocation capacity — especially the capacity to bear long-term illiquid positions — is determined by the predictability and non-interruptibility of the asset-side inflow, not directly by the DB/DC label or liability duration. Mandatory contributions construct, on the inflow side, a "synthetic long-duration asset" (a predictable, legally non-interruptible present value of net inflows), which functionally substitutes for DB's long liability-duration anchor, making illiquid allocation a sustainable equilibrium; the DB/DC label thereby degenerates into two special cases of this continuous mechanism. This perspective is entirely orthogonal to the preceding four chapters: Chapter 4 addresses the hardness of the liability-side put, Chapter 5 addresses the ontology of the continuous spectrum and buffer property rights, Chapter 6 addresses the measurement-visibility layer, Chapter 7 addresses system-layer complementarity, and this chapter is the only causal axis that enters from the asset-side inflow, and also the branch with the most robust Direction-A consequence-side identification among the five chapters.

The judgment has two tiers, with the following honest boundary. Consequence side (Direction A, the sole claim of this chapter's core assertion): given the nature of the inflow (degree of compulsion, duration, tail interruptibility, conditional variance), its causal effect on illiquid allocation is identifiable (institution→allocation), with the most robust observational identification source being the legislative-staggering instrumental variable. Cause side (Direction B, no clean identification, explicitly labeled as political-economy case and theory): "by what political economy is the nature of the inflow determined," "who has the right to interrupt or dilute mandatory inflows," and "does allocation pressure reshape the withdrawal institution in reverse" — the entire chapter has no clean Direction-B instrument for these questions; they are presented as case studies and theory, without identification-style language.

A convergence of wording is needed here to maintain consistency in the identification-status labeling across the paper. For the legislative-staggering instrumental variable, one natural but overly strong phrasing would call it "the cleanest identification source across the entire spectrum"; this paper does not adopt that phrasing, converging instead on "the most robust observational identification source on the Direction-A consequence side, with the highest evidentiary tier." The reason is that the compulsion-degree index contains a self-constructed exit-difficulty coding (a composite coding of statutory coverage rate times contribution-obligation rigidity times early-exit difficulty), whose pure identification cleanliness is weaker than the true randomized field experiment of Leg A in Chapter 6 — the latter, by experimental design, holds real constraints fixed and varies only the visible signal, a cleanliness that an observational instrumental variable cannot match. Hence, the advantage of the legislative-staggering instrumental variable lies not in having the cleanest coding, but in the strongest quasi-experimental exogeneity of the withdrawal event on which it relies (emergency withdrawals were exogenously triggered by the populist politics of a public-health event), together with already-published causal quantification (approximately $1.59 in future savings lost per $1 withdrawn) as the top-tier evidentiary support. This convergence may look like a mere wording adjustment, but it is in fact part of the paper's overall discipline of honest boundaries: the cleanliness of different identification sources must be accurately ranked, and one branch should not be broadly declared optimal in identification cleanliness merely because it is most complete in verifiable facts.

Related to this, an orthogonality argument must be added to clarify where this chapter's evidentiary weight actually rests. The compulsion-degree index characterizes the legal rigidity of the inflow contract — that is, "whether the contribution flow can be interrupted or diluted"; whereas the defining distinction of the DB/DC label is "who bears the benefit risk." These two dimensions are conceptually orthogonal: a purely mandatory DC system (in which the individual bears the benefit risk) can have extremely high inflow-contract rigidity (non-interruptible), while a voluntary DB (in which the employer bears the benefit risk) can have a soft inflow — the two dimensions can vary independently. Therefore, this chapter's evidentiary weight rests on the core coefficient of the legislative-staggering instrumental variable and the orthogonal counter-evidence, while the "collapse test of whether the label coefficient collapses after controlling for compulsion degree" is demoted to an auxiliary, non-load-bearing description, so as to avoid repeating the same tautological loop as the collapse test already deleted in Chapter 5.

8.2 Mechanism: The Three Moments of the Synthetic Asset

The mechanism of this chapter is a single causal chain — compulsion determines coverage rate, contribution rate, and wage growth, which in turn determine the present value and duration of the "synthetic asset," constituting the implicit collateral for illiquid holdings and enabling the illiquidity discount to be safely captured. This chain can be decomposed into three mutually independent, additive moments, each corresponding to a falsifiable Direction-A assertion.

The duration moment. A non-interruptible mandatory inflow is the asset-side equivalent of DB's long liabilities: the present value of a predictable net inflow is itself a long-duration asset, providing term support for illiquid holdings; even while the stock of assets is in a net-outflow phase, as long as the mandatory inflow from newly enrolled participants continues to regenerate, illiquid allocation can still be sustained. This is the key point distinguishing this chapter from the "tautological restatement" objection (see Counter-argument Two below) — the conventional explanation attributes the capacity to hold long-duration assets to long liability duration, whereas this chapter argues that a predictable, non-interruptible sequence of mandatory inflows itself constitutes a synthetic long-duration asset, whose present value provides collateral for illiquid holdings; as long as the mandatory inflow of new members keeps regenerating, this synthetic asset persists, and illiquid allocation can be sustained even while the stock is in net outflow.

The tail moment. The opening of withdrawal rights means the "non-interruptible" promise has been politically broken, re-pricing the tail probability of future politically driven redemptions, which endogenously raises the liquidity preference. The key is that the expectations channel dominates the contemporaneous cash-flow channel: money is still flowing in, contemporaneous cash flow is unchanged, yet merely the re-pricing of the tail risk that funds "might be politically siphoned off" is enough to depress the illiquid allocation target. The identification significance of this expectations channel is substantial — it means the effect of a withdrawal event cannot be simply explained as an extraction of contemporaneous cash flow, because the effect manifests through the expectations channel even before contemporaneous cash flow has changed; the opening of withdrawal rights, even if only a small fraction of participants actually withdraw, will — through the signal that the "non-interruptible promise has already been broken" — change all agents' pricing of future redemption tail risk, thereby systematically depressing the illiquid allocation target.

The variance moment. Informal employment and low coverage cause the contribution-density distribution to be right-skewed and high-variance, raising the conditional variance of net inflows, which raises precautionary liquidity demand and crowds out illiquid allocation. This channel operates through the cross-sectional volatility of contribution density forcing up liquidity needs, rather than through the total volume of inflows — low coverage is not simply "small in scale," but "high in inflow variance." This distinction matters because it strips the observation that "low-coverage systems allocate more conservatively" away from a scale effect: it is not that the system is conservative because it is small, but that the high variance of contribution density lowers the predictability of net inflows, thereby raising precautionary liquidity demand.

The three moments are mutually independent and additive; among them, the withdrawal event corresponding to the tail moment provides consequence-side quasi-experimental causal evidence for the duration moment, with the highest evidentiary tier — because the withdrawal event directly and exogenously shocks the core premise of the duration moment, namely "non-interruptibility," constituting a near-ideal natural experiment.

8.3 Identification Strategy: Three Direction-A Legs, Each with an Independent Exogenous Source

This chapter designs three identification legs for Direction A, each using an independent exogenous source to break collinearity. This design principle of "each leg with an independent exogenous source" directly responds to the collinearity objection concerning the three moments (see Counter-argument Two below): since systems with high compulsion tend simultaneously to have high coverage, difficult withdrawal, and a high share of formal employment, if the three moments were separated internally using only cross-sectional regression, it would be difficult to determine which moment is driving allocation; therefore, this chapter does not rely on internal cross-sectional separation, but instead uses three mutually orthogonal exogenous shocks to identify the three moments separately — the duration moment uses legislative staggering, the tail moment uses the withdrawal event, and the variance moment uses the informal-employment rate. Each shock moves only one moment and not the others, so collinearity is broken by design.

The legislative-staggering instrumental variable (main leg). The instrument is the timing stagger of the Netherlands' industry-by-industry mandatory enrollment effective from 1949, and Australia's schedule of statutory contribution-rate increases for compulsory superannuation phased in over successive years. The Dutch Industry-wide Pension Funds Act was established in 1949; at the request of social partners, the Minister of Social Affairs could order all employers in an industry to participate mandatorily, and today more than 75% of employees participate in industry funds owing to large and small compulsion (Chen and Beetsma 2014). Australia's mandatory employer contributions, effective from 1 July 1992, were raised in stages according to a nationally uniform legislative timetable (3%→9%→12%), with the rate steps imposed simultaneously on all funds, their timing determined by the federal legislative process — a near-exogenous dosage staircase (ATO 2024). The identification structure is: the exogeneity of the legislative timing renders the institutional form of the mandatory inflow exogenous, and one tests its causal effect on the illiquid share, that is, institution→allocation (Direction A). The pre-registered test of exogeneity is pre-adoption parallel trends (no abnormal trend in illiquid allocation prior to the effective date) plus placebo tests (no effect in non-mandated industries or non-rate-increase years); failure of either demotes the main leg to auxiliary description.

The withdrawal-event quasi-experiment (tail moment). Designed as a difference-in-differences and event study of the multiple rounds of withdrawal in Chile, Peru, and Australia in 2020–2021: the treatment group is mandatory systems that opened withdrawals, the control group is mandatory systems that did not open withdrawals in the same period (Netherlands, Nordic countries), estimated on a fund-by-time panel. Chile's three rounds of withdrawal (July 2020, approved December 2020, and a third round in April 2021) totaled approximately $48–50 billion, roughly equivalent to 19%–20% of gross domestic product, the largest pension withdrawal event globally as a share of gross domestic product, with assets under management falling from approximately 83% of gross domestic product before the withdrawals to approximately 62% by 2024 (IMF 2021). Published research further confirms that every $1 withdrawn produces approximately $1.59 in future retirement-savings loss, with monthly pension benefits falling by approximately 7.26% (Madeira 2022). The identification landing point is that the magnitude of the downward revision in the illiquid target is proportional to that system's historical withdrawal scale (or number of rounds), and is strongest in the most aggressive multi-fund tiers; exogeneity rests on the fact that the emergency withdrawals were exogenously triggered by the populist politics of a public-health event, independent of any fund's investment decisions. This leg has the highest evidentiary tier for two reasons: first, it imposes a direct exogenous shock on the core premise of the duration moment, namely "non-interruptibility" — the opening of withdrawal rights is precisely the breaking of the "non-interruptible" promise, making it a near-ideal natural experiment for the tail-moment mechanism; second, its consequences already have published causal quantification (approximately $1.59 in future savings lost per $1 withdrawn), giving this leg not only clean design but also verifiable quantitative evidentiary support. It must be emphasized again that, in this chapter's identification, the withdrawal event serves only as an exogenous shock to the "shadow price of non-interruptibility" (Direction-A consequence side), not as a causal analysis of "why withdrawal rights were politically broken" — the latter belongs to Direction B and is presented as a political-economy case, without identification-style language.

The contribution-density-variance instrumental variable (variance moment). The instrument is the informal-employment rate by industry or region (International Labour Organization data), exogenous to fund investment decisions; the structure is that informal employment raises the variance of contribution density, which in turn raises precautionary liquidity demand and depresses the illiquid share. This leg is orthogonal to the preceding two: the informal-employment rate changes neither the institutional form of the mandatory inflow (the duration moment) nor whether withdrawal rights are opened (the tail moment), affecting only the cross-sectional volatility of contribution density, thereby separately identifying the variance moment. In specification, the illiquid share is regressed on the Gini coefficient or coefficient of variation of contribution density, controlling for the log of total assets, maturity, and the DB/DC label; the expected coefficient is negative (the greater the density variance, the lower the illiquid share), and the coefficient of precautionary cash buffers on density variance is expected to be positive. Chile's contribution density, stratified by income (density here refers to the ratio of months with contributions to months owed, significantly lower for low-income earners than for high-income earners), shows significant cross-sectional differences, which can directionally supply a source of density variation for this leg; but comparable density-distribution data are not available under this study's conditions, and cross-jurisdiction splicing is especially difficult, so only a directional statement is offered here, with specific stratified figures pending verification against primary regulatory materials before entering the main text.

8.4 Robustness Counter-arguments and Treatment

Counter-argument One (dispute over the direction of the instrumental variable): the legislative-staggering instrumental variable in fact identifies Direction B (the cause side), and treating it as Direction A is self-deception; or, conversely, it identifies neither direction cleanly, because legislation itself is endogenous to the allocation and political pressures of the time. This counter-argument touches on a subtle methodological issue, and this chapter's treatment proceeds in two steps. As to direction, the direction of an estimand is defined by the arrow, not by "whether the legislation has a political cause." This is the key point: the legislative-staggering instrumental variable uses the legislative timing as an exogenous source of variation in the institutional form of the mandatory inflow, testing "institutional form (already exogenously given)→allocation," that is, institution→allocation (Direction A); why the legislation occurred at that particular time (its political-economic cause) is a separate question (Direction B, the cause side), explicitly labeled in this chapter as a case study, with no identification claimed. Acknowledging that legislation has a political cause does not mean this instrument identifies Direction B — it precisely treats "the legislation having already occurred" as a given exogenous shock, and tests its downstream consequences; this is a different matter from "identifying the cause of the legislation." Conflating the two would lead one to mistakenly believe that any variation with a political cause cannot be used to identify Direction A — which is plainly false: almost all institutional variation has a political cause, but as long as its timing is exogenous relative to the contemporaneous pressure on the downstream dependent variable, it can be used to identify the downstream consequence. As to exogeneity, what instrument validity requires is not "legislation without political cause," but that "the legislative stagger (who comes first, the specific timing of the year-by-year steps) is independent of each fund's contemporaneous allocation pressure": in the Netherlands, staggered by industry, the correlation between the timing of an industry's mandatory effective date and any individual fund's contemporaneous investment decision is far weaker than its correlation with the industry's labor-relations history, and including industry fixed effects absorbs industry heterogeneity, with identification resting on the timing stagger; in Australia, the phased rate increase follows a nationally uniform timetable, with the rate steps imposed simultaneously on all funds, their timing determined by the federal legislative process rather than by any fund's allocation pressure — a near-exogenous dosage staircase. If the pre-adoption parallel trends are significant and exogeneity is in doubt, then the legislative-staggering instrumental variable is demoted from the most robust identification source to auxiliary descriptive evidence, and this chapter's consequence-side identification falls back to the withdrawal-event quasi-experiment (whose exogeneity is even stronger: emergency withdrawals were exogenously triggered by the populist politics of a public-health event, independent of any fund's investment decisions), and the core assertion does not thereby lose support.

Counter-argument Two (collinearity of the three moments and the "tautological restatement" objection): first, systems with high compulsion tend simultaneously to have high coverage, difficult withdrawal, and a high share of formal employment, so the three moments are highly correlated in the cross-national cross-section, and a regression cannot separate which moment is driving the effect. Second, the "synthetic asset" is in fact merely a tautological restatement of the liability-side story — "predictable mandatory inflow" is nothing but DB's liability duration viewed from the other side, and does not provide a new causal axis orthogonal to the preceding four chapters. This chapter's treatment is as follows. For the collinearity of the three moments, identification separation relies on different exogenous sources of variation to strike different moments: the duration moment uses the legislative-staggering instrumental variable, the tail moment uses the withdrawal event (a direct exogenous shock to "non-interruptibility," orthogonal to coverage and variance), and the variance moment uses the informal-employment rate as an instrument for contribution-density variance; the three moments are identified by three orthogonal shocks, breaking collinearity by design, with a pre-registered additivity test — if any two moments cannot be marginally separated, they are explicitly labeled as merged into a single channel, rather than being forced into three. As for the "tautological restatement" objection (this is the sharpest thrust, requiring a direct response), the falsifiable core of this chapter's orthogonality is: if the "synthetic asset" equals a tautological restatement of liability duration, then after controlling for the DB/DC label and liability duration, "the ratio of mandatory inflow to stock assets" should no longer predict the illiquid share. This chapter's falsifiable assertion is precisely the opposite — controlling for the label and maturity, the illiquid share rises monotonically with the institutional compulsion degree (statutory coverage rate times contribution-obligation rigidity), and within purely mandatory DC systems (with no DB liability) it rises with the ratio of newly added mandatory inflow to stock assets. The logical crux of this assertion lies in the DC system: a DC system by definition has no DB liability duration; if, within a purely mandatory DC system, the illiquid share still rises with the ratio of mandatory inflow to stock, then this effect cannot be attributed to the liability side (because there is fundamentally no DB liability duration to attribute it to), and the mechanism must come from the asset-side inflow itself — the tautological-restatement objection is thereby falsified. Furthermore, a decisive criterion can be used to pit the two explanations against each other in a way that permits adjudication: if some mature mandatory DC fund, during a net-outflow phase (new inflow smaller than outflow), still maintains high illiquidity without deleveraging, this shows that the mechanism cannot be explained by "contemporaneous inflow" alone and must be explained by "the sustainable regeneration of new members' mandatory inflow," supporting the asset-side inflow mechanism; conversely, if net outflow immediately triggers illiquid de-allocation, then the "synthetic asset" narrative is weakened, and the phenomenon is closer to conventional liquidity management. This criterion is decisive precisely because it places the "asset-side inflow mechanism" and the "tautological restatement of liability-side duration" explanations under opposing predictions, allowing the data to adjudicate between them. This chapter registers this counter-argument as a pending item, not as a closed assertion: its complete resolution requires precise measurement of liability duration within purely mandatory DC systems (to serve as a clean control) and net-outflow-phase samples (while most mandatory DC systems remain in the accumulation phase, so net-outflow samples are scarce); pending acquisition of these two, this chapter's claim of "orthogonality to the liability side" is presented in the form of a falsifiable hypothesis with a pre-registered criterion.

8.5 Evidentiary Status and Pre-registration Comparison

The verifiable layer of this chapter has been confirmed, and its evidentiary tier is relatively the highest among the five chapters. Australia's phased statutory superannuation contribution-rate timetable (from 1992, rising from 9% in fiscal year 2002–03 to 12% by July 2025, with a freeze at 9.5% from 2014 to 2021 in between) has been verified as a publicly verifiable fact (ATO 2024); the compulsory mechanism of the Netherlands' 1949 Industry-wide Pension Funds Act (under which the Minister of Social Affairs may order all employers in an industry to participate mandatorily, today covering more than 75% of employees) has been verified as a publicly verifiable fact (Chen and Beetsma 2014); the scale of Chile's three rounds of withdrawal in 2020–2021 (approximately $48–50 billion, roughly 19%–20% of gross domestic product, with assets under management falling from approximately 83% to approximately 62% by 2024) has been verified as a publicly verifiable fact (IMF 2021); the quantification that "every $1 withdrawn produces approximately $1.59 in future savings lost" has been verified as a publicly verifiable fact, directly confirming the multiplier cited in this chapter (Madeira 2022). In addition, one orthogonality argument must be added: the compulsion-degree index characterizes the legal rigidity of the inflow contract (statutory coverage rate, contribution-obligation rigidity, early-exit difficulty) — that is, "whether the contribution flow can be interrupted or diluted" — which is conceptually orthogonal to the defining distinction of the DB/DC label, "who bears the benefit risk": a purely mandatory DC system can have extremely high inflow-contract rigidity, and a voluntary DB can have a soft inflow, the two dimensions being independently variable. Therefore, this chapter's evidentiary weight rests on the core coefficient of the legislative-staggering instrumental variable plus the orthogonal counter-evidence, with the collapse test demoted to auxiliary description, non-load-bearing (avoiding repetition of the loop already deleted in Chapter 5's collapse test).

The micro-identification layer of this chapter could not be executed due to data-availability constraints. The causal estimates for the three identification legs (the panel instrumental variable and difference-in-differences for legislative staggering, the fund-by-time difference-in-differences for the withdrawal event, and the contribution-density-variance instrumental variable) all require micro-panel data — detailed multi-fund-tier panels, cross-national allocation panels, spliced contribution-density distributions, and a self-constructed compulsion-degree index and exit-difficulty coding — which are not available under this study's conditions, so the effect magnitudes, significance, and sample sizes of the three legs are uniformly not reported. It must be stressed that, although this chapter is the branch with the relatively highest evidentiary tier among the five chapters (its verifiable institutional facts are the most complete, and the tail moment has published causal quantification support), this "highest tier" remains confined to the layer of verifiable aggregate facts; the micro causal coefficients of the three legs themselves remain unestimated, and this chapter's Direction-A consequence-side assertion likewise remains, under current conditions, at the stage of "identification design in place, evidence pending." The identification design is already in place, including the pre-adoption parallel-trends-plus-placebo (no effect in non-mandated industries or non-rate-increase years) exogeneity gate, whose execution awaits the initiation of micro-data projects. Furthermore, since most mandatory DC systems remain in the accumulation phase, net-outflow-phase samples are scarce, so the decisive counter-evidence for "the synthetic asset as a tautological restatement of liability duration" (whether a mature mandatory DC fund in a net-outflow phase still maintains high illiquidity without deleveraging) is retained as a falsifiable hypothesis with a pre-registered criterion, not as a closed assertion. The pre-registration thresholds are fully retained: for the duration moment, the compulsion-degree coefficient is expected to be positive (this is the load-bearing threshold); whether the label coefficient collapses is only an auxiliary descriptive observation, non-load-bearing, and does not by itself constitute a decision gate — whether or not it holds does not separately trigger a retreat to the liability-side narrative, since the adjudication between the asset-side inflow mechanism and the liability-side narrative is borne by the orthogonal counter-evidence and the decisive criterion in the net-outflow phase; for the tail moment, the illiquid target is expected to be revised downward after withdrawal, with the magnitude increasing with historical withdrawal scale, operating through the expectations channel (the target has already fallen while contemporaneous cash flow is unchanged); for the variance moment, the illiquid share is expected to be negative with respect to contribution-density variance; for the orthogonal counter-evidence, the illiquid share within purely mandatory DC systems is expected to rise with the ratio of mandatory inflow to stock (if the effect is entirely borne by the DB subsample, the orthogonality claim is demoted to a pending hypothesis); for the additivity test, the three moments are expected to be marginally separable and additive; the exogeneity pre-gate must be passed, failing which the duration moment is instead supported by the withdrawal-event quasi-experiment. The cause side (Direction B) has no pre-registered hypothesis; the Dutch labor unions' insistence on mandating collective buffers to counter the financial sector, and the populist politics of Chile's withdrawal rights, are presented as political-economy cases and theory, without identification-style hypotheses; in this chapter's identification, the withdrawal event serves only as an exogenous shock to the "shadow price of non-interruptibility" (Direction A), not as a causal analysis of "why withdrawal rights were politically broken."

8.6 Summary

This chapter enters from the contractual nature of the asset-side inflow, arguing that the capacity for long-term illiquid allocation is determined by the predictability and non-interruptibility of the inflow, and characterizes its mechanism using the three moments of the synthetic asset (the duration moment, the tail moment, and the variance moment), with the three moments identified separately by three mutually orthogonal exogenous shocks (legislative staggering, the withdrawal event, and the informal-employment rate). This chapter is the branch with the most robust Direction-A consequence-side identification and the highest evidentiary tier among the five chapters, with the withdrawal-event quasi-experiment having the strongest exogeneity and published causal quantification support; but it must be reiterated that this "highest tier" remains confined to the layer of verifiable aggregate facts, the micro causal coefficients of the three legs themselves remain unestimated, and this chapter's consequence-side assertion likewise remains, under current conditions, at the stage of "identification design in place, evidence pending." This chapter's asset-side inflow perspective is orthogonal to Chapter 4's liability-side put-hardness perspective: the former asks "how the contractual nature of the inflow determines allocation capacity," the latter asks "who bears the risk, and how put hardness determines the sign of allocation"; at this point all five analytical axes have been laid out, and the next chapter will synthesize them across axes.

Chapter 9 Cross-Axis Synthesis

9.1 The Orthogonality and Coexistence of the Five Axes

The five analytical axes of this paper delineate five mutually orthogonal margins of the same institutional phenomenon — the divergence in DB/DC asset allocation under low interest rates and its political-economic origins. Chapter 4 enters from the option position on the liability side, arguing that the hardness of the put writer determines the sign of the allocation; Chapter 5 enters from the ontology of property rights, arguing that the degree of risk-sharing and the distributional core are the first-order state variables and that the ownership of the buffer property right determines the distributional consequences of de-risking; Chapter 6 enters from measurement visibility, arguing that the discount benchmark, the recognition tier, and the statistical moment order shape the visible signal at the margin and thereby shape allocation; Chapter 7 enters from system-level complementarity, arguing that the direction of fragility is a systemic property of the larger institutional configuration; Chapter 8 enters from asset-side inflows, arguing that the contractual nature of the contribution flow determines the capacity for long-term illiquid allocation.

These five axes are able to remain orthogonal precisely because each anchors a different causal dimension of the institutional phenomenon: Chapter 4 anchors the option structure on the liability side, Chapter 5 anchors the ontology and property rights of institutional form, Chapter 6 anchors the measurement rendering of information, Chapter 7 anchors the systemic configuration within which the institution is embedded, and Chapter 8 anchors the contractual nature of asset-side inflows. No two axes share a core mechanism — put hardness is not the degree of risk-sharing, visibility is not the degree of complementarity, and the contractual nature of inflows is not liability duration — so the five axes are not five formulations of the same mechanism but separate depictions of five distinct mechanisms. This orthogonality carries important methodological weight: it means that the failure of any one axis does not implicate the failure of the others (for example, the entire axis of Chapter 7 being downgraded to a theoretical lens does not undermine the standing of the Direction A identification in Chapter 8), and it allows the five axes to stand separately on different evidentiary tiers.

At the same time, the relationship among the five axes is not one of mutually competing, exclusive explanations, but one of coexistence at different margins. Take "why is DC allocated this way" as an example: the soft put of DC's legal choice (Chapter 4), the insufficient visibility (Chapter 6), and the nature of inflows (Chapter 8) act on different margins; the three coexist rather than being strictly orthogonal and mutually irreducible — the equity center of gravity of a DC default glide path may simultaneously be influenced by the legal liability-shielding structure (Chapter 4), the mode of information display (Chapter 6), and the contractual nature of mandatory inflows (Chapter 8), with these three jointly shaping the allocation at different margins rather than any one of them exclusively determining it. This coexistence is itself an embodiment of this paper's methodological stance: an institutional phenomenon is the result of multiple mechanisms superimposed at different margins, and any exclusive claim by a single mechanism is difficult to sustain; this paper therefore does not claim that any one axis is the "true" explanation, but rather claims that each of the five axes depicts a genuine margin, and their superposition constitutes a more complete understanding of the institutional phenomenon.

9.2 The Coherence of the Governing Judgment

The five axes are unified by the governing judgment at two levels. First, at the ontological level, the five axes share the premise that "institutions form a continuous spectrum rather than a discrete classification": put hardness is continuous (Chapter 4), the degree of risk-sharing is continuous (Chapter 5), the visibility dial is continuously adjustable (Chapter 6), the degree of complementarity is continuous (Chapter 7), and the degree of compulsion is continuous (Chapter 8). The DB/DC dichotomy degenerates, on every axis, into a corner solution of that continuous variable. This sharing is not a rhetorical coincidence, but the consistent application, across five dimensions, of this paper's stance that institutions should be understood as continuous state variables — regardless of which margin one enters from, DB and DC are not two natural categories but the two ends of some continuous variable; once this is accepted, "the DB/DC difference" is transformed from a question about categories into a question about the position of a continuous variable, and position is something that can be explained, moved, and determined by political bargaining.

Second, at the normative level, the five axes jointly serve the inquiry "who bears the risk that has been transferred away," which is the core focus of judgment in this paper. Chapter 4 asks at which writer the risk comes to rest — when the residual put is written by silent taxpayers or absent generations, the bearers of the risk are those subjects who cannot currently speak or claim compensation. Chapter 5 asks into whose account the buffer value is written — when de-risking writes the collective buffer into the accounts of the present generation (especially those near retirement with high political organization), the present generation obtains a one-off wealth gain, the price of which is the implicit expropriation of absent generations. Chapter 8 asks how risk flows back to individuals once an uninterruptible commitment is broken — when the opening of withdrawal rights breaks the uninterruptibility of mandatory inflows, risk that had already been smoothed across time through synthetic assets flows back to individual accounts, and this depresses the long-term allocation capacity of the entire system through the expectations channel. These three axes, from three different margins, point to the same normative conclusion: risk does not disappear, it is merely reallocated along different margins, and those who bear this reallocation are often the subjects who are absent or silent in the present political bargaining. This is this paper's unified answer to its focal judgment, and it is also what fundamentally distinguishes this paper from any research that treats "the rising share of DC" as the endpoint of analysis.

9.3 The Consistency of Honest Boundaries

It must be reiterated that the identification status of the five axes is consistent: every axis strictly distinguishes the identifiable Direction A consequence side from the Direction B causal side that cannot be cleanly identified; the core assertion of every axis claims only Direction A (or, in the case of Chapter 7, explicitly labels the entire axis as a theoretical lens); and the causal side of every axis is presented as a political-economy case study and theory. This consistency is reflected in the fact that, although the five axes sit on different evidentiary tiers — Chapter 8 is the most solid (withdrawal-event quasi-experiments plus published causal quantification), Chapter 4 is next (staggered legislation affecting different stakeholders), the identifiable core of Chapter 6 depends on prospective experiments not yet obtained, the identifiable leg of Chapter 5 depends on reserve panel data not yet obtained, and the entire axis of Chapter 7 is a theoretical lens — no axis states its unidentified portion in the language of identification.

Owing to data-availability constraints, the micro-causal coefficients on which the core assertions of the five axes depend have, under the conditions of this study, not been estimated for any of them; this paper therefore honestly reports, for every axis, that "the identification design is in place, the source of exogenous variation genuinely exists in reality, and the directional expectation is consistent with published evidence," while retaining falsifiability thresholds established prior to the data. It must be specially noted that this downgrade is an execution constraint arising from "data unavailability," not a repudiation of the identification design itself: the pre-registered thresholds of the five axes have not been loosened by the absence of data, and once the corresponding micro-panel and experimental data become available in the future, they can be executed according to the original design, at which point the falsifiability thresholds fixed at the outset will still constrain the judgment. This consistency of honest boundaries is not a coincidental convergence of five independent analyses, but the inevitable result of this paper's steadfast adherence, from beginning to end, to the stance that "judgment precedes data" — precisely because judgment was designed and fixed prior to data, the temporary absence of data does not shake the structure of the judgment; it only postpones the testing of the judgment.

9.4 Summary

This chapter has synthesized the five analytical axes, explaining how they coexist at different margins, how they are unified by the governing judgment at both the ontological and normative levels, and how they share a consistent honest boundary. It must be emphasized that the synthesis of the five axes is not a simple summation but a form of coordination: they share the ontological premise of a continuous spectrum, jointly serve the normative inquiry into risk reallocation, and maintain consistent honesty in the labeling of identification status. Precisely because of this, the limitations of any single axis (such as the entire axis of Chapter 7 being downgraded to a theoretical lens, or the s-first-orderness of Chapter 5 resting at the level of a theoretical proposition) do not undermine the overall structure — indeed, being honestly labeled as such enhances the credibility of the whole. Together, the five axes depict a complete picture: the pension system is a continuous spectrum, "de-risking" is a redistribution of risk that moves along that spectrum, and the direction, mechanism, and consequences of this redistribution are precisely where this paper's incremental contribution of judgment, as distinct from a textbook-style comparison, resides.

Chapter 10 Conclusion

10.1 Main Conclusions

Centered on the divergence in asset allocation between DB and DC pensions under a low-interest-rate environment, this paper proposes and substantiates a governing judgment: the contingency of benefits (the degree of risk-sharing) is a continuous institutional state variable, and DB/DC are degenerate corner solutions of that continuous spectrum; "de-risking," which moves along the spectrum, constitutes an intergenerational redistribution of wealth toward the present generation, rather than the mere elimination of risk. Around this judgment, this paper develops five mutually orthogonal analytical axes, which respectively delineate: the drive exerted on the sign of allocation by the hardness of the writer of the residual guarantee option; the status of the degree of risk-sharing and the distributional core as first-order state variables; the marginal shaping of allocation by the measurement-visibility layer; the property of systemic institutional complementarity as the direction of fragility; and the determination of long-term illiquid allocation capacity by the contractual nature of asset-side inflows.

Although these five axes are each independent and mutually orthogonal, they are unified by a single governing judgment at two levels: at the ontological level, the five axes share the premise that "institutions form a continuous spectrum rather than a discrete classification," with the DB/DC dichotomy degenerating, on every axis, into a corner solution of the corresponding continuous variable; at the normative level, the five axes jointly serve the inquiry "who bears the risk that has been transferred away." This unification renders the five axes not five isolated analyses, but five coordinated facets of the same institutional phenomenon.

The core contribution of this paper lies in placing the political-economic inquiry "who bears the risk that has been transferred away" at the center of the analysis. "The rising share of DC" is not the endpoint of analysis but a phenomenon awaiting explanation; risk has not disappeared with the shrinkage of DB — it has merely been reallocated along different margins: from the employer's balance sheet to individual accounts, from defined benefits to a buffer in a property-rights vacuum, from uninterruptible mandatory inflows to commitments that can be broken politically. Those who bear this reallocation are often the subjects absent or silent in the present political bargaining: future taxpayers, absent generations, workers not yet in the system. The direction, mechanism, and consequences of this transfer constitute this paper's incremental contribution of judgment, as distinct from any textbook-style DB/DC comparison — a textbook comparison concludes once it states that "DB bears risk, DC transfers risk," whereas this paper begins precisely there, to inquire into the political economy of risk transfer, and carries this inquiry through all five analytical axes.

10.2 Policy Implications

The analysis in this paper carries several implications for institutional design, derived directly from the judgments of the five analytical axes. First, institutional choice should not be understood as a trade-off between the two discrete options of DB and DC, but rather as the parametric setting of the degree of risk-sharing and the distributional core on a continuous spectrum. This implication derives from the continuous-spectrum framework of Chapters 3 and 5: given that the vast majority of real-world institutions fall within the spectrum, and given that the degree of risk-sharing can be explicitly parameterized as an adjustable dial, the true question of institutional design is not "choose DB or choose DC," but "where on the spectrum to set the degree of risk-sharing, and how to set the various parameters of the distributional core." The Netherlands' new system, which explicitly writes the cap on the solidarity reserve and the drawdown rules into legislation, is precisely an institutional practice of this understanding (DNB 2023) — it transforms institutional design from a binary choice of categories into an explicit specification of a set of continuous parameters. Second, the distributional consequences of "de-risking" depend on the specific design of the distributional core — whether the buffer value is written into present-generation accounts or retained as a cross-generational reserve depends on the representational structure of the governance body and the allocation of design authority; any de-risking reform should therefore explicitly examine its intergenerational distributional consequences, rather than being justified solely under the nominal goal of "reducing risk." This implication derives from the normative insight of Chapter 5: de-risking, in a property-rights vacuum, is substantively a distributional decision, and if its distributional dimension is not explicitly examined, a reform may complete an implicit wealth transfer to the present generation under the banner of "risk reduction." Third, the transplantation of mature-market experience cannot be done wholesale: the direction of fragility is a systemic property of the larger institutional configuration, and if a country's pension system is transplanted into an environment where complementary institutions have not yet co-evolved into place, the consequences may run in the opposite direction. This implication derives from the systemic-complementarity lens of Chapter 7, and it must be honestly labeled as a theoretical lens rather than an identified causal threshold — it provides a structural reason for why wholesale transplantation may fail, not a quantifiable probability of failure. Fourth, the uninterruptibility of mandatory inflows is the foundation of long-term illiquid allocation capacity; opening up withdrawal rights, even if harmless to current-period cash flow, will depress illiquid-allocation targets through the expectations channel. This implication derives from the tail-moment mechanism of Chapter 8: once the "uninterruptible" commitment is broken, the tail probability of future politically motivated redemptions is repriced, and the illiquid-allocation target is revised downward accordingly — a downward revision that occurs through the expectations channel, prior to any actual change in current-period cash flow. The quantification from Chile's withdrawal episode — that "each dollar withdrawn is associated with a loss of approximately 1.59 dollars in future savings" — provides cautionary evidence for this implication (Madeira 2022).

10.3 Research Limitations

This paper has several limitations that must be honestly stated. First, and most fundamentally, the empirical conclusions reported in this paper are confined to publicly verifiable institutional aggregate facts; the micro-causal identification on which the core assertions of the five axes depend — owing to the unavailability, under the conditions of this study, of the required fund-level panels and experimental data (such as a fund-level put-hardness panel, a micro-panel of reserves and allocation, participant behavioral logs from a visibility-randomization experiment, a multi-fund-tier, multi-period panel of withdrawal events, and self-constructed codings of compulsion and exit difficulty) — is uniformly reported as "identification design in place, evidence pending," and this paper does not claim that any causal effect has been established. This means that the contribution of this paper lies primarily at the level of theory, judgment, and method, and that empirical causal testing must await the initiation of micro-data projects. This limitation is fundamental, and this paper does not conceal it in any way — representing a design being in place as identification having been achieved, or representing directional consistency as causal proof, are slippages that this paper explicitly rejects. Second, the sample sizes of several key cases are inherently thin: the UK CDC currently has only one instance (Royal Mail); the history of extreme events at the system level in Chapter 7 is sparse (the UK in 2022 is the primary case, with the Netherlands and Japan not triggering during the same period); and the panel window for the Netherlands' solidarity reserve in Chapter 5 remains short (taking effect in 2023, with the transition to be completed by 2028 and the reserve cap to be met by 2037) — all of which limit the cross-sectional statistical power of the related propositions; even once micro-data become available in the future, early estimates of these propositions must be treated with caution, as confidence intervals may be wide. Third, several numerical figures, though corrected by this paper, still await verification against primary sources: the share of the one-off payment in a certain Danish plan's wholesale transition (which this paper verifies as approximately 20% rather than the commonly circulated 25%); the trajectory of Royal Mail's affordability threshold (this paper verifies its actual contribution rate as 19.6%, whereas the commonly circulated "50%" is in fact the benefit-efficiency ratio of CDC relative to DC, not a contribution rate); and the distinction in basis between UK pension funds' share of UK equity ownership and their in-fund equity allocation (32.4%→1.5% is the ownership share, while in-fund equity allocation fell from approximately 60%–70% to being predominantly liability-matching) — this paper has honestly labeled and corrected all three of these. Fourth, the causal side (Direction B) is presented throughout this paper as theory and case study, without clean identification having been obtained; the core question of the focal judgment — how allocation pressure reshapes institutions in reverse — remains an open avenue for research: what this paper is able to establish in a falsifiable manner is the consequence side of Direction A, whereas the causal side of Direction B, although it is the soul of the focal judgment, can, owing to the limitations of identification technique, only be presented as political-economic theory and case study. A consolidated discussion of this paper's various limitations and the corresponding research boundaries is provided separately in the closing section, "Data Availability and Research Boundaries."

10.4 Future Research

Building on the limitations above, this paper identifies several directions for future research, each corresponding one-to-one to this paper's identification designs and constituting a natural continuation of the research that follows this paper. First, the acquisition of micro-data and the execution of identification designs: this paper has established an executable identification blueprint and pre-registered thresholds for each of the five axes; once the corresponding fund-level panels and experimental data are obtained (regulators' fund-level annual reports, record-keepers' participant behavioral panels, rating agencies' methodology-revision libraries, multi-fund-tier, multi-period panels of withdrawal events, and so forth), they can be executed according to the original design, at which point the pre-registered falsifiability thresholds will still be binding, and this paper's judgments will face the tests fixed at the outset, unrelaxed by the data. Second, the search for independent exogenous variation on Direction B: the soul of this paper's focal judgment — how allocation pressure reshapes institutions in reverse — can, for lack of clean identification, only be presented through theory and case study; any exogenous variation discovered in the future that is independent of contemporaneous allocation pressure, with the arrow genuinely running from allocation pressure to institutional form, could elevate the causal side from case study to identification, thereby filling the most important gap in this paper. Third, the construction of orthogonal state-variable measures: if a measure of the degree of risk-sharing independent of benefit contingency could be constructed (built purely from buffer property-rights structure and bargaining structure), or a measure of complementarity that precedes the crisis and excludes LDI exposure (built purely from pre-crisis financial structure), then the first-order proposition of the corresponding axis (Chapter 5) or the causal proposition (Chapter 7) could be elevated from theory to empirics. Fourth, the continued tracking of hybrid and transitional forms: hybrid forms such as the UK CDC and the Netherlands' new system are still in an early stage; as samples accumulate and panel windows lengthen, this paper's many propositions concerning the middle segment of the continuous spectrum will gain stronger cross-sectional statistical power, at which point this paper's directional expectations can be subjected to more stringent testing. Together, these directions sketch a research path running from judgment to identification, from design to execution.

10.5 Closing Remarks

This paper's position, from beginning to end, has been that the core of economic research is judgment, not mechanical data statistics; that the combination of humanistic and quantitative thinking requires that implementation and testing be addressed only after a sufficiently good analytical framework and identification strategy have been designed, rather than proceeding through undisciplined, ad hoc data exploration. This paper's re-parameterization of the pension-system continuous spectrum, its political-economic inquiry into "who bears the risk that has been transferred away," and its honest labeling of the identification status of the five analytical axes, are all embodiments of this position. This position also explains why, in numerous places, this paper would rather honestly label "identification design in place, evidence pending" than use identifying language to paper over gaps in evidence — because the value of a piece of research lies first in whether its judgment is sharp and falsifiable, not in whether its data have already been piled up; a judgment fixed and falsifiable prior to the data is closer to the true meaning of economics than a conclusion tailor-made for the data after the fact and incapable of falsification.

Returning to this paper's focal judgment: risk does not disappear, it is only redistributed. From the employer's balance sheet to the individual's account, from defined benefits to a buffer in a property-rights vacuum, from uninterruptible mandatory inflows to commitments that can be broken politically — every act of "de-risking" is an act of redistribution, and those who bear this redistribution are often the subjects absent or silent in the present political bargaining. "The rising share of DC" has never been a neutral technical trend; it is the cumulative result of countless redistributions of this kind. Seeing clearly the direction, mechanism, and consequences of this redistribution, and re-exposing the distributional dimension obscured by prudential discourse, is precisely where the value of judgment relative to data lies, and is also the fundamental contribution that distinguishes this paper from any textbook-style DB/DC comparison.

Data Availability and Research Boundaries

This section consolidates the downgrading statements scattered throughout the chapters of this paper into a standardized statement of research limitations, so that readers may survey in one place the precise boundaries of this paper's empirical conclusions. The empirical statements throughout this paper strictly distinguish two layers: first, publicly verifiable institutional aggregate facts, each labeled with an MLA in-text citation to its source, which constitute the portion of this paper that can be supported by a factual foundation; second, causal identifications that could not be executed owing to constraints on the availability of micro-data, uniformly presented as "identification design in place, the source of exogenous variation genuinely exists in reality, and the directional expectation is consistent with published evidence," without being upgraded to established causation.

Specifically, the micro-causal coefficients on which the core assertions of the five analytical axes depend have, under the conditions of this study, not been estimated for any of them. The data that would be needed but are unavailable include: a fund-level put-hardness panel (self-constructed coding of contractual trigger ladders, safe-harbor strength, and implicit public backstops, Chapter 4); a fund-level panel of reserve thickness and allocation and a series of interest-rate shocks (Chapter 5, Model B); a sponsor-level panel and form-conversion events (Chapter 5, Model C); participant behavioral logs from a field experiment on display randomization (Chapter 6, Leg A); a library of rating-methodology revision dates and a tier panel (Chapter 6, Leg B); a hand coding of cross-jurisdictional recognition stringency and complementarity (Chapters 6 and 7); and a multi-fund-tier, multi-period panel of withdrawal events together with self-constructed codings of compulsion and exit difficulty (Chapter 8). For all of these, this paper reports no regression coefficient, significance level, or sample size.

The standardized limitation statements for each chapter may be summarized as follows: the causal effect of put hardness in Chapter 4 remains at "identification design in place, evidence pending"; the three empirical legs of Chapter 5 (the normative insight, reserve absorption, and affordability) all remain at the design stage owing to the unavailability of micro-data, with s-first-orderness confined to the status of a theoretical proposition; the marginal adjustability of the visibility layer in Chapter 6 remains at "design in place" owing to the unavailability of prospective experimental data, with the visibility-based account of historical bifurcation confined to the status of a hypothesis; the entire axis of Chapter 7 is confined to the status of a theoretical and descriptive lens, because identification is in principle unattainable; and although Chapter 8 sits on the relatively highest evidentiary tier, the micro-causal coefficients of its three legs have likewise not been estimated, and its consequence-side assertions likewise remain at "identification design in place, evidence pending." In addition, the sample sizes of several key cases are inherently thin (the UK CDC has only one instance, extreme events at the system level are sparse, and the panel window for the Netherlands' solidarity reserve remains short), limiting the cross-sectional statistical power of the related propositions; three numerical figures, though corrected by this paper, still await verification against primary sources (the one-off payment in a certain Danish plan's transition is approximately 20% rather than 25%; Royal Mail's actual contribution rate is 19.6% rather than the commonly circulated contribution trajectory; and the distinction in basis between UK pension funds' share of UK equity ownership and their in-fund equity allocation). None of these boundaries loosen any pre-registered threshold on account of the temporary absence of data: once the corresponding micro-data become available in the future, they can be executed according to the original design, at which point the falsifiability thresholds established prior to the data will still constrain the judgment.

Numbered Figures and Tables

Table 1 Core Mechanisms Across the Five Analytical Axes

Table 2 Competing Hypotheses and Identification Strategies

Table 3 Data Sources and Availability of Verifiable Facts

References (Works Cited)

Atif, Muhammad, Sivathaasan Nadarajah, and Grant Richardson. "Staggered Adoption of Stakeholder Constituency Statutes and Corporate Cash Holdings in the U.S." Economic Modelling, vol. 124, 2023, www.sciencedirect.com/science/article/pii/S0264999323001372.

Australian Taxation Office (ATO). "Super Guarantee Rates and the Superannuation Guarantee (Administration) Act 1992." ato.gov.au, Australian Government, 2024, www.ato.gov.au/rates/key-superannuation-rates-and-thresholds/.

British Progress. ISA and Pension Reform: Why Forced Investment Is No Panacea. British Progress, 2024, britishprogress.org/reports/isa-and-pension-reform-why-forced-investment-is-no.

Chen, Damiaan, and Roel M. W. J. Beetsma. Mandatory Participation in Occupational Pension Schemes in the Netherlands and Other Countries. CESifo Working Paper No. 4593, Ifo Institute, 2014, www.ifo.de/DocDL/cesifo1_wp4593.pdf.

De Nederlandsche Bank (DNB). "Solidariteitsreserve en risicodelingsreserve." Open Boek Toezicht, De Nederlandsche Bank, 2023, www.dnb.nl/voor-de-sector/open-boek-toezicht/sectoren/pensioenfondsen/verzamelpagina-transitie-wet-toekomst-pensioenen/contract/solidariteitsreserve-en-risicodelingsreserve/.

Estévez-Abe, Margarita, Torben Iversen, and David Soskice. "Social Protection and the Formation of Skills: A Reinterpretation of the Welfare State." Varieties of Capitalism: The Institutional Foundations of Comparative Advantage, edited by Peter A. Hall and David Soskice, Oxford University Press, 2001, pp. 145–83.

Gao, Huasheng, Kai Li, and Yujing Ma. "Stakeholder Orientation and the Cost of Debt: Evidence from State-Level Adoption of Constituency Statutes." Journal of Financial and Quantitative Analysis, vol. 56, no. 6, 2021, pp. 1908–44, www.cambridge.org/core/journals/journal-of-financial-and-quantitative-analysis/article/abs/stakeholder-orientation-and-the-cost-of-debt-evidence-from-statelevel-adoption-of-constituency-statutes/534204BB689CD6894ED6EE60236E2FF7.

House of Commons Library. Collective Defined Contribution (CDC) Pension Schemes. Briefing Paper CBP-8674, UK Parliament, 2024, commonslibrary.parliament.uk/research-briefings/cbp-8674/.

ii.co.uk. "Big Final Salary Pension Schemes Dump UK Shares." Interactive Investor, 2023, www.ii.co.uk/analysis-commentary/big-final-salary-pension-schemes-dump-uk-shares-ii528043.

International Monetary Fund (IMF). Assessing Chile's Pension System: Challenges and Reform Options. IMF Working Paper WP/21/232, International Monetary Fund, 2021, www.imf.org/en/Publications/WP/Issues/2021/09/17/Assessing-Chiles-Pension-System-Challenges-and-Reform-Options.

La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny. "Law and Finance." Journal of Political Economy, vol. 106, no. 6, 1998, pp. 1113–55.

Madeira, Carlos. "The Impact of the Chilean Pension Withdrawals during the Covid Pandemic on the Future Savings Rate." Journal of International Money and Finance, vol. 126, 2022, www.sciencedirect.com/science/article/abs/pii/S0261560622000535.

NautaDutilh. "The Dutch Future Pensions Act: Five Things You Need to Know." nautadutilh.com, NautaDutilh, 2023, www.nautadutilh.com/en/insights/the-dutch-future-pensions-act-five-things-you-need-to-know/.

Pensions Age. "New Entrants Bolster Pension Risk Transfer Market Volumes in 2023." Pensions Age, 2024, www.pensionsage.com/pa/New-entrants-bolster-pension-risk-transfer-market-volumes-in-2023.php.

Pensions in Denmark. Wikipedia, Wikimedia Foundation, 2024, en.wikipedia.org/wiki/Pensions_in_Denmark.

Sharpe, William F. "Corporate Pension Funding Policy." Journal of Financial Economics, vol. 3, no. 3, 1976, pp. 183–93.

Treynor, Jack L. "The Principles of Corporate Pension Finance." The Journal of Finance, vol. 32, no. 2, 1977, pp. 627–38.

[^1]: For the purposes of this paper's citations, "ownership share" and "within-fund allocation ratio" are two bases that must not be conflated: the former characterizes the proportion of the entire UK equity market held by pension funds, while the latter characterizes the proportion of a single fund's assets allocated to equities. Conflating the two would lead to serious misreadings of the magnitude of UK DB de-risking; this paper therefore strictly distinguishes between them in every relevant citation.