Alternative Assets and Globalized Allocation in Mature Pension Markets Under Low Interest Rates: A Comparison Based on Canada, Australia, Japan, and the Netherlands

Chapter 1 Introduction

1.1 Statement of the Problem

Since the end of the twentieth century, real interest rates in major developed economies have declined persistently, and after the 2008 global financial crisis they entered an unusually prolonged period of low and even negative interest rates. For pension institutions whose core obligation is long-term benefit promises, this interest-rate environment constitutes a fundamental challenge. Traditional asset allocations centered on sovereign bonds and listed equities have seen their expected returns decline systematically once the risk-free rate was compressed; at the same time, the present value of liabilities for most defined benefit (hereafter DB) plans has risen as discount rates have fallen. This bidirectional squeeze — contraction on the return side and expansion on the liability side — has forced mature pension funds to seek new sources of risk premium while honoring existing payout commitments.

Against this backdrop, so-called alternative assets — private equity, infrastructure, private credit, unlisted real estate — together with cross-regional global diversification, have become the primary avenues through which mature pension funds have rebuilt their sources of return. Large pension funds in Canada, Australia, and the Netherlands have, over the past two decades, raised the share of alternative assets to a substantial proportion of their portfolios and established operating models characterized by internalized direct investment; Japan's Government Pension Investment Fund (hereafter GPIF), under the pressure of ultra-low interest rates, has been compelled to diversify globally, but its allocation to alternatives remains tightly bound by institutional authorization. The public annual reports of these institutions show that alternative assets have shifted from a marginal, diversifying supplement to a substantive component of portfolio return structure.

This phenomenon gives rise to the core questions of this paper. First, has low interest rates indeed driven a systematic migration of pension funds from traditional to alternative assets? Second, do alternative assets genuinely raise the return and diversification of the portfolio, or do they merely raise risk exposure — how much of the apparent improvement in the Sharpe ratio is a true premium, and how much is a statistical illusion produced by valuation smoothing? Third, what differences exist among the Canadian, Australian, Japanese, and Dutch models in governance structure and allocation depth, and under what constraints does each hold, and under what shocks does each fail? Fourth, what governance capabilities does alternative investing by pension funds require, and are these capabilities a precondition for allocation or a consequence of it? Fifth, does alternative-asset allocation give rise to valuation lag, liquidity mismatch, and governance complexity that, under stress, erode the very return advantage it claims to deliver?

1.2 Gaps in the Literature and the Limits of the Existing Narrative

Existing discussion of alternative allocation by mature pension funds can broadly be divided into two categories. The first is institutionally descriptive comparative research, which introduces, country by country, the organizational form, asset structure, and governance arrangements of pension funds, and on this basis summarizes a so-called mature-market experience. The value of this literature lies in the rich institutional detail it has accumulated, but its limitations are equally evident: it tends to line up the positive features of a given model in parallel, without probing the conditions under which that feature fails to hold or breaks down. When such research writes the high alternative-asset share of mature markets as a one-dimensional summary of advanced experience, it effectively treats the act of allocation itself as proof of capability.

The second category is quantitative portfolio-performance research, which attempts to use historical return data to assess the contribution of alternative assets to a portfolio's risk-return profile. This body of research faces a systematic methodological hazard: private-asset valuations are mostly based on lagged book values set quarterly by valuation committees, rather than on continuous market pricing (Institutional Investor 2022). If volatility, correlation, and Sharpe ratios are calculated directly from un-desmoothed private-return series, the low volatility produced by valuation smoothing will be misread as genuinely low correlation, systematically overstating the diversification benefit of alternative assets. Moreover, counting only surviving funds while ignoring liquidated and failed samples introduces further survivorship bias. Existing quantitative research has often failed to adequately address these three problems — net-of-fee accounting, desmoothing, and survivorship filtering.

A more fundamental gap lies at the level of causal structure. The implicit logic of the existing narrative is often: a given country allocates a higher proportion to alternative assets, hence its portfolio is more advanced, and is therefore worth learning from and replicating. This reasoning treats governance capability as an exogenously given background condition, rather than as an endogenous variable requiring explanation. What truly deserves scrutiny is this: is the high alternative-asset share of Canada's large funds the cause of their independent governance, internalized investment capability, and tolerance for long-horizon evaluation — or is it the consequence? If governance capability is a precondition for allocation depth rather than a consequence of it, then treating a high alternative-asset share itself as a replicable success paradigm constitutes a reversal of cause and effect. It is precisely on this gap that the present paper is built.

1.3 The Paper's Core Judgment and Contributions

The focal judgment of this paper concerns the endogeneity of governance capability. The causal direction this paper advances runs contrary to the paean to advanced experience: it is not that allocating to alternative assets makes a fund advanced, but that sufficiently strong governance capability is what allows a fund to dare to hold, and to deserve, alternative assets. The high alternative-asset share of large funds in mature markets is the consequence of their independent governance, internalized investment capability, and tolerance for long-horizon evaluation — not the cause.

Under this overarching judgment, the paper's specific contributions are three. First, the paper advances the somewhat general conclusion of governance-capability endogeneity into a set of more falsifiable, deeper propositions. The paper argues that investment, valuation, and risk-control capabilities are themselves intermediate goods tradable in factor markets — they can be hired, they can be divested, and they can be capped by political tolerance; what is truly endogenous and non-transferable is the allocation of governance control rights over capability — who holds decision rights over project origination, who holds veto rights, and whose authorization cannot be revoked by a political process. The moat is not capability, but control rights. This distinction separates, at the conceptual level, the replicable component from the non-replicable one, providing a more precise basis for judgments of transferability.

Second, the paper systematically incorporates three issues that existing research has often avoided — valuation smoothing, fees, and survivorship bias — into a critical examination of the true worth of the alternative premium. The paper argues that a considerable portion of the true premium displayed by mature markets is a statistical illusion produced by valuation smoothing; book-value smoothing has demand-side value for plans with weaker governance, and its speed of error correction rises with governance quality. At the level of private-credit instruments — the hardest to monitor — what institutions are buying is a product of actively manufactured low default rates, not a genuine reading of asset quality.

Third, the paper organizes the comparison of the four models as a variable-controlled examination revolving around a single judgment, rather than as four parallel expository accounts. The Canadian model represents the capability-ceiling sample of internalized direct investment; the Australian model tests whether capability can grow endogenously with scale; the Japanese model is a counter-example in which the desire to allocate is constrained by authorization limits that prevent heavy positioning; the Dutch model is a constraint sample in which alternative investment is subordinated to asset-liability management rather than to reaching for yield. The comparison of the four ultimately converges on a single mapping from governance-capability conditions to feasible alternative-allocation depth, specifying under what shock each model fails. The paper's endpoint is therefore the boundary of transferability of the Canadian model — a judgment that is the technical precondition for subsequent research on whether late-developing markets can learn from the Canadian model.

1.4 Methodological Stance and an Honest Declaration of Identification Boundaries

This paper holds a clear methodological stance: the core of economic research is the sharpness of judgment and the clarity of causal structure, not the scale of data. On this basis, for every load-bearing proposition, the paper specifies its identification strategy and falsification threshold prior to any encounter with data — that is, it fixes in advance, before seeing the data, what result would support the judgment and what result would falsify it, so as to preclude ex post rationalization.

At the same time, the paper makes an honest declaration of its own identification boundaries. The micro-level identification data this paper would require — including fund-level panels pairing governance characteristics with returns, panels of private-credit defaults and extensions, desmoothed private-return series, secondary-market discount series, and the final realized losses at fund liquidation — are not available within the scope accessible to this research. Accordingly, wherever this paper labels a proposition as identifiable within an identification design, the identification design is in place, but the causal coefficient is not estimable within the research environment available here. The paper therefore strictly distinguishes two levels: first, publicly verifiable institutional facts and institution-level aggregate facts, used to calibrate the direction of judgment; second, micro-level causal signs, which this paper does not treat as identified causal conclusions extrapolated from aggregate facts. Wherever the body of the paper cites verifiable institution-level or event-level aggregate facts, it marks the public source and year in an in-text parenthetical; wherever a given proposition serves a specific falsifiable judgment, the corresponding section states this binding relationship through a natural argumentative back-reference. This distinction is not a retreat from judgment but an expression of identification honesty — an honestly labeled identification boundary is of greater value to policy judgment than an overreaching claim disguised as proven causation.

1.5 Structure of the Paper

The remainder of the paper is organized as follows. Chapter 2 sets out the institutional background and alternative-allocation landscape of the four countries' pension funds, bringing the liability structures and asset allocations of Australia, Japan, and the Netherlands into the comparative frame. Chapter 3 establishes the paper's theoretical framework, clarifying the endogeneity of governance control rights and the single political constraint that runs through both the asset and liability sides. Chapters 4 through 8 each develop one of five propositions in a dedicated chapter, each organized in the sequence of judgment, mechanism, identification strategy, evidentiary status, and summary. Chapter 9 offers a cross-chapter synthesis, stitching the five propositions into a single whole and comparing the failure conditions of the four models. Chapter 10 is the conclusion, discussing policy implications and the limitations of this paper.

Chapter 2 Institutional Background: The Liability Structures and Alternative-Allocation Landscapes of Four Countries' Pension Funds

This chapter sets out the institutional background and alternative-asset allocation landscape of pension funds in four countries, providing the empirical coordinates for the theoretical analysis that follows. The selection of the four models is not an arbitrary geographic juxtaposition but a comparative framework constructed around the judgment of governance-capability endogeneity: Canada represents the capability ceiling, Australia tests the endogenous growth of capability with scale, Japan is a counter-example of constrained authorization, and the Netherlands is a constraint sample dominated by liability constraints.

2.1 The Canadian Model: Internalized Direct Investment under Independent Governance

Canada's large public pension funds, with their governance independence and internalized direct-investment capability, constitute what this paper calls the capability-ceiling sample. Take CPP Investments (Canada Pension Plan Investment Board) as an example: this institution was established under the 1997 CPP Investment Board Act, operates independently of the Canada Pension Plan itself, maintains an arm's length relationship with the federal and provincial governments, and is governed by an independent professional board. In its key governance arrangements, the Act mirrors the model of the Ontario Teachers' Pension Plan (OTPP), established in 1989 (World Bank 2018), which was Canada's first arm's-length, professionally operated pension fund. The board approves investment policy, appoints and removes the chief executive officer, and sets management compensation (CPP Investments 2024b; World Bank 2018). This governance architecture gives the Canadian model, at the institutional level, one pole of what this paper subsequently calls the independence of control-rights allocation.

In asset allocation, the Canadian model is characterized by a high proportion of private and real assets. As of fiscal year 2024, CPP Investments' net assets reached C$632.3B (CPP Investments 2024a); within its strategic portfolio, private equity accounts for approximately 23%, real assets approximately 26%, and credit approximately 14% (CPP Investments 2024a). This allocation structure shows that alternative assets already constitute the core of the portfolio in the Canadian model, rather than a mere supplement. It is also worth noting that, to operate this portfolio, CPP Investments paid approximately C$1,449M in external investment management fees in fiscal year 2024, plus approximately C$2,067M in performance fees (CPP Investments 2024a). The scale of these external fees corroborates, from one angle, a point this paper will later argue: investment capability itself is an intermediate good that can be priced and hired in factor markets.

It should be noted that the governance independence of the Canadian model is not an irrevocable institutional constant. In 2024, the government of Alberta dissolved the entire board of the Alberta Investment Management Corporation (AIMCo), an event widely interpreted in the industry as a case of political interference eroding the independence of the Canadian model (Mergers and Inquisitions 2024). This real-world event suggests that even the most independent arm's-length institutions may have their control-rights authorization revoked by a political process. This observation bears directly on the proposition of a cap on political tolerance that Chapter 3 of this paper will establish.

2.2 The Australian Model: Late-Developing Catch-Up in Internalized Capability Driven by Scale

Australia's pension system is dominated by defined contribution (hereafter DC) plans, which have accumulated a vast pool of assets through the mandatory superannuation system, and which operate under the supervision of the Australian Prudential Regulation Authority (APRA). Unlike the historical evolution path of the Canadian model, Australia's large superannuation funds (such as AustralianSuper and other member-account-based funds) have gradually built internalized investment capability in the course of rapid scale expansion, and thus constitute the sample for testing whether capability can grow endogenously with scale. A potentially confusing classification must be clarified here: Australia has a separate, genuine sovereign wealth fund — the Future Fund. The trend recorded in public reporting — internalizing the management of domestic infrastructure and real estate and increasing allocation to alternatives — pertains to this Future Fund (Markets Group 2024), not to the superannuation funds; the Future Fund carries no pension liabilities and therefore does not belong to the DC superannuation system discussed in this section. As a liability-free sovereign fund, the Future Fund falls, on the dimension of liability hardness, into the category discussed in Chapter 8 of this paper — "no liabilities, absent commitment device" — and its capacity implications are discussed there; this section does not conflate it with the DC superannuation funds as part of the same "Australian model" sample.

The analytical value of the Australian model lies in pushing the distinction between capability and control rights to a critical test. On one hand, scale expansion has indeed provided a material basis for the growth of internalized capability: a sufficiently large asset pool can amortize the fixed costs of building an in-house investment team, making internalization economically feasible, so that investment capability can grow endogenously as a partial consequence of scale. On the other hand, the defined-contribution institutional form imposes two constraints on this growth. The first is a liability-side constraint: under defined contribution, the ultimate investment risk is borne by the individual account holder, who in principle retains the right to switch between funds; this transferability weakens the commitment device that hard liabilities provide, making the liability hardness of the Australian model weaker than the DB traditions of Canada and the Netherlands. The second is a constraint at the level of control rights: scale can buy capability, but it does not necessarily and automatically secure control rights immune to political revocation — Chapter 3 of this paper will argue that control rights are rooted in institutional-level authorization arrangements, not a function of the scale of capital. The Australian model therefore offers a crucial vantage point: what it tests is not whether capability can grow with scale (which largely holds), but whether weak liability hardness combined with scale-driven control rights can sustain an illiquid allocation as deep as Canada's. This paper's judgment is that scale can bring capability, but cannot substitute for a commitment device and control rights — and it is precisely this distinction that marks the divergence between the Australian and Canadian models.

2.3 The Japanese Model: Globally Diversified Yet Authorization-Constrained under Ultra-Low Interest Rates

Japan's GPIF is one of the largest public pension funds in the world, and under Japan's long-standing ultra-low interest-rate environment, it has been forced to diversify globally in order to sustain the portfolio's expected return. However, GPIF's allocation to alternative assets is tightly constrained by institutional authorization: the statutory ceiling for its alternative assets is 5% of total assets, while in fiscal year 2024, the actual share of alternative assets was only about 1.63% (GPIF 2024). At the same time, GPIF has established an alternative-asset database and has been gradually expanding its alternative exposure (Chief Investment Officer 2023), indicating a gap — created by authorization constraints — between its willingness to allocate and its actual capability.

The Japanese model thus constitutes a crucial counter-example: it is an institution that wishes to allocate, and possesses ample capital scale, yet cannot take heavy positions in alternatives owing to governance and authorization constraints. GPIF's authorization ceiling is an important reference point for understanding this paper's core judgment — even the world's largest public pension fund has its capacity for illiquid assets compressed by authorization (an institutionalized expression of political tolerance), with the capacity boundary decoupled from the scale of capital.

The counter-example significance of the Japanese model lies precisely in the fact that it falsifies the apparent positive correlation between capital scale and allocation depth. If a high alternative-asset share were truly determined solely by capital scale or market development, then GPIF, as the world's largest public pension fund, ought to rank at the top in allocation depth; but the opposite is true — GPIF's actual alternative-asset share is the lowest among the four countries. This contrast pushes the explanatory variable away from capital scale and toward governance and authorization: GPIF's operating structure, which relies primarily on external mandates, together with its 5% statutory ceiling, together constitute an authorization ceiling that, even under the pressure of ultra-low interest rates and a strong desire to allocate, prevents it from raising its alternative exposure to a level that would substantively change the portfolio's return structure. At a deeper level, this authorization ceiling is not a purely technical risk limit but an institutionalized cap that political tolerance places on the bearing of illiquid, opaque risk by public funds — it specifies the degree to which a public trustee institution is permitted to deviate from a traditional, easily accountable asset structure. Chapter 8 of this paper will argue that this phenomenon is the projection of the political-tolerance constraint onto the liability side and the authorization side, and that it is, in fact, the same constraint manifesting itself a second time, alongside the asset-side control-rights ceiling revealed by the AIMCo event in Canada. Japan and Canada thus form a pair: the former caps its willingness to allocate through an ex ante statutory ceiling shaped by political tolerance, while the latter threatens existing control rights through an ex post board dismissal shaped by political tolerance — but the two constrain the same variable.

2.4 The Dutch Model: Asset-Liability Management under the Dominance of Liability Constraints

The Netherlands' pension system has traditionally been dominated by collective defined benefit plans, allocating assets under strong liability constraints, with its alternative allocation subordinated to the logic of asset-liability management (ALM) rather than pure reaching for yield. The Dutch model therefore constitutes a constraint sample. In recent years, driven by the Dutch pension reform (Wet toekomst pensioenen, hereafter WTP), the Netherlands' large pension funds have been significantly expanding their allocation to illiquid assets. Taking APG — the Dutch asset manager that invests on behalf of ABP, the largest sectoral pension fund — as an example, it plans to raise its private-market allocation from about 26% to over 30%, with real estate at about 10%, infrastructure rising from 5%–6% to a target of 10%, private equity at about 8%, and private credit rising from about 1.5% to a target of 2%–4%; ABP's alternative-asset target is approximately 28% (IPE 2024). The Dutch pension reform is driving the transition of collective defined benefit plans toward individualization (PGIM 2024), and this institutional change itself is a real-world instantiation of the category — institutionally exogenous sources of liability hardness — that this paper discusses later.

The distinctive significance of the Dutch model is that it brings to the fore the constraint that liability structure places on feasible alternative-allocation depth. Within the framework of collective defined benefit, a hard, predictable liability constitutes a commitment device that locks in capital and strips away the right of early redemption, thereby enabling deeper illiquid allocation. The WTP reform's shift toward individualization, however, may loosen this commitment device. Chapter 8 of this paper will use the Dutch model as an example to argue that liability hardness and governance are co-endogenous, both rooted in the same political constraint.

2.5 Summary of Institutional Background

Taken together, the institutional backgrounds of the four countries show that the alternative-allocation landscape of mature pension funds exhibits significant cross-sectional differences: Canada supports a high proportion of internalized direct investment through independent governance; Australia is catching up in internalized capability amid scale expansion; Japan's allocation is constrained by authorization limits; and the Netherlands' alternative allocation is subordinated to liability constraints. These differences cannot be explained by capital scale or market development alone — Japan is precisely the sample with the largest capital scale yet the lowest alternative-asset share. The explanatory variable behind these differences points to governance structure and authorization arrangements — that is, what this paper calls the allocation of control rights and political tolerance. The institutional facts set out in this chapter provide the empirical coordinates for the theoretical framework of the next chapter; the comparison of the four models will converge, in Chapter 9, into a matrix of failure conditions.

Chapter 3 Theoretical Framework: The Endogeneity of Governance Control Rights and the Single Political Constraint

This chapter establishes the theoretical framework of this paper. At the core of the framework is the reconstruction of the broad conclusion that governance capability is endogenous into two more precise propositions: first, what is truly endogenous is not capability but the allocation of control rights over capability; second, what acts on control rights (the asset side) and what acts on liability commitments (the liability side) is one and the same political tolerance constraint. These two propositions provide the unified theoretical foundation for the chapter-by-chapter arguments of the following five chapters.

3.1 From the Transmission Mechanism to the Governance Precondition

The starting point of this paper's analysis is the mechanism chain through which low interest rates transmit into alternative-asset allocation. This chain can be stated as follows: the decline in the risk-free rate lowers the expected returns of traditional portfolios; the return side therefore requires a new source of risk premium, and allocation consequently shifts toward alternative assets and global positioning. This transmission chain has already been fully argued in existing research on interest-rate mechanisms; this paper accepts its conclusions without repeating them, and instead focuses on the terminal end of the chain — what, exactly, is the threshold for carrying alternative assets.

The first theoretical judgment of this paper is: the threshold for carrying alternative assets is not whether investment capability can be acquired, but who controls that capability. Investment, valuation, and risk-control capabilities are intermediate goods that can be priced and hired in the factor market. An institution can wholesale-recruit a top investment team from outside, and can purchase the most advanced valuation and risk-control systems. Capability itself therefore does not constitute a moat — on the contrary, capability can be replicated and copied. What truly determines whether the alternative premium can be captured is the allocation of control rights: to whom does the decision-making power over deal origination belong; is the veto power held by an internal investment committee or by a politically appointed body; and can the institution's mandate be revoked through budgetary interference or political process.

3.2 Control Rights as an Endogenous Variable: Definition and Operationalization

This paper defines control rights as a vector composed of three institutional arrangements: first, whether the chain of board appointment and removal must go through a political process; second, whether veto power and deal-origination decision rights are, as a matter of charter, legally vested in an internal investment committee or in a politically appointed body; third, the historical incidence of mandate revocation and budgetary interference. These three dimensions can each be coded independently from institutional charters, relevant legislative texts, and governance event histories, and their common feature is that they can be measured independently of performance outcomes.

The operationalization here is not an optional technical detail, but is critical to whether this paper's theoretical framework can hold. An obvious objection is: if control rights are defined as the residual that cannot be copied and that determines success or failure, then any replicable factor, once shown to be replicable, is removed from the definitional domain of control rights, while any non-replicable residual is named control rights — rendering the proposition tautologically true and therefore unfalsifiable. This objection is sharp, and this paper confronts it directly. This paper's response is: to establish for control rights a pre-performance, independently codable proxy vector that is orthogonal to performance — namely the three institutional arrangements above — and to require that these proxies be measured prior to the observation of performance. Only in this way can the circular reasoning of inferring the residual from performance and then explaining performance by the residual be cut off. If, in the end, it proves impossible to establish a proxy vector for control rights that is independent of performance, this paper explicitly acknowledges that the claim of endogenous control rights will degenerate into a definitional classification framework that does not bear the task of causal identification. This honest boundary runs through the entire paper.

3.3 Capability Can Be Outsourced, but Control Rights Cannot Be Transferred

Under the definition above, this paper arrives at a counterintuitive judgment: capability can be outsourced, but control rights cannot be transferred. A buyer can wholesale-poach an investment team, achieving the transfer of capability; but the buyer cannot transfer, along with it, a mandate that is immune to political revocation. The robustness of a mandate is rooted in a specific institutional and political environment, and it does not transfer along with the movement of personnel. This distinction is central to understanding the transferability of the Canadian model: a late-developing market can learn capability, but it cannot learn control rights.

From this a further, equally counterintuitive proposition follows: stronger capability does not necessarily mean more direct investment. The prevailing assumption is that a fund that manages better must necessarily do more internalized direct investment, such that a high internalization rate is equated with strong capability and with being advanced. This paper holds that this monotonic assumption is mistaken. Given control rights, the rational choice of a high-capability institution is often to do less direct investment at the peak of the cycle — it treats the internalization rate as a timing variable rather than as a monotonic function of capability. Reading a high internalization rate directly as strong capability, as advanced, is therefore a reversal of causality. This proposition is carefully positioned within this paper's evidentiary system as a theoretical proposition; its identification awaits a cross-country comparable panel of internalization rates, and this paper does not disguise it as an already-identified conclusion.

3.4 Political Tolerance: A Single Constraint Running Through Both the Asset and Liability Sides

The second pillar of this paper's theoretical framework is the unity of the political tolerance constraint. Control rights are capped by political tolerance: when veto power is held by a politically appointed body, and when the mandate can be revoked, then even if the institution has hired a top team, its capability is capped by political tolerance, and the institution can only obtain exposure to the risk of alternative assets, without being able to capture the premium. This is the projection of the political constraint onto the asset side.

But the same political constraint also acts on the liability side. A hard, predictable fixed-benefit liability, which locks in funds and strips away the right of early redemption, constitutes a commitment device that enables the institution to carry deeper illiquid allocations. However, this commitment is fragile in a procyclical way: when interest rates rise or funding ratios change, the duration used for performance assessment endogenously shortens, the redemption option is politically reactivated, and long-term capital becomes short-term during periods of stress. This is the same political constraint's projection onto the liability side.

This paper therefore emphasizes a unifying judgment: the political tolerance that acts on the capping of control rights and the political tolerance that acts on the activation of commitments are two projections of the same endogenous variable, rather than two independent, mutually orthogonal constraints. The asset side and the liability side are not two dual, opposing pillars, but two facets of the same central axis. This unifying judgment carries an honest methodological implication: treating the asset-side proposition and the liability-side proposition as two independent judgments to be added together would artificially inflate the density of judgments; this paper explicitly merges them into a single axis in order to avoid such inflation. Political tolerance is the single endogenous axis running through both the asset and liability sides — this formulation does not weaken the judgment but rather renders it more unified and more forceful.

3.5 The Logical Progression of the Five Propositions

Under the framework above, this paper's five load-bearing propositions form a structure of logical progression rather than a parallel enumeration. The first proposition establishes the backbone starting point that capability can be outsourced while control rights are endogenous. The second proposition, building on this, goes a step further, asking just how genuine the premium captured by capability really is, and reveals the valuation-smoothing component within it. The third proposition moves to the level of dynamic pricing, arguing that the stronger the capability, the faster the premium is drained away by reflexivity, so that capability is an accelerant rather than a moat. The fourth proposition brings the first three propositions down to the level of the private credit instrument — the hardest to supervise — for in-depth dissection, revealing manufactured non-default. The fifth proposition closes at the liability side, arguing that the true boundary of capacity is jointly determined by the hardness of liabilities and the procyclicality of patient capital, and shares the same political constraint with the first proposition. The five propositions are interlocked, together serving the overarching judgment of the endogeneity of governance capability.

3.6 Overview of the Identification Strategy and Status of the Five Propositions

To help the reader grasp the identification designs of the following five chapters and their honestly labeled identification status, this section first presents, in a summary table, the core identification leg, its exogenous source, and the current identification status for each of the five load-bearing propositions. For each proposition in the table, the identification design is in place, but its causal coefficient is not estimable within the range accessible to this study; the institutional facts listed in the table are used only to calibrate the direction of the judgment.

Table 1 Core identification legs, competing hypotheses, and identification status of the five load-bearing propositions

3.7 Summary of the Theoretical Framework

This chapter reconstructs the endogeneity of governance capability into two more precise propositions: what is truly endogenous is the allocation of control rights over capability, not capability itself; and what acts on control rights on the asset side and on commitments on the liability side is the same political tolerance constraint. This chapter also establishes for control rights a pre-performance proxy vector independent of performance, in order to avoid unfalsifiable circular reasoning, and honestly states that, should this vector prove impossible to establish, the relevant claims will degenerate into a classification framework. The following five chapters will, in accordance with this framework, take up the arguments for the five load-bearing propositions one by one.

Chapter 4 The Endogeneity of Governance Control Rights: Capability Can Be Outsourced, What Is Endogenous Is the Allocation of Control Rights over Capability

This chapter argues the first load-bearing proposition of this paper: what is endogenous and non-transferable in a pension fund's ability to carry alternative assets is not capability itself, but the allocation of control rights over capability. This chapter is organized in the order of judgment, mechanism, identification strategy, evidentiary status, and summary.

4.1 Judgment

The core judgment of this chapter is falsifiable: investment, valuation, and risk-control capabilities can be priced, hired, and divested in the factor market, and can be capped by political tolerance; what is truly endogenous and non-transferable is the allocation of control rights over capability — who holds the decision-making power over deal origination, who holds veto power, and whose mandate is not subject to political revocation. The argument of this chapter serves precisely this judgment, and the mechanism, identification, and evidence throughout the chapter are organized around it.

This judgment divides the correct but ambiguous conclusion that governance capability is endogenous into a deeper, falsifiable proposition. Its counterintuitive aspect lies in this: capability is precisely what can be copied, and what cannot be copied is control rights. The prevailing narrative treats capability as a moat, holding that possessing superior investment capability can sustainably protect the premium; this paper instead argues that capability is a tradable intermediate good, and that the moat is not capability but control rights. A closely related, almost unreflective monotonic assumption — that a fund that manages better must necessarily do more direct investment — is likewise mistaken: given control rights, the rational choice of a high-capability institution is often to do less direct investment at the peak of the cycle.

Control rights are exogenously capped by political tolerance, and require the freedom of countercyclical timing; these two, together with deal-origination decision rights, combine to form governance control rights as the genuinely endogenous variable. But this paper emphasizes an honest boundary: this proposition holds only if a pre-performance, independently codable proxy can be established for control rights. If control rights are defined as the residual that cannot be copied and that determines success or failure, the proposition becomes tautologically true and unfalsifiable. This chapter therefore compulsorily introduces a test of prior operationalization, and specifies a concrete downgrade-trigger condition: if the triple pre-performance proxy (whether the board appointment-and-removal chain goes through a political process, the charter-based allocation of veto power and deal-origination decision rights, and the historical incidence of mandate revocation), once coded, cannot be shown to have measurable independent variance separate from performance outcomes — that is, if the proxy vector cannot be assigned a value prior to observing performance, or if its values are so collinear with performance readings that they cannot be separated — then the claim of this chapter is explicitly downgraded to a definitional classification framework that does not bear causal identification, and its identifiable portion retreats to the politically-capped discontinuity leg, which stands independently.

4.2 Mechanism

The causal chain of this chapter is as follows. Low interest rates compress the returns of traditional portfolios; the demand side therefore needs a new source of risk premium, and allocation shifts toward alternative assets; the threshold for carrying alternative assets is not whether capability can be acquired (capability can be outsourced), but who controls that capability; the allocation of control rights determines whether the premium can be captured.

Specifically, when control rights are strong — when deal-origination decision rights and veto power reside with the internal investment committee, board appointment and removal does not go through a political process, and the mandate is not revoked through budgetary interference — the institution can time countercyclically, can decline transactions that have been commoditized at the tail end, and thereby capture a genuine premium. When control rights are weak — when veto power resides with a politically appointed body and the mandate can be revoked — then even if the institution has hired a top team, its capability is capped by political tolerance, and the institution obtains only exposure to the risk of alternative assets, without the premium. This is precisely what this paper emphasizes: an institution lacking governance independence, even if it replicates the same allocation structure, obtains only mature-market-style risk exposure, not a mature-market-style premium.

This mechanism contains two counterintuitive nodes. First, capability can be copied while control rights cannot be transferred: a buyer can wholesale-poach a team, achieving the transfer of capability, but cannot transfer along with it a mandate immune to political revocation. Second, high capability does not equal more direct investment: given control rights, high-capability institutions often do less direct investment at the peak of the cycle, treating the internalization rate as a timing variable rather than a monotonic function of capability; hence equating a high internalization rate with strong capability, with being advanced, is a reversal of causality. This latter node is positioned in this paper as a theoretical proposition.

4.3 Identification Strategy

The identification strategy of this chapter distinguishes three tiers — the primary strategy, the backup strategies, and the downgrade exit — and honestly labels the identification status of each tier.

Primary strategy — the politically-capped discontinuity (identification design in place). The main load-bearing identification leg of this chapter uses the effective date of a compensation-disclosure regulatory rule as a discontinuity, employing an event-time difference-in-differences estimation to examine whether the compensation competitiveness and retention rate of institutions subject to disclosure constraints show a step-like decline following the disclosure shock. Its source of exogenous variation lies in the fact that the effective date of this regulatory rule is exogenous to the condition of any single institution — the regulatory agenda is not equivalent to any single fund's circumstances. The treatment variable is whether an institution or compensation line is subject to disclosure constraints or not. The falsifiable placebo test is: transaction types or compensation lines not subject to this rule should not exhibit a discontinuity at the same point in time. The design of this identification leg is in place, and its exogenous source holds in principle.

Backup strategy — capability separability (theoretical proposition). The naive design of using the sale or wholesale lift-out of an internal investment team as a natural experiment, examining whether performance migrates with the original institution after capability changes hands, fails on grounds of endogeneity: the sale or lift-out of a capability team is not an exogenous shock but an endogenous choice — which teams get sold or poached is systematically correlated with the strength of the institution's control rights and its performance outlook. Therefore, the observation that the original institution's performance does not migrate after capability changes hands is at least as consistent with three alternative explanations equivalent to the observation of endogenous control rights: first, reverse causality, in which the original institution's governance collapses first, which is what causes the team to leave and performance to decline; second, selection bias, in which what gets sold off is precisely the codable capability that has already been commoditized and sits at the tail end of value, so its transferability is an artifact of sample selection and cannot be extrapolated to core deal-origination capability; third, omitted variables, in which the scale and synergy of the buyer's platform are the true source of the performance difference. Given that this paper's identification design contains no exogenous instrument that cuts through this selection function of why a team gets traded, this paper abandons treating capability transactions as an exogenous shock, explicitly acknowledges it as an endogenous choice, and labels the outsourcability and separability of capability as a theoretical proposition. A shock that is genuinely exogenous to the capability team — such as an antitrust-mandated forced divestiture, or a passive spin-off from a parent company's bankruptcy — supplemented by buyer-platform fixed effects, could potentially separate the outsourcability of capability from selection or reverse causality; samples of such passive divestiture events are sparse and unavailable within this research environment.

Backup strategy — timing freedom (theoretical proposition). Another backup strategy of this chapter is, after controlling for asset-side attractiveness, to examine whether the procyclical shift of the internalization rate with interest rates and the cycle is independent of the capability stock. However, the internalization rate, the capability stock, and the interest-rate cycle are highly collinear, and all must be manually constructed from annual reports; this is observationally equivalent to the competing hypothesis of outsourcing driven by the cycle, regulation, and scale, so this paper downgrades it to a theoretical proposition pending data.

Downgrade exit (specified in advance). This chapter specifies in advance the concrete observations that would trigger a downgrade, so as to prevent the downgrade exit from becoming an empty formality. The trigger condition is any one of the following verifiable circumstances: first, the pre-performance proxy vector for control rights cannot be assigned a value prior to observing performance (charters, legislative texts, and governance event histories are insufficient to independently code the triple proxy); second, although the proxy vector can be assigned a value, its values are highly collinear with performance readings across the four-country cross-section and independent variance cannot be separated out (i.e., the proxy does not satisfy the requirement of orthogonality with performance); third, the placebo test for the compensation-disclosure discontinuity fails — transaction types or compensation lines not subject to the rule also exhibit an equivalent discontinuity at the same point in time, indicating that the discontinuity is not driven by the disclosure constraint. If any one of these is triggered, this chapter is downgraded to a definitional classification framework that does not bear causal identification, and its identifiable portion retreats to the single leg of the compensation-disclosure discontinuity — a leg that does not depend on the control-rights proxy and stands independently. It must be honestly noted that, within the data environment accessible to this study, the raw panel for the compensation-disclosure discontinuity is itself unavailable (see the section on evidentiary status), so this fallback leg, under the current environment, provides a design-level identification commitment rather than an immediately realizable estimate; in other words, what the downgrade exit guarantees under the conditions of this study is the logical closure of the identification strategy, while its empirical realization likewise awaits the availability of the corresponding micro-level panel. In sum, the identification strategy of this chapter contains two legs downgraded to theoretical propositions (capability separability, timing freedom), and one leg with the identification design in place (the compensation-disclosure discontinuity).

4.4 The Strongest Counterarguments and Their Treatment

This paper directly confronts the two strongest counterarguments to the propositions of this chapter.

The first counterargument points to unfalsifiability. It argues that the step of retreating the endogenous variable from tradable capability to control rights manufactures an unfalsifiable tautology: control rights are defined as that which cannot be copied and which determines success or failure, so that any replicable factor, once shown to be replicable, is removed from the definitional domain of control rights, while any non-replicable residual is named control rights. The flagship capability-transfer test can only falsify the binding of capability, not the endogeneity of control rights: if performance migrates, this is called capability being outsourceable and control rights being what is endogenous; if performance does not migrate, this is called control rights being non-transferable — two opposite observations are absorbed by the same proposition. This paper's treatment consists of three parts: first, establishing for control rights a pre-performance, independently codable proxy vector orthogonal to performance — namely, whether the board appointment-and-removal chain goes through a political process, the charter-based allocation of veto power and deal-origination decision rights, and the historical incidence of mandate revocation — all three constructed from charters, legislation, and governance event histories, independent of performance outcomes, thereby cutting off the circularity of inferring the residual from performance; second, pre-registering two opposite-direction observations as falsification conditions (detailed in the next section); third, specifying a downgrade exit in advance — if it ultimately proves impossible to establish a pre-performance proxy vector independent of performance, this chapter is downgraded to a classification framework, and the focus of judgment is taken over by the single leg of the compensation-disclosure discontinuity.

The second counterargument points to reverse causality and identification. It argues that the treatment in the flagship natural experiment (a capability team being sold or poached) is not an exogenous shock but an endogenous choice, observationally equivalent to the three alternative explanations noted above, so this event can neither confirm nor falsify the control-rights proposition. This paper accepts this counterargument, abandons treating capability transactions as an exogenous shock, and instead turns to shocks genuinely exogenous to the capability team (antitrust-mandated forced divestiture, passive spin-off from a parent company's bankruptcy), absorbing performance differences with buyer-platform fixed effects; until such a source of exogenous passive divestiture is obtained, the outsourcability and separability of capability is labeled a theoretical proposition, and the main text does not present it as an already-identified natural experiment.

4.5 Pre-Registered Expectations and Falsification Thresholds

This paper specifies the falsification thresholds for this chapter prior to any data analysis, so as to prevent post-hoc rationalization.

The pre-performance proxy vector for control rights is: whether the board appointment-and-removal chain goes through a political process; the charter-based legal allocation of veto power and deal-origination decision rights; and the historical incidence of mandate revocation and budgetary interference. All three must be measured prior to observing performance. The proposition of this chapter is falsified if and only if either of the following occurs: first, the control-rights proxy is strong but alternative-asset performance does not rise accordingly (the pre-performance proxy has no explanatory power); second, control rights are transferred along with a change in institutional control and performance migrates in tandem (in which case control rights are not, in fact, non-transferable).

On the side of the compensation-disclosure discontinuity, the pre-registered expectation is: following the disclosure shock, the compensation competitiveness and retention rate of constrained institutions show a step-like decline relative to unconstrained institutions; if no discontinuity exists, or if the discontinuity is absorbed by a contemporaneous common shock, then the politically-capped mechanism does not hold on this leg.

4.6 Evidentiary Status

This section faithfully reports the evidentiary status of this chapter within the research environment accessible to this study, strictly distinguishing verifiable institutional facts from unobtainable micro-level causal estimates.

At the level of verifiable facts, several institutional facts are consistent with the direction of this chapter's judgment. The governance independence of the Canadian model is an institutional fact rather than a performance reading: CPP Investments was established under legislation enacted in 1997, operates at arm's length from the government, and is governed by an independent board of directors, which approves investment policy, appoints and removes the chief executive officer, and sets management compensation (CPP Investments 2024b; World Bank 2018). The capping of control rights by political tolerance is borne out by a real event of political revocation: in 2024 the government of Alberta dissolved the entire board of AIMCo (Mergers and Inquisitions 2024), providing a genuine governance-event observation for two of the pre-performance proxies — whether the board appointment-and-removal chain goes through a political process, and the historical incidence of mandate revocation — namely, that even a top-tier arm's-length institution's control-rights mandate can still be revoked through a political process. That capability is outsourced at factor-market prices and can be separated from external managers is likewise corroborated institutionally: in fiscal year 2024, in order to generate C$46.4B in net income, CPP Investments paid external investment management fees of C$1,449M and performance fees of C$2,067M (CPP Investments 2024a) — that is, capability is explicitly priced in the factor market and can be hired and divested, consistent with this chapter's judgment that capability is a tradable intermediate good.

At the level of micro-level causal estimates, the identification legs of this chapter have causal coefficients that are not estimable within the range accessible to this study. The difference-in-differences identification design for the compensation-disclosure discontinuity is in place, but the panel of proxies for constrained institutions' compensation and retention must be manually constructed from annual reports and disclosure documents; its raw panel is unavailable due to data-availability constraints, so the core coefficient of this discontinuity is not estimable. The test of the pre-performance proxy vector for control rights requires self-constructed coding of charters and legislative texts and a governance event history, and a comparable scale across the four countries is likewise unavailable, so whether the pre-performance proxy predicts premium capture prior to and independently of performance cannot be tested; the AIMCo event provides only a single-point qualitative observation and does not constitute a panel. The samples of passive divestiture events required for capability separability are similarly sparse and unavailable, so it retains the status of a theoretical proposition.

It must be emphasized that the verifiable facts above serve only to calibrate the direction of judgment, and are by no means to be extrapolated as the sign of a micro-level causal estimate. The scale of CPP's external fees corroborates that the fee wedge genuinely exists, but does not constitute a causal test of the endogeneity of control rights; the AIMCo event supports the direction of the pre-performance proxy, but neither falsifies nor confirms the control-rights proposition. Accordingly, the falsification condition of this chapter has, within the range accessible to this study, been neither triggered nor ruled out, and the judgment maintains the status of a theoretical assertion with attached falsification conditions, and is not disguised as already identified. This is an honest boundary of identification, not a retreat of judgment.

4.7 Chapter Summary

This chapter argues that the endogenous threshold for a pension fund to carry alternative assets is not capability but control rights. Capability can be priced, hired, and divested in the factor market, and is capped by political tolerance; what is truly endogenous and non-transferable is the allocation of control rights over capability. This chapter establishes for control rights a triple pre-performance proxy vector independent of performance, so as to avoid unfalsifiable circularity, and honestly labels capability separability and timing freedom as theoretical propositions. In terms of evidentiary status, institutional facts such as Canadian governance independence, the AIMCo political-revocation event, and the scale of CPP's external fees calibrate the direction of the judgment, but the relevant micro-level causal coefficients are not estimable within the range accessible to this study. The identifiable portion of this chapter honestly contracts to the identification design of the compensation-disclosure discontinuity and the directional calibration provided by institutional facts. The capability–control-rights distinction established in this chapter is the starting point for the next chapter's inquiry into the genuineness of the premium.

Chapter 5 The Watered-Down True Premium via Valuation Smoothing: Part of the Diversification Return Is Just Outsourcing Book Volatility to a Smoothing Mechanism

This chapter argues the second load-bearing proposition: a considerable portion of the true premium exhibited by mature pension markets is a statistical artifact of valuation smoothing; book-value smoothing has demand-side value for plans with weaker governance, and its error-correction speed increases with governance quality. This chapter builds one further level on Chapter 4's finding that capability can be outsourced — even the premium captured by capability itself turns out to be watered down.

5.1 Proposition

The core proposition of this chapter is falsifiable: a considerable slice of the true premium exhibited by mature markets is a statistical artifact of valuation smoothing; book-value smoothing has demand-side value for plans with weaker governance, and its error-correction speed increases with governance quality; whether this behavior stems from an intent of active collusion, or even a willingness to pay a negative premium for book-value smoothing, remains a hypothesis awaiting pre-registration, to be tested via an identification design with separation — absent that, what this chapter can identify is confined to the difference in error-correction speed along the governance gradient. The mechanism, identification, and evidence in this chapter all serve this conditional proposition, without making a settled commitment regarding unobservable intent.

This proposition builds on and deepens Chapter 4. Chapter 4 established that capability can be outsourced; this chapter further shows that even the premium captured by capability itself is watered down: valuation smoothing in private assets suppresses realized volatility and correlation, systematically inflating Sharpe ratios and diversification metrics — this is a quantifiable accounting-fact layer. Further, book-value smoothing has demand-side value for pension plans with weaker governance — under pressure from funded-ratio assessment on the liability side, and out of a motive to avoid triggering rebalancing via realistic marking-to-market, such plans may actively choose low-transparency instruments, and may even be willing to pay a negative premium for book-value smoothing. This behavior is exposed when forced to liquidate: the secondary-market discount is the error-correction clock of truth, and governance quality determines whether that clock is allowed to run (fast correction) or is stopped (delayed correction).

This chapter particularly emphasizes an honesty boundary. The core of this chapter is not a universal claim asserting that pension funds as a class are all colluding buyers — that would conflict with the clean identification leg showing that better governance corresponds to faster discount convergence. This chapter converges its claim into a conditional proposition: the degree of collusion-seeking smoothing decreases with governance quality. At the same time, the intent of active collusion is not asserted as settled, but is downgraded to a hypothesis awaiting pre-registration. This convergence flips the identification leg — better governance, faster correction — from an observation that would weaken this chapter's claim into the positive identification leg for this chapter's conditional proposition.

5.2 Mechanism

The causal chain of this chapter is as follows. Private-asset valuation uses lagged book net asset value or appraised value, which artificially suppresses realized volatility and cross-sectional correlation, thereby inflating Sharpe ratios and diversification returns, so the true premium contains a smoothing component. Plans with weaker governance need this right to invisible volatility because of funded-ratio assessment, and so actively choose low-transparency instruments, possibly paying a negative premium. When institutions are forced to liquidate, the secondary-market discount exposes the truth, and governance quality determines whether the error-correction clock runs fast or slow.

This mechanism contains three counterintuitive nodes. First, smoothing is not a pure accounting flaw but has positive demand-side value for plans with weaker governance — it buys such plans the right to invisible volatility, and is thus an actively demanded product rather than passively accepted noise. Second, the profile of pension funds reverses along the governance dimension: those with weaker governance tend toward smoothing and resist correction, while those with stronger governance allow the secondary-market discount clock to run. Hence "pension funds are all colluding to seek smoothing" is a mistaken universal claim; the truth is a decreasing function along the governance gradient. Third, part of the diversification return is the outsourcing of book volatility to a smoothing mechanism, rather than genuinely low correlation.

5.3 Identification Strategy

This chapter's identification strategy likewise distinguishes tiers and reports status faithfully.

Primary strategy — desmoothed residual (identification design in place). The first identification leg of this chapter is to desmooth private-asset return series (using Getmansky- or Geltner-type moving-average structure reversal), quantifying the magnitude of the collapse in Sharpe ratio and correlation after desmoothing. This identification leg does not require an exogenous source of variation, because it asserts only the accounting-fact layer that a smoothing component exists and is quantifiable, without over-claiming causality on that basis. Its falsifiable placebo test is: a publicly traded control series that is already continuously marked to market should show no significant collapse after desmoothing. Sensitivity to the order of the moving average is a matter of parameter sensitivity, not identification failure.

Primary strategy — error-correction clock (identification design in place). The second identification leg of this chapter is an event-time regression using forced-liquidation events as the break point, examining whether governance characteristics (board independence, assessment horizon) predict the speed of discount convergence. Its source of exogenous variation is forced-liquidation events (liquidity shocks) as the break point. The treatment variable is plans with high versus low governance independence. The falsifiable placebo test is: non-forced (voluntary) liquidations should not exhibit a difference in convergence speed along the governance gradient. This identification leg has the cleanest cut, and its role flips from an observation in tension with the universal claim into the core positive leg of the conditional proposition (collusion decreases with governance).

Backup strategy — collusive intent and the sign of the negative premium (hypothesis awaiting pre-registration and theory pending direction). The intent of active collusion and the sign of the negative premium can only be elevated to identification via a separation design and an external anchor. The separation design for collusive intent is: under the same intensity of constraint, compare whether plans with high versus low governance independence differentially pursue low-transparency instruments; or use an exogenous change in regulatory standards (a jurisdiction suddenly requiring mark-to-market for private assets) as a difference-in-differences, observing whether plans reduce holdings (evidence of passivity or victimhood) or shift toward even less transparent instruments (evidence of actively seeking smoothing). Testing the sign of the negative premium requires an external ground-truth anchor that does not rely on self-reported private-asset net asset value — a pairing of public REITs and private real estate for the same asset, or the secondary-market discount rate — to yield the sign without circularity. Prior to obtaining an external ground-truth anchor and a separation design, paying a negative premium for collusion is not presented as a core incremental claim, and what this chapter can identify retreats to the desmoothed residual and the error-correction clock. In sum, this chapter's identification strategy contains two downgrades (collusive intent downgraded to a hypothesis awaiting pre-registration, the sign of the negative premium downgraded to a theory pending direction), with two legs where identification design is in place.

5.4 Strongest Counterarguments and Responses

This paper addresses three counterarguments to this chapter's proposition.

The first counterargument points to the unfalsifiable intent behind collusion. The counterargument holds that collusion, active selection of low-transparency instruments, and willingness to pay a negative premium for smoothing are all assertions about unobservable intent; the sign — that plans closer to the regulatory red line are more willing to accept private-asset exposure with lower net returns — is at least equally predicted by three competing hypotheses: passive regulatory arbitrage (standards permit lagged pricing, and plans under funded-ratio constraints are forced to choose instruments that pass the assessment, with no collusive intent), adverse selection or lack of capability (plans under greater pressure have weaker governance and are sold low-transparency, high-fee products by general partners — precisely the victim version), and legitimate duration matching (long-duration liabilities should properly be matched with long-duration illiquid assets). This paper accepts this counterargument, downgrading the intent of active collusion in its entirety to a hypothesis awaiting pre-registration, and feeds this downgrade back into both the title level and the proposition level; the separation design is as described in the preceding section; prior to obtaining the separation design, what this chapter can identify retreats to the desmoothed residual and the error-correction clock.

The second counterargument points to the non-identifiability of the sign of the negative premium. The counterargument holds that to demonstrate a negative liquidity premium, one must cleanly estimate that private assets' net expected return is lower than that of publicly traded assets of the same risk, knowingly and voluntarily accepted — but three obstacles stand in the way: private assets' expected return is itself a smoothed, self-reported net asset value, so using a contaminated measure to sign the effect is circular — one cannot simultaneously claim that net asset value is a smoothing artifact and use net-asset-value-based realized returns to measure the negative premium; lower net returns are difficult to distinguish cross-sectionally from higher expected returns whose volatility is smoothed down; and paired panels of funded ratio and exposure are not comparable across countries. This paper accepts this counterargument, downgrading the sign of the negative premium to a theoretical proposition pending direction, and advocates using an external ground-truth anchor that does not rely on self-reported private-asset net asset value to break the circularity; prior to obtaining an external anchor, the negative premium is not presented as a core incremental claim.

The third counterargument points to the tension between the downgrade statements and the load-bearing claim. This counterargument notes that even though active collusion and the negative premium have already been thoroughly downgraded at the level of the identification legs, this chapter's core claim should not, at the title level, still present active collusion as a settled characterization. This paper accepts this critique, feeding the core claim of this chapter back to the identified truth — that book-value smoothing has demand-side value for plans with weaker governance, and that error-correction speed increases with governance quality, while whether this stems from an intent of active collusion is a hypothesis awaiting pre-registration — thereby eliminating any settled commitment to unfalsifiable intent, and bringing this chapter's epistemic status into alignment with the level of its identification legs.

5.5 Pre-Registered Expectations and Falsification Thresholds

This paper commits to this chapter's falsification thresholds in advance of the data. On the two identification paths where identification design is in place: first, if the magnitude of the collapse in Sharpe ratio and correlation after desmoothing is not significant (the smoothing component is near zero within the robust range of moving-average orders), then the claim that the true premium contains a smoothing component is falsified; second, if the speed of discount convergence after forced liquidation does not vary with governance characteristics (no difference in convergence speed between plans with high versus low governance independence), then the claim that error-correction speed increases with governance is falsified, and the positive leg of the conditional proposition collapses.

On the separation thresholds for the hypothesis awaiting pre-registration: if, under the same intensity of constraint, plans with high governance independence are not less often forced yet still actively choose smoothing (i.e., governance does not predict the choice of smoothing when the degree of constraint is held equal), then active collusion is falsified, reverting to a passive regulatory-arbitrage or victim explanation; if, after regulatory standards exogenously require mark-to-market pricing, plans reduce holdings (rather than shifting to even less transparent instruments), this proves passivity or victimhood, and active smoothing-seeking is falsified. On the threshold for the sign of the negative premium: if an external anchor measures that plans obtain a positive premium rather than a negative one from packaging into private assets, then the direction of a negative premium and willingness to pay it is falsified.

5.6 State of the Evidence

This section faithfully reports the state of evidence for this chapter. At the verifiable level, smoothing in private-asset valuation is an observable accounting fact, and 2022 provides direct corroboration: that year public equity benchmarks fell by double digits (the S&P 500 fell about 20% over the full year), while private equity valuations over the same period declined only mildly, by about 3.2% (Institutional Investor 2022); critics have explicitly pointed out that private equity's apparent outperformance is illusory, because private-firm valuations are set quarterly by valuation committees rather than priced continuously by the market; the transmission of value from public comparables to private assets lags by at least three quarters or more, with fewer markdowns in 2022 and less catch-up in 2023 — the smoothing mechanism is clearly visible (SEI Investments 2022; Neuberger Berman 2022). This fact provides direct directional corroboration for the accounting-fact layer of this chapter's claim that the true premium contains a smoothing component. In addition, the demand-side value of smoothing for plans with weaker governance has institutional-mechanism corroboration: in 2022, the decline in public-market valuations lowered the asset base and pushed up the reading of institutional private-asset allocation shares (the so-called denominator effect), forcing institutions to pull back new commitments (RVK 2022) — corroborating the mechanism whereby funded-ratio and allocation-share assessments generate demand for the right to invisible volatility. This is evidence of the mechanism's existence, not of active collusive intent.

At the level of micro-causal estimation, the estimators for this chapter's two identification legs are unavailable owing to data-availability constraints. The method of desmoothed residuals is mature and its specification declarable, but the raw private-asset return series must be constructed from annual reports or databases, which is unavailable within the scope accessible to this study, so the magnitude of the collapse cannot be quantitatively estimated — the negative 3.2% versus negative 20% in 2022 is aggregate directional evidence, not a desmoothing estimator. The error-correction clock has the cleanest cut, but forced-liquidation events and secondary-market discount-rate series require dedicated event histories, which are likewise unavailable, so the governance gradient in convergence speed cannot be estimated. The paired panels of funded ratio and exposure needed for collusive intent and the sign of the negative premium are not comparable across countries, and external ground-truth private-asset pairings are also unavailable, so these retain the status of a hypothesis awaiting pre-registration and a theory pending direction, and are not presented as core incremental claims.

It must be emphasized that the negative 3.2% versus negative 20% figure for 2022 corroborates the direction of the existence of a smoothing component, but is by no means a desmoothing estimator. This chapter's falsification conditions cannot be executed within the scope accessible to this study; each judgment retains the status of identification design in place with attached falsification conditions, and is not disguised as an already-estimated causal effect. In addition, the specifications for desmoothing and valuation in this chapter must remain consistent with subsequent research on the narrative of decumulation-phase liquidation — this is a cross-study discipline of specification.

5.7 Chapter Summary

This chapter argues that a considerable portion of the diversification return exhibited by mature markets is the outsourcing of book volatility to a smoothing mechanism. Book-value smoothing has demand-side value for plans with weaker governance, and its error-correction speed increases with governance quality — this is what the two identification paths with identification design in place (the desmoothed residual and the error-correction clock) claim in this chapter. This chapter honestly downgrades the intent of active collusion to a hypothesis awaiting pre-registration and the sign of the negative premium to a theory pending direction, in order to avoid over-asserting unfalsifiable intent. In terms of the state of evidence, facts such as the lag in private-asset valuation and the denominator effect in 2022 calibrate the direction of the judgment, but the estimators for the two identification paths are unavailable within the scope accessible to this study. The static accounting-side smoothing illusion revealed in this chapter will, in the next chapter, turn into the reflexivity draining on the dynamic pricing side.

Chapter 6 Capability Is an Accelerator, Not a Moat: The Forward Premium Is Drained by Its Own Reflexivity

This chapter argues the third load-bearing proposition: the stronger the governance capability and the faster the execution of allocation, the faster the forward premium is drained by its own reflexivity — so capability is an accelerator rather than a moat; first movers capture a one-time competitive rent, while latecomers capture the beta that remains after competition. This chapter moves from the static accounting side of Chapter 5 to the dynamic pricing side.

6.1 Proposition

The core proposition of this chapter is falsifiable: the stronger the governance capability and the faster the allocation, the faster the forward premium is drained by its own reflexivity, so capability is an accelerator rather than a moat; first movers (such as institutions that pioneered Canadian-style internalization of direct investment) capture a one-time competitive rent, while latecomers capture the beta that remains after competition. Characterizing this distribution of rent and beta is the core contribution of this chapter relative to the existing literature on competitive decay. The mechanism and identification of this chapter both serve this judgment about the distribution of rent and beta.

This chapter builds on and pivots from Chapter 5. Chapter 5 revealed the smoothing illusion on the static accounting side; this chapter turns to the dynamic pricing side: large pension funds are marginal price setters, and their buying behavior endogenously changes prices — entry pushes up valuations, and the historical returns of already-executed trades are retroactively lifted (thereby producing good track records), so the premium expected going forward is drained in the present. The counterintuitive core is this: the stronger the capability and the faster the execution, the faster the draining — capability is not a moat (protecting the premium) but an accelerator (accelerating the dissipation of the premium). At the aggregate level, if all large funds simultaneously pursue global diversification, the targets become crowded; first movers capture a one-time competitive rent, while latecomers who replicate obtain the beta that remains after crowding and competition — this is standard competitive decay, not a fallacy.

This chapter explicitly states an honesty boundary: this paper abandons the over-labeling of "fallacy of composition." That crowding leads to declining returns is a normal manifestation of competitive alpha decay and price discovery, not a fallacy whereby individual rationality produces collective irrationality; it is observationally equivalent to, and cannot be distinguished from, the decay observed in any crowded trade. The core contribution of this chapter is therefore repositioned as quantifying the distribution of first movers' one-time rent and latecomers' post-competition beta — who enters at what point in time, and how much they capture. As for genuine coordination failure (externalities from homogeneous holdings), that belongs to the domain of liquidity and systemic risk, and is handed off to subsequent research on the liquidity-spiral interface; this chapter does not pass itself off here as demonstrating that the premium is non-aggregable.

6.2 Mechanism

The causal chain of this chapter is as follows. Large pension funds, as marginal price setters, have buying behavior that pushes up target valuations; the historical returns on already-held trades are retroactively lifted (producing a good track record), so the premium expected going forward is drained in the present; the stronger the capability and the faster the execution, the faster the draining. At the aggregate level, as all large funds simultaneously pursue global diversification, the targets become crowded; first movers capture a one-time competitive rent, while latecomers obtain the beta that remains after competition.

This mechanism contains three counterintuitive nodes. First, the moat turns into an accelerator: the conventional narrative treats capability as a moat protecting the persistence of the premium, whereas the truth is that capability accelerates the premium's self-dissipation — the better the allocation, the faster the premium dies. Second, first movers are not a generalizable paradigm but one-time rent claimants on a first-come-first-served basis: latecomers who replicate obtain the post-crowding beta, not the first movers' rent. Third (handed off to subsequent research), the homogenization of holders causes the liquidity of these assets to evaporate in the same direction during periods of stress; the premium (an asset-side benefit) and the liquidity liability (a liability-side risk) are in fact two sides of the same mechanism — but this belongs to the topic of the liquidity spiral, and is not passed off as such here.

6.3 Identification Strategy

This chapter's identification strategy is as follows.

Primary strategy — reflexivity (identification design in place). The first identification leg of this chapter examines the reflexive draining effect of large pension funds' buying. A naive correlation using large funds' position-building or position-adding events directly as a shock does not hold, because the intensity of position-building is endogenous — funds add positions precisely because they expect high returns, or are driven by indices and liabilities. This chapter therefore adopts an instrument that is genuinely exogenous to position-building: mechanical buying triggered by passive index inclusion or weight rebalancing, driven by index rules rather than return expectations, constituting non-informational buying pressure. The treatment variable is the intensity of mechanical buying triggered by index inclusion. The falsifiable placebo test is: comparable non-included targets, and parallel trends prior to inclusion. If forward returns decline after exogenous mechanical buying (a negative residual), this supports reflexive draining; if the exogenous source of index inclusion cannot be cleanly constructed (inclusion itself being correlated with fundamentals), then this leg retreats to a theoretical proposition, and reflexivity is presented only as a mechanistic narrative rather than as identified.

Primary strategy — first-mover rent and latecomer beta (identification design in place, direction). The second identification leg of this chapter constructs a sector-level (rather than fund-level) crowding indicator, regressing sector crowding on subsequent cross-sectional premia; the premium gap between first movers and latecomers is the evidence for rent versus beta. The direction of this conclusion (first movers are one-time rent claimants on a first-come-first-served basis, latecomers capture post-competition beta) is supported by standard competitive alpha decay alone. The falsifiable placebo test is: sectors without rising crowding should not show premium decay. It must be honestly noted that the sector-level crowding indicator has no corresponding data source within the evidentiary base on which this study relies, and the identification leans optimistic.

Backup strategy — crowding as liability (interface proposition). The third leg of this chapter examines the contemporaneous correlation between the liquidity discount of homogeneously held assets during periods of stress and rebalancing sell-off pressure. Because the load-bearing part of this claim falls within the dedicated chapter of subsequent research on the liquidity spiral, and its specification has not yet been frozen, this chapter establishes here only a weak version that does not rely on decumulation-phase liquidation (using publicly available holding-concentration measures for a cross-sectional-plus-stress-period contemporaneous correlation), leaving the strong version to subsequent research. In sum, this chapter's identification strategy, in the strict sense, contains no leg downgraded to a theoretical proposition, but does contain one leg handed off as an interface proposition to subsequent research (crowding as liability) and one abandoned label (fallacy of composition recast as competitive decay); the legs with identification design in place number two.

6.4 Strongest Counterarguments and Responses

This paper addresses two counterarguments to this chapter's proposition.

The first counterargument points to the relabeling of the fallacy of composition. The counterargument holds that if the non-aggregability at the sector level is simply a matter of many people doing the same thing, with the strategy's excess return competed away, this is common knowledge about competitive markets, and renaming it the fallacy of composition produces no new identifiable claim; a distinction must be drawn between two things that are conflated — alpha-crowding decay (capital inflows drive excess returns toward zero, a normal function of efficient markets, in which first movers capturing rent and latecomers capturing beta is correct pricing rather than a fallacy) versus a genuine fallacy of composition (each individual's diversifying behavior renders the system as a whole less diversified and more fragile, a matter of liquidity and systemic risk, which this chapter itself acknowledges as the subject of subsequent research); the two are observationally equivalent in their testable implications. This paper accepts this counterargument, abandons the label "fallacy of composition," honestly acknowledges that this is competitive alpha decay, and repositions the core contribution of this chapter as quantifying the distribution of first movers' one-time rent and latecomers' post-competition beta, handing genuine coordination failure off to the interface with subsequent research; the direction that first movers are not replicable is retained, as it is supported by standard competitive decay.

The second counterargument points to reverse causality and identification. The counterargument holds that the intensity of large funds' position-building is not an exogenous shock but is endogenous to the fund's forward expectation of the asset's return, so using endogenous position-building intensity to predict forward returns yields an ambiguous direction: if a fund adds positions because it anticipates favorable fundamentals and the return is subsequently realized, then position-building intensity positively predicts returns (the opposite of the desired negative sign); if a negative sign is observed, a competing explanation is mean reversion or performance chasing (passive index inclusion or momentum buying at the top followed by a pullback) — in which case it is buying at the top, rather than the buying itself, that drains the premium. This paper accepts this counterargument, abandons the correlation with raw position-building intensity, and instead uses an instrument that is genuinely exogenous to position-building — mechanical buying from passive index inclusion or weight rebalancing, driven by index rules rather than return expectations — using the index-inclusion event as non-informational buying pressure, to separate the draining of the premium by the buying itself from fund timing or private information; prior to obtaining the exogenous source, this leg is labeled as identifiable once an exogenous source is obtained.

6.5 Pre-Registered Expectations and Falsification Thresholds

This paper commits to this chapter's falsification thresholds in advance of the data. On the reflexivity leg: position-building intensity driven by exogenous sources such as index inclusion should negatively predict the residual of the target's forward return; if forward returns do not decline after exogenous mechanical buying (a non-negative residual sign), then the claim that the buying itself drains the premium is falsified, reverting to an explanation of timing or private information. On the leg of first-mover rent and latecomer beta: sector crowding should negatively predict subsequent cross-sectional premia, and first movers' premium should be significantly higher than latecomers'; if the premium does not decay after crowding rises, or there is no premium gap between first movers and latecomers, then first-mover rent is falsified. In addition, if the exogenous source of index inclusion cannot be cleanly constructed (inclusion itself being correlated with fundamentals), then the reflexivity leg retreats to a theoretical proposition and is not presented as identified.

6.6 State of the Evidence

This section faithfully reports the state of evidence for this chapter. At the verifiable level, the institutional background of large public funds mechanically and at scale buying alternative assets holds, supporting the direction of competitive decay: GPIF has established an alternative-asset database and expanded its alternatives exposure (Chief Investment Officer 2023); Australia's sovereign wealth fund the Future Fund and the Netherlands' APG are both increasing allocations to private assets and alternatives (Markets Group 2024; IPE 2024) — the fact that large public funds as marginal buyers are flocking into alternatives in the same direction holds, supporting the direction of competitive alpha decay whereby first movers capture one-time competitive rent and latecomers capture post-competition beta. In addition, mechanical buying from index inclusion, as an exogenous source of non-informational buying pressure, exists institutionally: passivization and index-inclusion rules are transparent, and mechanical buying driven by rules rather than return expectations is a recognized category of non-informational demand shock in the literature (Bank for International Settlements 2025); this paper consumes upstream conclusions on the synthetic scarcity created by index inclusion by citation, without repeating the exposition.

At the level of micro-causal estimation, the estimators for this chapter's two identification legs are unavailable owing to data-availability constraints. The complete panel of target forward returns needed for the reflexivity leg is unavailable, so whether forward returns decline after exogenous buying (the sign of the residual) cannot be estimated. The sector-level crowding indicator needed for the first-mover-rent-and-latecomer-beta leg has no corresponding data source within the evidentiary base on which this study relies, so the negative prediction of crowding on the premium, and the premium gap between first movers and latecomers, cannot be estimated. The weak version of crowding-as-liability, which relies on publicly available holding concentration, is partly available, but the object of decumulation-phase liquidation belongs to subsequent research and is likewise unavailable, so it retains the status of an interface proposition, with the strong version outsourced to subsequent research.

It must be emphasized that the aggregate fact of large funds flocking into alternatives in the same direction supports the direction of competitive decay, but is by no means an estimator. This chapter's mechanism holds as a narrative but is not passed off as identified; the falsification conditions cannot be executed within the scope accessible to this study, and the judgment retains the status of identification design in place, not disguised as identified.

6.7 Chapter Summary

This chapter argues that capability is an accelerator rather than a moat: the stronger the governance capability and the faster the allocation, the faster the forward premium is drained by its own reflexivity; first movers capture a one-time competitive rent, while latecomers capture the post-competition beta. This chapter honestly abandons the over-labeling of "fallacy of composition," repositioning the core contribution as the distribution of rent and beta, and leaves the strong version of genuine coordination failure to subsequent research. In terms of the state of evidence, the aggregate fact of large funds flocking into alternatives in the same direction and the institutional existence of the index-inclusion exogenous source calibrate the direction of the judgment, but the estimators for the two identification legs are unavailable within the scope accessible to this study. This chapter reveals the self-dissipation of capability at the aggregate level on the asset side, preparing the way for the next chapter's descent to the instrument layer that is hardest to supervise.

Chapter 7 The Manufactured Non-Default Thesis: What the Private-Credit Layer Buys Is Not Credit Quality but Actively Manufactured Non-Default

This chapter argues the fourth load-bearing proposition: at the private-credit and continuation-fund instrument layer—the hardest layer to monitor—what pension funds buy is not credit quality but non-default itself, an actively manufactured product; the net-return advantage attributed to internalization owes to a substantial degree to fee arbitrage, with capability being a sufficient rather than a necessary condition. This chapter carries the judgments of the preceding three chapters down to the hardest instrument layer for detailed dissection.

7.1 The Judgment

The core judgment of this chapter is falsifiable: at the hardest private-credit and continuation-fund instrument layer, what pension funds buy is not credit but this actively manufactured product of non-default—deferred default must manifest through the ex post truth-event of realized losses at liquidation, failing which this chapter concedes demotion to a theoretical proposition; and the cross-chapter tension between "internalization advantage equals fee arbitrage" and Chapter 4's "internalization equals governance control rights" is placed under this chapter's jurisdiction and resolved via a pre-registered competing test. The mechanism, identification, and competing tests of this chapter all serve this judgment, with the truth-anchor as its falsifiable precondition.

This chapter carries Chapter 4's governance-control-rights judgment and Chapter 5's smoothing judgment down to the hardest instrument layer for detailed dissection. Private credit's low default rate is not a reading of asset quality but an output produced by the general partner: through payment-in-kind (PIK) conversion, liability management exercises (LME), and rollover timing, defaults are hidden inside a stockpile of forbearance; the genuine governance capability is the capacity to monitor this forbearance stockpile, and the pension fund, as a limited partner, has no means of monitoring it. The institutional order attempts to patch governance with fairness opinions, but a compliance credential substitutes for rather than supplements genuine judgment, diluting the locus of responsibility—a governance-layer version of Gresham's law. Another institutional attempted solution is internalization (the Canadian model), but the net-return advantage of internalization owes in large part to the saved fee wedge and the allocation of carried interest (carry), rather than to stronger credit capability.

This chapter especially emphasizes one point of intellectual honesty regarding boundaries. This paper cannot treat unobservability simultaneously as its core argument and as an alibi—PIK, rollovers, and liability management exercises all have entirely legitimate counter-explanations (standard contractual structure, the normal course of refinancing, ordinary debt restructuring), and a large deviation is read as accumulation of a forbearance stockpile while a small deviation is read as evidence that the quantity is inherently unobservable—both outcomes are non-falsifying. Hence there must be an ex post verifiable truth-event (realized losses at liquidation) that makes the forbearance stockpile visible; absent that, this core proposition is demoted to a theoretical proposition.

7.2 Mechanism

The causal chain of this chapter runs as follows. Private credit's low default rate, taken by limited partners as a reading of asset quality, is in fact the output of the general partner using PIK conversion, liability management exercises, and rollover timing to hide defaults inside a forbearance stockpile; the genuine governance capability is monitoring this forbearance stockpile, and limited partners have no means of monitoring it—hence what is bought is manufactured non-default (a product, not credit quality); the forbearance stockpile is ultimately exposed as realized losses at liquidation.

This mechanism contains three counter-intuitive nodes. First, patient capital and low default rates are not a laudable asset advantage but an actively manufactured product—the low default rate is the general partner's output, not a reading of the asset. Second, the mandatory proliferation of the institutional patch of fairness opinions in fact lowers genuine scrutiny—a compliance credential substitutes for genuine judgment, diluting the locus of responsibility; this is a governance-layer version of Gresham's law, not more rigorous governance. Third, the Canadian internalization advantage owes in large part not to capability but to the saved fee wedge and the allocation of carried interest (fee arbitrage), with capability being a sufficient but not necessary condition. This directly opposes Chapter 4's judgment that internalization equals governance control rights or capability, and this tension is placed under this chapter's jurisdiction, to be resolved via a pre-registered competing test.

7.3 Identification Strategy

The identification strategy of this chapter is as follows.

Primary strategy—fairness-opinion outsourcing (identification design in place). The first identification leg of this chapter takes as its break point the effective date of a securities-regulatory rule mandating fairness opinions, conducting an event-time observation of whether real secondary-market discounts and dissent rates rise or fall after the rule takes effect. The source of exogenous variation lies in the fact that the rule's effective date is exogenous to any single transaction. The treatment variable is the transaction type subject to versus not subject to the rule. The falsifiable placebo test is the transaction type not subject to the rule. This paper further tightens the labeling, adding observations that distinguish "useless" from "harmful": the ex post incidence of dissent and litigation, and the difference in long-run realized losses between transactions that adopted fairness opinions and those that did not; only if adopters show higher ex post losses (i.e., the fairness opinion endorsed a bad deal) does this support the strong label of governance-outsourcing degradation—a mere failure of discounts to fall supports only the weak version of "no additional information."

Primary strategy—fee arbitrage versus capability (identification design in place). The second identification leg of this chapter conducts a shadow fee re-estimation for internalized programs (restoring the saved fees using comparable external fee schedules), examining how much of the net-return advantage survives after stripping out the fee wedge. This leg carries sole claim to the attribution of internalization (Chapter 4's corresponding claim is revised to cite this chapter's competing-test conclusion). At the identification-design level, this paper pre-registers a competing test: whether the residual—after shadow fee re-estimation of internalized programs and stripping out the fee wedge—correlates with proxies for governance control rights (governance independence, program-sponsorship ownership, i.e., Chapter 4's antecedent proxies); if the residual is close to zero, this chapter's fee-arbitrage explanation holds (fee arbitrage predominates and capability is not necessary; Chapter 4 must then concede that the internalization advantage can be replicated by any scale player able to compress fees, and does not reflect governance control rights); if the residual is significant and increases with the governance-control-rights proxy, then Chapter 4's governance-control-rights explanation holds (this chapter's claim that the advantage is largely fee-driven is overturned). This pair of oppositely directed predictions renders the two chapters' evidence mutually independent rather than mutually contradictory.

Backup strategy—manufactured non-default (theoretical proposition). Using the deviation in the proportion of PIK, liability management exercises, and rollover frequency as a proxy for the forbearance stockpile cannot be falsified, because the target concedes itself to be unobservable—using an incomplete proxy to measure a quantity that concedes itself unobservable admits of no falsification. This chapter therefore adopts an ex post truth-anchor: the gap between the final realized loss at fund maturity/liquidation and the low default rate reported during the fund's life (liquidation being the moment of truth that can no longer be deferred); or the difference in default recognition for the same batch of loans held on a bank balance sheet (which must be marked) versus held in a private-credit vehicle (which can forbear). Until a realized-loss-at-liquidation anchor is obtained, manufactured non-default is labeled a theoretical proposition, and the main text does not present it as an identified mechanism. In sum, this chapter's identification strategy includes one item demoted to a theoretical proposition (manufactured non-default), with two legs whose identification design is in place.

7.4 Strongest Counter-Arguments and Responses

This paper addresses three counter-arguments to this chapter's propositions.

The first counter-argument points to constructive unobservability. It holds that this chapter's principal load-bearing core proposition concedes itself to be inherently unobservable, and its falsification point depends on the deviation between headline default rates and the forbearance-stockpile proxy—if the forbearance stockpile is inherently unobservable, then any proxy is incomplete, and a large deviation is read as accumulation of the forbearance stockpile while a small deviation is read as the quantity being inherently undetectable; neither outcome is falsifying. More fundamentally, PIK, rollovers, and liability management exercises have legitimate counter-explanations, and their use amounting to manufactured non-default requires proof that it exceeds normal levels and conceals real losses—yet real losses are precisely what is unobservable. This paper accepts this counter-argument, and no longer treats unobservability simultaneously as core argument and alibi; instead it adopts a truth-anchor—the gap between the final realized loss at fund maturity/liquidation and the low default rate reported during the fund's life, or the difference in default recognition for the same batch of loans held by a bank versus held privately—using ex post realized losses to anchor deferred default, converting manufactured non-default from unobservable to falsifiable; until a realized-loss-at-liquidation anchor is obtained, this core proposition is labeled a theoretical proposition, and what this chapter can identify retreats to the two identification paths of fairness-opinion outsourcing and fee arbitrage.

The second counter-argument points to an unresolved counter-explanation and self-contradiction. It notes that this chapter's fee-arbitrage-versus-capability claim, and Chapter 4's claim that high-capability rationally invests less directly and that program-sponsorship ownership equals governance capability, offer opposing causal attributions for the same explained variable (the net-return advantage of internalization): as to why the Canadian internalization model's net return is high, Chapter 4 answers governance control rights and capability, while this chapter answers saved fees—these are not two complementary mechanisms but competing explanations contesting the same residual; if this chapter holds (the advantage is mainly fees), Chapter 4's claim that internalization equals non-transferable capability is weakened, and vice versa. This paper accepts this counter-argument and makes a cross-chapter reassignment: this chapter assumes sole claim to the attribution of the internalization advantage (Chapter 4's corresponding claim is revised to cite this chapter's competing-test conclusion rather than repeating the claim); and at the identification-design level pre-registers a competing test (whether the residual increases with the governance-control-rights proxy), using oppositely directed predictions so that the two chapters' judgments are mutually independent rather than mutually contradictory.

The third counter-argument points to endogenous treatment, a comparatively minor scope risk. It notes that the fairness-opinion break-point estimates the total effect of mandatory fairness opinions on discounts and dissent rates, and reading this as "a compliance credential substitutes for genuine judgment" requires an additional assumption—discounts failing to fall might also occur because the rule is too new for the market to have adjusted, or because fairness opinions genuinely improve pricing so that discounts should already be low. This paper accepts this comparatively minor criticism, retains the fairness-opinion break point's status as this chapter's strongest identification leg (its identification validity is not undermined by this counter-argument), and merely tightens the labeling by adding distinguishing observations (the ex post incidence of dissent and litigation, and the difference in long-run realized losses between adopting and non-adopting transactions): only if adopters show higher ex post losses does this support the strong label of governance-outsourcing degradation; a mere failure of discounts to fall supports only the weak version of "no additional information."

7.5 Pre-Registered Expectations and Falsification Thresholds

This paper fixes this chapter's falsification thresholds prior to the data. On the fairness-opinion-outsourcing leg, distinguishing "useless" from "harmful": if the long-run realized losses of transactions adopting fairness opinions are not higher than those of non-adopters (and the ex post incidence of dissent and litigation does not rise), then the strong label of governance-outsourcing degradation is falsified, retreating to the weak version of "no additional information"; if discounts fall after the mandate (a positive information increment), then "compliance substituting for genuine judgment" is falsified. On the fee-arbitrage-versus-capability leg, an oppositely directed bidirectional falsification: if the residual, after stripping out the fee wedge, is close to zero, then Chapter 4's "internalization equals governance control rights" is falsified here and this chapter's fee arbitrage holds; if the residual is significant and increases with the governance-control-rights proxy, then this chapter's claim that the advantage is largely fee-driven is falsified and Chapter 4 holds—whichever side it falls on, the result is an identifiable, oppositely directed conclusion, so the two chapters' judgments coexist in parallel rather than contradicting one another. On the manufactured-non-default leg: if the difference between the fund's final realized loss at liquidation and the default rate reported during its life is not significant (i.e., losses at liquidation match the reported default rate), then deferred default and manufactured non-default are falsified, reverting to the position that the low default rate is genuine asset quality.

7.6 Evidentiary Status

This section faithfully reports this chapter's evidentiary status, and honestly discloses one new identification risk that lies outside this paper's original identification design. At the verifiable level, the scale and growth rate of the private-credit market—the institutional background of this instrument layer—holds up: the International Monetary Fund's 2024 estimate puts private credit at roughly USD 2 trillion (rapidly growing, requiring closer supervision) (International Monetary Fund 2024); figures from the Bank for International Settlements and the Alternative Investment Management Association put it at roughly USD 3 trillion to 3.5 trillion in early 2025 (up from roughly USD 2 trillion in 2020), with 2024 capital deployment of roughly USD 592.8 billion, a year-on-year increase of roughly 78% (Alternative Investment Management Association 2025; Bank for International Settlements 2025)—low-default private credit is precisely the object pension funds are increasing allocations to, so the real-world scale of this chapter's instrument-layer critique holds. Furthermore, market concern over private-layer forbearance stockpiles and deferred default has been independently documented by the Financial Stability Board and the Federal Reserve: the Financial Stability Board's report on private-credit vulnerabilities and related Federal Reserve research flag vulnerabilities concerning the opacity of private-credit valuation, bank exposure to private credit, and financial-stability implications (Financial Stability Board 2026; Federal Reserve 2025)—the problem domain this chapter addresses, namely that monitoring the forbearance stockpile is the genuine governance capability and limited partners lack the means to monitor it, has been independently confirmed at the regulatory level (this establishes the existence of the problem, not the causal proof of this chapter's core proposition).

Regarding the fairness-opinion break-point identification leg, a major correction to the evidentiary record must be made and honestly reported. The securities regulator adopted a private fund adviser rule on August 23, 2023, under which secondary or continuation transactions led by a general partner were mandatorily required to obtain a fairness opinion or valuation opinion from an independent third party (Securities and Exchange Commission 2023)—precisely the institutional object of this chapter's "compliance credential substituting for genuine judgment." However, this rule was vacated in full by a federal appellate court on June 5, 2024, which ruled that the regulator had exceeded its statutory authority (Morgan Lewis 2024). This judicial reversal of fact means that the treatment period in this chapter's fairness-opinion difference-in-differences design in effect never took effect, or was retroactively vacated; the event-time break point's sample window collapses, and the source of exogenous variation is itself impaired by the judicial vacatur. This new identification risk arose outside the point at which this paper's identification design was frozen; this paper records it honestly and accordingly downgrades its identification confidence in this leg, rather than concealing it to sustain the core claim. This chapter's identified foundation therefore effectively retreats to a directional calibration combining identification design with market scale and problem-domain evidence.

At the level of micro-causal estimation, the estimators for each identification leg of this chapter are unavailable owing to data limitations. Beyond the damage from the judicial vacatur noted above, the fairness-opinion break point's panel data on secondary discounts, dissent rates, and long-run realized losses are unavailable, so the core coefficient cannot be estimated. The shadow fee re-estimation required for fee-arbitrage-versus-capability needs internalized-program net returns and external comparable fee schedules constructed from annual reports, which are unavailable within the scope accessible to this research, so whether the residual is close to zero (supporting the fee-arbitrage explanation) or increases with the governance-control-rights proxy (supporting the governance-control-rights explanation) cannot be determined—the scale of CPP's FY2024 external fees (management fees of C$1,449 million plus performance fees of C$2,067 million) merely corroborates that the fee wedge is real and sizable (CPP Investments 2024a), and by no means constitutes the estimator for the shadow fee re-estimation. The gap between realized loss at liquidation and the default rate reported during the fund's life, required for manufactured non-default, requires full-fund-lifecycle data that is hardest to obtain at the private layer, and is likewise unavailable; it retains theoretical-proposition status, and this paper does not pass off a deviation measure as an identified finding.

7.7 Chapter Summary

This chapter argues that at the private-credit instrument layer—the hardest to monitor—what pension funds buy is the actively manufactured product of non-default, not credit quality; the internalization advantage owes to a substantial degree to fee arbitrage, with capability being a sufficient but not necessary condition. This chapter honestly demotes the core proposition of manufactured non-default to a theoretical proposition, pending the truth-anchor of realized losses at liquidation, and assigns the cross-chapter tension over internalization attribution to this chapter's jurisdiction, to be resolved through an oppositely directed competing test. As to evidentiary status, this chapter honestly discloses one new identification risk lying outside this paper's original identification design—the fairness-opinion rule was judicially vacated in June 2024, impairing the exogenous source for the corresponding identification leg; this paper accordingly downgrades its identification confidence, rather than concealing it to sustain the claim. This chapter negates, from the instrument layer, the technocratic prescription that a stronger team, more compliance, and more internalization equal better governance, and prepares for the next chapter's conclusion, which returns to the liability side.

Chapter 8 The True Boundary of Capacity: The Capacity Ceiling Is Jointly Determined by Liability Hardness and the Procyclicality of Patient Capital

This chapter argues the fifth load-bearing proposition: the true capacity ceiling for alternative assets is jointly determined by liability hardness and the procyclicality of patient capital, rather than by the size of the capital pool; but liability hardness and allocation depth are endogenous to a common cause—namely governance, as discussed in Chapter 4—so this chapter is reframed as the marginal contribution of liability hardness given governance, sharing the same political constraint with Chapter 4. This chapter closes out the liability side.

8.1 Judgment

The core judgment of this chapter is falsifiable: the true capacity ceiling for alternative assets is jointly determined by liability hardness and the procyclicality of patient capital, rather than by the size of the capital pool; but liability hardness is endogenous to a common cause together with allocation depth, governance, and political tolerance (precisely the main thread of Chapter 4), so this chapter is reframed as the marginal contribution of liability hardness given governance, sharing the same political constraint with Chapter 4 rather than constituting an independent dual. The mechanism, identification, and evidence of this chapter all serve this judgment regarding the marginal contribution of liability hardness given governance.

This chapter closes out the liability side. The conventional metric treats capacity in terms of the size of the capital pool (or the presence or absence of liabilities), but this paper holds that the true constraint on capacity is whether an institution possesses an opportunistic option to redeem or exit: a hard, predictable, fixed-payout liability locks up capital and strips away the right of early redemption, constituting a commitment device, and can therefore support deeper illiquid allocation. Conversely, the absence of liabilities at a sovereign wealth fund means the sovereign can withdraw at any time—there is no commitment device—so its illiquid capacity is correspondingly compressed. This is precisely the mechanism behind the conservative paradox whereby the institutions with the most capital allocate the least deeply. But commitment is procyclically fragile: rising interest rates or changes in funding ratios endogenously shorten the assessment duration horizon, reactivating the redemption option; long-term capital becomes short-term precisely during periods of stress, and patient capital therefore cannot serve as a countercyclical stabilizer.

This chapter especially emphasizes two honest boundaries. First, liability hardness and the depth of illiquid allocation are mutually causal and endogenous to a common cause: the institutions capable of sustaining deep illiquid allocation are precisely those that already possess strong governance and political protection, and can therefore also sustain hard liability contracts—governance (precisely the endogenous protagonist of Chapter 4) is the common cause determining both simultaneously. Treating liability hardness as an exogenous independent variable and regressing allocation depth on it is tantamount to treating an endogenous variable from Chapter 4 as an exogenous treatment. This chapter therefore reframes itself as the marginal contribution of liability hardness given governance, and abandons the demographic-structure instrumental variable whose exclusion restriction fails via the real-interest-rate back door. Second, Chapter 4 and this chapter are not mutually independent orthogonal duals—their endogenous variables are in fact one and the same (political tolerance, or extractability), being two projections of the same constraint onto the asset side (the ceiling on governance control rights) and the liability side (activation of commitment). Treating them as two orthogonal pillars would artificially inflate the density of the judgment, so this paper honestly reframes the argument around a single main axis.

8.2 Mechanism

The causal chain of this chapter is as follows. The same constraint of political tolerance or extractability is projected, on the asset side, as a ceiling on governance control rights (Chapter 4), and on the liability side, as the activation of commitment (this chapter). A hard, predictable, fixed-payout liability, as a commitment device that strips away the redemption right, locks up capital and can support deeper illiquid allocation; but rising interest rates or changes in funding ratios endogenously shorten the assessment duration horizon, politically reactivating the redemption option; long-term capital becomes short-term precisely during periods of stress, and patient capital therefore cannot behave countercyclically.

This mechanism contains four counterintuitive nodes. First, the absence of liabilities is not a capacity advantage but a disadvantage—the absence of liabilities at a sovereign wealth fund means the sovereign can withdraw at any time (there is no commitment device), compressing illiquid capacity (i.e., the conservative paradox whereby the institutions with the most capital allocate the least deeply). Second, patient capital is not a countercyclical stabilizer—commitment is procyclically fragile, and long-term capital becomes short-term during periods of stress, retreating precisely when stability is most needed. Third, the capacity boundary is decoupled from the size of the capital pool—the true constraint is liability hardness (commitment) and the procyclicality of patient capital, not the size of the capital pool; and liability hardness itself is not an exogenous cause but a projection of governance (marginal contribution can only be discussed given governance). Fourth, in defined contribution or semi-liquid vehicles, subscription flow itself may become a self-referential support for the underlying valuation (this point is flagged as a hypothesis).

8.3 Identification Strategy

The identification strategy of this chapter is as follows.

Primary strategy—procyclicality of patient capital (identification design in place). The first identification leg of this chapter, after controlling for asset-side attractiveness, examines whether governance events or redemption behavior are independently predicted by interest rates and funding ratios; it uses the 2022 UK liability-driven investment (LDI) crisis as a natural experiment (rising interest rates endogenously shortened the assessment duration horizon, and long-term capital became short-term during the period of stress). The source of exogenous variation is the exogenous timing of the 2022 crisis (the interest rate shock was not the choice of any single fund). The treatment variable is UK defined benefit plans constrained by this type of leveraged structure. The falsifiable placebo test is: plans not constrained by this type of leveraged structure should not exhibit the same degree of duration shortening. It must be noted that even though this natural experiment is specific to the UK's leveraged structure and not easily generalizable, this does not undermine its validity as a proof of existence.

Secondary strategy—liability hardness (correlational, pending an institutionally exogenous source). The second leg of this chapter is the primary load-bearing leg still awaiting repair. A naive regression using the rigidity of the liability contract (the hard commitment of fixed payouts, versus the extractability available to sovereign wealth funds) as the core independent variable, controlling for size, and using demographic structure as an exogenous instrument for liability duration, cannot stand, for three reasons: endogeneity to a common cause (governance is the common cause, precisely the main thread of Chapter 4); the dummy variable for the absence of liabilities at sovereign wealth funds is likewise endogenous; and the exclusion restriction of the demographic-structure instrument fails via the back-door path from demographics to the real interest rate to the attractiveness of illiquid assets. This chapter therefore takes a threefold approach: first, explicitly incorporating and controlling for the governance and political protection of Chapter 4 as a confounder; second, switching to an institutionally exogenous source that moves only the rigidity of the liability contract without moving governance or asset-side attractiveness—jurisdictional differences in statutory fixed-payout guarantee rules, and differences in how bankruptcy law treats pension claims (which institutionally lock in liability hardness, exogenous to the fund's own governance choices); third, abandoning the demographic-structure instrument whose exclusion restriction has failed. The chapter is reframed as the marginal contribution of liability hardness given governance. Until the institutionally exogenous source is in place, the core regression is flagged as correlational rather than causal; if the marginal coefficient on liability hardness collapses to zero after controlling for the governance common cause, this chapter merges into Chapter 4 and does not contest a separate claim from it.

Secondary strategy—self-referential valuation (hypothesis, pending an exogenous subscription shock). The third leg of this chapter examines the self-referential support that subscription flow provides to the underlying reported net asset value. The naive approach of directly measuring the synchronicity between subscription flow into defined contribution or semi-liquid vehicles and the underlying reported net asset value, and measuring whether secondary-market discounts widen with channel size, cannot stand, because this narrative is trivially true at all times, synchronicity is observationally equivalent to performance-chasing, and the discount widening with channel size is endogenous. This chapter therefore switches to a test that is directionally identifiable and does not rely on systemic collapse: using an exogenous subscription shock (e.g., a default defined contribution plan being passively enrolled into a semi-liquid vehicle due to a regulatory change, with the capital inflow unrelated to the vehicle's fundamentals) to see whether it pushes up the reported net asset value (an exogenous inflow pushing up net asset value would prove that subscriptions support valuation, ruling out performance-chasing); for the secondary-market discount, one must control for asset-class sentiment and cycle conditions and then examine whether the residual widens with channel size. In the absence of an exogenous subscription shock, self-referential valuation is downgraded to a hypothesis, serving only as mechanism narrative. In sum, the identification strategy of this chapter contains two downgrades (the core regression on liability hardness downgraded to correlational, pending an institutionally exogenous source; self-referential valuation downgraded to a hypothesis, pending an exogenous subscription shock); one leg has identification design in place (the 2022 LDI crisis); and the demographic-structure instrument has been abandoned.

8.4 Strongest Counterarguments and Responses

This paper addresses three counterarguments to the propositions of this chapter.

The first counterargument points to reverse causality and omitted variables. The counterargument holds that liability hardness and the depth of illiquid allocation are mutually causal and endogenous to a common cause—the institutions capable of sustaining deep illiquid allocation are precisely those that already possess strong governance and political protection, and can therefore also sustain hard liability contracts, with governance being the common cause determining both simultaneously (precisely the claim of Chapter 4 of this paper); the positive coefficient from liability hardness to allocation depth could be entirely produced by the common cause of independent governance, or could even run in reverse (a pre-existing culture of committing to illiquid asset allocation could itself, in turn, entrench the rigidity of the liability contract); controlling for size controls for the wrong variable, when the true confounder is governance and political protection; this chapter therefore collides with Chapter 4, treating an endogenous variable from Chapter 4 as an exogenous treatment; and the exclusion restriction of the demographic-structure instrument fails via the back-door path from demographics to the real interest rate. This paper accepts this sharp counterargument and takes the threefold approach described above (explicitly incorporating and controlling for the confounder of governance and political protection, switching to an institutionally exogenous source, and abandoning the demographic-structure instrument), reframing the chapter as the marginal contribution of liability hardness given governance; until the institutionally exogenous source is in place, the core regression is flagged as correlational rather than causal; if the marginal coefficient on liability hardness collapses to zero after controlling for the governance common cause, this chapter merges into Chapter 4.

The second counterargument points to unfalsifiable self-reinforcement. The counterargument holds that the narrative whereby subscription flow supports net asset value and assets are never forcibly liquidated is a self-reinforcing narrative that is falsifiable only in the event of systemic collapse and otherwise trivially true at all times: an unfalsifiable time sequence (as long as there is no collapse, the proposition cannot be falsified, and once there is a collapse, the proposition is deemed confirmed); an opposing explanation for the synchronicity between subscription flow and net asset value (capital chasing recently high net asset value vehicles, i.e., net asset value driving subscription flow, the opposite direction, and a matter of common sense); and the endogeneity of the widening secondary-market discount (large channels happen to expand precisely when sentiment peaks). This paper accepts this counterargument, downgrades self-referential valuation to a hypothesis, and switches to a test that is directionally identifiable and does not rely on systemic collapse (whether an exogenous subscription shock pushes up the reported net asset value; the residual in the secondary-market discount after controlling for cycle conditions and sentiment). In the absence of an exogenous subscription shock, self-referential valuation serves only as mechanism narrative and is not treated as identified.

The third counterargument points to the tension of duality with Chapter 4, a comparatively minor structural issue. The counterargument notes that this chapter claims a duality between the asset side (Chapter 4) and the liability side (this chapter), but the endogenous variables of the two are in fact one and the same (governance, or political tolerance): the ceiling on control rights in Chapter 4, imposed by political tolerance, and the procyclical fragility of commitment in this chapter, with the redemption option politically reactivated, are two projections of the same political constraint onto the asset side and the liability side, rather than two independently dual pillars; duality implies two orthogonal foundations, when in reality it is a single variable counted twice, artificially inflating the density of the judgment through dualistic rhetoric. This paper accepts this comparatively minor criticism and honestly reframes the framework around a single main axis: Chapter 4 and this chapter share the same underlying variable of political tolerance or extractability; the framework changes from two orthogonal dual pillars to a single political constraint acting separately on the asset side (the ceiling on control rights) and the liability side (activation of commitment)—this is in fact a stronger and more unified judgment (political tolerance is the single endogenous main axis running through both the asset and liability sides), but no longer claims that the two are independently dual so as to artificially inflate structural density.

8.5 Pre-registered Expectations and Falsification Thresholds

This paper sets out the falsification thresholds of this chapter prior to the data. On the leg concerning the procyclicality of patient capital: during the 2022 LDI shock, the assessment duration horizon and the retreat of long-term capital at plans constrained by this type of leveraged structure should be independent of asset-side signals and predicted by interest rates and funding ratios; if the retreat of long-term capital is not independent of asset-side signals (i.e., driven purely by the asset side rather than by the reactivation of commitment), then the procyclicality of patient capital and the fragility of commitment are falsified. On the primary load-bearing leg concerning liability hardness: after controlling for governance and political protection, or using an institutionally exogenous source, the marginal coefficient of liability hardness on allocation depth should be significantly positive; if the marginal coefficient on liability hardness collapses to zero after controlling for the governance common cause (its explanatory power fully absorbed by the governance common cause), then the independent marginal contribution of liability hardness is falsified, and the capacity boundary is attributed entirely to governance (in which case this chapter merges into Chapter 4); if allocation depth does not change after liability hardness is locked in using an institutionally exogenous source, then the direction from hardness to depth is falsified. On the leg concerning self-referential valuation: if the reported net asset value does not rise following an exogenous subscription shock (a passive inflow unrelated to fundamentals), then subscription-supported valuation and self-reference are falsified, reverting to the performance-chasing explanation.

8.6 State of the Evidence

This section faithfully reports the state of the evidence for this chapter. At the level of what can be verified, the procyclical fragility of patient capital has, as a proof of existence, ironclad evidence from the 2022 UK LDI crisis: following the mini-budget of September 23, 2022, gilt yields rose sharply, and within the week to September 27, yields across multiple maturities rose by more than 100 basis points; leveraged LDI defined benefit plans were forced into fire sales of long-duration gilts to meet margin and collateral calls, and these forced sales produced a discount of approximately 10%, accounting for nearly half of the total decline in gilt prices (Bank of England 2023a); the Bank of England purchased £19.3 billion of gilts (comprising £12.1 billion conventional and £7.2 billion index-linked) between September 28 and October 14 as a backstop (Bank of England 2023b). On this basis, the proposition that long-term capital becomes short-term during periods of stress, and that patient capital cannot serve as a countercyclical stabilizer, obtains ironclad existence evidence (the interest rate shock was exogenous to the choice of any single fund; the fact that the LDI crisis is specific to the UK's leveraged structure does not undermine its validity as a proof of existence).

Regarding the direction of an institutionally exogenous source for liability hardness, the Dutch pension reform driving private-market allocation provides directional corroboration for this category of institutionally exogenous source: the Dutch APG scheme raised its private-market allocation from approximately 26% to more than 30% as a result of pension reform, with private credit rising from approximately 1.5% toward a target of 2% to 4%, and ABP's target allocation to alternatives at approximately 28% (IPE 2024); the transition of Dutch collective defined benefit arrangements toward individualization (PGIM 2024) is precisely a real-world instance of this chapter's category of institutionally exogenous sources that lock in liability hardness via jurisdictional or rule differences. In addition, there is direct evidence that the absence of liabilities leads to compressed capacity and mandate ceilings: GPIF's statutory ceiling for alternative assets is 5% of total assets, while the actual allocation in fiscal year 2024 was only approximately 1.63% (GPIF 2024)—even the world's largest public pension fund, with an enormous capital pool, has its illiquid capacity compressed by a mandate ceiling; the capacity boundary is decoupled from the size of the capital pool and is determined by governance and mandate (political tolerance). This phenomenon, together with the ceiling on control rights imposed by political tolerance in Chapter 4, is the same political constraint's projection onto the liability side, rather than an independent dual.

At the level of micro-causal estimation, the estimators for each identification leg of this chapter are unobtainable owing to data availability constraints. The complete plan-level panel of duration and redemption behavior required for the leg on the procyclicality of patient capital is unavailable, so whether the retreat of long-term capital is independent of asset-side signals and is predicted by interest rates and funding ratios cannot be estimated; the proof of existence stands, but the panel-level causal coefficient cannot be estimated. The four-country panel of allocation depth and the calibration of the governance and political-protection confounder required for the liability hardness leg are unavailable, and the sample of sovereign withdrawal histories at sovereign wealth funds is limited, so whether the marginal coefficient on liability hardness collapses after controlling for the governance common cause, and whether this chapter merges into Chapter 4, cannot be determined—the status remains correlational rather than causal, and the demographic-structure instrument has been abandoned. The defined contribution subscription flow, underlying net asset value, and secondary-market discount data required for the self-referential valuation leg are likewise unavailable, so it retains hypothesis status, serving only as mechanism narrative.

It must be emphasized that the event-level aggregate facts of the 2022 LDI crisis (more than 100 basis points within a week, an approximately 10% discount, a £19.3 billion intervention) constitute a proof of existence for procyclical fragility, but the plan-level causal coefficient cannot be estimated; the Dutch pension reform driving private-market allocation provides directional corroboration for this category of institutionally exogenous source, but is by no means an estimator; GPIF's 5% ceiling versus its actual 1.63% provides direct directional evidence that mandate ceilings compress capacity, but is by no means the marginal coefficient of liability hardness.

8.7 Chapter Summary

This chapter argues that the true capacity ceiling for alternative assets is jointly determined by liability hardness and the procyclicality of patient capital, rather than by the size of the capital pool; but liability hardness and allocation depth are endogenous to a common cause, namely governance, so this chapter is reframed as the marginal contribution of liability hardness given governance. This chapter honestly downgrades the core regression on liability hardness to correlational, pending an institutionally exogenous source, downgrades self-referential valuation to a hypothesis, pending an exogenous subscription shock, and abandons the demographic-structure instrument whose exclusion restriction has failed. This chapter also reframes Chapter 4 and this chapter, from an independent duality, into two projections of the same political constraint, so as to avoid artificially inflating the density of the judgment. On the state of the evidence, the 2022 UK LDI crisis provides ironclad existence evidence for procyclical fragility, and GPIF's mandate ceiling provides direct directional evidence for the decoupling of capacity from size, but the causal coefficients for each identification leg remain unestimable within the scope accessible to this study. At this point, the five load-bearing propositions have each been argued in turn; the next chapter stitches them into a single whole and provides a comparison of failure conditions across the four modes.

Chapter 9 Cross-Proposition Synthesis: A Single Political Axis, Four Layers of Illusion, and the Failure Conditions of the Four Modes

The preceding five chapters each argued a load-bearing proposition. The task of this chapter is not to repeat the conclusions of each chapter but to reveal the deep structure connecting them—the five propositions are not five isolated judgments but five facets of a unified whole organized around the overarching judgment of the endogeneity of governance capability. This chapter synthesizes from three angles: first, it identifies the single political axis running through all five propositions; second, it organizes the five propositions into the successive layers of the premium illusion; third, it converges the comparison of the four modes into a matrix of failure conditions.

9.1 A Single Political Axis: The Unified Constraint from the Asset Side to the Liability Side

The deepest common structure of this paper's five propositions is a political-tolerance constraint that runs throughout. This constraint appears in different guises across the different propositions, but its essence is the same endogenous variable.

In Chapter 4, political tolerance manifests as a ceiling on governance control rights: when veto power is held by politically appointed bodies and when mandates can be revoked, then even if an institution hires a top-tier team, its capability is capped by political tolerance—the institution obtains only risk exposure without the premium. The Alberta provincial government's 2024 dissolution of AIMCo's entire board (Mergers and Inquisitions 2024) is precisely the real-world manifestation of this cap—even the most arm's-length institution's mandate remains revocable through political process.

In Chapter 8, the same political tolerance manifests as the activation of liability commitments: hard liabilities, as commitment devices, are supposed to lock in capital and bear deep illiquidity, but rising interest rates or changes in funding ratios endogenously shorten the assessment horizon, the redemption option is politically reactivated, and long-term capital turns short in periods of stress. GPIF's statutory 5% cap on alternative assets, against an actual allocation of only 1.63% (GPIF 2024), is the direct expression of this constraint on the mandate side—capacity is compressed by the mandate (the institutionalization of political tolerance), not by the scale of assets under management.

This paper therefore emphasizes a key synthetic judgment: the control-rights cap in Chapter 4 and the commitment activation in Chapter 8 are not two independent, mutually orthogonal constraints, but two projections of the same political-tolerance constraint onto the asset side and the liability side. This judgment carries an honest methodological implication—treating the two as separate, additive judgments would inflate the apparent density of the argument; this paper explicitly consolidates them into a single axis. That political tolerance is the single endogenous axis running through both the asset and liability sides is a formulation that does not weaken the judgment but rather renders it more unified and more forceful. Canada's AIMCo incident (Mergers and Inquisitions 2024) and Japan's mandate ceiling (GPIF 2024), though seemingly belonging to two different phenomena—governance intervention and allocation constraint—are in fact the manifestation of the same political constraint in two countries and on two sides.

9.2 The Four Layers of the Premium Illusion: From Accounting to Pricing, From Instruments to Aggregation

Chapters 2 through 5 of this paper can be read as a layer-by-layer dissection of the true quality of the alternative-asset premium displayed by mature markets. These four propositions are not parallel critiques but successive presentations of the illusion at four distinct layers.

The first layer is the smoothing illusion on the static accounting side (Chapter 5). Private-asset valuations use lagged book net asset values, suppressing realized volatility and correlation, thereby exaggerating diversification benefits—part of the diversification benefit is simply book volatility outsourced to a smoothing mechanism. In 2022, private equity valuations declined by only about 3.2%, while public equity benchmarks fell by about 20% over the same period (Institutional Investor 2022)—a direct manifestation of this smoothing in an extreme year.

The second layer is the reflexive draining on the dynamic pricing side (Chapter 6). Large pension funds, acting as marginal price-setters, see their own buying behavior push up valuations and retroactively generate strong performance, such that the forward-looking premium is drained in the present; the stronger the capability and the faster the execution, the faster the draining. The static smoothing illusion and the dynamic reflexive draining are two layers of the same premium illusion—the former concerns the accounting presentation of realized returns, the latter concerns the pricing dissipation of forward-looking returns.

The third layer is manufactured non-default on the instrument side (Chapter 7). At the private credit instrument layer—the hardest to monitor—low default rates are the output of general partners using payment-in-kind conversions, liability management transactions, and opportunistic maturity extensions to hide defaults in a forbearance inventory, rather than a reading of asset quality. Manufactured non-default is the credit-dimension counterpart of manufactured smoothing—Chapter 5's smoothing conceals volatility, Chapter 7's forbearance inventory conceals default, and the two share the same accounting-convention origin. The private credit market grew from approximately $2 trillion in 2020 to approximately $3 trillion to $3.5 trillion in early 2025 (International Monetary Fund 2024; Alternative Investment Management Association 2025); the expansion of its scale gives this instrument-layer illusion systemic significance, and the Financial Stability Board and the Federal Reserve have independently warned of its fragility (Financial Stability Board 2026; Federal Reserve 2025).

The fourth layer is competitive decay on the aggregation side (the aggregation portion of Chapter 6). If all large funds simultaneously pursue global diversification, the targets become crowded, first-movers capture a one-time competitive rent, and latecomers obtain only the post-competition beta. This layer pushes the individual illusion to the collective level: even if a given first-mover once genuinely captured the premium, a latecomer's replication can only obtain the beta that remains after competition. GPIF's establishment of an alternative-assets database and expansion of exposure, and the Netherlands' APG increasing its private-asset allocation (Chief Investment Officer 2023; IPE 2024), are precisely the real-world picture of this simultaneous rush.

The four layers jointly point to a synthetic judgment: the alternative-asset premium displayed by mature markets is of far lower quality than its book presentation suggests—smoothed in accounting, drained in pricing, manufactured at the instrument level, and decayed by competition in aggregation. This synthetic judgment is a systematic rebuttal of the panegyric to advanced-market experience: it does not deny that alternative assets once brought genuine premiums to first-movers, but it points out that a substantial part of this premium is a statistical illusion, and that it cannot simply be replicated by latecomers.

9.3 The Distinction Between Capability and Control Rights Running Throughout

The second thread running through the five propositions is the distinction between capability and control rights. This distinction is established in Chapter 4 and recurs repeatedly in the other chapters, constituting the conceptual core of this paper's judgment on transferability.

Chapter 4 establishes that capability can be priced, hired, and unbundled in the factor market, and that the moat is not capability but control rights. CPP's payment of external management fees of C$1,449M and performance fees of C$2,067M in fiscal year 2024 (CPP Investments 2024a) directly confirms that capability is a tradeable intermediate good. Chapter 6 further points out that capability is not only purchasable—once purchased, it accelerates the dissipation of the premium—capability is an accelerator rather than a moat, which further confirms that control rights (not capability), are the endogenous variable. Chapter 7 offers a sharp corollary at the instrument level: most of Canada's internalization advantage comes from fee arbitrage rather than capability, and capability is a sufficient but not necessary condition; the real-world anchor of this corollary remains CPP's considerable fiscal-year-2024 external fee wedge (CPP Investments 2024a)—internalization is replicable (fee arbitrage), while control rights are not. Chapter 8 then points out that liability hardness itself is not an exogenous factor either, but a projection of governance—the institutions capable of bearing deep illiquidity are precisely those that already possess strong governance and political protection.

This recurring distinction provides a clear boundary for this paper's judgment on transferability: what can be learned is capability (investment teams, valuation systems, risk-control processes, and even allocation structures); what cannot be learned is control rights (mandates that cannot be politically revoked, veto power independent of political process, the freedom of countercyclical timing) and the commitment device of hard liabilities. A late-developing market that replicates only the learnable parts without possessing the unlearnable parts will obtain a mature-market-style risk exposure, not its premium.

9.4 A Failure-Condition Matrix for the Four Modes

This paper's comparison of the four modes is not four parallel expositions but a controlled-variable examination organized around the judgment of the endogeneity of governance capability. This section converges the comparison of the four modes into a failure-condition matrix, indicating under what kind of shock each mode fails. A methodological honesty boundary must first be declared: the failure conditions listed in the table below are directional judgments based on this paper's theoretical framework and verifiable institutional facts, not causal conclusions estimated from micro-panel data; each failure condition corresponds to a falsifiable threshold set out in advance of data in the respective chapters of this paper, and its empirical testing awaits the availability of corresponding micro-data.

Table 2 Matrix of Governance Characteristics, Liability Structure, and Failure Conditions Across the Four Modes

This matrix reveals the core judgment of the four-mode comparison. First, the differences among the four modes cannot be explained by the scale of assets under management—Japan is precisely the sample with the largest assets under management yet the shallowest allocation, and its constraint comes from the mandate rather than from capital. Second, each of the four modes has its own failure shock, and none is an unconditional advanced-market experience: the Canadian mode fails when transplanted to a locale lacking governance independence, the Australian mode is limited by the weak liability hardness of defined contribution, the Japanese mode is limited by the mandate ceiling, and the Dutch mode sees its commitment device loosen after the shift toward individualization. Third, the failure conditions of all four modes can be traced back to this paper's two main threads—the endogeneity of governance capability (control rights) and the unity of the political constraint. This is precisely the direct correction of the common defect of writing the four modes as four parallel expositions: the comparison of the four modes must produce a cross-cutting judgment of who fails under what conditions, rather than a juxtaposed introduction.

9.5 The Ranking of Transferability and Where the Judgment Lands

Synthesizing the failure conditions above, this paper offers a directional ranking judgment on the transferability of the four modes. The least transferable is the core of the Canadian mode—the allocation of control rights—because it is rooted in a specific governance independence and political environment and cannot be transferred through the replication of personnel or structure. More transferable are the elements at the capability level (investment teams, valuation systems, risk-control processes), because they can be purchased in the factor market. In between lies liability hardness—it is partly determined by institutional design (statutory defined-benefit guarantees, the treatment of pension claims under bankruptcy law) and can therefore be partly obtained through institutional design, but it is at the same time a projection of governance, so its transferability is constrained by whether governance independence can be established first.

The final landing point of this paper's judgment is therefore clear: there exists a hard boundary to the transferability of the Canadian mode. A late-developing market that merely raises its alternative-asset allocation ratio without first possessing governance independence, an internal investment team, market-based compensation, and long-horizon assessment will obtain not Canadian-style premium capture but Canadian-style risk exposure—valuation opacity, liquidity mismatch, principal-agent leakage—without its premium. Treating a high alternative-asset share itself as a replicable formula for success is a causal inversion that mistakes the act of allocation for proof of capability. This technical judgment constitutes the policy premise for subsequent research on whether and how late-developing markets should learn from these models.

9.6 Chapter Summary

This chapter has synthesized the arguments of the preceding five chapters from three angles. First, the five propositions are threaded through by a single political-tolerance axis, which projects onto the asset side as a control-rights cap and onto the liability side as commitment activation—the two are two projections of the same constraint rather than independent counterparts. Second, Chapters 2 through 5 can be read as a layer-by-layer dissection of the premium illusion across four layers—accounting, pricing, instruments, and aggregation—jointly revealing that the quality of the alternative-asset premium displayed by mature markets is far lower than its book presentation. Third, the comparison of the four modes converges into a failure-condition matrix indicating under what shock each mode fails, and on this basis a ranking of transferability is given. The landing point of this paper's judgment is the hard boundary on the transferability of the Canadian mode. The next chapter will turn from this technical judgment to policy implications and set out the limitations of this paper.

Chapter 10 Conclusion

10.1 Main Findings

This paper, focused on the causal structure underlying mature pension funds' alternative and global allocation in a low-interest-rate environment, has proposed and substantiated a set of interlocking propositions. This section summarizes its main findings.

First, this paper advances the somewhat generic conclusion that governance capability is endogenous into a more precise causal structure: what is truly endogenous and non-transferable is not investment, valuation, and risk-control capability itself, but the allocation of governance control rights over that capability. Capability can be priced in factor markets, hired, and unbundled; the moat is not capability but control rights — who holds the decision-making power to originate projects, who holds veto power, and whose mandate cannot be politically revoked. This distinction separates, at the conceptual level, the replicable component (capability) from the non-replicable component (control rights), providing a precise basis for judging transferability. To prevent this distinction from degenerating into an unfalsifiable tautology, this paper establishes a threefold set of pre-performance, independently codable proxy vectors for control rights.

Second, this paper systematically incorporates valuation smoothing, fees, and survivorship bias into a critique of the true quality of the alternative premium, revealing that a considerable portion of the diversification returns displayed by mature markets is a statistical illusion. This illusion unfolds progressively across four layers: valuation smoothing on the accounting side, reflexive draining on the pricing side, manufactured non-default on the instrument side, and competitive decay on the aggregation side. At the private credit instrument layer — the hardest to monitor — institutions are buying a product of actively manufactured low default rates, not a reading of asset quality; the internalized net-return advantage derives substantially from fee arbitrage, with capability being a sufficient but not necessary condition.

Third, this paper reveals a single political-tolerance constraint running through both the asset and liability sides. It projects onto the asset side as a control-rights ceiling, and onto the liability side as commitment activation; the two are two projections of the same endogenous variable, not independent duals. Hard liabilities, as a commitment device, could in principle support deep illiquidity, but commitment is procyclically fragile — during stress periods, long-term capital turns short, and patient capital cannot serve as a countercyclical stabilizer. The true capacity ceiling for alternative assets is jointly determined by liability hardness and the procyclicality of patient capital, not by the scale of capital.

Fourth, this paper organizes the comparison of the four models into a variable-controlled examination centered on a single judgment, converging on a matrix of failure conditions: the Canadian model fails in transplant settings lacking governance independence, the Australian model is constrained by the weak liability hardness of defined contribution (DC) arrangements, the Japanese model is constrained by a mandate ceiling, and the Dutch model has seen its commitment device loosen following the shift toward individualization. None of the four models constitutes an unconditionally advanced practice; each has its own failure shock.

Synthesizing the above findings, this paper offers the following answers to the five core questions posed at the outset. Low interest rates have indeed driven pension funds' migration toward alternative and global allocation, but the depth of that migration is constrained by governance and liability structure, not determined solely by allocation intent. Whether alternative assets raise returns or merely raise risk exposure depends on whether the institution possesses the governance preconditions to capture the premium; for institutions with insufficient governance, alternatives mainly raise risk exposure. The differences among the four models are rooted in governance structure and mandate arrangements, and their transferability has hard boundaries. Governance capability is a precondition for allocation, not its result — treating a high alternative allocation share as a replicable success paradigm is a reversal of cause and effect. Alternative allocation does indeed produce valuation lag, liquidity mismatch, and governance complexity, and it undercuts its claimed return advantage during stress periods.

10.2 Policy Implications

This paper's judgments carry several policy implications for late-developing markets — particularly public pension funds and sovereign reserve funds currently considering emulating the Canadian model. It must be reiterated that this paper offers a technical judgment on whether emulation is feasible, not a policy judgment on whether or how to emulate; the implications below are a natural extension of the technical judgment toward the policy level, and their full policy argumentation awaits future research.

First, the allocation ratio is not an independent lever that can be raised in advance. This paper's central judgment is the endogeneity of governance capability — governance capability is a precondition for allocation depth, not its result. Therefore, if a late-developing market simply raises its alternative allocation ratio without governance independence, an internal investment team, market-based compensation, and long-horizon performance evaluation, what it will obtain is mature-market-style risk exposure rather than the premium. The policy sequence should be to first build the governance preconditions and only then discuss allocation depth, not the reverse. Treating the act of raising the alternative share itself as a policy goal mistakes an allocation action for proof of capability.

Second, governance independence is a non-outsourceable core that must be established through institutions rather than personnel. This paper argues that capability can be outsourced, but control rights cannot be transferred. The threefold pre-performance proxies for control rights — whether the board appointment-and-removal chain follows a political process, the statutory ownership of veto power and project-origination decision rights, and the historical incidence of mandate revocation — are all institutional-level arrangements. Therefore, if a late-developing market wishes to obtain the control rights needed to capture the premium, it must establish, at the institutional level, a mandate that cannot be revoked through political process, a veto power independent of political appointment, and freedom for countercyclical timing — rather than relying merely on recruiting talent or replicating organizational structures. The lesson of the Canadian AIMCo incident is that even the most arm's-length institution's control-rights mandate can be politically revoked (Mergers and Inquisitions 2024) — the robustness of control rights must be sustained through continuous institutional safeguards.

Third, liability structure design precedes deepening of asset allocation. This paper argues that the true capacity for alternatives is determined by liability hardness, and that hard liabilities, as a commitment device, must be established first. For late-developing markets, if the liability structure lacks a hard commitment device (such as redemption rights under defined contribution (DC) arrangements, or sovereign withdrawability under a public reserve fund), then illiquid capacity is inherently limited, and forcibly deepening alternative allocation under such conditions will amplify liquidity-mismatch risk. Japan's GPIF mandate ceiling (GPIF 2024) and the loosening of the commitment device following the Dutch pension reform (PGIM 2024) illustrate, from opposite directions, how liability structure constrains allocable depth.

Fourth, assessment of the alternative premium must strip out accounting illusions. This paper reveals that the true quality of the alternative premium in mature markets is far lower than its book-value presentation — smoothed on the accounting side, drained on the pricing side, manufactured on the instrument side, and decayed by competition on the aggregation side. Therefore, when late-developing markets assess the expected return of alternative allocation, they must de-smooth private return series, adopt after-fee metrics, and strip out survivorship bias, rather than directly inferring true diversification returns from book-value low volatility. This is especially true at the rapidly expanding private credit instrument layer (International Monetary Fund 2024; Alternative Investment Management Association 2025), where a low default rate may be a manufactured product rather than a reading of asset quality; the Financial Stability Board and the Federal Reserve have independently warned of its fragility (Financial Stability Board 2026; Federal Reserve 2025), and late-developing markets must remain vigilant on this point.

Fifth, as latecomers, they must squarely confront the distributional consequences of competitive decay. This paper argues that first movers capture a one-time competitive rent, while latecomers capture post-competition beta. Therefore, if a late-developing market replicates a class of alternative assets only after mature markets have already fully entered it, what it will obtain is the beta that remains after crowding, not the rent that first movers once captured. This means that simply replicating the allocation structure of mature markets is, by timing alone, already destined to fail to reproduce their historical returns. Policy consideration should distinguish the one-time rent within historical returns from sustainable beta, rather than taking the first mover's historical performance as the basis for latecomers' expectations.

10.3 Limitations of This Paper

This paper has several important limitations that must be honestly disclosed. Some of these limitations concern the boundaries of identification, while others concern the domain of applicability of the judgment itself.

First, and most importantly, the micro-level data required for this paper's identification are not obtainable within the scope accessible to this research. This paper has provided identification strategies and falsifiability thresholds specified in advance of the data for each of its five load-bearing propositions, but the fund-level governance–return matched panel, the private credit default-and-extension panel, the de-smoothed private return series, the secondary-market discount series, and the final realized losses at fund liquidation are all unobtainable. Consequently, for every proposition this paper labels as having an identification design in place, the causal coefficient is not estimable within this research's environment. This paper therefore strictly distinguishes verifiable institutional and aggregate facts (used to calibrate direction) from micro-level causal signs (for which no identification claim is made). This distinction reflects identification honesty, but it also means that this paper's core propositions await empirical testing once the corresponding micro-level data become available. The verificatory facts reported in this paper — such as the fact that private equity valuations fell only about 3.2% in 2022 while public benchmarks fell about 20% (Institutional Investor 2022; SEI Investments 2022), and that during the 2022 UK liability-driven investment crisis, gilt yields rose by more than 100 basis points within a week (Bank of England 2023a) — are all institutional and event-level aggregate facts that corroborate the direction of the judgment, and are by no means micro-level causal estimates.

To make clear, at a glance, the verifiable aggregate facts, the judgment direction each calibrates, and the micro-level data required but unobtainable for each proposition, the table below summarizes the sources, metrics, and calibrated judgment directions of the institutional and event-level facts used in this paper, and indicates the availability of the micro-level data required for identification of each proposition.

Table 3 Verifiable Aggregate Facts, Sources, and Micro-Data Availability

Second, several propositions in this paper have the status of theoretical propositions or hypotheses rather than identified causal conclusions, as this paper honestly labels. In Chapter 4, the separability of capability and freedom of timing are theoretical propositions; in Chapter 5, collusive intent is a hypothesis awaiting pre-registration, and the negative-premium sign is a theory whose direction remains undetermined; in Chapter 6, crowding-as-liability is an interface proposition; in Chapter 7, manufactured non-default is a theoretical proposition; in Chapter 8, the liability-hardness-as-core regression is correlational rather than causal, and self-referential valuation is a hypothesis. These downgrades are honest boundaries of identification; the falsifiability thresholds specified in advance of the data for each proposition are retained, but this paper does not disguise them as already identified. It should be specially noted that the core exogenous source for the fair-opinion discontinuity in Chapter 7 — a securities regulatory rule — was judicially vacated in its entirety in June 2024 (Morgan Lewis 2024), impairing the exogenous source of what should have been the cleanest identification leg; this paper honestly reports this new identification risk, which lies outside its original identification design, and has accordingly lowered its confidence in identification.

Third, this paper's judgments have a domain of applicability. This paper takes Canada, Australia, Japan, and the Netherlands — four mature markets — as its sample, and extrapolation of its conclusions must be undertaken with caution. Each of the four models has its own specific institutional and political environment; although the endogeneity of governance capability and the unity of the political constraint distilled in this paper are general in nature, their concrete manifestation in other institutional environments must be examined separately in light of local institutions. Furthermore, although this paper's existence proofs (such as the proof, via the 2022 UK liability-driven investment crisis, of the procyclical fragility of patient capital) hold, they are specific to the UK's leverage structure, and extrapolation to other markets must be undertaken with caution — an existence proof establishes that a phenomenon can occur, but does not establish its prevalence in other environments.

Fourth, this paper's metrics are subject to a cross-study consistency constraint. This paper's figures on alternative allocation share, allocation structure, and the like share the same source as upstream research and were not independently collected as an aggregate baseline; the metrics for private valuation and de-smoothing must remain consistent with the narrative of subsequent research concerning clearing during the decumulation (payout) phase. This metric discipline ensures cross-study consistency, but it also means that part of this paper's quantitative foundation depends on the corresponding output of upstream and downstream research.

10.4 Directions for Future Research

Based on the above limitations, this paper points to several directions for future research.

First, and most directly, once micro-level data become available, empirical testing of the five load-bearing propositions should be conducted according to the identification strategies and falsifiability thresholds specified in advance of the data by this paper. One of this paper's contributions lies precisely in operationalizing its judgments in advance into executable identification designs — the construction of the control-rights pre-performance proxy vector, the estimation of de-smoothed residuals and the error-correction clock, the test of reflexivity under an index-inclusion exogenous source, the test of manufactured non-default under a true anchor of realized losses at liquidation, and the test of the marginal contribution of liability hardness under an institutional exogenous source. Once these designs are connected to data, they can either elevate this paper's theoretical propositions into identified causal conclusions, or falsify them.

Second, is the search for exogenous sources needed for each of this paper's propositions. Chapter 4 awaits passive divestiture events from antitrust-forced divestment or passive spin-offs following parent-company bankruptcy; Chapter 7 awaits full-life-cycle fund liquidation data; Chapter 8 awaits jurisdictional differences in statutory fixed-benefit guarantees and bankruptcy-law treatment. Systematic collection of these exogenous sources is the key precondition for grounding this paper's identification designs. It is especially worth noting that, since the exogenous source for the fair-opinion discontinuity in Chapter 7 has been impaired by judicial vacatur, future research must seek an alternative exogenous source to identify the governance-outsourcing effect whereby compliance credentials substitute for genuine judgment.

Third, is the full extension of this paper's judgments to the policy level. This paper offers a technical judgment on whether emulation is feasible; its extension to a policy judgment on whether and how to emulate must be completed in conjunction with the specific institutional environment of the late-developing market, the geographic and political availability of controllable-cash-flow entities, and the specific arrangements at the account and constraint layers. The mapping from governance-capability conditions to allocable alternative depth, and the failure-condition matrix provided by this paper, are the technical preconditions for this policy argumentation.

10.5 Closing Remarks

This paper's fundamental stance is that the core of economic research is the sharpness of judgment and the clarity of causal structure, not the scale of data. This paper's examination of mature pension funds' alternative allocation is not meant to deny the significance of alternative assets in a low-interest-rate environment, but to correct a reversal of cause and effect that mistakes an allocation action for proof of capability. The high alternative allocation share of mature markets is the result of their independent governance, internalized capability, and tolerance for long-horizon evaluation — not a cause that can be directly replicated. What is truly endogenous and non-transferable is the allocation of control rights over capability, and the single political-tolerance constraint running through both the asset and liability sides. If a late-developing market does not first possess these preconditions and merely raises its alternative allocation ratio, what it will obtain is not a mature-market-style premium, but its risk exposure. This judgment, together with the falsifiable identification designs this paper has specified for it, constitutes this paper's contribution to the question of the endogeneity of governance capability; and this paper's honest labeling of its own identification boundaries is an inseparable part of that contribution — an honestly labeled identification boundary is worth more to judgment than an overreaching claim disguised as proven causation.

References (Works Cited)

Alternative Investment Management Association. Strong Growth Sees Private Credit Market Reach US$3.5 Trillion. AIMA, 2025, www.aima.org/article/press-release-strong-growth-sees-private-credit-market-reach-us-3-5-trillion.html.

Bank for International Settlements. The Rise of Private Credit. BIS Quarterly Review, Bank for International Settlements, Mar. 2025, www.bis.org/publ/qtrpdf/r_qt2503b.htm.

Bank of England. An Anatomy of the 2022 Gilt Market Crisis. Working Paper, Bank of England, 2023a, www.bankofengland.co.uk/-/media/boe/files/working-paper/2023/an-anatomy-of-the-2022-gilt-market-crisis.pdf.

Bank of England. Financial Stability Buy/Sell Tools: A Gilt Market Case Study. Quarterly Bulletin, Bank of England, 2023b, www.bankofengland.co.uk/quarterly-bulletin/2023/2023/financial-stability-buy-sell-tools-a-gilt-market-case-study.

Chief Investment Officer. Japan's GPIF Creates Alts Database as It Expands Exposure to Asset Class. CIO, 2023, www.ai-cio.com/news/japans-gpif-creates-alts-database-as-it-expands-exposure-to-asset-class/.

CPP Investments. F2024 Annual Report. Canada Pension Plan Investment Board, 2024a, www.cppinvestments.com/wp-content/uploads/2024/07/CPP-Investments-F2024-Annual-Report.pdf.

CPP Investments. Governance: Independence. Canada Pension Plan Investment Board, 2024b, www.cppinvestments.com/about-us/governance/independence/.

Federal Reserve. Bank Lending to Private Credit: Size, Characteristics, and Financial Stability Implications. FEDS Notes, Board of Governors of the Federal Reserve System, 2025, www.federalreserve.gov/econres/notes/feds-notes/bank-lending-to-private-credit-size-characteristics-and-financial-stability-implications-20250523.html.

Financial Stability Board. Report on Vulnerabilities in Private Credit. Financial Stability Board, 2026, www.fsb.org/uploads/P060526.pdf.

Government Pension Investment Fund. Annual Report Fiscal Year 2024. GPIF, 2024, www.gpif.go.jp/en/performance/annual_report_fiscal_year_2024.pdf.

Institutional Investor. PE Valuations Drop but Still Remain Far above Public Markets. Institutional Investor, 2022, www.institutionalinvestor.com/article/2bstor7aqjglivlq87dhc/portfolio/pe-valuations-drop-but-still-remain-far-above-public-markets.

International Monetary Fund. The Fast-Growing $2 Trillion Private Credit Market Warrants Closer Watch. IMF Blog, International Monetary Fund, 2024, www.imf.org/en/blogs/articles/2024/04/08/fast-growing-usd2-trillion-private-credit-market-warrants-closer-watch.

Investment & Pensions Europe. APG to Considerably Expand Illiquid Investments Unit. IPE, 2024, www.ipe.com/apg-to-considerably-expand-illiquid-investments-unit/10033669.article.

Markets Group. Future Fund Adds to Alternatives. Markets Group, 2024, www.marketsgroup.org/news/future-fund-adds-to-alternatives.

Mergers and Inquisitions. Canadian Pension Funds. Mergers & Inquisitions, 2024, mergersandinquisitions.com/canadian-pension-funds/.

Morgan Lewis. Fifth Circuit Vacates SEC Private Fund Adviser Rules in Full. Morgan, Lewis & Bockius LLP, 2024, www.morganlewis.com/pubs/2024/06/fifth-circuit-vacates-sec-private-fund-adviser-rules-in-full.

Neuberger Berman. CIO Weekly Perspectives: Private Equity Valuations Withstand the Volatility of 2022. Neuberger Berman, 2022, www.nb.com/en/global/insights/cio-weekly-perspectives-private-equity-valuations-with-stand-the-volatility-of-2022.

PGIM. Countdown Clock Begins: Dutch Pension Reform. PGIM, 2024, www.pgim.com/be/en/institutional/insights/emea/countdown-clock-begins-dutch-pension-reform.

RVK. Market Insights: The Denominator Effect and Private Markets. RVK, Inc., 2022, www.rvkinc.com/pdf/insights_20220909.pdf.

Securities and Exchange Commission. SEC Enhances the Regulation of Private Fund Advisers. Press Release 2023-155, U.S. Securities and Exchange Commission, 2023, www.sec.gov/newsroom/press-releases/2023-155.

SEI Investments. Asset Owners and Private Equity Valuation Lag. SEI, 2022, www.seic.com/institutional-investors/our-insights/asset-owners-and-private-equity-valuation-lag.

World Bank. The Evolution of the Canadian Pension Model: Practical Lessons for Building World-Class Pension Organizations. World Bank Group, 2018, documents1.worldbank.org/curated/en/780721510639698502/pdf/121375-The-Evolution-of-the-Canadian-Pension-Model-All-Pages-Final-Low-Res-9-10-2018.pdf.