Entities — Research Map

Every load-bearing claim in the essay, followed by the research that supports or connects to it.

Load-Bearing Claims

Four contract primitives (pay basis, risk exposure, decision rights, information rights) determine behavior
Four roles (providers, allocators, middlemen, mercenaries) are emergent equilibria, not imposed categories
Other combinations are unstable — they generate contradictions and decay toward one of the four
Discretion without risk exposure is unstable
Completion pay with meaningful risk exposure is unstable (attribution across time horizons)
Outcome pay without information is unenforceable — the agent games the metrics
Time pay with broad discretion enables free-riding
Of 36 primitive combinations, only 4 are self-sustaining under agent-controlled information
Agent information control is the proximate stabilizer of the landscape
Monitoring cost is the reason agent control persists
The GP is a trust aggregator — bundles functions, captures rent on the whole package because the LP cannot disaggregate contribution
The LP’s fee reflects information disadvantage, not service value
Reinforcing mechanisms (centralized liability, regulation, reputational insurance, discretion as default) further stabilize the landscape
Some regions of the contract space remain degenerate even with perfect monitoring
Bundle drift — the same role can shift toward different bundles over time as contract terms change
When information control transfers from agents to systems, the attractor landscape reshuffles
Trust aggregation becomes contestable when per-component contribution is observable
The Accountable Executor becomes stable under system-controlled information
The Direct Principal becomes stable under system-controlled information
The Accountable Intermediary becomes stable under system-controlled information

Claim-by-Claim Research Mapping

1. Four contract primitives determine behavior

Jensen & Meckling (1976) — Establishes that the firm is a nexus of contracts and that agency costs (monitoring, bonding, residual loss) are unavoidable frictions of delegation. The primitives framework extends this by decomposing “the contract” into four specific dimensions rather than treating it as a monolith. Jensen & Meckling show that ownership structure and capital structure interact with agency costs — the primitives formalize what those interactions look like at the individual contract level.

Holmstrom (1979) — The informativeness principle: any signal containing information about effort should be included in the contract. This directly supports why information rights are a primitive — the contract’s enforceability depends on what is observable. Holmstrom also establishes the fundamental tradeoff between incentive strength and risk exposure for the agent, which is the interaction between the pay basis and risk exposure primitives.

Grossman & Hart (1986) — Ownership matters because contracts are incomplete: the party with residual control rights has bargaining power and stronger investment incentives. This is the theoretical foundation for decision rights as a primitive — who holds residual control determines behavior in unforeseen states, not just in the states the contract specifies.

2. Four roles are emergent equilibria

Schelling (1978) — Macro-level patterns emerge from micro-level individual incentives in ways no one intended. Small, individually rational contract choices produce large, stable aggregate structures through feedback loops and tipping points. The four roles are a Schelling-type emergent outcome: no one designed the GP/LP/broker/lawyer structure, but it self-organizes from the primitive constraints.

Williamson (1975, 1985) — Organizational forms (markets, hierarchies, hybrids) exist to minimize transaction costs under bounded rationality and opportunism. The four bundles are analogous to Williamson’s governance structures — stable organizational responses to specific transactional conditions. The key drivers (asset specificity, uncertainty, frequency) map loosely onto the primitive dimensions.

Coase (1937) — Firms exist because market transaction costs can exceed internal coordination costs; boundaries shift when technology changes either cost. The four roles are the capital-markets equivalent of Coase’s question: why do these specific organizational boundaries exist? The answer is the same — they minimize total contracting costs given the information regime.

3. Unstable combinations decay toward the four attractors

Holmstrom & Milgrom (1991) — When agents perform multiple tasks but only some are measurable, strong incentives on measured tasks distort effort away from unmeasured ones. This is the mechanism by which unstable combinations fail: a contract that tries to pay on outcomes for one dimension while leaving another unmeasured creates distortion. The contract either adds measurement (and stabilizes) or drops outcome pay (and decays to time-based).

Schelling (1978) — The decay dynamic is a tipping/feedback process. An unstable combination doesn’t sit still — small perturbations push it toward the nearest attractor. Schelling’s framework explains why the landscape has discrete stable points rather than a continuum.

4. Discretion without risk exposure is unstable

Jensen & Meckling (1976) — The asset substitution / risk shifting problem: when agents can make consequential decisions without bearing downside, they take excessive risk or optimize for personal convenience. Jensen & Meckling formalize this as the agency cost of equity — discretion without risk is the structural condition that generates it.

Holmstrom (1979) — Optimal contracts must tie pay to observable outcomes to counterbalance unobservable effort. Discretion without risk means the agent’s effort is unobservable and their pay is unaffected by outcomes — the worst possible incentive configuration. The contract is either refused by the principal or collapses to a different form.

5. Completion pay with meaningful risk exposure is unstable

Rau (2000) — M&A advisory fees are overwhelmingly contingent on deal completion, and acquirers advised by banks with stronger completion incentives exhibit weaker long-term performance. This empirically demonstrates the instability: completion pay creates a bias toward closing regardless of quality. Adding long-term risk exposure would fix the bias but creates an attribution problem — did the deal fail because of the banker or the market?

McLaughlin (1990, 1992) — Contingent fees are not necessarily bad incentives — they are an efficient response to deal uncertainty and information asymmetry. But they are only stable when combined with reputation effects and repeat relationships, not with formal risk exposure. The contract works without risk because the repeat game substitutes for it.

Golubov, Petmezas & Travlos (2012) — Top-tier advisors offset completion bias through reputation effects and repeated-game discipline. This confirms the instability claim from the other direction: completion pay is only tolerable when reputation (not contractual risk) provides the counterweight. Formal risk exposure is not the solution — it creates worse problems.

6. Outcome pay without information is unenforceable

Holmstrom (1979) — The informativeness principle directly: if the principal cannot observe a signal correlated with effort, outcome-based pay degenerates. The agent games whatever metric is available, or the contract collapses to flat pay. This is the formal version of “outcome pay without information is unenforceable.”

Holmstrom & Milgrom (1991) — The multitask extension: when only some dimensions are measurable, outcome pay on the measured dimensions pulls effort away from the unmeasured ones. The agent isn’t gaming in a dishonest sense — they’re rationally allocating effort toward what is rewarded. The information gap makes the contract self-defeating.

Grossman & Stiglitz (1980) — Prices cannot be perfectly informationally efficient because no one would pay to produce the information that makes them efficient. Applied to contracts: outcome pay requires information, but producing that information is costly. If the cost exceeds the benefit, the contract reverts to a cruder form.

7. Time pay with broad discretion enables free-riding

Holmstrom & Milgrom (1991) — Time-based pay is optimal when tasks are hard to measure and the principal wants balanced effort. But when combined with broad discretion, the agent can allocate effort to personally convenient tasks rather than valuable ones. The principal cannot distinguish good from bad discretionary choices, so the agent free-rides.

Gibbons & Murphy (1992) — Career concerns create implicit incentives that can substitute for explicit performance pay early in a career, but fade with tenure. Time pay with broad discretion works when career concerns are strong (junior agents trying to build reputation) but degrades as the agent becomes established and implicit incentives weaken.

8. Of 36 combinations, only 4 are self-sustaining under agent-controlled information

Williamson (1975, 1985) — Williamson’s “discriminating alignment” framework: not all governance structures work for all transaction types; there is a limited set of viable configurations. The 36-to-4 reduction is the capital-markets equivalent — most theoretically possible contract configurations are not viable under the prevailing information regime.

Holmstrom (1979) + Holmstrom & Milgrom (1991) + Jensen & Meckling (1976) — Together, these three papers provide the instability mechanisms that eliminate 32 of the 36 combinations. Holmstrom provides the incentive-risk tradeoff, Holmstrom & Milgrom provide the multitask distortion, Jensen & Meckling provide the agency cost logic. Each unstable combination fails for a reason traceable to one of these three mechanisms.

9. Agent information control is the proximate stabilizer

Hayek (1945) — The central economic problem is coordinating dispersed knowledge that exists in fragments across individuals. Hayek shows that prices aggregate this dispersed knowledge. In the entities framework, agent information control is the reason prices (fees, carry structures) cannot aggregate performance information — the agent sits between the principal and the relevant data, preventing the price mechanism from working.

Polanyi (1966) — Much knowledge is tacit — embodied in skills and context rather than expressible in rules. This provides a structural reason why agent information control persists: some of what the agent knows genuinely cannot be codified, which means even a willing agent cannot fully transfer information to a system. The stabilizer is not just strategic withholding but genuine illegibility.

Walsh & Ungson (1991) — Organizational memory exists but retrieval fails when it matters. Information that could theoretically reduce agent control is distributed across retention facilities (individuals, routines, structures) and cannot be accessed at the moment of decision. The agent controls information partly because the organization cannot retrieve its own.

10. Monitoring cost is the reason agent control persists

Jensen & Meckling (1976) — Agency costs include monitoring costs explicitly. The principal-agent equilibrium is not perfect — it is optimal given the cost of monitoring. Reduce monitoring costs and the equilibrium shifts. This is the direct theoretical foundation for why cheaper monitoring (via AI) changes the landscape.

Coase (1937) — The boundary of the firm is set where the marginal cost of internal coordination equals the marginal cost of using the market. Monitoring cost is a component of both. Technology that reduces monitoring cost shifts firm boundaries — and by extension, the boundaries between the four roles.

Brooks (1975) — Communication overhead grows combinatorially with team size. This places a hard constraint on monitoring: even if the principal wants to monitor every agent, the coordination cost of doing so scales faster than the team. The pension-fund example (five investment staff, forty GP relationships, two hours per week each) is a Brooks’ Law problem.

Dunbar (1992, 1993) + Hill & Dunbar (2003) — Human social networks are layered (5/15/50/150) and constrained by cognitive resources, not technology. Active reciprocal relationships are the binding limit. This explains why the middleman’s information advantage (the relationship graph in their head) is structurally durable — it is bounded by cognitive limits that technology has not historically overcome.

11. The GP is a trust aggregator

Kaplan & Schoar (2005) — PE performance persists strongly across successive funds by the same GP, unlike in mutual funds. But capital flows are sticky — top GPs raise somewhat larger funds but not enough to compete away returns, while poor performers continue raising capital. This is direct evidence of trust aggregation: LPs pay for the bundle because they cannot disaggregate, and the market does not efficiently price GP skill even when persistence data exists.

Metrick & Yasuda (2010) — About two-thirds of expected GP revenue comes from fixed components (management fees) insensitive to performance. Fund size, not fee rates, is the primary lever for GP wealth. This empirically confirms that the GP’s fee structure reflects trust aggregation rather than performance pricing — the majority of GP economics come from the commitment to the bundle, not from demonstrated alpha.

Phalippou & Gottschalg (2009) — After correcting for biases, the average PE fund underperforms the S&P 500 by ~3% per year net of fees. Total GP fees represent over 25% of invested capital. The industry’s opacity allows underperforming managers to continue raising. This is the empirical cost of trust aggregation: the information disadvantage is large enough that the median LP is paying for negative alpha and cannot detect it.

12. The LP’s fee reflects information disadvantage

Grossman & Stiglitz (1980) — Informed traders must be compensated for the cost of information through the gap between price and true value. Applied to PE: the LP pays a premium (2/20) that reflects the cost of the GP’s information advantage. If information were free, the premium would compress. The fee is not a service price — it is an information rent.

Glosten & Milgrom (1985) — Bid-ask spreads arise from adverse selection: the spread is a direct measure of information asymmetry. The GP’s bundled fee is the PE equivalent of a bid-ask spread — it reflects the LP’s inability to distinguish informed from uninformed components of the GP’s service.

13. Reinforcing mechanisms further stabilize the landscape

Cyert & March (1963) — Organizations rely on standard operating procedures, aspiration levels, and organizational slack that create inertia. The traditional GP/LP fund structure benefits from thick institutional infrastructure (lawyers, LPs, legal precedent, HR systems) that all assume the current model — a Cyert & March-style ecosystem of routines that resists change even when the underlying conditions shift.

March & Simon (1958) — Organizations satisfice rather than optimize, relying on routines and hierarchy to manage complexity. The current capital markets structure is a satisficing equilibrium — it works well enough that the coordination cost of switching to a better structure exceeds the perceived benefit.

14. Some regions remain degenerate even with perfect monitoring

Grossman & Hart (1986) — Even with complete information, some combinations of control rights and investment incentives are structurally incoherent. Full residual risk with no discretion is punishment without agency — the agent bears consequences for decisions they did not make. This is a property rights constraint, not an information constraint.

15. Bundle drift

Gibbons & Murphy (1992) — Career concerns create dynamic incentive shifts: the same role changes its effective contract over time as implicit incentives strengthen or weaken. Bundle drift is partly a Gibbons & Murphy dynamic — the explicit contract stays the same but the implicit contract (reputation, career stage, market conditions) shifts the effective primitive values.

Teece (1980, 1982) — Economies of scope are conditional and dynamic: the value of bundling multiple activities changes as the fit between activities and organizational capabilities evolves. Bundle drift occurs when the scope economics of a role change — what was once worth bundling becomes worth unbundling as external conditions shift.

16. When information control transfers to systems, the landscape reshuffles

Gorton & Pennacchi (1990) — The emergence of the loan sales market is explained by technological progress that reduced information asymmetries between sellers and buyers, making previously non-marketable assets marketable. This is a direct precedent: information technology transferred control from the originating bank to the market, and new contract types (loan sales) emerged that were previously impossible.

Hayek (1945) — If the economic problem is coordinating dispersed knowledge, and system-controlled information aggregates that knowledge more effectively than agent-mediated reporting, then the coordination mechanism changes. The system becomes the price mechanism for performance information that was previously agent-controlled.

Walsh & Ungson (1991) + Argote (1999, 2013) — Organizational memory fails at retrieval, and knowledge depreciates through turnover and inactivity. System-controlled information solves the retrieval problem (context is available at the moment of decision) and the depreciation problem (the system does not forget or leave). Both mechanisms that sustained agent information control are weakened.

17. Trust aggregation becomes contestable when per-component contribution is observable

Holmstrom (1979) — The informativeness principle again: if new signals about effort become available, the optimal contract changes to incorporate them. System-controlled information provides new signals about each component of the GP’s bundle. The optimal contract — which previously had to treat the bundle as monolithic — can now price components separately.

Kaplan & Schoar (2005) + Metrick & Yasuda (2010) — GP performance persists and fees are mostly fixed, but capital flows don’t efficiently respond. If per-component attribution becomes feasible, the market can finally price GP skill at the component level. The sticky capital / persistent underperformance equilibrium that Kaplan & Schoar and Phalippou document breaks when LPs can see which components generate value.

Sirmans, Sirmans & Turnbull (1999) — Property management fee structures already show predictable behavioral consequences — base fees encourage different behavior than leasing commissions than renewal fees. This is micro-level evidence that contract design at the component level changes behavior. If the GP’s bundle is decomposed into components with tailored contracts, each component’s behavior shifts accordingly.

18. The Accountable Executor becomes stable under system-controlled information

Holmstrom (1979) — The informativeness principle: outcome pay is only viable when effort-correlated signals are available. Under agent-controlled information, task-level outcome attribution was impossible — the contract collapsed to time pay (mercenary). System-controlled information provides the granular signals that make outcome pay enforceable at the executor level. The Accountable Executor is the contract Holmstrom’s framework always implied was optimal but that information constraints made impractical.

Holmstrom & Milgrom (1991) — The multitask problem previously killed this combination: paying a bounded executor on one measured outcome distorted effort away from unmeasured dimensions. System-controlled information resolves this by making multiple dimensions measurable simultaneously. The contract can reward balanced performance rather than single-metric optimization.

Sirmans, Sirmans & Turnbull (1999) — Property management contracts already show embryonic versions of this bundle: leasing commissions and renewal fees tie pay to specific outcomes rather than time. The Accountable Executor formalizes and extends this — a PM paid on system-verified NOI contribution, with partial risk and continuous monitoring. The primitive combination exists in nascent form; system-controlled information is what makes it stable.

19. The Direct Principal becomes stable under system-controlled information

Kaplan & Schoar (2005) — The largest, most sophisticated LPs already invest directly, bypassing GP intermediation where they have sufficient scale and internal capability. The Direct Principal equilibrium exists at the top of the market. What prevented it from extending downmarket was the information gap: smaller LPs lacked the staff and data to make direct investment decisions competently. System-controlled information closes that gap.

Grossman & Stiglitz (1980) — The cost of information acquisition determines who can afford to be informed. If system-controlled information reduces the cost of asset-level visibility from “large internal team” to “platform subscription,” the set of principals who can act as Direct Principals expands. The equilibrium that was previously restricted to sovereign wealth funds becomes accessible to mid-market LPs.

20. The Accountable Intermediary becomes stable under system-controlled information

Rau (2000) + McLaughlin (1990, 1992) — Completion pay without risk is the standard middleman bundle. Adding post-close risk was previously unstable because attribution across the deal window was too expensive — you could not determine whether a bad outcome was the broker’s fault or the market’s. System-tracked post-close outcomes with sufficient deal volume make statistical attribution feasible, stabilizing a contract with holdbacks and clawbacks that would previously have shed risk and reverted to the standard middleman.

Golubov, Petmezas & Travlos (2012) — Reputation currently substitutes for contractual risk in moderating completion bias among top-tier advisors. The Accountable Intermediary adds a structural enforcement mechanism alongside reputation. It works in high-volume contexts where the system can accumulate enough data for attribution — not in bespoke, one-off transactions where the sample size is too small for statistical comparison.