Implications of Mature Pension Markets' Asset Allocation Experience for China's Multi-Pillar Pension System: An Institutional Adaptation Study Based on Long-Term Capital, Asset-Liability Management, and Risk Governance

Chapter 1 Introduction

Section 1 Statement of the Problem

Population aging and the downward shift of the interest-rate center are two structural pressures that global pension systems currently face in common. For mature pension markets, these two pressures have driven the continuous evolution of asset allocation strategies: strategic allocation to equity assets, global diversification, the introduction of alternative assets, the refinement of asset-liability management (ALM) frameworks, and the institutionalization of decumulation (payout) phase risk governance — these constitute the mainline of pension investment experience over the past several decades. Among these, long-term holding of equities and broad diversification as benchmark strategies for long-term capital have been systematically expounded in the classic investment literature (Malkiel 2018). For China, which is still in the phase of system-building, these experiences are naturally viewed as ready-made solutions worth learning from. Existing policy discussions and academic research have mostly proceeded along the path of summarizing overseas experience, comparing it against China's current conditions, and proposing improvement recommendations, with the implicit premise that allocation logics that have proven effective in mature markets can, after checking for institutional differences, be appropriately adjusted and transplanted to the corresponding tiers of China's pension system.

This study raises a fundamental challenge to this premise. The starting point of this challenge lies in a distinction that existing research has systematically overlooked: the transferability of allocation techniques and the availability of allocation preconditions are two different matters. That an asset allocation strategy proves effective in mature markets is often not merely because the strategy itself is well designed, but because the entire set of institutional preconditions on which the strategy depends — independent governance structures, an exogenously stable risk-free rate benchmark, a unified accounting basis, mature default contribution mechanisms, and an annuitization market backed by adequate reserves — already exists locally and is taken for granted. When this set of institutional preconditions is abstracted away and only the allocation technique is extracted and transplanted on its own, the success or failure of the transplant no longer depends on the merits of the technique, but on whether these taken-for-granted institutional preconditions can likewise be independently obtained at the target tier of the target environment.

This distinction is by no means mere wordplay. It directly determines how the core policy question of which experiences can be drawn upon and which cannot be simply copied is to be answered. Without this distinction, "cannot be simply copied" easily degenerates into an empty conclusion unsupported by any institutional constraint, or worse, degenerates into a shield that can be arbitrarily raised against any concrete proposal; while "can be drawn upon" easily degenerates into a fast track that skips institutional verification, directly pasting an overseas template onto some tier in China. This study attempts to draw a unified, item-by-item verifiable criterion for what can and cannot be transplanted: whether a given experience can be transplanted depends on whether the institutional preconditions that underpin it can be independently obtained at the target tier — if the preconditions are available, transplantation is possible; if the preconditions are absent, transplantation is not possible, and the precondition must first be built. This criterion reconstructs "cannot be simply copied" from a mere conclusion into a verifiable proposition about the availability of preconditions, thereby simultaneously constraining both the abuse of "cannot be simply copied" and the abuse of direct borrowing — two opposite tendencies.

Section 2 The Literature Gap

Existing research on China's pension asset allocation can broadly be classified into three types, each of which, to varying degrees, has overlooked the institutional-precondition problem that concerns this study.

The first type of research focuses on optimizing allocation ratios. This type of research typically builds on some objective function to argue that China's pensions should raise the allocation share of equity, overseas, or alternative assets, and provides suggested target ranges. Its contribution lies in quantification; its limitation lies in reducing the allocation problem to a technical problem of solving for an optimal ratio under given constraints, while lacking examination of the institutional origins of the constraints themselves, the heterogeneity of the constraints across different tiers, and the political-economic conditions required to change the constraints. When this type of research offers recommendations such as steadily raising equity allocation, it often fails to answer how much to raise it to, why that particular number, what would happen without the increase, and it does not ask whether the same allocation logic can be applied across different pension tiers.

The second type of research focuses on surveying and drawing on overseas experience. This type of research systematically introduces the governance and allocation practices of pensions in Canada, Japan, Norway, the United Kingdom, the Netherlands, Australia, the United States, and other countries, and proposes directions that China may draw upon. Its contribution lies in broadening perspective; its limitation lies in the fact that what is drawn upon often remains at the level of allocation outcomes and institutional forms, lacking penetration into the deep institutional preconditions on which these outcomes and forms depend. When this type of research recommends that a particular tier in China draw on a particular overseas template, it typically checks surface-level differences in fund attributes and regulatory frameworks, but fails to verify whether deep preconditions such as governance independence, the exogeneity of the risk-free anchor, and the carrier of the default mechanism are available at the target tier.

The third type of research focuses on introducing asset-liability management frameworks. This type of research emphasizes that China's pensions should establish tools such as duration matching and risk budgeting. Its contribution lies in introducing liability-oriented allocation thinking; its limitation lies in the fact that its analytical framework presumes the existence of an exogenously stable risk-free interest rate benchmark as the anchor for duration matching, without examining whether, in an environment where fiscal policy is deeply involved in the pricing of safe assets, this benchmark itself might become an endogenously shifting variable. When the risk-free anchor itself shifts along with fiscal operations, an asset-liability management framework premised on a stable risk-free anchor loses its footing.

The common gap across the three types of research is that they all treat institutional preconditions as a given backdrop rather than as a variable to be examined. The contribution of this study lies precisely in bringing this backdrop to the foreground, treating whether institutional preconditions can be independently obtained at the target tier as the core variable running through the entire analysis, and reconstructing the diagnosis of China's pension asset allocation problem on this basis.

Section 3 Research Contributions and Structure

The contributions of this study can be summarized at three levels.

At the framework level, this study establishes a tiered adaptation framework that fills in, cell by cell, the four entity tiers of China's pension system — the Basic Pension Insurance Fund, the National Social Security Fund, enterprise annuities and occupational annuities, and personal pensions — across six dimensions: account tier and its corresponding fund duration, liability constraints, risk budget, allocation strategy, and regulatory assessment, thereby demonstrating the incommensurability among the four tiers. Pension target funds do not constitute an independent liability entity; rather, they serve as the investment vehicle of the fourth tier — personal pensions — and as the landing point of the bidirectional linkage between pensions and capital markets, and are treated as an appendix tier attached to the fourth tier and the payout end, rather than listed as an independent judgment cell. The value of this framework lies not in the enumeration of dimensions, but in the judgment content of each cell and the incommensurability between cells — that is, why what the National Social Security Fund can do, the Basic Pension Insurance Fund cannot; why the allocation logic applicable to the second pillar cannot be applied to the third pillar.

At the judgment level, this study distills five sets of judgments that are mutually independent and share no core mechanism, corresponding respectively to five problem domains: the governance carrier, the risk-free anchor, accounting visibility, behavioral architecture, and payout-end absorption. Each set of judgments is presented in a falsifiable, non-obvious propositional form, accompanied by a rigorous identification strategy and pre-registered falsification conditions, thereby drawing a clear line against policy slogans devoid of empirical content. For ease of cross-chapter reference, these five problem domains are hereafter referred to as the first through fifth sets of judgments; the five sets share no core causal mechanism among them.

At the closing level, this study explicitly builds on the conclusions of four prior studies on the interest-rate transmission mechanism, institutional differences, the endogeneity of allocation-tool governance, and decumulation-phase risk governance, bringing the general conclusions of overseas experience down to the specific circumstances of China's four entity tiers and the pension target fund appendix tier, thereby demonstrating why China's approach holds, and thus closing the complete thread running from macro interest-rate shocks, institutional differences, allocation transition, and risk exposure to China's adaptation.

The remainder of this study is organized as follows. Chapter 2 lays out the institutional background of China's three-pillar pension system, providing the institutional-factual foundation for the subsequent analysis. Chapter 3 establishes the theoretical framework, elucidating three basic principles: the incommensurability of the four tiers, the endogeneity of governance, and the impossibility of simple copying. Chapters 4 through 8 respectively analyze the five problem domains of the governance carrier, the risk-free anchor, accounting visibility, behavioral architecture, and payout-end absorption, each chapter following the structure of judgment, mechanism, identification strategy, honest labeling of evidence, and summary. Chapter 9 provides a cross-paper closing synthesis, explicitly building on the conclusions of the four prior papers to demonstrate the China-adaptation framework. Chapter 10 is the conclusion, presenting policy implications and research limitations.

Chapter Summary

This chapter has presented the core question of this study: that the transferability of pension asset allocation experience depends on whether the institutional preconditions underpinning that allocation logic can be independently obtained at the target tier, rather than on the sophistication of the allocation technique itself. This chapter has pointed out the common gap in existing research across the three directions of allocation-ratio optimization, surveying of overseas experience, and the introduction of asset-liability management frameworks — namely, treating institutional preconditions as a given backdrop rather than as a variable for examination — and on this basis has established this study's analytical stance of treating the availability of institutional preconditions as the core variable. Finally, this chapter has laid out the study's three-level contribution and overall structure, establishing the framework for the chapters that follow.

Chapter 2 The Institutional Background of China's Three-Pillar Pension System

Section 1 The Structure of Three Pillars, Four Entity Tiers, and an Appendix Tier

China's pension system follows the internationally common three-pillar classification at the overall architectural level, but the complexity of its internal structure far exceeds what a three-way division can capture. The first pillar is basic pension insurance, comprising basic pension insurance for urban employees and basic pension insurance for urban and rural residents; it is mandated by national legislation, adopts a model combining social pooling with individual accounts, and is the pillar with the broadest coverage, providing basic pension security. The second pillar is supplementary pension insurance, comprising enterprise annuities for enterprise employees and occupational annuities for staff of government agencies and public institutions; it is jointly funded by employers and employees and adopts an account-based accumulation model. The third pillar is the personal pension, which individuals voluntarily join, in which individuals independently choose their investments, which enjoys tax preferences, and which is an account-based personal pension savings arrangement.

In the analysis of this study, in order to fully unfold the tier-level differences in allocation logic, the three pillars described above are further subdivided into four entity account tiers and one appendix carrier tier. The first tier is the Basic Pension Insurance Fund, whose fund attributes prioritize safety and payment capacity, with allocation objectives subject to constraints of fiscal sustainability, matching of payment maturities, and risk floors. The second tier is the National Social Security Fund, which, as the nation's long-term strategic reserve, has room for long-term equity investment, global diversified allocation, and prudent engagement with alternative assets, and is equipped with specialized governance. The third tier is enterprise annuities and occupational annuities, which as second-pillar occupational pensions involve allocation logic concerning default mechanisms, life-cycle design, risk stratification, and long-term assessment. The fourth tier is the personal pension, which as a third-pillar account-based arrangement depends for its effective operation on tax incentives, low-fee defaults, investor education, and long-term holding. The first four tiers constitute the four entity tiers that this study analyzes cell by cell. The pension target fund and its linkage with capital market development do not constitute an independent account tier, but rather the investment vehicle through which the fourth tier — the personal pension — enters the market: pensions here are simultaneously a source of long-term capital for the capital market and dependent on the capital market's supply of assets suitable for long-term allocation; this bidirectional pension-capital market linkage is the market-side extension of the fourth tier's allocation strategy, and this study therefore treats it as an appendix tier attached to the fourth tier, with its judgment content unfolding together with the fourth tier's behavioral architecture and payout-end absorption, rather than being listed as an independent judgment cell. The structural differences among the four entity tiers, together with the positioning of the appendix tier, can be presented collectively in the table below.

Table 1 Comparison of Fund Duration, Liability Constraints, and Assessment Mechanisms Across China's Four Pension Entity Tiers (with the Pension Target Fund Appendix Tier)

The fund attributes, liability constraints, and risk tolerance of the four entity tiers differ enormously. The National Social Security Fund, as a strategic reserve free of rigid short-term payment pressure, can bear greater short-term volatility in exchange for long-term returns; the Basic Pension Insurance Fund, as pooled capital bearing immediate payment obligations, must place payment capacity above returns. The liability structure of second-pillar occupational annuities lies between defined benefit and defined contribution, while third-pillar personal pensions are entirely defined-contribution individual accounts. This profound heterogeneity across tiers constitutes the institutional basis for this study's fundamental stance that the same asset allocation framework cannot be uniformly applied across all tiers.

This heterogeneity is also reflected in the tier-level differences in regulatory and assessment mechanisms, and differences in assessment mechanisms in turn shape the allocation behavior of each tier. The Basic Pension Insurance Fund is subject to the strictest safety-oriented assessment, and its allocation behavior is highly prudent; the National Social Security Fund uses long-term returns as its assessment benchmark, tolerating greater short-term volatility and a longer investment horizon; second-pillar occupational annuities are subject to relatively short-term assessment pressure, and their allocation behavior is often more conservative than what their strategic liability structure would otherwise allow. The mismatch between the assessment cycle and the liability duration is an important institutional clue for understanding why the allocation behavior of each tier deviates from its theoretical optimum. This clue is repeatedly invoked in subsequent chapters: in the chapter on the governance carrier, administrative bans synthesize patient capital, which in substance replaces voluntary long-term holding with assessment constraints; in the chapter on behavioral architecture, downward migration is in substance participants' assessment-driven response to the gap in benefits across tiers. Precisely because the differences in assessment mechanisms across entity tiers are so large, any attempt to apply a unified allocation framework across all tiers will fail by ignoring the incommensurability of assessment mechanisms across tiers.

Section 2 The Low Interest Rate Environment and the Compression of Safe-Asset Returns

In recent years, China's risk-free interest rate center has undergone a systemic downward shift, delivering a profound shock to the asset side of the pension system. This shock is concentrated in the successive reductions of the guaranteed interest rate on life insurance products. The cap on the guaranteed interest rate for ordinary life insurance products, previously 3.5 percent, was lowered to 3.0 percent, then to 2.5 percent effective September 1, 2024, and further to 2.0 percent effective August 31, 2025; in January 2025 the regulatory authorities established a dynamic adjustment mechanism linking the guaranteed rate to market interest rates, and the research value for July 2025 fell to 1.99 percent, a new low since 1999 (National Financial Regulatory Administration 2024). Although the guaranteed interest rate is a pricing anchor on the liability side, its successive reductions directly reflect the systemic compression of safe-asset returns: the government bond yield curve, serving as the duration-matching benchmark, has correspondingly broken downward.

The compression of safe-asset returns means far more for the pension system than a mere decline in return levels. It has brought to the surface, one by one, a series of institutional contradictions that were implicitly masked in a higher interest rate environment. The wedge between the administratively set crediting rate and the continuously declining market rate has widened; the liquidity cost implicit in closed accounts has become expensive as opportunity cost rises; and the put option embedded in guaranteed-floor arrangements has raised true liabilities as its in-the-money probability rises. These contradictions were not created by low interest rates, but were pushed from implicit to explicit by low interest rates. The reason the empirical implications of the many mechanisms analyzed in the subsequent chapters of this study have become testable in the current environment is precisely that low interest rates have brought these originally latent costs to the surface. How shifts in the interest rate environment expose latent fragilities in the asset-liability arrangements of long-term institutional investors is also a systemic issue continuously discussed in the Bank for International Settlements' annual economic report (Bank for International Settlements 2023).

Section 3 The Deep Entanglement of Fiscal Policy and Pensions

A prominent institutional feature of China's pension system is the deep entanglement between fiscal policy and the asset side of pensions, a feature that is uncommon in international comparison yet decisive for allocation logic.

This entanglement is first reflected in the support that state capital provides to the first pillar. To fill the historical gap formed by deemed contribution years in basic pension insurance for enterprise employees, the state implemented a policy of transferring a portion of state-owned capital to replenish the social security fund, with the transfer ratio set at 10 percent of the relevant state-owned equity; the transfer work was essentially completed by 2020, with a transfer scale of 1.68 trillion yuan, and a book value of approximately 2.1 trillion yuan in 2024 (Ministry of Finance 2021). This means that the solvency of the first pillar is, to a considerable degree, supported by a contingent stream of state-owned capital dividends, intertwining fiscal contingent liabilities with pension payment obligations.

This entanglement is secondly reflected in the fiscal-commitment nature of the crediting mechanism. In occupational annuities, the employer contribution portion for units with full fiscal funding is handled via a crediting method, accruing interest at a specified crediting rate, with only the remaining portion subject to market-based operation; the crediting rate standard is determined by local human resources and social security departments based on actual investment returns (Ministry of Human Resources and Social Security 2015). The crediting rate for individual accounts in basic pension insurance for urban and rural residents is administratively set and decoupled from the market, with a level significantly higher than the contemporaneous deposit rate. These administratively set crediting rates are, in essence, redistributive prices of fiscal commitments rather than investment returns actually realized on the asset side. When market rates decline while crediting rates remain elevated due to countercyclical commitments, the difference between the two is converted into an off-balance-sheet fiscal contingent liability.

This entanglement is finally reflected in the institutional arrangements for pooling and absorption. The Interim Measures for the Operation and Management of Cash Proceeds from the Transferred State-Owned Equity Replenishing the Social Security Fund, issued in 2024, stipulates that local absorbing entities shall entrust no less than 50 percent of the cumulative cash proceeds as of the end of the prior year to the National Council for Social Security Fund for investment and operation, with the remainder operated independently by local absorbing entities within a defined scope, the defined scope including bank deposits, purchases of government bonds in the primary market, and capital injections into the transferred enterprises and their controlled enterprises (Ministry of Finance 2024). This system both specifies the intensity of pooling and sets boundaries for it, behind which lies a complex contest between central and local governments over control rights of pension assets.

The deep entanglement between fiscal policy and pensions means that any overseas allocation experience whose implicit premise is fiscal neutrality and the exogeneity of the risk-free anchor must, when transplanted to China, first verify the institutional constraints brought about by this entanglement. This judgment will be repeatedly invoked in subsequent chapters.

Chapter Summary

This chapter has laid out the institutional background of China's three-pillar pension system. It has explained the subdivision of the three pillars into four entity tiers (with the pension target fund appendix tier) and the profound heterogeneity among the tiers in fund attributes, liability constraints, and risk tolerance, providing the institutional basis for the stance that the same allocation framework cannot be uniformly applied across all tiers. It has described the systemic compression of safe-asset returns under the low interest rate environment, pointing out that the primary effect of low interest rates is to push originally implicit institutional contradictions into visibility. Finally, this chapter has characterized the deep entanglement between fiscal policy and the asset side of pensions, including the support that state capital provides to the first pillar, the fiscal-commitment nature of the crediting mechanism, and the institutional arrangements for pooling and absorption — an entanglement that constitutes a common source of institutional constraints for the analysis in subsequent chapters.

Chapter 3 Theoretical Framework: Four-Tier Incommensurability, Governance Endogeneity, and Non-Transferability

Section 1 The Tiered Adaptation Framework and Four-Tier Incommensurability

The tiered adaptation framework is the basic structure organizing all judgments in this study. Its basic thesis is: different pension accounts correspond to different fund horizons, different fund horizons correspond to different liability constraints, different liability constraints correspond to different risk budgets, different risk budgets correspond to different allocation strategies, and different allocation strategies in turn correspond to different regulatory and assessment mechanisms. This chain is not a simple juxtaposition of six dimensions but a deductive structure that transmits step by step from fund characteristics to regulatory assessment. The framework requires that these six dimensions be filled in cell by cell across four institutional tiers, and strictly prohibits applying the same allocation logic across tiers; pension target funds, as a supplementary tier, do not constitute an independent liability entity but rather serve as the investment vehicle for the fourth-tier individual pension and the point of convergence for bidirectional linkage between pensions and capital markets, with their judgment content presented as a supplement alongside the fourth tier and the decumulation (payout) phase, rather than as a separate independent judgment cell.

The core value of the framework lies in the incommensurability between cells. By incommensurability we mean that an allocation judgment valid at one tier cannot be transposed to another tier, because the institutional preconditions supporting that judgment do not hold at the other tier. The reason the National Social Security Fund can undertake long-term equity investment and globally diversified allocation is that, as a strategic reserve, it faces no rigid immediate payout pressure and is equipped with specialized, independent governance; the basic old-age insurance fund cannot do the same, not because its managers lack competence, but because it bears immediate payout responsibility and does not possess independent governance. The reason the second-pillar occupational annuity can introduce default mechanisms and life-cycle design is that an employer exists as an automatic contribution-bearing entity; the third pillar, when facing the flexibly employed population, cannot do the same, not because that population lacks contribution capacity, but because the bearing entity on which the default mechanism depends is absent for that population. Identifying the judgment content of each cell and revealing the institutional roots of incommensurability between cells is the central task of this study.

Section 2 The Principle of Governance Endogeneity

Governance endogeneity is the first basic principle on which this study judges what is transferable and what is not. This principle holds that: the truly scarce component of pension allocation capacity — the component that truly determines the feasibility of allocation — is not technical capability that can be outsourced or built, but governance control rights that are endogenous to institutional structure and cannot be transferred across entities.

The principle of governance endogeneity can be broken down into two dimensions. On the divide between capability and control rights: the professional technical capabilities required for allocation — investment research, risk measurement, asset valuation — can in principle be acquired by recruiting talent, purchasing services, or engaging professional institutions, and are thus transferable; whereas the governance control rights on which allocation depends — the ultimate decision-making authority over asset disposal, the institutionalized arrangement of risk attribution — are endogenous to an entity's position within the institutional structure and cannot be independently acquired across entities or tiers. On the divide between bearing entity and technique: the operating technique of a mechanism can be transplanted, but the institutional bearing entity on which that mechanism depends for its operation may not be transferable across environments. The automatic deduction software for a default contribution mechanism can be purchased externally, but the employer payroll withholding relationship on which the default mechanism depends for its validity cannot be created out of nothing.

The principle of governance endogeneity provides the underlying logic of the criterion for the tiered adaptation framework: a given allocation experience is transferable if, and only if, the governance control rights or institutional bearing entity that carry it can be independently obtained at the target tier. This principle will be developed concretely across the chapters on governance bearing entities, behavioral architecture, and decumulation-phase absorption.

Section 3 The Argumentative Criterion for Non-Transferability

Non-transferability is the second basic principle on which this study constrains policy recommendations, and also the watershed distinguishing this study from policy sloganeering. This principle requires: any conclusion of non-transferability must not remain at the level of a bare conclusion but must be supported on two fronts.

The first is a specific institutional constraint. Every conclusion of non-transferability must specify precisely which specific institutional precondition is absent at the target tier, thereby depriving the overseas template of its foothold. For example, the transferability of the Canadian pension model presupposes its governance independence, whereas the corresponding tier's governance independence in China depends on administrative rank rather than legal guarantee — the absence of this precondition constitutes an institutional constraint against direct transplantation; the transferability of U.S. Individual Retirement Accounts and employer-sponsored defined contribution (DC) plans presupposes decades of account accumulation and mature default mechanisms, whereas the account maturity and default-mechanism bearing entity for China's individual pension are not yet in place — the absence of this precondition constitutes another institutional constraint.

The second is a judgment of political-economic feasibility. Every policy recommendation must be bound to three elements: the implementing entity, an analysis of resistance, and the sequencing of implementation — who will drive this reform, who will oppose it and why, and what should be done first along with what the preconditions are. A recommendation lacking an implementing entity and resistance analysis is tantamount to evading feasibility; a recommendation that only states what should be done, without stating who will do it, where it will get stuck, or what step comes first, is tantamount to a policy slogan.

The principle of non-transferability and the principle of governance endogeneity mutually reinforce each other: governance endogeneity reveals why certain preconditions cannot be obtained out of nothing, while non-transferability requires that this unobtainability be translated into concrete institutional constraints and feasibility judgments. Together the two constitute the methodological foundation on which this study assesses the transferability of overseas experience.

Section 4 Four Types of Argumentative Traps the Framework Avoids

The tiered adaptation framework is not only a positive organizing structure but also a checklist for avoiding argumentative traps. With respect to research on China's pension asset allocation, four types of argumentative traps are most common, and this study uses the framework as a mirror to avoid each of them in turn.

The first type is generic recommendations lacking a counterfactual. Recommendations such as "steadily raise the equity allocation ratio" or "increase allocation to alternative assets" — which lack a counterfactual, lack a boundary, and apply everywhere — do not answer how much to raise it, why that particular number, or what would happen without the increase, nor do they ask whether the same logic can apply across different tiers. This study avoids this trap through a framework in which every cell must be filled in: any allocation recommendation must be bound to three elements — the account tier, the specific constraint, and the antecedent conclusion — so that the recommendation is anchored to a specific constraint at a specific tier rather than floating at a level that applies everywhere.

The second type is the direct-mapping shortcut of wholesale transplantation. Pasting the Canadian model directly onto the National Social Security Fund, or pasting the U.S. Individual Retirement Account directly onto the individual pension, while skipping the check on institutional constraints, is the typical form of this trap. This study avoids this trap through the two principles of governance endogeneity and non-transferability: any mapping must first verify whether the institutional preconditions carrying the overseas template can be independently obtained at the target tier; if the precondition is absent, direct transplantation is impermissible.

The third type is decoupling from prior research and starting anew on one's own. Setting aside the conclusions of the preceding four papers and re-summarizing overseas experience before turning to China would reduce the concluding paper to an island, severing the progressive main line running through interest-rate mechanisms, institutional differences, allocation tools, risk boundaries, and the China solution. This study avoids this trap by requiring that every conclusion drawn from overseas experience be traced back to an antecedent conclusion, so that the concluding paper becomes the landing point of the preceding four papers rather than a fresh start.

The fourth type is evading feasibility. Recommendations without an implementing entity, without resistance analysis, and without an implementation sequence — stating only what should be done without stating who will do it, where it will get stuck, or what step comes first — are the trap into which policy research most easily slides. This study avoids this trap through the three elements of political-economic feasibility: every recommendation must specify the implementing entity, the source of resistance, and the implementation sequence, so that the recommendation becomes an executable path rather than a normative slogan.

The common feature of these four traps is treating institutional preconditions as a given background to be passed over without question. The value of the tiered adaptation framework lies precisely in bringing this background to the foreground, requiring every judgment and recommendation to confront directly the question of whether institutional preconditions are obtainable.

Chapter Summary

This chapter has established the theoretical framework of this study. This chapter has clarified the six-dimensional deductive structure of the tiered adaptation framework and the incommensurability between cells, pointing out that the framework's value lies in revealing why an allocation judgment valid at one tier cannot be transposed to another. This chapter has proposed the principle of governance endogeneity, distinguishing transferable technical capability from non-transferable governance control rights, and transplantable operating technique from non-transplantable institutional bearing entities. This chapter has established the argumentative criterion of non-transferability, requiring that every conclusion of non-transferability be supported by a specific institutional constraint and a judgment of political-economic feasibility. Finally, this chapter has used the framework as a mirror to avoid four types of argumentative traps: generic recommendations lacking a counterfactual, the direct-mapping shortcut of wholesale transplantation, decoupling from prior research, and evading feasibility. The three principles together constitute the methodological foundation for the analysis in the following five chapters, in which governance endogeneity and non-transferability jointly underpin the unified criterion that transferability depends on the obtainability of institutional preconditions.

Chapter 4 The Governance Bearing Entity: Governance Capacity as a Stock Endowment Granted by Constitutional Position

This chapter corresponds to the first set of judgments in this study, situated on the axis of governance control rights between the basic old-age insurance fund and the National Social Security Fund.

Section 1 The Judgment

The core judgment proposed in this chapter is: pension governance capacity is neither a transplantable best practice nor a flow that can be gradually accumulated through construction, but a stock endowment granted once and for all by constitutional position. This judgment comprises two testable sub-propositions. First, governance quality declines in a step fashion across the three trustee tiers — proximate to the central government, provincial consolidation, and local trusteeship — this is a discrete step proposition, not a continuous function decaying with distance. Second, patient capital synthesized through administrative means is not equivalent to genuine patient capital: patient capital synthesized through means such as sale bans, mandatory long-term holding, and forced capital injection, by stripping the entity of its rebalancing rights and stop-loss rights, results in a self-destruction of property rights, producing an institutionalized reverse rebalancing whereby the lower the return, the greater the addition to positions. The step-wise comparison of governance quality across the three trustee tiers, and the change in rebalancing discipline of the same fund before and after the sale-ban breakpoint, are the identification entry points this chapter uses to test the two sub-propositions above.

It is necessary to clarify at the outset of this judgment the boundary of what this chapter identifies. What this chapter can cleanly identify is the step-wise difference in governance indicators across the three trustee tiers, and the rise in left-tail risk under sale-ban constraints. As for the causal separation between granted stock and constructible flow, because there is no counterfactual compensating design — such as a randomly imposed incentive-neutralization treatment on some province — this cannot be tested within the observation period, and this chapter therefore retains it only as a limited assertion: that no rise in governance independence with construction effort has been observed within the observation period, explicitly flagged as a pre-registration hypothesis awaiting further testing. This limited assertion must never be read as proof that the granted-stock account has been established. This self-imposed identification boundary will be maintained throughout this chapter.

This chapter's position within the tiered adaptation framework lies on the axis of governance control rights between the basic old-age insurance fund and the National Social Security Fund. The specific institutional constraint it anchors on is: the absence of statutory firewalling, the dependence of governance independence on administrative rank rather than law, and the lock-in of pooling by rents that localities obtain through the three tools of surplus fiscal rollover, credit leverage, and investment platforms. The bearing entity of governance — statutory firewalling and constitutional position — cannot be independently obtained across tiers; this is the hardest institutional anchor for judging that the relevant experience is non-transferable.

Section 2 The Mechanism

The causal chain of this chapter can be stated as a single transmission path: the absence of statutory firewalling makes governance independence dependent on administrative rank rather than law; better governance — that is, centralized pooling by the council — is rationally locked in by the rents of the three tools under local incentives; this gives rise to the pooling paradox, whereby the very localities that ought to diffuse better governance are the ones that cannot be pooled; consequently, from the top down, there is an attempt to synthesize patient capital through sale bans, mandatory long-term holding, and forced capital injection; but this synthesis is in fact a retained obligation that strips away rebalancing and stop-loss rights; ultimately producing an institutionalized reverse rebalancing whereby the lower the return, the greater the addition to positions — that is, self-destruction of property rights.

The first critical link in this causal chain is a three-tier step rather than continuous decay. The Social Security Fund Council, proximate to the central government, simultaneously holds a higher administrative rank, stronger monitoring capacity, greater political weight, and more ample fiscal support. Within China's institutional environment, distance from the center is almost perfectly collinear with the degree of incentive distortion, so distance does not constitute an independent variable that can vary orthogonally to the incentive structure. Precisely for this reason, this chapter asserts only that governance quality declines in a step fashion across the three tiers — proximate to the central government, provincial consolidation, and local trusteeship — which is the strongest assertion that three discrete tier points can support — and does not assert that governance quality decays as some continuous function of distance. Three points cannot identify any continuous functional form, let alone an exponential one.

The second critical link in this causal chain is the divide between retained obligation and patient capital. Genuine patient capital refers to an entity voluntarily holding assets at market lows while retaining the right to stop losses and rebalance; synthesized patient capital, by contrast, locks positions through administrative sale bans, strips away rebalancing rights, and raises left-tail risk under a mechanism whereby the lower the return, the greater the addition to positions. The distinction between the two can be measured on the time series of the same fund: by comparing the rebalancing discipline of the same fund before and after the institutional breakpoint, one can test whether self-destruction of property rights has occurred.

The third critical link in this causal chain is the political-economic floor of the pooling paradox. Resistance to pooling is not a matter of cognition or capability — that is, it is not that localities fail to understand that centralized management is superior — but a matter of an incentive floor: localities rationally benefit from surplus rents, and pooling means cutting off that rent. Precisely for this reason, the very localities that ought to diffuse better governance are the ones that cannot be pooled.

Section 3 Identification Strategy

The identification in this chapter consists of two primary strategies, one backup strategy, and one identification fallback design, all arranged at the design level.

The first primary strategy is cross-sectional step-difference identification. Taking the three trustee tiers — proximate to the central government, provincial consolidation, and local trusteeship — as discrete treatments, it compares the step in governance indicators across tiers. The chosen governance indicators — the number of political-procedural tiers in board appointment and removal, the statutory attribution of veto rights and project initiation rights, and the historical incidence of authorization revocation — are all predetermined relative to performance outcomes and independently codable, thereby avoiding the circularity of defining governance quality in reverse from performance outcomes. The identifying assumption is that the governance differences across the three tiers are systematic and not attributable to trustee capability, an assumption supported by using occupational annuities and enterprise annuities — which share the same environment — as controls to absorb differences in capability and asset pools. This strategy asserts only a step, without fitting a functional form.

The second primary strategy is a time-series event study on the same fund. Taking the three-year sale ban and use restrictions stipulated by the 2024 Measures for the Operation of Transferred State-Owned Capital as the institutional breakpoint, it examines the change in rebalancing discipline and left-tail risk of the same recipient entity before and after the sale-ban constraint was imposed. The identifying assumption is that the timing of the breakpoint is exogenous to the contemporaneous allocation pressure of any single fund — that is, the formation of the policy agenda does not depend on the contemporaneous condition of a single fund — an assumption supported by a pre-breakpoint parallel-trends test. The advantage of this strategy is that comparing the same fund before and after avoids the incomparability across funds in trustee capability and asset pools.

The backup strategy is to seek exogenous variation in incentive intensity within the same administrative tier, using differences in the policy accessibility of a given rent-generating tool across same-tier provinces to separate the incentive account from the governance-endowment account. This strategy deliberately avoids using cross-tier distance, because distance is collinear with incentives, and using distance to falsify the opposing hypothesis would be a false dichotomy.

The identification fallback design was fixed at the design stage. If the governance indicators across the three trustee tiers cannot be independently coded prior to performance — that is, if the coding is contaminated by performance outcomes — then the step proposition is downgraded to a taxonomic framework, and the main text will be presented on the basis of that taxonomic framework, with the identifiable core identification link falling back to the single-fund event study around the 2024 breakpoint alone, which does not depend on three-tier coding and stands independently. If the compensating-design counterfactual remains unobtainable, the causal claim of granted stock is permanently flagged as pending pre-registration, and this does not falsify the step proposition.

Section 4 Honest Labeling of Evidence

At the level of verifiable aggregate facts, there are several institutional facts that can provide directional corroboration for this chapter's judgment, but their nature as an aggregate two-point difference, rather than a micro cross-sectional regression coefficient, must be made explicit.

The National Social Security Fund — directly managed and entrusted for management by the Social Security Fund Council, proximate to the central government — has achieved an average annual investment return of 7.36 percent since its establishment, with cumulative investment gains of 1,682.576 billion yuan, and total fund assets of 3,014.561 billion yuan at the end of 2023; whereas the local basic old-age insurance funds entrusted to the Council for operation — distant from the center, with funds belonging to localities — have achieved an average annual investment return of 5.06 percent since entrustment in December 2016, 5.52 percent in 2024 alone, with total assets of 2,839.652 billion yuan at the end of 2024 (National Council for Social Security Fund 2024). Both share the same trustee — the Social Security Fund Council — and the same regulator, and their approximately 2.3 percentage-point return gap co-varies with the difference in the principal's constitutional position, consistent in direction with governance quality declining in a step from proximate to the center to distant from the center. It must be stressed, however, that this is an aggregate two-point difference rather than a three-tier cross-sectional regression, one that conflates multiple factors such as years since establishment and asset duration; it therefore serves only as directional corroboration, and the step coefficient across tiers still awaits micro data.

The institutional anchor of the pooling paradox also has verifiable policy evidence. The aforementioned 2024 Measures mandate pooling of only no less than 50 percent of cash returns, leaving the remainder to localities to operate independently within a defined scope (Ministry of Finance 2024) — an institutional arrangement consistent in direction with the pooling paradox, whereby the localities that most ought to diffuse are the ones that cannot be pooled, and with the political-economic floor of localities rationally benefiting from surplus rents. The institutional breakpoint of property-rights self-destruction is likewise real and verifiable: the same Measures stipulate a three-year sale ban on transferred state-owned capital and restrict its use to deposits, government bonds, and capital injections into the transferring enterprise, disallowing free rebalancing or stop-loss switching — this is precisely the confirmation of the existence of the institutional fact underlying the proposition that patient capital synthesized through sale bans, mandatory long-term holding, and forced capital injection strips away rebalancing rights and thereby leads to self-destruction of property rights.

At the micro-identification level, owing to data availability constraints, the core identification links of this chapter can only present the identification design, not fill in estimation results. The cross-sectional step regression on governance indicators across the three trustee tiers cannot be executed, because a tiered coding panel of the political-procedural chain of board appointment and removal, the statutory attribution of veto and initiation rights, and the historical incidence of authorization revocation is unavailable — some attribution clauses are confidential and require clause-by-clause coding of legislative texts; the identification design is in place, and the directional expectation is that the governance indicators of the tier proximate to the center are systematically higher than those of the tier distant from the center, with the falsification condition being no systematic difference across the three tiers. The same-fund event study on self-destruction of property rights cannot be executed, because the holdings, rebalancing, and left-tail sequences of the same recipient entity before and after the 2024 breakpoint are unavailable — the holdings detail of the recipient entity is not public; its sign is falsifiable while its absolute magnitude awaits data, with the falsification condition being that the risk-adjusted return of the sale-ban group is not lower than that of the free-rebalancing group. The difference-in-differences identification of pooling resistance — that is, comparing the collection elasticity of shortfall provinces versus surplus provinces — cannot be executed because the provincial collection-elasticity panel is internal collection-administration data that is unavailable; the direction can be established while the magnitude awaits data.

An honest boundary that must especially be observed is: the rents from the three tools — surplus fiscal rollover, credit leverage, and investment platforms — are unobtainable in explicit ledgers owing to their off-balance-sheet nature, but the absence of an explicit rent ledger must never be read as evidence of no rent; the absence merely lowers the strength of the claim about the scale of rent, and must never be reversed into positive evidence of no rent. Likewise, the absence of observed increases in governance independence with construction effort within the observation period must never be read as proof that the granted-stock account has been established.

Section 5 Counterarguments and Their Treatment

Two of the strongest possible objections to this chapter's judgment must be addressed directly, because the manner of response itself defines the boundary of this chapter's identification.

The first objection targets the form of identification. If this chapter's core identification claimed that independence decays exponentially with distance from the center, then the cross-section of the three trustee tiers has only three discrete tier points, which cannot identify any continuous functional form, let alone an exponential one; moreover, distance from the center and governance tier are almost perfectly collinear in China — the closer to the center, the higher the administrative rank, the stronger the monitoring, the greater the political weight, and the more ample the fiscal support — so distance is not an independent variable that can vary orthogonally to the incentive structure. Accordingly, the self-declared falsification test of "decay by distance rather than by incentive" is a false dichotomy: observing that pooling becomes harder the farther away supports both distance decay and incentive distortion equally — the two are observationally equivalent. In response to this objection, this chapter's treatment is an honest downgrade: abandoning the claim of a continuous functional form and downgrading the core identification claim to the discrete step proposition — that governance quality declines in a step across the three tiers of proximate to the center, provincial consolidation, and local trusteeship — the strongest assertion three points can support; while acknowledging that distance and incentive are collinear, no longer using distance as the falsification test, and instead using exogenous variation in incentive intensity within the same administrative tier to separate the incentive account from the governance-endowment account. This chapter does not adopt any continuous distance-law formulation.

The second objection targets falsifiability. The structure whereby "template success proves precisely that it cannot be replicated" is a heads-I-win, tails-you-lose structure: template failure proves governance does not work, template success proves that success cannot be generalized — no observed outcome can ever refute the proposition, and any favorable evidence is absorbed as evidence of non-generalizability, which is precisely the pathology of non-transferability being used as a shield. In response to this objection, this chapter's treatment is to adopt no absorptive rhetoric of this kind, retaining only testable sub-propositions bound to ex-ante falsification criteria: first, the step difference in governance indicators across the three tiers, with the falsification condition being no systematic difference across the three tiers; second, that the risk-adjusted return of the sale-ban portfolio is lower than that of the free-rebalancing portfolio, with the falsification condition being that the risk-adjusted return of the sale-ban portfolio is not lower than that of the free portfolio. Any unfalsifiable, synthesizable-away formulation that cannot be bound to an ex-ante success criterion is not retained in this chapter, which binds itself only to the falsifiable proposition of rising left-tail risk in the same-fund event study.These two treatments together ensure that this chapter's judgment is directionally falsifiable, rather than a shield that can deflect any evidence.

This chapter's two main identification designs and their falsification conditions can be presented together in the table below, set against the treatment of the two objections.

Table 2 Competing Hypotheses and Identification Strategies for the First Set of Judgments

Chapter Summary

This chapter has argued for the judgment that pension governance capacity is a stock endowment granted once and for all by constitutional position. This chapter has shown that governance quality declines in a step across the three trustee tiers rather than decaying continuously, because distance from the center and incentive distortion are almost perfectly collinear in China, and three discrete points can support only a step proposition; this chapter has shown that patient capital synthesized through administrative sale bans results in self-destruction of property rights by stripping away rebalancing rights. At the evidentiary level, the approximately 2.3-percentage-point return gap between the National Social Security Fund and locally entrusted funds, the 2024 Measures' mandatory pooling of 50 percent, and the institutional breakpoint of the three-year sale ban and use restrictions all provide directional corroboration for the judgment; while the micro-identification links — the three-tier step regression, the property-rights self-destruction event study, and the difference-in-differences on pooling resistance — are, owing to data availability constraints, honestly flagged as having the identification design in place while the absolute magnitudes await data. The causal separation between granted stock and constructible flow is permanently flagged as a pre-registration hypothesis, owing to the absence of a compensating-design counterfactual.

Chapter 5. The Risk-Free Anchor: Endogenous Benchmarks on the Asset Side and the Absorption of Debt Resolution

This chapter corresponds to the second group of judgments in this study, situated on the fiscal axis of the asset side of the Basic Pension Insurance Fund and the National Social Security Fund.

Section 1. The Judgment

The core judgment advanced in this chapter is: China's pension risk-free anchor is not an exogenously given constant, but a target endogenously moved by fiscal authority. First, the risk-free anchor shifts as rigid payout guarantees break down, pinning category-locked assets to the wrong side of a depreciating basket. Second, the first and second pillars have been institutionalized as absorption pools for fiscal contingent liabilities — the notional accounts of occupational annuities are contingent pension liabilities recorded on the fiscal balance sheet, not personal assets, and the positive wedge in the crediting rate is an unrecorded implicit intergenerational tax. Separating an absorption mandate from prudent asset-liability management requires evidence of duration overweighting plus an unnamed peer control group, rather than merely observing whether increased holdings coincide with issuance windows. The identification in this chapter therefore revolves around these two links: duration overweighting and the unnamed control comparison.

The deeper significance of this judgment is that it elevates the allocation problem from how much duration to hold to whose definition of the risk-free anchor serves as the benchmark. The traditional asset-liability management (ALM) framework assumes an exogenously stable risk-free anchor — the government bond yield curve — as the benchmark for duration matching. But when fiscal authority moves the definition and pricing of safe assets through debt resolution, implicit guarantees, and countercyclical crediting-rate commitments, the risk-free anchor itself becomes an endogenous variable subject to fiscal manipulation. The meta-level question for the allocation problem is thus no longer how much duration, but who holds the right to define the risk-free anchor.

What this chapter can cleanly identify comprises three threads: the shift in the risk-free anchor, the duration-overweighting fingerprint of debt-resolution absorption, and the upper-bound argument that negates the 7 percent return and actuarially neutral accounting convention. As for the absolute magnitude of the intergenerational tax wedge, this is unattainable because it involves the net present value of intergenerational transfers, and is offered only as a directional (sign) judgment — the direction can be established, but the absolute magnitude awaits data. This chapter's position within the tiered-adaptation framework falls on the fiscal axis of the asset side of the Basic Pension Insurance Fund and the National Social Security Fund, and the specific institutional constraint it anchors on is the absence of an exogenous risk-free anchor — the risk-free anchor is defined by fiscal authority, endogenously moved by fiscal authority, with the first and second pillars institutionalized as the ultimate absorbers of debt resolution, and the crediting rate is a fiscal redistribution price rather than an investment return. The fact that an exogenous risk-free anchor is unattainable independently in China is the hardest institutional anchor establishing that overseas ALM templates premised on a stable risk-free anchor cannot be directly transplanted.

Section 2. Mechanism

The causal chain of this chapter can be stated as: low interest rates and the asset shortage turn the floor of safe assets from a protection into a risk exposure; pension funds are designated to absorb debt-resolution duration; the 7 percent return and actuarially neutral accounting convention systematically overstate real returns; as a result, the fates of fiscal authority and pension funds become deeply entangled.

The first key link in this causal chain is the meta-level problem of the endogenous benchmark. Traditional ALM uses the government bond curve as an exogenously stable duration-matching benchmark, but China's risk-free anchor shifts as rigid payout guarantees break down. When fiscal authority moves the definition and pricing of safe assets, "risk-free" itself becomes an endogenous variable, and the meta-level question of the allocation problem thereby shifts from how much duration to who holds the right to define the benchmark.

The second key link in this causal chain is the difficulty of identifying the ultimate absorber of debt resolution. In a low-rate environment, a duration gap already calls for allocating to long bonds; when supply is expanded, long bonds have better liquidity and lower impact costs, so following issuance windows to increase holdings of long bonds is normal behavior under rational ALM, not the fingerprint of an absorption mandate. The true fingerprint of an absorption mandate is overweighting that ALM would not buy but a mandate would: first, continuing to add holdings even after liability duration is already fully matched; second, buying duration or credit tranches that significantly deviate from what one's own liability structure requires. Only this kind of overweighting can separate an absorption mandate from ordinary duration management.

The third key link in this causal chain is notional accounts as contingent liabilities. The return locked in by the occupational annuity crediting rate is not an investment return but a fiscal redistribution price promised by the state — it is recorded on the fiscal balance sheet as a contingent pension liability, and the positive wedge of the crediting rate — the crediting rate minus the true asset-side return — is an unrecorded implicit intergenerational tax. The 7 percent return and actuarially neutral accounting convention systematically overstate real returns; even estimated under the most favorable accounting convention, their upper bound remains below fiscal authority's cost of financing.

Section 3. Identification Strategy

Identification in this chapter consists of two primary strategies, one backup strategy, and one identification fallback design.

The first primary strategy is a difference-in-differences identification of debt-resolution absorption. Institutions named as absorbers of debt resolution serve as the treatment group, and unnamed peer institutions of the same type serve as the control group, comparing the difference in the slope of holdings increases between the two during the same issuance window. The identifying assumption is that being named is exogenous to an institution's own contemporaneous ALM needs — that is, the debt-resolution list is determined by policy, not by an institution's duration gap. The key to this strategy lies in the overidentification restriction: only overweighting that continues after liability duration is already matched, or buying duration significantly exceeding one's own liability duration, counts as the fingerprint of a mandate; merely aligning with issuance windows is insufficient to carry the identification, because ALM and mandate predictions coincide here.

The second primary strategy is a quasi-natural experiment based on the upper-bound argument. Using crediting-rate adjustment events as a discontinuity, this strategy estimates the reassessment of the spread between the crediting rate and the cost of financing. It asserts that, even under the most favorable accounting convention, the upper bound of the real-return component of the crediting rate remains below fiscal authority's cost of financing, and reports confidence intervals and test power. This strategy explicitly does not use an insignificant result as evidence — an insignificant coefficient cannot be read as support for the opposing hypothesis, because failure to reject the null does not mean the null is true; in a context where data are unattainable, low test power is precisely the most likely reason for an insignificant coefficient.

The backup strategy uses demographic cohort shocks as an instrumental variable to estimate the sign of the crediting-rate wedge — the direction can be established, but the magnitude awaits data.

The identification fallback design is fixed at the design stage. If institutions' liability duration and the duration tranche of their purchases cannot be constructed, debt-resolution absorption downgrades to a sign-only judgment, and the identifiable core identification retreats to two threads — the event study of the risk-free-anchor shift and the upper-bound argument — neither of which depends on precisely constructing duration overweighting. If the testable implication of a sign reversal — namely, a negative cross-sectional correlation between compliance ratings and true exposure — is not found, it downgrades to the conventional interest-rate-risk description of nominal safe assets' real returns turning negative, and is not overstated as a meta-level sign reversal.

Section 4. Honest Labeling of the Evidence

At the level of verifiable aggregate facts, several institutional facts lend directional support to this chapter's judgment.

The systematic decline of the risk-free rate has already been reflected in the successive reductions of the assumed interest rate for life insurance discussed earlier, falling from 3.5 percent through 3.0 percent and 2.5 percent to 2.0 percent, with the research value declining to 1.99 percent in July 2025 (金融监管总局 2024); the return on safe assets has correspondingly broken downward, consistent in direction with the floor of safe assets turning from a protection into a risk exposure and the risk-free anchor shifting as rigid payout guarantees break down. The fiscal-commitment nature of the occupational annuity's notional-account system is also institutionally confirmed: for units with full fiscal funding, the unit's contribution portion is handled on a notional basis and accrues interest at the crediting rate, with only the remaining portion operated on a market basis (人力资源和社会保障部 2015) — this institutionally confirms the judgment that the crediting rate is a shadow rate set by fiscal authority rather than a market return, and that notional accounts are contingent pension liabilities recorded on the fiscal balance sheet. This judgment's direction is also consistent with the international literature: if public pension liabilities are measured using a discount rate commensurate with their risk properties, their scale will be systematically larger than what official actuarial conventions show (Novy-Marx and Rauh 2011).

At the level of micro-identification, owing to data limitations, this chapter's core identification links can only be presented as identification designs. The difference-in-differences for debt-resolution absorption cannot be executed, because grouping into named and unnamed institutions depends on disclosure of the debt-resolution list, which is incomplete, and because institutions' liability duration and the duration tranche of purchases would need to be constructed from annual reports and holdings, which are partly unattainable; the identification design is in place, identifying only the sign of the slope difference, not point-estimating the total scale of absorption — the sign is falsifiable, the absolute magnitude awaits data, and the falsification condition is that no overweighting exists, or that increases occur only when duration is unmatched. The upper-bound argument cannot be executed, because the reassessment of the spread in the real-return component of the crediting rate and data on crediting-rate adjustment events are unattainable — this involves pairing fiscal financing cost conventions with realized annuity asset-side returns; the identification design is in place, the sign of the upper bound is falsifiable, the absolute magnitude awaits data, and the falsification condition is that the upper bound exceeds the cost of financing. The testable implication of a sign reversal cannot be executed because the cross-sectional pairing of compliance ratings with exposure on the wrong side of the depreciating basket is unattainable; if no negative correlation is found, it downgrades to the conventional interest-rate-risk description. The intergenerational tax wedge, because its absolute magnitude involves the net present value of intergenerational transfers, is unattainable and is offered only as a sign-type judgment — the direction can be established, the absolute magnitude awaits data.

The honest boundary that must be especially observed is: this chapter never reads any insignificant coefficient as positive evidence for the opposing hypothesis. The claim that notional accounts are contingent liabilities is carried by positive evidence such as point estimates of the spread reassessment and event studies of crediting-rate adjustments, and never by the inferential fallacy of insignificance supposedly supporting such a claim. The absence of data only lowers the strength of the claim; it never reverses into support for the opposing hypothesis.

Section 5. Counterarguments and Responses

Two of the strongest possible objections to this chapter's judgment must be addressed directly.

The first objection concerns the separability of identification. The opposition between issuance windows and liability duration collapses precisely during a low-rate debt-resolution period: a pure ALM manager, subject to no absorption mandate whatsoever, facing falling rates and a safe-asset shortage, would in any case be expected to buy long bonds during windows of expanded long-bond supply — because the duration gap calls for allocating to long bonds, and because expanded supply means better liquidity, lower impact costs, and higher availability. Accordingly, following issuance windows to increase holdings of long bonds is normal behavior under rational ALM, not the fingerprint of an absorption mandate; alignment of increased-holdings timing with issuance windows cannot falsify the ALM hypothesis, because ALM and an absorption mandate make coincident predictions of buying long bonds in a low-rate environment and buying during supply windows, constituting a false opposition. This chapter's response to this objection is to supplement with an overidentification restriction: redefine the mandate fingerprint as overweighting that ALM would not buy but a mandate would — continuing to add holdings even after liability duration is already fully matched, or buying duration and credit tranches that significantly deviate from what one's own liability structure requires — layered with an unmandated peer control group, so that the difference in the slope of holdings increases between named and unnamed institutions in the same issuance window separates out the mandate effect. Alignment of timing alone is insufficient to carry the identification.

The second objection concerns the legitimacy of the inference. If the original falsifiable criterion is stated as: if actuarial neutrality holds, the real-return component of the crediting rate should be significantly positive, and the data show it is insignificant, and this insignificance is instead taken to support the thesis, this constitutes a statistical inference fallacy — reading an insignificant coefficient as support for the opposing hypothesis. Failure to reject the null does not mean the null is true, still less that the opposing hypothesis is true; an insignificant coefficient may result from a small sample, high noise, conflated conventions, or low test power, and in a context where Chinese data are explicitly noted as unattainable, low power is precisely the most likely reason. This chapter's response to this objection is to entirely abandon any formulation that confirms the thesis via insignificance, and instead adopt the upper-bound argument — replacing an absence of information with the presence of information: asserting that, even under the most favorable accounting convention, the upper bound of the real-return component of the crediting rate remains below fiscal authority's cost of financing, and reporting confidence intervals and test power; the claim that notional accounts are contingent liabilities is instead carried by positive evidence such as point estimates of the spread reassessment and event studies of crediting-rate adjustments, never relying on an insignificant result. This response ensures that this chapter's core claims rest on positive evidence, rather than treating a lack of refutation as evidence of support.

Chapter Summary

This chapter has argued for the judgment that there is no exogenous risk-free anchor on the asset side of China's pension system, and that the first and second pillars have been institutionalized as absorption pools for fiscal contingent liabilities. This chapter shows that the risk-free anchor shifts as rigid payout guarantees break down, elevating the meta-level allocation question from how much duration to who holds the right to define the benchmark; this chapter shows that separating an absorption mandate from prudent ALM requires duration overweighting and an unnamed control comparison, rather than the alignment of increased-holdings timing with issuance windows, because the latter is a false opposition in which ALM and mandate predictions coincide; this chapter shows that notional accounts are contingent liabilities recorded on the fiscal balance sheet, and that its upper-bound argument replaces an absence of information with the presence of information, avoiding the inferential fallacy of confirming a claim via an insignificant coefficient. At the evidentiary level, the successive reductions in the assumed interest rate and the fiscal-commitment nature of the occupational annuity's notional-account system provide directional support for the judgment, while the micro-identification links — the difference-in-differences for debt-resolution absorption, the upper-bound argument, and the sign-reversal cross-section — are, owing to data limitations, honestly labeled as having identification designs in place with absolute magnitudes awaiting data; the intergenerational tax wedge is offered as a sign-type judgment.

Chapter 6. Accounting Emergence: The Emergence-Driven Divergence in Second-Pillar Allocation

This chapter corresponds to the third group of judgments in this study, situated on the axis of the crediting mechanism for liability pricing in enterprise annuities and occupational annuities.

Section 1. The Judgment

The core judgment advanced in this chapter is: low interest rates per se do not change the economic substance of second-pillar assets; what changes is when, and at what book-value speed, these assets are seen. The dual-track structure of fiscal contributions and notional accounting in occupational annuities causes the same contribution to see two risk-free rates — the fiscally funded portion is locked at a higher level by the crediting rate, while the market-based portion faces a real curve that has already fallen below 2 percent; trustees use the higher crediting rate as an implicit opportunity-cost anchor, causing the market-based portion to undergo defensive contraction. At the same time, differences in discount-rate smoothing rules — life insurance uses a 750-day moving average, while annuities are revalued daily — create an emergence time lag of roughly 1.5 to 2 years and a misalignment in allocation timing. Identification in this chapter is carried by two threads: a cross-section with identical cash flows but dual crediting rates, and a lead-lag test during the period of parallel accounting standards.

The core of what this chapter can cleanly identify is the cross-section of fiscal contributions versus notional-accounting dual crediting rates. The same contribution, with completely identical cash flows, differing only in the book risk-free rate, gives rise to differing allocations that are purely an emergence effect — this is the most irrefutable identification of accounting-emergence-driven allocation, because the identity of cash flows rules out the competing explanation of true cash-flow change as the driver. A secondary-level identification is the lead-lag relationship during the period in which old and new standards run in parallel: a standard switch changes no cash flow whatsoever, only the point of emergence; the appearance of an arbitrage gap is itself proof of emergence-driven behavior.

A boundary must be drawn at the very outset of the judgment. Differences in duration and left-tail behavior for arrangements with guaranteed floors must be attributed to a genuine rise in option-liability exposure, not merely to accounting seeing things faster. A guarantee itself constitutes a genuine economic liability — the embedded short put position's in-the-money probability rises under low rates, which is a genuine increase in cash-flow obligation, not an acceleration of accounting emergence. Accordingly, the behavioral divergence of the guaranteed-floor group belongs to the proposition of economic substance, and is relocated to Section 5 of Chapter 9, in the risk-governance discussion of the decumulation (payout) phase, as closing evidence on the short guarantee-option position — it must never be recycled back as a falsification point for emergence-driven behavior. If the behavioral divergence of the guaranteed-floor group were used to argue for accounting-emergence-driven behavior, this would precisely prove the very thing this chapter seeks to negate — genuine cash-flow-change-driven behavior — constituting self-undermining. The emergence-driven evidence in this chapter is carried only by two positive findings: the cash-flow-identical cross-section and the arbitrage gap during the parallel-standards period.

This chapter's position within the tiered-adaptation framework falls on the axis of the crediting mechanism for liability pricing in enterprise annuities and occupational annuities, and the specific institutional constraint it anchors on is China's dual-track notional accounting, the difference in discount-rate smoothing rules, and the crediting rate serving as an implicit opportunity-cost anchor. The accounting-emergence rule is unique to China's dual-track notional accounting — this is the hardest institutional anchor establishing that overseas second-pillar experience, premised on a single accounting convention, must first be checked against the accounting carrier.

Section 2. Mechanism

The causal chain of this chapter can be stated as: differences in accounting recognition rules cause the same asset to have different emergence speeds and illusory rates across different tiers and different standards; trustees allocate according to book value rather than economic substance; the result is conservative contraction, arbitrage gaps, and a thickened left tail.

The first key link in this causal chain is the dual-rate illusion, in the spatial dimension. The fiscally funded portion of occupational annuities is locked at a higher crediting rate, while the market-based portion faces a genuinely declining curve. The same contribution is split by accounting rules into two risk-free rates, and trustees use the higher crediting rate as an implicit opportunity-cost anchor, making the market-based portion appear to underperform relatively, thereby triggering defensive contraction. The key here is that cash flows are entirely identical, and the allocation difference is driven purely by the difference in book rates, constituting a pure emergence effect.

The second key link in this causal chain is the emergence time lag, in the temporal dimension. Life insurance discount rates are smoothed using a 750-day moving average, while annuities are revalued daily; the same interest-rate-decline shock emerges on the books of the two at speeds differing by roughly 1.5 to 2 years. During a standard switch or parallel period, a lead-lag arbitrage gap appears: arbitrageurs exploit the difference in emergence timing for the same cash flow under the two sets of standards, thereby confirming emergence-driven behavior.

The third key link in this causal chain is smoothing as a buried landmine rather than a mitigant. Discount-rate smoothing does not eliminate the interest-spread loss; it merely delays its book emergence. Under low rates, a higher nominal crediting rate produces worse actual allocation — the higher the crediting rate, the more severe the contraction of the market-based portion, and the thicker the left tail.

Section 3. Identification Strategy

Identification in this chapter consists of one primary strategy, one secondary strategy, and one identification fallback design.

The primary strategy is the cash-flow-identical cross-section of the dual-rate illusion. This regresses the contraction in market-based-portion allocation on the provincial cross-section of the ratio of fiscally funded to notional-accounted contributions. Its core identification logic is: the same contribution, with completely identical cash flows, differing only in the book risk-free rate — the allocation difference is driven purely by accounting rules, constituting a pure emergence effect. The identifying assumption is that the ratio of fiscal funding to notional accounting is exogenous to the allocation preference for the market-based portion — that is, this ratio is determined by fiscal funding rules and historical accumulation, not by contemporaneous allocation choices. This strategy is the most irrefutable identification of accounting-emergence-driven behavior, because the identity of cash flows directly rules out the competing explanation of true cash-flow change. This strategy also resolves the concern of circular reasoning: the measured variable — the book rate — and the explained variable — allocation — measure different things, and the ratio precedes the allocation outcome and can be coded ex ante.

The secondary strategy is a lead-lag identification of the emergence time lag during the parallel-standards period. During the period in which old and new standards run in parallel, this tests whether the difference in emergence timing between life insurance and annuity books, in response to the same interest-rate shock, produces an arbitrage gap. The identifying assumption is that a standard switch changes no cash flow, only the point of emergence; the appearance of a lead-lag arbitrage confirms emergence-driven behavior. This strategy uses the exogenous timing of the standard switch to pin the arrow of causation from the emergence-timing difference to the arbitrage gap, without relying on any ex post judgment of emergence speed.

The identification fallback design is fixed at the design stage. If the ratio of fiscal funding to notional accounting is shown to be endogenous to allocation preference — that is, the ratio co-varies with contemporaneous allocation choices — then the primary identification retreats to the single thread of lead-lag during the parallel-standards period, which does not depend on the exogeneity of the ratio, since the exogeneity of the standard switch is stronger. The difference-in-differences on guaranteed-floor arrangements is permanently labeled as a proposition of economic substance and consigned to the closing discussion in Section 5 of Chapter 9 on risk governance in the decumulation phase; it is not recycled back as a falsification point for emergence-driven behavior.

Section 4. Honest Labeling of the Evidence

At the level of verifiable aggregate facts, the second pillar — owing to the relative availability of book data — is the domain among this study's five problem areas with relatively the most solid evidence, but its hard core still lies in the cash-flow-identical cross-section, not in the economic substance of guaranteed-floor arrangements.

The dual-track structure of fiscal contributions and notional accounting, constituting two risk-free rates for the same contribution — this core identification link has clear institutional confirmation of its institutional foundation: for units with full fiscal funding of occupational annuities, the unit's contribution portion is handled on a notional basis and accrues interest at the crediting rate, with only the remaining portion operated on a market basis, which institutionally causes the same occupational annuity plan to internally span a notionally locked defined benefit (DB) segment and a market-based defined contribution (DC) segment (人力资源和社会保障部 2015). This confirms the institutional existence of the same contribution seeing two risk-free rates, with cash flows identical and only book rates differing. There is also an aggregate shadow of divergence between notional-account and market-based returns: for occupational annuities dominated by the notional-account mechanism, the average annual investment return since 2019 has been 4.42 percent, with an investment operation scale of RMB 3.11 trillion at the end of 2024; while for enterprise annuities, which are fully market-based with no notional lock, the weighted average return in 2024 was 4.77 percent, with a scale of RMB 3.64 trillion, and an average annual return of 6.17 percent since 2007 (人力资源和社会保障部 2025). Both are second-pillar occupational pension schemes with highly overlapping asset classes, and the difference in their return structures co-varies with the differing notional versus market-based emergence mechanisms, directionally consistent with the crediting rate serving as an implicit opportunity-cost anchor that causes defensive contraction of the market-based portion. It must be emphasized, however, that the return gap between enterprise annuities and occupational annuities is confounded by multiple factors including years since establishment, trustee structure, and member profile; this serves here only as directional corroboration, not as a point estimate of a pure emergence effect. The difference in discount-rate smoothing rules also genuinely exists: life insurance reserve discount rates use a 750-day moving average smoothing, while annuities are revalued daily or based on actual returns — this rule difference is the institutional precondition for the emergence time lag. That changes in accounting recognition rules can genuinely alter the risk-taking of pension asset allocation has already been given causal evidence by empirical research under international accounting-standard transitions, which supports this emergence mechanism (Anantharaman and Chuk 2018).

At the level of micro-identification, owing to data limitations, this chapter's core identification links can only be presented as identification designs. The dual-rate-illusion cross-section cannot be executed, because the provincial panel of fiscal-funding-to-notional-accounting ratios and market-based-portion allocation structure is unattainable — provincial occupational-annuity book structures are not publicly disclosed; the identification design is in place, the sign is falsifiable, the absolute magnitude awaits data, and the falsification condition is that the ratio is unrelated to contraction. The lead-lag identification of the emergence time lag cannot be executed, because the series of emergence-timing differences between life insurance and annuity books and arbitrage-gap price spreads during the old-new standard parallel period are unattainable; the identification design is in place, awaiting data, and the falsification condition is that no gap appears during the parallel period.

The honest boundary that must be especially observed is: this chapter never reads the behavioral divergence of any guaranteed-floor group as evidence of emergence-driven behavior, because that divergence is also compatible with the economic-substance explanation of a rational response to a genuinely enlarged option liability — precisely the opposite of this chapter's claim. Emergence-driven evidence is carried only by the positive findings of the cash-flow-identical cross-section and the parallel-period arbitrage gap; the absence of divergence or of a gap constitutes falsification, not self-confirmation. Likewise, the absence of observed allocation divergence must never be read as support for the proposition that the emergence time lag has simply not yet arrived while the thesis still holds.

Section 5. Counterarguments and Responses

Two of the strongest possible objections to this chapter's judgment must be addressed directly, the first of which bears directly on the choice of this chapter's core identification link.

The first objection concerns the self-undermining risk of using guaranteed-floor arrangements to argue for emergence-driven behavior. If the behavioral divergence between guaranteed-floor and non-guaranteed-floor groups were used as the flagship evidence for emergence-driven behavior, two flaws would arise. First, the treatment variable — whether a floor guarantee is included — is non-random and strongly endogenous: choosing to include a guarantee is, under legal constraints, a practical choice, and plans that choose to include a guarantee differ systematically in sponsor type, member age structure, risk preference, and bargaining power; observing different duration and left-tail behavior in the guaranteed-floor group may stem from these plans being inherently different, rather than from the convexity of the guarantee option. Second, and more fundamentally, this test cannot separate accounting emergence from genuine cash-flow change precisely where a guaranteed floor is present — a guaranteed floor itself constitutes a genuine economic liability; the short put position's in-the-money probability rises under low rates, which is a genuine increase in cash-flow obligation, not merely something accounting sees faster. Using a guaranteed floor to argue for accounting-emergence-driven behavior would precisely prove the very thing this chapter seeks to negate — genuine cash-flow-change-driven behavior. This chapter's response to this objection is to replace the core identification link: using the cross-section of fiscal-contribution versus notional-accounting dual crediting rates as the core identification — the same contribution, with completely identical cash flows, differing only in the book rate, so that the allocation difference is a pure emergence effect — this is the most irrefutable identification of emergence-driven behavior; guaranteed-floor arrangements are instead reassigned to the proposition of economic substance and relocated to Section 5 of Chapter 9, in the risk-governance discussion of the decumulation phase, as closing evidence on the short guarantee-option position, permanently not to be recycled back as a falsification point for emergence-driven behavior.

The second objection concerns falsifiability. If the proposition that economic substance is unchanged and only the emergence speed changes were treated as a bidirectional claim that can both explain allocation divergence and provide an excuse when data are insufficient, then the proposition survives both when emergence is fast and divergence is observed and when emergence is slow or the lag has not yet arrived and divergence is not observed — a bidirectional self-confirmation in which falsifiability approaches zero; and if the claim that the crediting rate produces poor allocation lacks a measure of the crediting rate independent of the allocation outcome, then the illusory rate and the explained allocation may be measuring the same thing, constituting circularity. This chapter's response to this objection is to resolve the circularity via the cash-flow-identical cross-section: the ratio of fiscal contributions to notional accounting precedes the allocation outcome and can be coded ex ante, being determined by fiscal funding rules and historical accumulation, so that the measured variable — the book rate — and the explained variable — allocation — measure different things; and to use the exogenous timing of the parallel-standards period to pin the arrow of causation from the emergence-timing difference to the arbitrage gap, without relying on any ex post judgment of emergence speed. The emergence-driven evidence is carried only by the divergence in the cash-flow-identical cross-section and the positive finding of the parallel-period arbitrage gap; their absence constitutes falsification.

Chapter Summary

This chapter has argued for the judgment that the divergence in second-pillar allocation is driven mainly by accounting-emergence rules rather than by genuine cash-flow change. This chapter shows that the dual-track structure of fiscal contributions and notional accounting causes the same contribution to see two risk-free rates, with cash flows identical while allocation diverges, constituting a pure emergence effect — the most irrefutable identification of emergence-driven behavior; this chapter shows that the difference in discount-rate smoothing rules produces an emergence time lag of roughly 1.5 to 2 years and an arbitrage gap. At the evidentiary level, the occupational annuity dual-account system, the return gap — positive in direction and consistent with the direction implied by the crediting-rate-anchor hypothesis — between notional-account-dominated occupational annuities and fully market-based enterprise annuities (the precise point spread is not asserted numerically because disclosure periods for the two types of plans are not comparable), and the 750-day discount-rate smoothing rule provide directional support for the judgment, while the micro-identification links — the dual-rate-illusion cross-section and the lead-lag emergence time-lag test — are, owing to data limitations, honestly labeled as having identification designs in place with absolute magnitudes awaiting data. Guaranteed-floor arrangements, being a genuine economic liability, are relocated to the closing discussion in Section 5 of Chapter 9 on risk governance in the decumulation phase, and are never recycled back as a falsification point for emergence-driven behavior.

Chapter 7 Behavioral Architecture: The Institutional Regressivity Driving Cold Contributions in the Three-Pillar System

This chapter corresponds to the fourth group of judgments in this study, situated on the behavioral-architecture axis of individual pensions, and crosses the two layers of the basic pension insurance fund and individual pensions at the flexible-employment fault line.

Section 1 Judgment

The core judgment advanced in this chapter is: the cold-contribution phenomenon in the three-pillar system is driven mainly by institutional features, not mainly by individual literacy. The unified tax-preference nudge constitutes a regressive nudge for low- and middle-income earners — the arrangement of deferring taxation on contributions amplifies this regressivity; the closed account is an unpriced prepayment of a liquidity option under low interest rates; the dominant form of the coverage gap is not non-enrollment but enrollment in the worst tier — low interest rates have reversed the relative attractiveness of the Employee Pension Insurance and the Urban-Rural Resident Pension Insurance, triggering a cascade of downward migration; and the auto-enrollment mechanism is structurally absent for the flexibly employed population, since this population has no employer, no payroll, and no default pathway — the coverage gap arises at the site of decision costs and institutional carriers, not at the site of contribution capacity. The identification in this chapter accordingly rests on three components: a comparison of effect sizes between mechanism and literacy, quantile-controlled price sensitivity, and a quasi-natural experiment on default carriers.

At the outset of this judgment, a methodological self-constraint must be made explicit: this chapter makes no identifiable claim about the rationality attribute of resident behavior. The phenomenon whereby contribution downgrading varies with the crediting rate identifies sensitivity to price, not rationality — sensitivity to price is simultaneously compatible with rational optimization, framing effects, herding behavior, adaptive heuristics, and myopia under loss aversion; variation with the crediting rate is not a sufficient criterion for rationality. Furthermore, the rational-response framework itself is nearly unfalsifiable — rational choice theory has ample degrees of freedom to restate any observed behavior as an optimal response to some set of constraints and preferences. This chapter therefore converts the question from the unidentifiable question of attribute — whether residents are in fact rational — into the identifiable question of mechanism — which policy lever is more effective at the margin. What this chapter can cleanly identify is: the price sensitivity reflected in the fact that, after controlling for income quantile, downgrading still varies with the crediting rate; the institutional attribution reflected in the fact that the contribution effect of introducing a default or automatic deduction far exceeds the effect of strengthening investor education; and the difference in contributions between purely closed accounts and accounts with limited liquidity windows.

This chapter's position within the layered-adaptation framework lies on the behavioral-architecture axis of individual pensions, crossing the two layers of the basic pension insurance fund and individual pensions at the flexible-employment fault line. The specific institutional constraints it anchors to are the regressivity of the unified nudge, the unpriced nature of the closed account's liquidity-option prepayment, and the absence of a carrier for the auto-enrollment mechanism among the flexibly employed. That the institutional carrier of the default mechanism is not independently available at the flexible-employment layer is the hardest institutional anchor for the judgment that the default mechanism is transferable if and only if a default carrier exists.

Section 2 Mechanism

The causal chain of this chapter can be stated as follows: low interest rates weaken savings incentives, and combined with the unpriced nature of closure costs and the regressivity of the unified nudge, this causes low- and middle-income earners to migrate downward or defer contributions, from which the coverage gap is endogenously generated.

The first key link in this causal chain is regressive nudging. The unified tax preference nudge is of high value to high-income earners — the deduction is large under high marginal tax rates — and of low or even negative value to low- and middle-income earners; the arrangement of deferring taxation on contributions amplifies this regressivity. A unified nudge is a regressive nudge — it introduces distributive politics into the behavioral-manifestation layer.

The second key link in this causal chain is the downward-migration cascade. Low interest rates have reversed the relative attractiveness of the Employee Pension Insurance and the Urban-Rural Resident Pension Insurance: when the actual return on Employee Pension Insurance contributions declines, low- and middle-income earners choose to enroll in the worst tier — that is, they abandon the Employee Pension Insurance in favor of the Urban-Rural Resident Pension Insurance, or simply stop contributing altogether — triggering a cascade of downward migration. The gap is not non-enrollment but enrollment in the worst tier; it is an endogenous inter-tier disparity driven by incentives.

The third key link in this causal chain is the country of no defaults. The efficacy of the auto-enrollment mechanism depends on the employer as its carrier — that is, automatic deduction from payroll. For the flexibly employed population, who have no employer, no payroll, and no default pathway, the coverage gap arises at the site of decision costs and institutional carriers, not at the site of contribution capacity — even where contribution capacity exists, the absence of a default carrier means contributions will not occur automatically.

It must be emphasized that capacity constraints and institutional responses can coexist; this chapter does not construct a false dichotomy. Low-income individuals can simultaneously be constrained by capacity and be sensitive to meager actual returns. Precisely for this reason, this chapter does not ask the unidentifiable question of whether residents are rational, but asks the identifiable question of which lever is more effective at the margin.

Section 3 Identification Strategy

The identification in this chapter consists of three primary strategies, one backup strategy, and one identification fallback design.

The first primary strategy is a comparison of effect sizes between mechanism and literacy. This compares the contribution effect of introducing a default or automatic deduction against the contribution effect of strengthening investor education. The identifying assumption is random assignment between the default arm and the investor-education arm. Its core proposition is: if the default effect far exceeds the investor-education effect, then the gap lies at the institutional site rather than the literacy site. This comparison can identify the relative effect sizes of the institutional lever and the educational lever without making any claim about resident rationality.

The second primary strategy is quantile-controlled identification of price sensitivity. After controlling for income quantile, this tests whether downgrading and deferred contribution still vary with the crediting rate. The identifying assumption is that, after controlling for quantile, variation in the crediting rate is exogenous to individual literacy. Its core is: continued variation with the crediting rate after controlling for quantile proves sensitivity to price — here it must again be emphasized that what is identified is price sensitivity, not rationality.

The third primary strategy is a quasi-natural experiment on default carriers. This uses the introduction of per-order billing for occupational injury protection as a quasi-natural experiment, an arrangement that exogenously provided a contribution carrier for a portion of the flexibly employed population. The identifying assumption is that the introduction of per-order billing is exogenous to individual willingness to contribute.

The backup strategy is a comparison of closed windows on liquidity liabilities — that is, the difference in contributions between purely closed accounts and accounts with limited liquidity windows — which is a sign-only judgment, since the value of the prepaid option is difficult to price precisely though its direction can be established.

The identification fallback design is determined at the design stage itself. If the default arm and the investor-education arm cannot be randomly assigned, the mechanism-versus-literacy comparison falls back to the single quasi-natural experiment on per-order billing, identified via exogenous variation in the carrier. The attribute of rationality versus non-rationality is never subject to an identifiable claim, even where data are abundant — only the effect size of the mechanism is identified, never the attribute.

Section 4 Honest Labeling of the Evidence

At the level of verifiable aggregate facts, a number of institutional facts lend support to the judgments of this chapter.

The scale fact of "hot account-opening, cold contributions" provides direct evidence for the phenomenon this chapter seeks to explain: the individual pension system, having piloted in thirty-six cities, was rolled out nationwide on December 15, 2024; as of November 2024, 72.79 million people had opened accounts, but the proportion actually contributing was only 22%, with average per-capita contributions of about RMB 2,000 — far below the annual cap of RMB 12,000 (Securities Times 2024). This directly supports the reality and scale of cold contributions as the phenomenon to be explained, and indicates that the gap arises from contribution willingness rather than account-opening accessibility. The regressivity of the unified tax preference also has an institutional foundation: individual pensions employ an arrangement in which contributions and investment income are tax-exempt while withdrawals are taxed at 3% — under a unified tax-rate structure, this is of high deduction value to those with high marginal tax rates and of low or even negative value to low- and middle-income earners, since the 3% at withdrawal may exceed their marginal tax rate (Ministry of Finance, State Taxation Administration 2022) — this confirms the institutional existence of "a unified nudge is a regressive nudge." The structural fact that the flexibly employed lack a default carrier is likewise verifiable: flexibly employed individuals may choose to participate in the Employee Pension Insurance or the Urban-Rural Resident Pension Insurance in an individual capacity, with no employer for automatic deduction, no payroll withholding, and no default pathway; as of the end of 2024, the number of flexibly employed individuals participating in basic Employee Pension Insurance was 70.57 million (Ministry of Human Resources and Social Security 2023) — the employer carrier on which the auto-enrollment mechanism depends is structurally absent for this population. The institutional gap driving the downward-migration cascade also has empirical support: the average monthly basic pension under the Urban-Rural Resident Pension Insurance is about RMB 214, with the national minimum standard rising from RMB 103 in 2023 to RMB 123 in 2024, giving a replacement rate of about 12-13%, whereas the replacement rate under the Employee Pension Insurance is systematically much higher, reaching as much as 80% for government and public-institution employees (Ministry of Civil Affairs of the People's Republic of China 2024) — when the actual return on Employee Pension Insurance declines under low interest rates, marginal participants who enroll in the worst tier are driven by a real benefit gap.

At the level of micro-identification, owing to data-availability constraints, the core identification components of this chapter can only be presented as identification designs. The effect-size comparison between mechanism and literacy cannot be executed, because a three-arm controlled experiment has not been implemented and the data are unavailable; the identification design is in place, pending data, with the falsification condition being that the investor-education effect proves larger. The quantile-controlled identification of price sensitivity cannot be executed, because a panel of migration between Employee Pension Insurance and Urban-Rural Resident Pension Insurance enrollment, crediting rates, and income quantiles — internal enrollment-migration micro-data — is unavailable; the identification design is in place, the sign is falsifiable, the absolute magnitude is pending data, with the falsification condition being that no variation with the crediting rate remains after controlling for quantile. The difference-in-differences design on default carriers using per-order billing cannot be executed because a panel of contribution sequences is unavailable; the identification design is in place, the sign is falsifiable, the absolute magnitude is pending data. The liquidity-liability comparison is treated as a sign-only judgment because the value of the prepaid option is difficult to price precisely.

The honest boundary that must especially be observed is: this chapter acknowledges that capacity constraints and institutional responses can coexist, and does not construct a false dichotomy — observing price sensitivity does not rule out capacity constraints, and therefore continued sensitivity after controlling for quantile is never to be read as purely institutional, with no capacity constraint; this chapter identifies only which lever is more effective at the margin. The attribute judgment of resident rationality versus non-rationality — since price sensitivity is compatible with multiple explanations including rational optimization, framing, herding, myopia, and loss aversion — is permanently labeled as non-identified; this chapter retains only mechanism attribution.

Section 5 Counterarguments and Responses

Two of the strongest possible objections to the judgment of this chapter must be addressed directly, the first of which directly delimits the wording boundary of this chapter's judgment.

The first objection concerns the non-identifiability of the rationality label. If the core assertion of this chapter were about the rationality attribute of behavior, then the falsification design provided could only identify responsiveness to price, not rationality. That downgrading still varies with the crediting rate after controlling for income quantile demonstrates sensitivity to relative returns, but sensitivity to price is simultaneously compatible with rational optimization, salience and framing effects — i.e., the crediting rate being more conspicuous — herding, adaptive heuristics, and myopia under loss aversion; variation with the crediting rate is not a sufficient criterion for rationality, since non-rational agents also respond to salient price signals. At a deeper level, the rational-response framework itself is nearly unfalsifiable — rational choice theory has ample degrees of freedom to restate any observed behavior as an optimal response to some set of constraints and preferences: cold contributions can be described as a rational response to a regressive architecture, and were residents to contribute enthusiastically, that too could be described as a rational response to the tax preference. And the falsification point that non-rational or capacity-constrained behavior should not vary with the crediting rate is itself a false dichotomy, since in reality capacity constraints and sensitivity to returns can coexist. In response to this objection, this chapter's disposition is to withdraw the core assertion from the domain of rationality versus non-rationality to the identifiable domain of institution versus individual deficiency — cold contributions are driven mainly by institutional features rather than mainly by individual literacy, which is what the three-arm experiment and the tax-preference/default comparisons can identify, without any need to claim resident rationality; and to substitute the effect-size comparison of mechanism attribution for the rationality label — if the contribution effect of introducing a default far exceeds the effect of strengthening investor education, this supports the conclusion that the gap lies at the institutional site rather than the literacy site. This chapter does not adopt "rational response" as an identifiable claim.

The second objection concerns a boundary slippage between the transferability of the default mechanism and what cannot simply be copied over. This chapter, on one hand, asserts that the default mechanism is transferable, while on the other hand emphasizing that the auto-enrollment mechanism is structurally absent for the flexibly employed. A tension exists here: if the population lacking an employer carrier is precisely the population that cannot carry the default mechanism, then the transferability of the default mechanism does not hold for the flexibly employed population that most needs expanded coverage — transferability is hollowed out by its own scope of applicability, falling into the mirror problem of selectively claiming that this can be transferred but that cannot, absent a unified criterion for drawing the boundary. In response to this objection, this chapter's disposition is to attach a carrier condition to transferability and thereby establish a unified criterion: for populations with an employer carrier, what is transferable is the default mechanism, not the rate of return; for populations without an employer carrier — that is, the flexibly employed — the carrier for the default must first be resolved, whether through platform withholding or automatic aggregation via the tax system, since the absence of a carrier is itself a more fundamental constraint than the rate of return. On this basis, the unified criterion for what can and cannot be transferred is whether an institutional carrier for the default exists — where a carrier exists, transfer is possible; where none exists, the carrier must first be built. Transferability thereby becomes a verifiable proposition about the availability of a precondition, eliminating selective boundary-drawing.

Chapter Summary

This chapter has argued for the judgment that cold contributions in the three-pillar system are driven mainly by institutional features rather than mainly by individual literacy. This chapter has shown that the unified nudge constitutes a regressive nudge, that low interest rates reverse the relative attractiveness across tiers and trigger a cascade of downward migration, and that the auto-enrollment mechanism is structurally absent for the flexibly employed; this chapter has especially emphasized converting the question from the unidentifiable matter of resident rationality into the identifiable matter of which lever is more effective at the margin, identifying only price sensitivity and mechanism effect size, not the attribute of rationality. At the evidentiary level, the fact that over 70 million individual pension accounts have been opened while the contribution ratio is only 22%, the regressive tax structure arising from deferred taxation of contributions, the absence of a default carrier for the flexibly employed, and the benefit gap between the roughly RMB 214 basic pension under the Urban-Rural Resident Pension Insurance and the Employee Pension Insurance, all provide support for the judgment — while the micro-identification components, namely the effect-size comparison between mechanism and literacy, the quantile-controlled identification of price sensitivity, and the quasi-natural experiment on default carriers, are, owing to data-availability constraints, honestly labeled as identification designs in place with absolute magnitudes pending data.

Chapter 8 Decumulation-Phase Uptake: The Misallocation and Soft Constraint of Longevity-Risk Uptake

This chapter corresponds to the fifth and analytically strongest cluster of judgments in this study, situated on the cross-layer axis of longevity-uptake responsibility allocation and the urban–rural dual redistribution axis.

Section 1 Judgment

The core judgment advanced in this chapter is the analytically strongest of this study: institutional uptake of longevity risk is systematically absent as the system migrates from defined benefit (DB) or pay-as-you-go arrangements toward defined contribution (DC) or individual-account arrangements, with the uptake gap widening as the degree of contribution-definedness rises. First, because the payout phase does not mandate annuitization, longevity risk that could otherwise be pooled and diversified is instead disaggregated household by household and handed back, unchanged, to individuals and families — responsibility undergoes reverse socialization, with the uptake party unambiguously identified as the family. Second, for those covered by the non-annuitized defined contribution scheme who are also not covered by the first pillar, the family likewise lacks the capacity to absorb the risk, so it settles as a soft fiscal liability — a liability with an identified bearer but no pricing and no provisioning. The counter-cyclical commitment embedded in the bookkeeping interest rate of the Urban and Rural Resident Basic Pension Insurance converts interest-rate risk into an implicit fiscal liability, and the true regressivity lies in the spatial mismatch of payout capacity — the provinces with the highest bookkeeping rates are precisely those that are deeply aged and fiscally weak. The identification in this chapter accordingly takes as its backbone the interaction term for reverse socialization and the bidirectional falsification of spatial mismatch.

This chapter must first correct an institutional fact that needs clarifying, so that the judgment is not built on a mistaken foundation. China's first pillar — the Basic Pension Insurance for Urban Employees and the Basic Pension Insurance for Urban and Rural Residents — is, in its payout structure, essentially a pay-as-you-go defined benefit (DB) life annuity: once a participant reaches the statutory age, payments are disbursed monthly until death, and longevity risk is institutionally absorbed by the pooling fund and the fiscal system; the basic pension component of the Urban and Rural Resident Pension Insurance is likewise paid for life. Therefore, the pay-as-you-go system does not exit at the decumulation phase — quite the opposite, it is precisely at the decumulation phase that payout begins. The absence of uptake is not the first pillar's withdrawal, but rather something that occurs systematically as the system migrates from defined benefit or pay-as-you-go toward defined contribution or individual accounts. This stronger, more identifiable judgment locates the uptake gap precisely in the defined contribution account segment and in the population not covered by the first pillar, rather than referring broadly to a decumulation-phase vacuum across the entire five-pillar system.

This chapter accordingly abandons the negative framing of an "unowned vacuum" and instead adopts an identifiable misallocation of the uptake party and a soft constraint. The uptake party for the Urban and Rural Resident Pension Insurance is the fiscal authority; the problem is not that no one takes up the risk, but that the uptake party — namely the fiscal authority — has no pricing and no provisioning. What this chapter can identify cleanly is reverse socialization — the interaction term between payout method and account size, predicted here to be non-negative, i.e., under non-annuitization exposure rises rather than falls with account size — as well as spatial mismatch — the pass-through elasticity of the Urban and Rural Resident Pension Insurance bookkeeping rate is structurally near zero, compounded by a positive cross-section in which high bookkeeping rates, deep aging, and fiscal weakness coincide. As for the absolute magnitude of uptake under the soft constraint on the uptake party, because the fiscal authority's actuarial disclosure and budgetary arrangements for this contingent liability are partly unavailable, this is treated only as a pre-registrable hypothesis, established via a cross-sectional positive contrast between defined benefit life annuities and defined contribution accounts — falsified if found, supported only if not found — replacing a layer-by-layer negative search for absence.

This chapter's position within the layered adaptation framework lies on the cross-layer axis of longevity-uptake responsibility allocation and the urban–rural dual redistribution axis. The uptake party at the decumulation phase — an annuitizing entity with both pricing and provisioning — is not independently available at the defined contribution layer; this is the hardest institutional anchor establishing that overseas decumulation-phase experience, which presupposes a mature annuitized market, cannot be directly transplanted. This chapter upgrades each layer from an asset-allocation framework to a risk-uptake responsibility-allocation framework, representing the highest-order increment of judgment in this study.

Section 2 Mechanism

The causal chain of this chapter can be stated as follows: the migration of the decumulation phase from defined benefit to defined contribution causes longevity risk to be disaggregated household by household and returned to families; those not covered settle as soft fiscal liabilities; and the urban–rural spatial mismatch amplifies the differential in default probability.

The first critical link in this causal chain is the corrected institutional fact. China's first-pillar payout structure is essentially a pay-as-you-go defined benefit life annuity: payments are disbursed monthly from the statutory age until death, with longevity risk institutionally absorbed by the pooling fund and the fiscal system. Lifetime annuitization is the default form of a pay-as-you-go system; individuals cannot disaggregate it household by household. Therefore, the absence of uptake occurs systematically as the system migrates from defined benefit to defined contribution, with the uptake gap widening as the degree of contribution-definedness increases.

The second critical link in this causal chain is reverse socialization, in which the uptake party is unambiguously the family. The payout phase of the defined contribution account system does not mandate annuitization; longevity risk that the collective pool could otherwise diversify is instead disaggregated household by household and returned, unchanged, to individuals and families. This is the reverse socialization of responsibility — longevity risk that should have been socialized is instead handed back to families. Under non-annuitization, longevity exposure rises rather than falls with account size, meaning that those with larger accounts are handed back even greater exposure.

The third critical link in this causal chain is the soft-constrained fiscal authority — that is, a bearer exists but provisioning does not. For those covered by the non-annuitized defined contribution scheme who are also not covered by the first pillar, the family likewise lacks the capacity to absorb the risk, and it settles as a soft fiscal liability. The uptake party for the Urban and Rural Resident Pension Insurance is the fiscal authority; the problem is not that no one takes up the risk, but that the uptake party has no pricing and no provisioning — the counter-cyclical commitment of the bookkeeping rate converts interest-rate risk into an implicit fiscal liability.

The fourth critical link in this causal chain is the spatial mismatch of payout capacity. The true regressivity lies in the fact that the provinces with high bookkeeping rates are precisely those that are deeply aged and fiscally weak — where commitments are high is precisely where payout capacity is weak, and inequality is converted from a difference in benefit flows into a difference in default probabilities.

Section 3 Identification Strategy

The identification in this chapter is built from two primary strategies, one backup strategy, and one identification fallback design.

The first primary strategy is identification via the interaction term for reverse socialization. It centers on the interaction between payout method and account size. Its identifying logic is as follows: if annuitization does not affect exposure, this interaction term should be negative, i.e., under annuitization, large-account exposure is absorbed by annuitization; whereas this chapter predicts it to be non-negative, i.e., under non-annuitization, exposure rises rather than falls with account size — precisely reverse socialization. The identifying assumption is that the choice of, or constraint on, payout method is exogenous to longevity exposure, or that institutional constraints on payout method are used as variation. This strategy inherently identifies the uptake party as the family, and does not involve the threshold problem of a negative existence proposition.

The second primary strategy is bidirectional falsification of spatial mismatch, handled consistently with the chapter on the risk-free anchor. The fiscal-uptake hypothesis predicts that the pass-through elasticity of the Urban and Rural Resident Pension Insurance bookkeeping rate is structurally near zero — the fiscally set bookkeeping rate is mechanically unable to transmit asset-side returns — whereas the direct asset-pass-through hypothesis predicts this elasticity to be significantly greater than zero. This strategy must report confidence intervals and statistical power, treating a significantly positive elasticity as falsification of the fiscal-uptake hypothesis, rather than passively harvesting a near-zero estimate as support. Positive identification is instead carried by the parallel positive cross-section of spatial mismatch, namely that high bookkeeping rates, deep aging, and fiscal weakness jointly point toward high default probability.

The backup strategy is a positive cross-sectional contrast on misallocation of the uptake party. The uptake party is anchored ex ante as an entity whose balance sheet contains a liability item that varies with beneficiaries' survival status and a corresponding provisioning rule; on this basis, a cross-sectional contrast in visibility is drawn between defined benefit life annuities, which have longevity reserves, and defined contribution accounts, which do not — falsified if found, supported only if not found. The evidence separating a soft constraint on the uptake party from an outright vacuum is whether the fiscal authority has any actuarial disclosure or budgetary arrangement for this contingent liability. This strategy substitutes a positive contrast for a negative search, thereby avoiding the pitfall that the threshold for a negative existence proposition cannot be anchored.

The identification fallback design was fixed at the design stage. If payout method cannot be separated from longevity exposure, reverse socialization falls back to the single track of the positive cross-section of spatial mismatch, which is constructed from provincial bookkeeping rates, aging rates, and fiscal capacity, and does not depend on the exogeneity of payout method. The soft constraint on the uptake party is permanently flagged as pre-registered, and is established only via positive contrast, not via the negative judgment that an absence of trace on the liability side constitutes proof.

Section 4 Honest Labeling of the Evidence

At the level of verifiable aggregate facts, several institutional facts lend support to the judgment of this chapter.

The corrected institutional fact that the first pillar is a pay-as-you-go defined benefit life annuity that pays out precisely at the decumulation phase has been confirmed: the payout structure of the Basic Pension Insurance for Urban Employees and the Basic Pension Insurance for Urban and Rural Residents is essentially a pay-as-you-go defined benefit life annuity, disbursed monthly from the statutory age until death, with longevity risk institutionally absorbed by the pooling fund and the fiscal system; the basic pension component of the Urban and Rural Resident Pension Insurance averages roughly RMB 214 per person per month and is paid for life (Ministry of Civil Affairs of the People's Republic of China 2024), which confirms that the absence of uptake is not the first pillar's withdrawal but rather something that occurs systematically as the system migrates from defined benefit toward defined contribution. The institutional premise of reverse socialization — that the payout phase of defined contribution schemes does not mandate annuitization — is likewise verifiable: the Personal Pension, an account-based defined contribution arrangement, does not mandate annuitization at payout and permits lump-sum or installment withdrawal, so longevity risk that the collective pool could otherwise diversify is instead disaggregated household by household and returned to individuals and families; meanwhile 72.79 million accounts have been opened for the Personal Pension yet contribution activity remains cold (Securities Times 2024); at the same time, the supply of commercial annuities has contracted, with the guaranteed rate stepped down from 3.5% to 2.0% (National Financial Regulatory Administration 2024), compressing the supply of life annuities. The institutional foundation of spatial mismatch — that the counter-cyclical commitment of the Urban and Rural Resident Pension Insurance bookkeeping rate constitutes an implicit fiscal liability — is equally verifiable: the bookkeeping rate on individual accounts under the Urban and Rural Resident Pension Insurance is administratively set, decoupled from the market, and significantly higher than contemporaneous deposit rates, while the national minimum standard for the basic pension is administratively raised year after year and absorbed by the fiscal authority — the uptake party is the fiscal authority, but it has no pricing and no provisioning. In addition, the phased delay of the statutory retirement age, effective from January 1, 2025, raises the retirement age for male employees progressively from 60 to 63 and for female employees progressively from 50 or 55 to 55 or 58, while from 2030 the minimum contribution period is progressively raised from 15 to 20 years (Standing Committee of the National People's Congress 2024), constituting the institutional backdrop for adjustments to the timing of decumulation and to uptake responsibility; and the transfer of 10% of state-owned equity to replenish the social security fund, intended to close the gap arising from deemed contribution years (Ministry of Finance 2021), indicates that the payout side of the first pillar is supported by a contingent dividend flow, echoing the accumulation of soft fiscal liabilities for those not covered.

At the level of micro-identification, owing to data-availability constraints, this chapter's core identification steps can present only the identification design. The reverse-socialization interaction term cannot be executed because the panel of payout method, account size, and longevity exposure is unavailable — micro-data on individual account withdrawal behavior are not public — so the identification design is in place, the sign is falsifiable, but the absolute magnitude awaits data; the falsification condition is that the interaction term is significantly negative. The bidirectional falsification of spatial mismatch cannot be executed because the panel of provincial resident-insurance bookkeeping rates, aging rates, fiscal capacity, and asset-side returns is partly unavailable — the province-level pass-through elasticity of the bookkeeping rate requires internal data — so the identification design is in place, the sign is falsifiable, but the absolute magnitude awaits data; the falsification condition is that the elasticity is significantly greater than zero, or that the spatial-mismatch cross-sectional coefficient is not positive. In the cross-sectional contrast on misallocation of the uptake party, the longevity-reserve provisioning rules for defined benefit life annuities are verifiable and serve as the positive anchor, with the absence of reserves under defined contribution accounts serving as the contrast; however, because the fiscal authority's actuarial disclosure and budgetary arrangements for this contingent liability are partly unavailable, the magnitude of uptake is treated as a pre-registrable hypothesis.

The honesty boundary that must be especially observed is this: this chapter never passively harvests a near-zero pass-through elasticity as support for the fiscal-uptake hypothesis — a near-zero estimate is a directional judgment that can be overturned by positive direct pass-through, and a significantly positive elasticity would falsify fiscal uptake, with positive identification instead carried by the positive cross-section. Likewise, this chapter never reads the absence of trace on the liability side as establishing that the vacuum in attribution holds; instead it substitutes the positive contrast between life insurance and account-based systems, such that finding a liability item linked to survival status constitutes falsification. The absence of data only lowers the strength of the claim regarding uptake magnitude — it never reverses it.

Section 5 Counterarguments and Responses

Two of the strongest possible objections to the judgment of this chapter must be addressed directly, the first of which concerns an institutional fact that must be corrected.

The first objection concerns the accuracy of the institutional fact and the scope of the claim. If the first pillar of the judgment were that the first pillar withdraws at the decumulation phase, this would be a factual error, and this error would directly remove the foundation of the negative "vacuum" claim. China's first pillar — the Basic Pension Insurance for Urban Employees and the Basic Pension Insurance for Urban and Rural Residents — is, in its payout structure, essentially a pay-as-you-go defined benefit life annuity: participants receive monthly payments from the statutory age until death, with longevity risk absorbed by the collective pooling fund and the fiscal system, which is precisely the defining feature of a pay-as-you-go system, and the basic pension component of the Urban and Rural Resident Pension Insurance is likewise paid for life; for the population already covered by the first pillar, longevity risk is precisely what is institutionally absorbed, since lifetime annuitization is the default form of pay-as-you-go and individuals cannot disaggregate it household by household. Therefore, the pay-as-you-go system does not withdraw at the decumulation phase — quite the opposite, it is precisely at the decumulation phase that payout begins — and what truly lacks an uptake party is the defined contribution account segment and the population not covered by the first pillar, rather than a decumulation-phase vacuum across the entire five-pillar system; exaggerating a partial gap into an overall vacuum is an overclaim of scope. This chapter's response to this objection is to correct the institutional fact and narrow the scope: it explicitly acknowledges that the first pillar's payout structure is a pay-as-you-go defined benefit life annuity that pays out precisely at the decumulation phase, narrows the judgment from an overall vacuum across the five-pillar decumulation phase to a partial gap in the defined contribution account segment and among those not covered by the first pillar, and rewrites the core judgment as the stronger, more identifiable claim that longevity uptake is systematically absent as the system migrates from defined benefit or pay-as-you-go toward defined contribution. This chapter does not adopt the framing of the first pillar withdrawing or an overall vacuum across the five pillars.

The second objection concerns the fact that the threshold for a negative existence proposition cannot be anchored. The original approach — searching across the five pillars and treating failure to find an uptake party as establishing an attribution vacuum — is a negative existence proposition with an unanchored threshold: any candidate uptake party pointed to, whether the pooling fund, fiscal backstop, or family, can always be rebutted with "that doesn't count, it lacks precise pricing or explicit provisioning, it's merely an implicit backstop," and since the threshold for pricing and provisioning was never anchored ex ante, this constitutes a moving goalpost; failing to find a trace is equivalent to the trace being implicit or the data being unobservable, so falsifiability tends toward zero, treating absence of evidence as evidence of absence. Moreover, within the judgment, "vacuum" meaning "no one takes up the risk" and "reverse socialization" meaning "returned to the family" undercut each other on the very question of who bears longevity risk — the latter precisely identifies the uptake party as the family, contradicting the former. This chapter's response to this objection is to discard the "unowned vacuum" framework and replace it with an identifiable misallocation of the uptake party and a soft constraint: the uptake party for the Urban and Rural Resident Pension Insurance is the fiscal authority, and the problem is not that no one takes up the risk but that the uptake party has no pricing and no provisioning; it anchors an observable criterion ex ante — an entity counts as an uptake party if and only if its balance sheet contains a liability item that varies with beneficiaries' survival status and a corresponding provisioning rule — converting this into a positive cross-sectional contrast between defined benefit life annuities and defined contribution accounts, falsified if found, supported only if not found; it also reorders the primacy of identification, making reverse socialization — which inherently identifies the uptake party as the family — the core identification step, and demoting "vacuum" to a bounded inference restricted to those who are neither annuitized nor covered, for whom the family also lacks the capacity to absorb the risk and it therefore settles as a soft fiscal liability, so that the two claims move from mutual exclusion to a progressive relationship.

The five core judgments of the full paper, together with their identification strategies and honest-labeling status, are presented together in the table below, to display the full picture of identification design and data boundaries across the five problem domains.

Table 3 Overview of the core identification steps, identification strategies, and evidentiary status of the five sets of judgments

Chapter Summary

This chapter has argued for the analytically strongest judgment of this study: that institutional uptake of longevity risk is systematically absent as the system migrates from defined benefit toward defined contribution. The chapter first corrected an institutional fact — that the first pillar, as a pay-as-you-go defined benefit life annuity, pays out precisely at the decumulation phase and does not withdraw — thereby precisely locating the uptake gap in the defined contribution account segment and among those not covered by the first pillar, rather than referring broadly to an overall vacuum across the five pillars. On this basis, the chapter replaced the negative "unowned vacuum" with an identifiable misallocation of the uptake party and a soft constraint, showing that the absence of mandatory annuitization at payout causes longevity risk to be reverse-socialized back to families, that those not covered settle as unprovisioned soft fiscal liabilities, and that the urban–rural spatial mismatch in payout capacity amplifies the differential in default probability. At the evidentiary level, the institutional fact of the first-pillar lifetime annuity, the non-mandatory annuitization of the Personal Pension, the fiscal absorption of the Urban and Rural Resident Pension Insurance bookkeeping rate decoupled from the market, and institutional backdrops such as delayed retirement and the transfer of state-owned assets all lend support to the judgment, while the micro-identification steps — the reverse-socialization interaction term and the bidirectional falsification of spatial mismatch — are honestly labeled, owing to data-availability constraints, as having the identification design in place with the absolute magnitude awaiting data, and the absolute magnitude of uptake under the soft constraint on the uptake party is permanently flagged, within a positive-contrast framework, as pre-registered. This chapter upgrades each layer from an asset-allocation framework to a risk-uptake responsibility-allocation framework, providing the highest-order organizing perspective for the conclusion of the entire study.

Chapter 9 Cross-Paper Synthesis: From Overseas Experience to China Adaptation

Section 1 The Logic of Synthesis: The China Convergence of a Single Thread

This study is the concluding and grounding segment of the complete research thread running from interest-rate mechanisms, through institutional differences, allocation instruments, and risk boundaries, to a China solution. The preceding four papers respectively established the transmission-mechanism chain from interest rates to allocation, an institutional taxonomy and risk-bearing mapping, an allocation toolbox together with its governance-capacity conditions, and a risk-boundary and governance-reconstruction framework. The task of this chapter is to explicitly carry the general conclusions of these four papers down to the specific circumstances of China's four entity tiers and the pension target fund supplementary tier, thereby demonstrating why the China solution holds, and so bringing to convergence the single thread running from macro interest-rate shocks, through institutional differences, allocation transformation, and risk exposure, to China adaptation.

Synthesis is not a repetition of the preceding four papers, but a grounding of their general judgments in the specific forms of China's institutions. This chapter successively takes up the interest-rate mechanism paper, the institutional-differences paper, the paper on the endogeneity of governance in allocation instruments, and the paper on decumulation-phase risk governance, showing the concrete landed form each paper's core judgment takes within each tier in China, and revealing how the present study's five sets of judgments each supply a China-specific empirical interface for these four papers.

Section 2 Taking Up the Interest-Rate Mechanism Paper: The Institutionalized Channel of the Shadow Rate in China

The interest-rate mechanism paper established the transmission chain from interest rates to allocation, one of whose core concepts is the restructuring of the real-return structure of risk-free assets under the shadow rate. The chapter of the present study on the risk-free anchor is the primary synthesis interface for this interest-rate mechanism in China.

The derivation may be stated as follows. The premise is the judgment of the interest-rate mechanism paper: low rates are transmitted to allocation behavior via a restructuring of the real-return structure of risk-free assets. The reasoning is that, within China's institutional environment, low rates are not an exogenous shock but rather push from latent to manifest the misalignment whereby the risk-free anchor is endogenously shifted by fiscal authorities — the transmission of rates in China is realized through the institutionalized channel in which fiscal authorities shift the risk-free anchor and the first and second pillars absorb the duration of debt resolution; the concrete form the shadow rate takes in China is the wedge between the booking rate and the real return on the asset side, the booking rate being a shadow rate set by fiscal authorities rather than a market rate. The conclusion is that the precondition for China's pension asset-liability management reform is not the introduction of more refined duration-matching instruments, but first clarifying to whom the defining authority over the risk-free anchor belongs, because in an environment where the risk-free anchor is endogenously shifted by fiscal authorities, overseas asset-liability management templates that presuppose a stable risk-free anchor cannot be directly transplanted. The account tier to which this derivation is bound is the asset side of the Basic Pension Insurance Fund and the National Social Security Fund, its constraint is the absence of an exogenous risk-free anchor, and its antecedent conclusion is the transmission-mechanism chain of the interest-rate mechanism paper.

The interest-rate mechanism also resonates throughout the other chapters. In the chapter on governance vehicles, the interest rate is transmitted via the logic that lower returns trigger more position-adding — low rates render manifest, from latent, the self-destruction of property rights in synthetic patient capital, because it is only under low returns that the left-tail cost of the sale-restriction constraint becomes binding. In the chapter on accounting manifestation, the interest rate is transmitted via the dual-rate illusion between the booking rate and the real curve — the booking rate, partly locked in by fiscal contributions, is itself a shadow booking rate, and low rates render the dual-rate wedge manifest from latent. In the chapter on decumulation-phase absorption, the interest rate is transmitted via the countercyclical commitment of the booking rate for the Urban and Rural Resident Pension Insurance — under low rates, the countercyclical commitment converts interest-rate risk into an implicit fiscal liability. These four points of resonance show that the transmission chain depicted in the interest-rate mechanism paper does not appear in China in a single form, but unfolds across the five sets of judgments in institutionalized forms including the manifestation of property-rights self-destruction, the manifestation of the dual-rate wedge, and the accumulation of implicit fiscal liabilities — their common feature being that the primary effect of low rates is not to change economic substance but to push costs that were already latent within institutional arrangements into visibility.

Section 3 Taking Up the Institutional-Differences Paper: The China Mapping of the Defined-Benefit–Defined-Contribution Continuum

The institutional-differences paper established an institutional taxonomy and risk-bearing mapping grounded in the tripartite classification of defined benefit (DB), defined contribution (DC), and public reserves, emphasizing that institutional difference is a continuous spectrum rather than a discrete typology. Multiple chapters of the present study take this taxonomy as their coordinate system.

In the chapter on governance vehicles, institutional difference corresponds to the fiduciary-structure continuum — the three fiduciary tiers of proximity-to-center, provincial concentration, and local trusteeship are not discrete types but three sampled points on the continuum of governance control rights, isomorphic to the institutional-differences paper's judgment that institutional difference is a continuum rather than a discrete typology. In the chapter on accounting manifestation, institutional difference corresponds to intra-spectrum differentiation between funded contributions and booking — the funded-contribution portion is DB-locked by booking, while the market-invested portion is DC, such that a single occupational annuity plan spans the DB–DC continuum internally; the present study reveals that this intra-spectrum differentiation is driven by booking rules rather than by cash flows, deepening institutional difference into the dimension of the booking mechanism.

The derivation may be stated as follows. The premise is the judgment of the institutional-differences paper: DB absorbs risk, DC transfers risk, institutional difference is distributed along a continuum, and risk-bearing maps onto different agents. The reasoning is that China's occupational annuity notional accounts sit, on the DB–DC spectrum, as booking-system DB — the booking rate locks in the benefit, i.e., a DB-style commitment — while the funding side is DC-ized; this mismatch between booking-DB and funding-DC maps risk-bearing onto the fiscal authorities as a contingent liability rather than onto individuals. The decumulation phase is the terminus of this mapping: longevity risk is pushed onto households at the DC tier — that is, reverse socialization — and settles as a soft fiscal liability wherever coverage is absent. The conclusion is that China's pension system design cannot take a single booking convention as given and directly transplant overseas second-pillar experience, nor can it take a mature annuitized market as given and directly transplant overseas decumulation-phase experience, because the mapping of risk-bearing in China is systematically altered by the booking mechanism and by soft constraints on the absorbing agent. The account tier to which this derivation is bound is enterprise annuities and occupational annuities together with cross-tier longevity absorption, its constraint is dual-track booking and soft constraints on the decumulation-phase absorbing agent, and its antecedent conclusion is the risk-bearing mapping of the institutional-differences paper.

It is worth noting that the stopping-option logic in the institutional-differences paper — whereby residual risk is pushed, along the principal-agent chain, to different stopping points — is repeatedly invoked in the present study and reaches its terminus at the decumulation phase. The downward-migration cascade in the chapter on behavioral architecture is a stopping point at which risk and responsibility are pushed, across tiers, to the worst-off tier — from Employee Pension Insurance down to Urban and Rural Resident Pension Insurance; in the chapter on decumulation-phase absorption, the stopping point for longevity risk — pushed onto households at the DC tier and settling as a soft fiscal liability wherever coverage is absent — is the final link in this stopping-point logic.

Section 4 Taking Up the Paper on the Endogeneity of Governance in Allocation Instruments: Capacity Is Transferable, Control Rights Are Not

The allocation-instruments paper established an allocation toolbox and put forward the key judgment of governance-capacity conditions — namely, that the transferability of alternative and global allocation is premised on governance capacity. Its core judgment is that capacity can be learned but control rights cannot — governance capacity can be outsourced, unbundled, and capped by political tolerance, while what is endogenous is the non-transferable governance control right. The chapter of the present study on governance vehicles is the primary synthesis interface for this judgment in China.

The derivation may be stated as follows. The premise is the judgment of the allocation-instruments paper: allocation capacity can be outsourced and unbundled, while governance control rights are endogenous to institutional position and non-transferable. The reasoning is that governance control rights in China are granted once and for all by constitutional position — the step-function discontinuity between proximity to the center and distance from it — so the concrete form that the non-transferability of control rights (as established in the allocation-instruments paper) takes in China is a granted stock that cannot be replicated across tiers: whereas the allocation-instruments paper identifies overseas control rights via offshore governance discontinuities and shadow-fee re-estimation, the present study identifies China's granted stock via the cross-section of the three-tier fiduciary structure and the 2024 institutional discontinuity — the two being the landed forms, in China, of the same endogenous axis of control rights. The conclusion is that China's pension system cannot directly transplant a Canadian-style overseas allocation template on the premise of governance independence; any proposal to raise alternative and global allocation must first resolve the constitutional precondition of the granted stock — its implementing agents being the National Council for Social Security Fund and the fiscal authorities at the central level, its resistance coming from local agents that benefit from surplus rents, and its implementation sequence requiring the establishment of statutory ring-fencing and provisioning constraints before any discussion of raising allocation ratios. The account tier to which this derivation is bound is the Basic Pension Insurance Fund and the National Social Security Fund, its constraint is the absence of statutory ring-fencing together with governance independence that depends on administrative rank rather than law, and its antecedent conclusion is the governance-capacity condition of the allocation-instruments paper.

The governance-endogeneity judgment of the allocation-instruments paper also resonates throughout the other chapters. In the chapter on behavioral architecture, the technology of the default mechanism — i.e., automatic payroll-deduction software — can be outsourced, but the carrier of the default — the employer, the payroll system, the tax-collection apparatus — cannot be transferred across populations, isomorphic to the allocation-instruments paper's judgment that what is transferable is capacity while what is non-transferable is the carrier and the control right. Furthermore, the smoothing illusion revealed by the allocation-instruments paper — that the true premium contains a smoothing component, and valuation smoothing suppresses realized volatility — is, together with the present study's finding in the chapter on the risk-free anchor that the booking-rate convention systematically overstates real returns, and the finding in the chapter on accounting manifestation that discount-rate smoothing delays the manifestation of interest-spread losses, a shared instance of the divergence between book figures and economic substance in different applications — the present study's booking-rate wedge and discount-rate smoothing are the counterparts, at China's booking layer and second-pillar liability-discounting layer, of the smoothing illusion identified in the allocation-instruments paper.

Section 5 Taking Up the Paper on Decumulation-Phase Risk Governance: Risk Placement as an Institutionally Endogenous Choice of Attribution

The paper on decumulation-phase risk governance established a risk-boundary and governance-reconstruction framework, whose meta-judgment is that risk placement is an institutionally endogenous choice of attribution that must be designed endogenously and cannot be transplanted. The chapter of the present study on decumulation-phase absorption is the primary synthesis interface for this judgment in China, and the short put on guarantee options in the chapter on accounting manifestation is also brought to synthesis in this section.

The derivation may be stated as follows. The premise is the judgment of the decumulation-phase risk-governance paper: risk placement is an institutionally endogenous choice of attribution, and the risk boundary of the decumulation phase is determined institutionally as to attribution, requiring endogenous design and precluding transplantation. The reasoning is that the present study pushes this choice of attribution to the final link at the decumulation phase — the attribution of longevity risk is misaligned at the DC tier, being returned household-by-household, and is softly constrained at Urban and Rural Resident Pension Insurance, where the fiscal authorities bear nominal responsibility without provisioning; the self-cannibalizing mechanism of implicit guarantees identified in the decumulation-phase risk-governance paper finds its concrete instance in China precisely in the countercyclical commitment of the booking rate for Urban and Rural Resident Pension Insurance — where the localities that commit to higher rates have weaker payout capacity, constituting the self-cannibalization of the implicit guarantee; the pre-commitment rule-locking identified in the decumulation-phase risk-governance paper corresponds, in China, to the requirement that provisioning constraints on the absorbing agent be established first. The conclusion is that China's pension system cannot directly transplant overseas decumulation-phase experience such as target-date funds on the premise of a mature annuitized market; an absorbing agent at the decumulation phase — an annuitizing agent with pricing and provisioning capacity — cannot be independently obtained at the DC tier, and therefore the precondition for decumulation-phase reform is first establishing provisioning constraints on the absorbing agent and a credible no-bailout mechanism. The account tier to which this derivation is bound is cross-tier longevity absorption and urban-rural dual redistribution, its constraint is the misalignment and soft constraint of the decumulation-phase absorbing agent, and its antecedent conclusion is the governance-reconstruction framework of the decumulation-phase risk-governance paper.

The guarantee-option judgment of the decumulation-phase risk-governance paper finds direct resonance in the chapter on accounting manifestation. The essence of an arrangement with a guaranteed floor is a genuine short put option, whose economic liability truly increases under low rates as the in-the-money probability rises — this is a genuine increase in the cash-flow obligation, not an acceleration of accounting manifestation. The present study strictly separates this from accounting manifestation — accounting manifestation is a book-value dimension, while the hardness of the guarantee option is an economic-substance dimension — and brings the economic liability of guarantee-bearing arrangements to synthesis with the risk boundary of the decumulation-phase risk-governance paper: that is, a guarantee, as a genuine short put option, sees its economic liability genuinely rise under low rates, stitched together with the risk governance of the decumulation phase. This very separation itself constitutes a concrete corroboration of the decumulation-phase risk-governance paper's judgment that risk placement is an institutionally endogenous choice of attribution: how a liability is manifested in accounting, and who bears that liability economically, are two questions that must be treated separately.

The derivations and bound elements of the four synthesis exercises above may be presented together in the table below, so as to display the full picture of the synthesis thread.

Table 4 Antecedent Conclusions, China Landed Forms, and Synthesis Interfaces of the Four Cross-Paper Synthesis Derivations

Section 6 Cross-Tier Ordering of Political-Economic Feasibility

The four instances of synthesis not only supply the institutional constraints explaining why transplantation is impossible, but also implicitly yield a political-economic feasibility ordering across the reform agendas. Following the second criterion of the non-transplantability principle established by the present study, this section specifies, for each tier's reform, the implementing agent, the source of resistance, and the implementation sequence, so that the synthesis moves beyond diagnosis to an executable path.

For the reform of governance vehicles, the implementing agent is the National Council for Social Security Fund and the fiscal authorities at the central level; the resistance comes chiefly from local agents that rationally benefit from the triple rents of fiscal cash-flow management, credit leverage, and investment platforms drawn from surpluses — centralization means cutting off these rents, so local opposition is a matter of incentives rather than of cognition; the implementation sequence should be to first establish statutory ring-fencing and provisioning constraints before raising allocation ratios, because before the constitutional precondition of the granted stock is resolved, raising alternative and global allocation would only amplify the governance gap into a returns gap.

For the reform of the risk-free anchor and the booking mechanism, the implementing agents are the fiscal and monetary authorities together with pension regulators; the resistance comes chiefly from path dependence on endogenous intervention in risk-free asset pricing and from vested expectations locked in by the countercyclical commitment of the booking rate; the implementation sequence should be to first reduce fiscal endogenous intervention in risk-free asset pricing and unify the dual-track booking conventions, before introducing duration-matching instruments premised on a stable risk-free anchor, because asset-liability management tools cannot be operationalized while the risk-free anchor remains endogenously shifted and the dual-track booking conventions coexist.

For the reform of behavioral architecture and the decumulation phase, the implementing agents are the human resources and social security authorities, the tax authorities, and platform-economy employing entities; the resistance comes chiefly from the fragmentation of flexible-employment employment relationships and from the fiscal-visibility pressure implied by provisioning constraints on the absorbing agent; the implementation sequence should be to first establish a default carrier — platform withholding or automatic tax-based collection — for populations without an employer, and to establish provisioning constraints and a credible no-bailout mechanism for the decumulation-phase absorbing agent, before raising annuitization levels and expanding third-pillar coverage, because in the absence of a carrier and of provisioning for the absorbing agent, proposals to expand coverage and annuitization would be hollowed out by the absence of institutional preconditions.

The shared logic of this cross-tier ordering is: the precondition for reform is always the building of institutional carriers, not the adjustment of allocation parameters; any sequence that places parameter adjustment ahead of carrier-building will have the parameter adjustment rendered ineffective by the absence of its precondition.

Section 7 The Comprehensive Proposition of the China Adaptation Framework

Synthesizing the above instances of synthesis and ordering, the present study arrives at a unified proposition regarding China's pension adaptation: the transferability of asset-allocation experience from mature pension markets depends, without exception, on whether the institutional precondition carrying that allocation logic is independently obtainable at the target tier. This proposition unfolds cell by cell across the four entity tiers, and extends to the pension target fund supplementary tier as the fourth-tier investment vehicle, as detailed below.

At the National Social Security Fund tier, the experience of long-term equity holdings, global diversification, and prudent alternative allocation is relatively transplantable, because this tier possesses the strategic-reserve attribute of facing no rigid immediate payout pressure, along with relatively independent, professionalized governance — the allocation preconditions are relatively complete. At the Basic Pension Insurance Fund tier, the same allocation experience is not transplantable, because this tier bears immediate payout obligations and its governance independence depends on administrative rank rather than law, so its transplantability is constrained by the constitutional precondition of the granted stock. At the enterprise-annuity and occupational-annuity tier, the experience of default mechanisms, life-cycle allocation, and risk stratification is transplantable for populations with an employer carrier, but the dual-rate illusion created by dual-track booking must first be reckoned with, because overseas experience premised on a single booking convention would be distorted, under dual-track booking, by accounting-manifestation rules. At the personal pension tier, the experience of tax incentives and low-fee defaults has transplantability that depends on whether a default carrier is obtainable — transplantable for populations with an employer, but requiring the prior establishment of a platform-withholding or automatic tax-collection carrier for the flexibly employed.

The pension target fund, as a supplementary tier, has judgment content that is not independent of the four tiers above, but rather is the market-side extension of the fourth tier's personal pension allocation strategy; its conclusions are directly inherited from the two sets of judgments on behavioral architecture and decumulation-phase absorption, requiring no separate independent judgment cell. As regards its role as a source of long-term capital for capital markets, the transplantability of international practice in target-date and target-risk funds depends on the availability of the fourth tier's carrier — only once the default carrier is in place and contributions are no longer cold will the pension target fund have a stable flow of long-term capital available to enter the market; the present study's judgment regarding the absence of a default carrier for flexible employment therefore also defines the ceiling on the pension target fund's role as a source of long-term capital — in other words, unless cold contributions are resolved, the commitment of pension assets as a scale of long-term capital for capital markets will fail to materialize. As regards the other side of its dependence on capital markets supplying assets available for long-term allocation, the decumulation phase of the pension target fund remains constrained by the reverse socialization arising from the fact that DC accounts are not compulsorily annuitized — target-date funds, in mature markets, connect to the decumulation phase through annuitization, whereas in China the pricing and provisioning constraints of the absorbing agent are not yet independently obtainable; therefore, even once the pension target fund has accumulated long-term assets, its decumulation phase will still return longevity risk to households one by one, unless an absorbing agent with pricing and provisioning capacity is first established. It follows that both sides of the bidirectional linkage between pensions and capital markets — the capital-source side and the asset-supply side — are respectively constrained by the absence of the fourth tier's carrier and by the soft constraint on the decumulation phase's absorbing agent; the pension target fund does not constitute a new experiential channel independent of the four-tier judgments, but rather is the projection of the four-tier judgments onto the capital-market interface. This is precisely the ground on which the decumulation-phase experience of mature annuitized markets cannot be directly transplanted.

This unified proposition brings every entity cell of the tiered adaptation framework down to the single criterion of institutional-precondition availability, and brings the bidirectional linkage on both sides of the pension target fund supplementary tier down likewise to the two preconditions of the fourth tier's carrier and the decumulation phase's absorbing agent, thereby rendering transplantability or non-transplantability a verifiable proposition about precondition availability, while simultaneously constraining both of the opposite tendencies of over-borrowing directly and over-invoking non-transplantability.

Chapter Summary

This chapter has completed the explicit synthesis of the preceding four papers. Through four derivations, this chapter has respectively taken up the interest-rate mechanism, institutional differences, the endogeneity of governance in allocation instruments, and decumulation-phase risk governance: the interest-rate mechanism is transmitted in China via the institutionalized channel by which fiscal authorities shift the risk-free anchor; the risk-bearing mapping of institutional differences is altered in China by the booking mechanism and soft constraints on the absorbing agent; the governance endogeneity of allocation instruments manifests in China as the constitutional precondition of a granted stock; and the choice of attribution in decumulation-phase risk governance is pushed, in China, to the final link of longevity-risk absorption. This chapter concludes by synthesizing the four instances of synthesis into a unified proposition of China adaptation — namely, that the transferability of allocation experience depends, without exception, on whether the institutional precondition is independently obtainable at the target tier — and brings this criterion down to every cell of the four entity tiers and to the pension target fund supplementary tier, thereby converging the complete thread running from macro interest-rate shocks, through institutional differences, allocation transformation, and risk exposure, to China adaptation.

Chapter 10 Conclusion

Section 1 Principal Conclusions

The present study, taking as its subject the institutional adaptation of mature pension markets' asset-allocation experience to China's multi-pillar pension system, has proposed a systematic revision of the existing paradigm of experiential transplantation. The central conclusion of the study is: the transferability of pension asset-allocation experience does not depend on the technical sophistication of the allocation technique itself, but on whether the institutional precondition carrying that allocation logic is independently obtainable at the target tier. Around this central conclusion, the present study has built a tiered adaptation framework and distilled five mutually independent sets of judgments.

The first set of judgments holds that pension governance capacity is a stock endowment granted once and for all by constitutional position, whose quality declines in a step function across governance tiers, and that patient capital synthesized through administrative sale restrictions undergoes self-destruction of property rights because it is deprived of rebalancing rights. The second set of judgments holds that there is no exogenous risk-free anchor on the asset side of China's pension system, that the definition and pricing of risk-free assets are endogenously shifted by fiscal authorities, and that the first and second pillars are institutionalized as an absorbing pool for fiscal contingent liabilities. The third set of judgments holds that second-pillar allocation bifurcation is driven mainly by accounting-manifestation rules rather than by genuine changes in cash flows, such that a single contribution is assigned two risk-free rates owing to the booking mechanism. The fourth set of judgments holds that cold contributions to the third pillar are driven mainly by institutional regressivity rather than mainly by individual financial literacy, and that the coverage gap arises at the point of institutional carriers rather than at the point of contribution capacity. The fifth set of judgments holds that the institutional absorption of longevity risk is systematically absent as the system migrates from DB toward DC, such that the absence of compulsory annuitization causes risk to be reverse-socialized back onto households, while it settles, for the uncovered, as an unprovisioned soft fiscal liability.

The five sets of judgments each supply a China-specific empirical interface for the preceding four papers, and jointly converge on a unified proposition: the priority task of China's pension reform is not to raise the allocation ratio of some asset class, but first to resolve the availability of the institutional carriers on which the allocation logic of each tier depends.

Section 2 Policy Implications

The policy implications of the present study must strictly follow the aforementioned criteria for the non-transplantability argument — that is, every recommendation must be bound to an implementing agent, a resistance analysis, and an implementation sequence, rather than remaining at the level of a slogan about what should be done.

As regards governance vehicles, the recommendation to raise the ratio of alternative and global allocation must first resolve the constitutional precondition of the granted stock; its implementing agent is the National Council for Social Security Fund and the fiscal authorities at the central level; its resistance comes from local agents that rationally benefit from surplus rents; its implementation sequence is first to establish statutory ring-fencing and provisioning constraints, and only then to discuss raising allocation ratios. As regards the risk-free anchor, the recommendation to build an asset-liability management framework must first clarify to whom the defining authority over the risk-free anchor belongs; its precondition is reducing fiscal endogenous intervention in risk-free asset pricing, failing which duration-matching instruments premised on a stable risk-free anchor cannot be operationalized. As regards accounting manifestation, the recommendation to improve second-pillar allocation must first unify the dual-track booking conventions, because the booking rate, as an implicit opportunity-cost anchor, would distort the allocation of the market-invested portion. As regards behavioral architecture, the recommendation to expand third-pillar coverage must distinguish between populations: default mechanisms can be introduced for populations with an employer, whereas for the flexibly employed a platform-withholding or automatic tax-collection carrier must first be built, and the regressivity of uniform tax incentives must be corrected through differentiated design. As regards decumulation-phase absorption, the recommendation to raise annuitization levels must first establish provisioning constraints on the absorbing agent and a credible no-bailout mechanism, failing which the reverse socialization of longevity risk and the accumulation of soft fiscal liabilities cannot be contained.

The shared logic running through these policy implications is: the sequence of reform should be to first build institutional carriers and only then adjust allocation parameters, because in the absence of carriers, any adjustment of allocation parameters will be distorted by the absence of its institutional precondition.

Section 3 Limitations of the Study

The present study has three main limitations.

The first is the limited availability of micro-identification data. Although the core judgments of the present study are each accompanied by rigorous identification designs and pre-registered falsification conditions, owing to data-availability constraints, most micro-identification components — the discontinuity regression of governance indicators, the event study of property-rights self-destruction, the difference-in-differences of debt-resolution absorption, upper-bound argumentation, the cross-section of the dual-rate illusion, the comparison of effect sizes between mechanism and literacy, the interaction term for reverse socialization, and the pass-through elasticity of spatial mismatch — can present only the identification design as ready, without being able to fill in estimation results; the corresponding conclusions are accordingly presented in the form that the sign is falsifiable while the absolute magnitude awaits data. This limitation means that the judgments of the present study are falsifiable in direction, but their absolute magnitude awaits future testing once data become available. The present study has strictly adhered to the principle that the absence of data only lowers the strength of a claim and never reverses into positive evidence for the opposing hypothesis; accordingly, none of these judgments awaiting testing has been overstated as an already identified causal effect.

The second is the temporary non-testability of several causal separations. In the chapter on governance vehicles, the causal separation between the granted stock and the constructible flow, owing to the lack of a counterfactual compensation design in which incentive neutralization was randomly applied to some province, cannot be tested within the observation period, and is permanently marked as a pre-registered hypothesis awaiting a genuine quasi-natural experiment — the change in a given principal's governance indicators before and after a reform of the fiduciary structure — before it can be upgraded to identifiable. In the chapter on decumulation-phase absorption, the absolute magnitude of the absorbing agent's soft constraint, owing to the unavailability of actuarial disclosure of fiscal contingent liabilities, is likewise left as a pre-registered hypothesis.

The third is the difference in nature between aggregate corroboration and micro-identification. The verifiable aggregate facts cited in the present study — such as the return gap between the National Social Security Fund and locally trusteed funds, and the return gap between enterprise annuities and occupational annuities — are differences between two points or two groups in aggregate, confounded by multiple factors including years since establishment, fiduciary structure, asset duration, and member profile, and serve only as directional corroboration, never masquerading as pre-registered cross-sectional regression coefficients. This distinction in nature is reiterated repeatedly throughout the text, so as to avoid overclaiming micro-identification on the basis of aggregate differences.

The above limitations point toward directions for future research: as pension micro-data gradually become open, the identification designs of the present study can be put into execution, and their pre-registered falsification conditions can be tested one by one; and as reforms of the fiduciary structure and of the absorbing agent's provisioning system advance, several of the temporarily non-testable causal separations may also gain genuine quasi-natural experiments as sources of identification. It must be emphasized that these limitations do not diminish the value of the judgments of the present study. The core contribution of the present study lies in its judgments rather than in data-filling: the falsifiability of the five sets of judgments, the credibility of the identification strategies, and the explanatory power of the unified criterion of institutional-precondition availability, none of these depend on the present availability of micro-data. On the contrary, the present study's prior commitment — writing down judgments and falsification conditions through rigorous design in advance, and leaving data testing to the future — precisely embodies a research stance of design before implementation, judgment governing data: data are a tool for testing judgments, not a precondition for judgments to hold. On the day data become available, the falsification conditions preset by the present study will stand ready as the yardstick for testing these judgments.

Chapter Summary

This chapter has summarized the principal conclusions, policy implications, and limitations of the present study. The central conclusion of the present study is that the transferability of allocation experience depends on whether the institutional precondition is independently obtainable at the target tier; the five sets of judgments respectively characterize the five problem domains of governance vehicles, the risk-free anchor, accounting manifestation, behavioral architecture, and decumulation-phase absorption, and jointly point toward a policy sequence of first building institutional carriers and only then adjusting allocation parameters. The limitations of the present study lie mainly in the limited availability of micro-identification data, the temporary non-testability of several causal separations, and the difference in nature between aggregate corroboration and micro-identification; these limitations have all been faithfully flagged within the study, and point toward future research directions contingent on data openness and institutional reform.

Works Cited

Ministry of Finance. The Transfer of Partial State-Owned Capital to Enrich the Social Security Fund Has Been Basically Completed. Ministry of Finance of the People's Republic of China, 2021, http://m.mof.gov.cn/czxw/202101/t20210114_3643670.htm.

Ministry of Finance. Interim Measures for the Operation and Management of State-Owned Equity and Cash Income Transferred to Enrich the Social Security Fund (Cai She [2024] No. 10). Chinese Government Website, 2024, https://www.gov.cn/zhengce/zhengceku/202403/content_6941042.htm.

Ministry of Finance, State Taxation Administration. Announcement on Individual Income Tax Policies Relating to Individual Pensions (Announcement No. 34 [2022] of the Ministry of Finance and the State Taxation Administration). Ministry of Finance of the People's Republic of China, 2022, http://szs.mof.gov.cn/zhengcefabu/202211/t20221104_3849781.htm.

National Financial Regulatory Administration. Notice on Matters Relating to Establishing a Mechanism Linking and Dynamically Adjusting Predetermined Interest Rates to Market Interest Rates. National Financial Regulatory Administration, 2024, http://www.eeo.com.cn/2025/0721/739907.shtml.

Malkiel, Burton. A Random Walk Down Wall Street (11th Edition). Translated by Zhang Wei, China Machine Press, 2018. A classic work on long-term asset allocation and market efficiency.

Standing Committee of the National People's Congress. Decision on Implementing a Gradual Delay in the Statutory Retirement Age. Chinese Government Website, 2024, https://www.gov.cn/yaowen/liebiao/202409/content_6974294.htm.

National Council for Social Security Fund. Annual Report of the National Council for Social Security Fund (2023) and Annual Report on the Entrusted Basic Pension Insurance Fund. National Council for Social Security Fund, 2024, https://www.ssf.gov.cn/portal/xxgk/fdzdgknr/cwbg/sbjjndbg/webinfo/2024/10/1730355860076084.htm.

Ministry of Human Resources and Social Security. Measures for Occupational Annuities of Government Agencies and Public Institutions. Ministry of Human Resources and Social Security of the People's Republic of China, 2015, http://www.mohrss.gov.cn/.

Ministry of Human Resources and Social Security. Participation of Flexibly Employed Persons in Basic Pension Insurance for Employees. Chinese Government Website, 2023, https://www.gov.cn/yaowen/liebiao/202305/content_6855917.htm.

Ministry of Human Resources and Social Security. Summary of Enterprise Annuity Fund Business Data for 2024. Ministry of Human Resources and Social Security of the People's Republic of China, 2025, https://fund.eastmoney.com/a/202504023363840432.html.

Securities Times. The Individual Pension System Is Fully Rolled Out: Over 70 Million Accounts Opened, with an Actual Contribution Rate of About 20 Percent. Securities Times Online, 2024, https://www.stcn.com/article/detail/1426865.html.

Ministry of Civil Affairs of the People's Republic of China. Statistical Bulletin on the Development of Civil Affairs in 2024 and Data on Aging Programs. Ministry of Civil Affairs of the People's Republic of China, 2024, https://finance.sina.com.cn/roll/2024-12-14/doc-inczmtur4654080.shtml.

Anantharaman, Divya, and Elizabeth Chuk. "The Economic Consequences of Accounting Standards: Evidence from Risk-Taking in Pension Plans." The Accounting Review, vol. 93, no. 4, 2018, pp. 23–51.

Bank for International Settlements. Annual Economic Report 2023. BIS, 2023, https://www.bis.org/publ/arpdf/ar2023e.htm.

Novy-Marx, Robert, and Joshua Rauh. "Public Pension Promises: How Big Are They and What Are They Worth?" The Journal of Finance, vol. 66, no. 4, 2011, pp. 1211–1249.