Pooling Equilibrium¶
Sustain a signaling-game equilibrium in which every hidden sender type chooses the same on-path action and the receiver's posterior remains its prior.
Core Idea¶
A pooling equilibrium is a signaling-game equilibrium in which all privately informed sender types choose the same on-path action. Because each type sends that action, its observation gives the receiver no information that discriminates types: the receiver's on-path posterior remains the prior. But shared observed behavior alone does not establish equilibrium. The receiver must respond optimally to its beliefs, and every sender type must prefer its prescribed action to available deviations under a coherent sequential assessment, including responses to off-path signals.[1][2]
This entry uses pure full pooling. Some signaling models discuss partial or mixed pooling, where only a subset of types shares an action or types randomize; those cases should not be silently substituted for the all-types/same-action test. Off-path beliefs are part of the equilibrium assessment but are not universally fixed to “any deviation means low type.” That is one possible model-specific support, not an identity rule.[1][2]
Structural Signature¶
Sig role-phrases: hidden type → common sender action → prior-preserving on-path belief → sequential incentive-and-response assessment.
- Privately informed type space. Nature selects a sender type known to the sender but hidden from the receiver. Without that asymmetry, same-action play is not pooling over private types.[1]
- Shared on-path sender action. Every type selects one same action in the assessed profile. Type-specific actions define a separating or hybrid pattern instead.[1][2]
- Prior-preserving on-path belief. Observing the common action does not update relative type probabilities, so the receiver's posterior at that information set equals its prior.[1]
- Sequential strategy-and-belief assessment. Receiver responses and all sender-type choices must be sequentially rational under specified beliefs and payoffs, including contingencies after deviations. Otherwise the pattern is observed co-action, not equilibrium.[1][2]
An exact wage, breakfast, signal cost or off-path posterior is setting-specific, not an extra universal role.
What It Is Not¶
Pooling is not mere coincidence that different actors happened to choose the same act once. The concept concerns a full type-contingent strategy and belief assessment. It is not separating equilibrium, where distinct types send distinguishable signals, nor automatically a hybrid in which one type mixes and an observed action can change the posterior.[1][2]
It is also not live Risk Pooling, an aggregation concept from insurance/portfolio risk. Live Signaling and Signalling Economics are topical neighbors but their current definitions emphasize revealing or credibly conveying hidden information; a pooled on-path action does not reveal type. They cannot be imposed as strict parents without reconciling that definition. Live Nash Equilibrium describes a strategy profile and lacks the explicit belief-assessment object used here.
Scope of Application¶
In a labor-market model, workers privately know productivity and choose education. If both types choose the same level, the employer learns nothing new from observing it and can pay the prior-weighted expected productivity. The MIT notes derive no-deviation conditions and illustrate one set of off-path employer beliefs; the authors say other off-path assignments may also work in that model.[2]
In the Beer and Quiche game, an informed strong or weak sender chooses a breakfast, after which an uninformed receiver decides whether to challenge. MIT's notes identify pooling equilibria in the original game and contrast a revised version with a separating equilibrium. The physical breakfast is not the transferable object; the mapping of private type, common message, receiver belief and sequential incentives is.[1]
An observed common act in an organizational or biological analogy should not automatically be labeled a pooling equilibrium: a signaling-game model with types, payoffs, beliefs and sequential checks must be supplied.
Clarity¶
The receiver's unchanged on-path posterior does not mean the receiver ignores all evidence everywhere. A deviant action is off-path under the proposed profile, and its interpretation must be specified for the equilibrium test. That off-path interpretation may differ among admissible assessments; it is not generally identified from the common action alone.[2]
Likewise, “all types send the same signal” is necessary in the pure full-pooling case but insufficient. If one type would earn more by deviating under the stated receiver response, the proposed profile fails the equilibrium part even while its observed action pattern is pooled.[1][2]
Manages Complexity¶
Pooling compresses the receiver's on-path inference: the observed action does not split the hidden type distribution, so one prior-weighted response can apply there. The compression is local. To assess stability, the model still needs the type-dependent sender payoffs, receiver best response, alternative messages and the beliefs associated with those alternatives.[1][2]
Abstract Reasoning¶
For a candidate profile, first ask whether every type chooses the same action. If not, pure full pooling is absent. If yes, use Bayes' rule at that reached action: because all types choose it, the posterior is the prior. Then check receiver optimality and each sender type's deviation incentives given specified off-path responses. This ordered test keeps the observational pattern separate from equilibrium certification.[1][2]
Changing signal costs or belief responses can make a previously sustainable shared action fail or make a separating/hybrid outcome possible. That sensitivity is a model result, not a change in the definition of pooling.[2]
Knowledge Transfer¶
Education/wages and breakfast/challenge are unlike carriers. In both, hidden type determines incentives, types select one action, the receiver sees the action but not the type, and a sequential belief-and-response assessment determines whether the common action is stable. This is exact role transfer within signaling games; “pooling” in risk management has a different mechanism. The general game-theoretic notion of equilibrium is a conceptual ancestor, but no current live node defines the same strategy-plus-belief genus precisely enough for a strict edge.[1][2]
Examples¶
Labor-market education. The private types are high- and low-productivity workers. Under one proposed shared action, both choose education \(e^*\). The employer's on-path belief remains the prior and the offered wage reflects prior-weighted expected productivity. The sequential assessment also specifies wages after other education levels and checks that neither type gains by choosing them. The notes give one supporting belief scheme but not a universal one.[2]
Mapped back: the same education level is only the visible role; the prior-preserving wage and no-deviation check are what make it a pooling Equilibrium.
Beer and Quiche. The private types are strong and weak senders. In a source-described pooling profile, both choose one shared breakfast; the receiver's on-path type belief stays at the prior; and the subsequent challenge response plus beliefs after the unused breakfast enter the sequential assessment. The revised game in the notes has a different, separating outcome, showing that merely having the same action menu is not enough.[1]
Mapped back: breakfast and education are not equivalent objects, but both fill the common sender-message role under hidden-type inference and incentive tests.
Negative boundary. In MIT's hybrid education case, the low type chooses zero education and the high type mixes between zero and another level. Observing the second level reveals information, while the posterior after zero differs from the prior. This is not pure full pooling even though both types can sometimes choose zero.[2]
Structural Tensions¶
- On-path ignorance versus off-path discipline. The common signal simplifies inference at the reached action, but a stable equilibrium also needs behavior and beliefs after unused alternatives. Assigning punitive beliefs can sustain pooling in a model; unjustified beliefs can make a claimed equilibrium fragile. Diagnostic: Under the declared off-path response, does each type still prefer the common message?[2]
- Shared signal versus type revelation. A common act preserves the prior, while different actions can let the receiver infer type and choose a more tailored response. Signals costly enough for some types may support separation, but cost alone does not determine which equilibrium is selected. Diagnostic: Which type's incentive constraint fails first when the message cost or receiver response changes?[1][2]
- Model economy versus empirical identification. A pooling assessment reduces many type-contingent histories to one reached action, yet the same observed behavior can coexist with distinct off-path beliefs. Inferring the entire equilibrium from observed actions alone overstates what data reveal. Diagnostic: Which unused-action beliefs are actually identified, and which are model assumptions?[2]
Structural–Framed Character¶
Pooling Equilibrium is mixed-structural, leaning formal: its common-action/posterior condition is exact inside a specified signaling game, while payoffs and beliefs encode modeled human or strategic behavior. Its evaluative weight is not an endorsement of pooling as efficient or fair; equilibrium means sequential incentives are satisfied under the model. Its human-practice dependence appears in education or breakfast examples and the construction of payoff and information assumptions, but the same test could be applied to any valid signaling-game carrier. Its institutional origin is game-theoretic analysis, not a rule made true by an economic institution. Its vocabulary travel is legitimate between unlike signaling settings only when hidden types, common action, unchanged posterior and incentive assessment recur. Import versus recognition excludes calling ordinary risk aggregation or accidental same-action behavior a pooling equilibrium.
The portable skeleton is a strategic equilibrium under incomplete information, but the current live Nash Equilibrium node lacks the belief-assessment genus, and the current Signaling nodes emphasize information revelation. These are related live objects, not asserted parents. Its character: a precise signaling-game equilibrium pattern whose formal role test is portable within that domain but not prime-general.
Structural Core vs. Domain Accent¶
What is skeletal. Privately informed types choose actions, an uninformed receiver updates beliefs and responds, and strategies and beliefs must be sequentially rational. This is the conceptual candidate for a future signaling-game-equilibrium parent. Current live Nash Equilibrium captures no-profitable-unilateral-deviation at the strategy-profile level, but does not define the necessary belief assessment; therefore no strict live parent is staged.
What is domain-bound. Full pooling adds one common on-path sender action across all types and a prior-preserving posterior at that action. Education levels, wages, breakfasts, challenges and a particular off-path belief assignment are setting-specific. Remove the common action or the incentive/belief assessment and the identity becomes partial pooling, separating behavior, or mere co-action.
Why this is not a prime. Strategic equilibrium, information asymmetry and belief updating have broader abstractions, but this named form is recognized only in signaling games with private types and sequential assessment. Applying the phrase to combined insurance risks imports vocabulary without the sender/receiver mechanism. A future exact genus might connect it to the live supergraph; lack of that genus is not a reason to force a false strict edge now.
Instantiates / Related Primes¶
No strict typed parent relation is asserted in the current DAG.
Neighborhood in Abstraction Space¶
Pooling Equilibrium sits in a sparse region of the domain-specific corpus (64th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Network Security Vulnerabilities & Trust (26 abstractions)
Nearest neighbors
- Bayesian Nash Equilibrium — 0.86
- Bayesian Persuasion — 0.86
- Cheap Talk — 0.84
- Perfect Information — 0.84
- Noisy Channel Model — 0.84
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
- Separating equilibrium. Different sender types use different actions and the receiver may infer type. Tell: are all types prescribed one identical action?[1]
- Hybrid or partial pooling. A subset shares an action or a type mixes, so some observations can update beliefs. Tell: does the posterior after every reached action equal the prior under the claimed full-pooling profile?[2]
- Common behavior without equilibrium. Same acts can occur despite profitable deviations. Tell: have strategies, off-path beliefs and sequential best responses been checked?[1]
References¶
[1] MIT OpenCourseWare 14.12 F12, Dynamic Games with Incomplete Information, Chapter 16, PDF pp. 2–4 and 9–11, especially Definition 16.6 on p. 10. The PDF identifies itself as F12 despite a later course-directory URL. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p
[2] Mihai Manea, Non-Cooperative Games, lecture notes dated 19 January 2017, hosted by MIT OpenCourseWare 14.126, PDF pp. 38 and 57–58. Directly supports education-model pooling, prior-based wage, deviation conditions and hybrid contrast. registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r