Bargaining Power

Version

v4 · 2026-08-30 · History

Prime #

1322

Origin domain

Economics

Subdomain

negotiation → Economics

Related primes

Asymmetry

Core Idea

Bargaining power is a relational advantage that lets one party move the terms of an exchange toward itself because failing to reach agreement costs it less than it costs the other side. ^[1] The operative quantity is not wealth, size, or status but the gap between what each party receives under agreement and what it receives under its best available alternative to agreement; the side whose fallback sits closer to the deal can hold out longer and therefore claims more of the jointly created surplus. ^[2]

The canonical identity is narrower than the phrase's everyday use. Four commitments must hold together: at least two parties can accept, reject, delay, or revise the terms; each holds a disagreement payoff backed by an alternative it could actually take; the parties know or can infer how those payoffs compare; and the comparison changes which demands, concessions, or imposed terms are feasible. Drop any one and the classification fails. Power in this sense is comparative and situational — a property of a pair and a moment rather than a stock either party carries from room to room, which is why the same actor is formidable in one relationship and helpless in the next.

Structural Signature

Bargaining power encodes a structural pattern: two disagreement payoffs → unequal dependence on this particular agreement → a shifted division of the joint surplus, bounded at both ends by the parties' fallbacks. Every step of the pattern is comparative; nothing about either party's absolute resources enters it, only the ratio of what each stands to lose by walking away. ^[3]

Recurring features:

• A surplus that exists only under agreement. The joint value of settling exceeds the sum of what the parties obtain separately, so a range of terms exists that both prefer to no deal; without that range there is nothing to divide and no power to exercise. ^[1]
• A disagreement payoff attached to each side. The next-best buyer, the in-house alternative, the strike fund, the litigation track, the unchanged status quo — each party carries a value for the world in which this deal never closes.
• Unequal exposure to delay, replacement, or exit. The differential in these costs, rather than the absolute level of either party's resources, is what moves terms.
• A credibility gate. An alternative counts only insofar as the counterpart believes it would be taken; an option that is real but invisible, or visible but implausible, exerts no pull on the settlement.
• An adjustable settlement variable. Price, wage, delivery schedule, indemnities, scope, exclusivity, or the allocation of residual risk must be movable, or the asymmetry has nowhere to register.
• Latency. The advantage can sit entirely unused and still shape terms, because the counterpart is choosing against the option it knows exists rather than against the demand actually made.
• No table required. Posted prices, standard-form contracts, and take-it-or-leave-it offers qualify whenever refusal remains the counterparty's operative response.

What It Is Not

The abstraction claims considerably less than the everyday phrase suggests, and most misreadings come from over-reading it.

It does not claim that the stronger side gets everything. Power operates inside the range that both parties prefer to no deal; it moves the settlement within that range, and the weaker party's own fallback still sets a floor beneath which it walks. A prediction that one side captures the entire surplus is a prediction that the other side's alternative is worth nothing at all, which is a strong empirical claim rather than a consequence of disparity. ^[1]

It is not a claim about character, nerve, or skill. Preparation and composure affect how reliably a party realizes the terms its position already supports, but they are not the position. A negotiator who is calm, articulate, well briefed, and out of alternatives occupies a weak position that is being played well.

It is not a moral verdict, in either direction. The mechanism is silent on whether a division is just, on whether the option set was arrived at fairly, and on whether the disparity should be regulated away. Those judgements require premises the abstraction does not supply and cannot generate.

It does not require hostility, deliberate strategy, or awareness. Parties who like each other, who have never modelled anything, and who would sincerely deny exerting leverage still settle where their alternatives put them; the asymmetry works through what each side is willing to accept, not through anyone's intent. ^[4]

Finally, it is not a point prediction. The abstraction is directional. It says which way terms move when a fallback improves or a delay cost rises; it does not say what the number will be.

Broad Use

The same comparison of fallbacks organizes settings that share no vocabulary and no institutions.

Labor and employment. Wage and condition outcomes track the tightness of the external hiring market, the depth of the strike fund, the employer's cost per idle day, and the availability of replacement workers far more closely than they track the stated positions of either side. ^[5]

Procurement and supply chains. A buyer's terms improve with each qualified second source and degrade with each proprietary component, dedicated tooling investment, or certification that makes a switch slow.

Litigation and settlement. The settlement figure is set by each side's expected value at trial net of costs and delay, which is why a defendant with a fast, cheap, and predictable trial path concedes little regardless of the merits rhetoric exchanged in mediation.

Diplomacy and sanctions. A state's leverage tracks its capacity to endure the absence of an agreement — stockpiles, alternative trading partners, domestic tolerance for hardship — rather than its aggregate size or its formal standing.

Platform governance. The terms a platform imposes on sellers, creators, or developers tighten as the share of their revenue that arrives only through that platform rises, and loosen as soon as a credible second channel appears.

Households and care. Divisions of money, work, and time within a household respond to each member's independent income, exit options, and the cost of dissolution, even where the participants describe the arrangement entirely in the language of preference or fairness.

Biological exchange. Partners in mutualisms trade at rates that shift with the number of alternative partners available and the cost of remaining unpaired, producing a market-like adjustment of exchange rates with no negotiation, deliberation, or language anywhere in the system. ^[4]

Clarity

The abstraction earns its place by dissolving four persistent confusions.

The first is between being strong and being needed less. Everyday usage assigns power to whoever is larger, louder, richer, or more senior, which produces the standing puzzle of the enormous firm that cannot extract better terms from a tiny sole-source supplier. Read as a comparison of fallbacks, the puzzle dissolves: the supplier has alternative buyers and the firm has no alternative supplier. ^[6]

The second is between the size of the surplus and its division. Parties routinely argue about one while believing they are arguing about the other — a supplier defends its cost structure when the dispute is about the split, a buyer proposes efficiency measures when what it wants is a smaller share for the other side. Separating creation from division tells each side which of its arguments could possibly move anything.

The third is why concessions so often buy nothing. A concession that leaves both disagreement payoffs untouched changes the atmosphere and not the equilibrium, and the terms drift back toward where the alternatives put them. Only a move that changes what someone can do without the deal has purchase on the outcome. ^[3]

The fourth is among three very different sources of a lopsided result: one side had better options, one side was deceived, or one side was left no acceptable choice at all. These look alike from outside and call for different responses — building alternatives, compelling disclosure, or prohibiting the conduct. Naming the mechanism tells you which remedy is even relevant before anyone argues about whether to apply it.

Manages Complexity

What the abstraction buys is permission to stop tracking most of what a negotiation visibly consists of.

You can stop tracking positions. Opening demands, anchors, and declared red lines are moves in a sequence whose endpoint is set by the alternatives standing behind them, so the transcript is largely uninformative about where the deal lands. You can stop tracking style and personality for the same reason. You can also stop tracking the order in which concessions were made, because what binds at the close is the option set present at the close and not the path that produced it.

You can track justifications lightly. Each side supplies a rationale — cost pressure, fairness, precedent, prevailing market rates — and these rationales are shaped by the position at least as much as they shape it, though experimental work on self-serving fairness judgements finds they can also drive a negotiation to impasse rather than merely decorate one. ^[7]

What survives the compression is a short state vector: each party's disagreement payoff, each party's cost per period of continued delay, the credibility of each fallback in the other's eyes, and the range of terms both still prefer to no deal. Four quantities and a range replace an unbounded record of interaction, and all four can be estimated from outside the room and in advance by someone who knows the industry rather than the people. That is why a competent analyst can predict the shape of a settlement without observing the negotiation at all.

Abstract Reasoning

The abstraction licenses a specific diagnostic sequence rather than a general attitude.

Begin by naming the settlement variable and the parties who can move it. Then write, for each party, the world in which no agreement is ever reached: what it does on the first morning after talks fail, what that is worth, and how fast that value decays. Third, apply the credibility gate to each fallback — ask whether the party would actually take it, whether the counterpart can verify it, and what would have to happen for it to be abandoned under pressure. Fourth, compare the two decay rates, since the side that loses less per week of delay can convert time into terms. Fifth, state a direction rather than a number: an improvement in one party's fallback moves terms toward that party by an amount the mechanism does not specify. ^[8]

Two tests keep the diagnosis honest. The swap test exchanges the parties' alternatives while holding everything else fixed; if the predicted terms move, an option asymmetry is doing the work, and if they do not, something else — legitimate authority, deception, or outright force — is producing the outcome. The intervention test asks what would have to change for the settlement to change, and accepts only answers of the form "this party's alternative improves." An answer of the form "this party argues more forcefully" fails the test, which is the formal version of a familiar observation: a party whose fallback is worse than the deal in front of it, facing a counterpart who can leave at no cost, is bluffing rather than bargaining, however confident it looks.

Knowledge Transfer

What ports between substrates is the comparison and its consequences: the ranking of disagreement payoffs, the credibility gate that decides which alternatives count, the convertibility of patience into terms, and the prediction that outcomes respond to changes in options rather than to changes in argument. Those four survive translation from a union hall to a procurement portal to a mutualistic symbiosis, because none of them mentions money, speech, or intention. ^[4]

What does not port is more interesting. The closed-form division rules do not: arithmetic that splits a surplus in proportion to patience assumes a measurable pie, transferable value, and shared knowledge of both fallbacks, and most substrates supply none of the three. The unitary-party assumption does not: a side that is in fact a coalition runs an internal bargain of its own, and internal division can either destroy an external threat or make it unbreakable, inverting the prediction in either direction. The exogeneity of the options does not: where one party writes the rules that determine what the other can do instead, fallbacks are outputs of the very process the model treats them as inputs to.

The evaluative charge does not port either, and carrying it across is the commonest failure. In a commodity market a disparity in options is unremarkable; in a household, a clinic, or an asylum interview the identical structure is read as a problem demanding intervention. The mechanism is the same in all four settings, and the response is supplied by the setting rather than by the abstraction.

Examples

Formal/abstract

Alternating offers under unequal impatience. Two parties divide a surplus of size one. They alternate proposals: one proposes, the other accepts or counters in the following period, and so on without a deadline. Delay is costly, and the two parties discount the future at different rates. They are identical in every other respect — same information, same rules, same alternatives — so any difference in outcome is attributable to the cost of waiting alone. The equilibrium division favours the more patient party, and the advantage widens as the impatient party's per-period loss grows; in the limit where one side loses nearly nothing by waiting and the other loses a great deal, the patient side takes almost the whole surplus while still leaving the other just enough to prefer agreeing to walking.

Now attach an outside option to one party. If the option pays less than that party's equilibrium share, nothing changes at all: the option is real but does not bind, because the party would never exercise it. If it pays more, it truncates the division at the option's value, and the entire gain accrues at the moment the option becomes credible rather than at the moment it is invoked. ^[9]

Mapped back: the model isolates the two inputs the abstraction says matter and shows that they behave differently. Relative impatience acts continuously, shading the division by degrees. An outside option acts as a floor that is either binding or irrelevant, with nothing in between. The setup also shows why unexercised options are not idle: the patient party's advantage is realized in the very first accepted offer, and no delay ever occurs along the equilibrium path. Power that is never used is still the thing setting the terms.

Applied/industry

A cloud commitment renewal. A mid-market software company is eighteen months from the end of a three-year infrastructure commitment with a single provider. Its stated position is strong: the account is growing, competing quotes are in hand, and its executives are prepared to say that everything is on the table. Its actual disagreement payoff is poor. Its data sits in a proprietary managed database, a third of its services depend on provider-specific queues and identity primitives, moving several petabytes elsewhere carries egress charges, and an internal estimate puts migration at nine months of the platform team's entire capacity. The provider can reconstruct most of this from the account's own usage telemetry. The renewal lands close to list price, and the tone of the negotiation has nothing to do with it. ^[10]

The terms change only when the option set does, and that work is engineering rather than rhetoric. Over the following year the company moves its data layer onto a portable engine, wraps provider-specific services behind internal interfaces, runs a standing ten percent of production traffic on a second provider, and reserves capacity there. None of this is done in order to leave. It is done so that leaving becomes a nine-week project instead of a nine-month one, and so that the claim is verifiable by the counterpart rather than merely asserted by the company.

Mapped back: the case separates the three things the diagnostic keeps apart. The stated position was never the constraint. The credibility gate is what the migration estimate failed, which is why the competing quotes bought nothing. And the leverage was created between negotiations rather than during one, which is the practical form of latency. The case also shows the endogeneity problem in miniature: the provider's advantage came partly from architectural choices it had encouraged for years, so the fallback the model wants to treat as given was in part manufactured by the party the model says benefits from it.

Structural Tensions

T1 — The credibility gate punishes discretion. An alternative shifts terms only when the counterpart believes it, so a party that has quietly qualified a second source, secured a rival offer, or rehearsed a migration gains nothing until it discloses. Disclosure converts the option into terms and simultaneously exposes it: incumbents respond with pre-emptive discounts, exclusivity riders, litigation, or by acquiring the alternative outright. The option is worth most when it is known and safest when it is hidden, and no choice of timing satisfies both conditions at once.

T2 — Extraction erodes its own basis. The division that maximizes the current period pushes the counterpart to the edge of the terms it can still accept, which is exactly the condition under which investing in alternatives starts to pay for it. Buyers who press suppliers hardest fund the second source that will later discipline them; employers who capture the whole surplus of a tight market train the workforce that leaves. Deliberately leaving surplus on the table is not generosity but maintenance of the dependence the position rests on.

T3 — The inputs are usually outputs. The diagnostic wants each party's outside option as exogenous data, but in durable relationships the advantaged side has generally shaped it: exclusivity terms, proprietary formats, non-compete clauses, licensing gates, debt, and visa sponsorship all narrow what the counterpart could do instead. Once fallbacks are manufactured, asking who holds power and asking who wrote the option set become the same question, and an analysis that stops at the current alternatives will describe the symptom and miss the mechanism entirely.

T4 — Patience wins a prize it does not want. The party that loses least from delay claims the largest share, but a low delay cost usually means the agreement matters little to it. The relationships in which one side holds the most power are therefore the ones it cares least about, and what it extracts is small next to what the dependent side would have paid for terms it could live with. When both sides are patient, the mechanism predicts near-equal division and long delay, with the surplus burning quietly while they wait each other out.

T5 — Power is inferred from the thing it explains. Disagreement payoffs are rarely observable, so analysts read them backwards out of the settlement and then present the settlement as evidence of the disparity. The reasoning closes on itself unless an independent measurement is available: a documented competing offer, an observed walkaway, or a change in one side's options driven by something unrelated to the negotiation. Without one, the abstraction can absorb any outcome whatever, and its apparent explanatory reach is a warning rather than a strength.

T6 — One mechanism, two verdicts. That terms follow options is used as an indictment, since the division reflects dependence rather than contribution, and as an exoneration, since both sides preferred agreeing and nothing was taken by force. Both readings are consistent with the structure, because the abstraction is silent about how the option set arose and about which disparities a society should tolerate. Analysts routinely import the missing premise while presenting the result as neutral description, and the vocabulary's economic provenance makes that import hard to see.

Structural–Framed Character

Bargaining Power sits on the framed side of the structural–framed spectrum, labeled mixed-framed at an aggregate of 0.5, with all five diagnostics reading at half — an even balance rather than one criterion carrying the grade.

The skeleton that travels is comparative and option-dependent. At least two actors can accept, reject, delay or revise terms; each has a credible outside option describing what it obtains if no agreement is reached; the costs of delay, exit, replacement or nonagreement are unequal; and the less-dependent side can move price, obligations, timing, scope, risk or surplus division toward itself. Power here is not an intrinsic stock — not size, status, force or formal office — and can be latent without being exercised. Labor, diplomacy, procurement, household negotiation, platform governance and coalition formation preserve the same outside-option and disagreement-cost structure.

Human-practice-bound does the most work at its half weight. The roles are stated formally, in disagreement payoffs and surplus division, but each presupposes parties who can refuse — which is what the entry means when it calls the abstraction socially framed yet portable across institutional substrates. Institutional origin reads half for the same reason: the paradigm settings are negotiated, though no particular institution is constitutive. Vocabulary travels at half, bringing negotiation and bargaining-theory terms along. Evaluative weight is half, since power carries a distributive tone the mechanism itself does not assert. Import-vs-recognize is half.

The grade means the prime is portable but not free: confirm the outside options are real. Where meaningful refusal has been removed, what you have is coercion, not bargaining power.

Substrate Independence

Bargaining Power is a highly substrate-independent prime — composite 4 / 5 on the substrate-independence scale. The portable content is comparative, mentioning neither wealth nor size: each side holds a value for the world where no agreement is reached, the side whose fallback sits closer to the deal can wait longer, and the division of the jointly created surplus shifts toward it, bounded at both ends by those two fallbacks. Labor negotiation, diplomacy, procurement, household allocation, platform governance, and coalition formation instantiate it without adjustment. Its limit is written into the statement itself, which remains socially framed: the roles require parties that value outcomes, can refuse or delay, and can infer how their alternatives compare, so it crosses institutional substrates freely while staying inside the class of interacting agents.

• Composite substrate independence — 4 / 5
• Domain breadth — 4 / 5
• Structural abstraction — 4 / 5
• Transfer evidence — 4 / 5

Relationships to Other Abstractions

Current abstraction Bargaining Power Prime

Parents (1) — more general patterns this builds on

• Bargaining Power is a kind of Asymmetry Prime

  Bargaining power is an asymmetry specialized to unequal credible exit, delay, and replacement options that shifts attainable terms between actors.

Children (2) — more specific cases that build on this

• Cornering the market Domain-specific is a kind of Bargaining Power

  The proposed strict upward parent is prime:bargaining_power.

• Market power Domain-specific is a decomposition of Bargaining Power

  Market power is the price-theoretic specialization of bargaining power: finite substitution makes the counterparty's exit option costly and lets the actor move terms and surplus toward itself.

Hierarchy path (1) — routes to 1 parentless root

• Bargaining Power → Asymmetry

Neighborhood in Abstraction Space

Bargaining Power sits in a sparse region of abstraction space (81^st percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.

Family — Game-Theoretic Strategy & Equilibrium (23 primes)

Nearest neighbors

• Competition — 0.71
• Separation of Powers — 0.70
• Coordination — 0.69
• Property Rights — 0.68
• Preference Heterogeneity and Conflict — 0.68

Computed from structural-signature embeddings · 2026-09-10

Not to Be Confused With

Bargaining power must first be separated from Asymmetry, its parent. Asymmetry names any directed imbalance in a relation whose two sides are not interchangeable under swap, and it makes no commitment about what is imbalanced or about what the imbalance does. Bargaining power fixes both. The asymmetric quantity must be the cost and credibility of not agreeing, and the imbalance must register in terms that are actually settled — price, obligation, timing, scope, or the allocation of risk. A relation can be thoroughly asymmetric in attention, affection, information, or effort while the two sides hold equally good alternatives, and such a relation is not an instance of this prime. The specialization is what supplies the predictive content: knowing that a relation is asymmetric tells you almost nothing about the terms, while knowing which side can walk away more cheaply tells you which direction they will move.

Coercion is the nearest and most consequential confusion. Coercion shapes another agent's choice by manipulating the costs and threats attached to its options so that the agent itself selects the coercer's preferred action; bargaining power works through options the counterpart already holds. The dividing line is whether meaningful refusal survives. A supplier facing a buyer with three qualified alternatives accepts poor terms and could have refused them; a party facing a credible threat of ruin has had refusal removed rather than priced. The distinction is not academic, because the remedies diverge sharply: disparities in options are addressed by building alternatives, while the removal of refusal is addressed by prohibition. Deterrence sits adjacent and differs again, arranging consequences so that an action is not taken at all rather than dividing a surplus, and it operates perfectly well where no joint gain exists for anyone to split.

Monopoly, and the broader notion of market power it exemplifies, is a generator of bargaining power rather than the thing itself. Monopoly is a structural fact about a market: a single locus controls access to something for which there are no close substitutes. It usually confers bargaining power, but it carries apparatus the prime does not require — a residual demand curve, a marginal-cost benchmark, a price-cost wedge, quantity restriction, and a welfare accounting of the loss. Bargaining power needs none of these and appears routinely where there is no market, no price, and no quantity to restrict: a custody arrangement, a treaty text, a committee dividing a fixed budget. The converse case is equally instructive. A monopolist facing a single buyer who can do without the product entirely holds structural market power and very little bargaining power in that particular exchange.

Authority differs in the direction from which the binding comes. Authority is the recognized, legitimate right to issue binding decisions within a defined scope, and its force derives from that recognition rather than from what the bound party could otherwise do. A regulator's order binds a firm with excellent alternatives; a court's ruling binds the party with the stronger walkaway. Where authority operates there is frequently no surplus being divided and no acceptance to withhold, so the constitutive roles are simply absent. The two can coexist and point in opposite directions — an agency with unquestioned authority over a licence may have almost no bargaining power over the contract terms offered by its sole capable vendor — which is precisely why they must be separated before either is used to explain an outcome.

Credible Commitment is the most common category error in the other direction. Deliberately constraining one's own future choices is a technique for improving a bargaining position, not a synonym for holding one. A ratification requirement, a published price floor, or a most-favoured-nation clause raises the cost of one's own concession and therefore moves terms, but the commitment is an intervention on the disagreement structure while bargaining power is the comparative advantage that results. Collapsing the two conceals the cases where commitment is unavailable, is not believed, or backfires by making agreement impossible where it was previously merely expensive. Much the same holds for Switching Cost and Lock-In, which explain why one party's alternatives are costly, and for Information Asymmetry, which describes unequal private knowledge: each can generate a disparity in credible options, each can be present with no such disparity at all, and none of them is the disparity. Narrative Persuasion completes the set from the far side. It shifts what a party believes or wants while leaving the cost of nonagreement exactly where it was, which is why a persuaded counterpart concedes only until it recomputes its alternatives, and why the two produce identical-looking concessions with opposite durability.

Solution Archetypes

No catalogued solution archetypes reference this prime yet.

Notes

The prime's economic vocabulary is narrower than its reach. Nothing in the identity requires money, contracts, or markets, only parties who can refuse and alternatives that differ in value, which is why applications in kinship, clinical consent, and biological exchange are instances rather than metaphorical extensions.

Two open questions are worth flagging for later work. The first is measurement: outside options are rarely observed directly, and the diagnostic above depends on estimating them without reading them back out of the outcome they are meant to explain. The second is the endogeneity of the option set, which recurs in the tensions and in the applied example; a full treatment of how dependence is manufactured probably belongs in a neighbouring entry rather than in this one.

Contested claims

• [026] The formal bargaining literature and the behavioural literature disagree about whether stated rationales do any causal work. Non-cooperative bargaining theory (Rubinstein 1982; Muthoo 1999) predicts the settlement from disagreement payoffs and impatience alone, with justifications entering nowhere – which is the reading the sentence took. Experimental and behavioural work disagrees: Babcock and Loewenstein (1997) show that role-induced self-serving fairness judgements measurably cause impasse and costly delay, and that interventions acting only on those judgements change settlement rates.

References

[1] Nash, John F. "The Bargaining Problem". Econometrica 18(2), 1950. Establishes the disagreement point and the individual-rationality constraint that bound an attainable division of joint surplus, so each party's fallback sets a floor beneath which it will not settle and improving that fallback shifts the solution toward it. registry ↩a ↩b ↩c

[2] Fisher, Roger, and William Ury. Getting to Yes: Negotiating Agreement Without Giving In. Houghton Mifflin, 1981. Introduces the best alternative to a negotiated agreement (BATNA) and argues that relative negotiating strength depends primarily on how attractive not reaching agreement is to each side. registry ↩

[3] Rubinstein, Ariel. "Perfect Equilibrium in a Bargaining Model". Econometrica 50(1), 1982. Shows the unique subgame-perfect division of an alternating-offer bargain is fixed by the parties' relative impatience and disagreement payoffs rather than by absolute resources or by the sequence of concessions made along the way. registry ↩a ↩b

[4] Noe, Ronald, and Peter Hammerstein. "Biological markets: supply and demand determine the effect of partner choice in cooperation, mutualism and mating". Behavioral Ecology and Sociobiology 35(1), 1994. Shows exchange rates between partners shifting with partner availability and the cost of remaining unpaired, in systems with no deliberation, language or intent – the cross-substrate case for the comparison porting without money, speech or awareness. registry ↩a ↩b ↩c

[5] Hicks, John R. The Theory of Wages. Macmillan, 1932. Models the wage settlement as the intersection of an employer concession curve and a union resistance curve, both functions of the expected cost of a work stoppage, so terms track each side's cost of disagreement rather than the positions either side states. registry ↩

[6] Emerson, Richard M. "Power-Dependence Relations". American Sociological Review 27(1), 1962. Establishes that power in a relation resides in the other party's dependence – which varies inversely with the alternatives available to it – rather than in either party's size or resources. registry ↩

[7] Babcock, Linda, and George Loewenstein. "Explaining Bargaining Impasse: The Role of Self-Serving Biases". Journal of Economic Perspectives 11(1), 1997. Shows that assessments of a fair settlement are biased by the role a party occupies, and that this role-induced rationale is itself a measurable cause of costly delay and impasse. registry ↩

[8] Muthoo, Abhinay. Bargaining Theory with Applications. Cambridge University Press, 1999. Formal treatment of outside options, inside options, risk of breakdown and impatience, deriving the directional comparative statics by which an improved fallback or a lower cost of delay moves the settlement toward its holder without fixing a number. registry ↩

[9] Binmore, Ken, Avner Shaked, and John Sutton. "An Outside Option Experiment". The Quarterly Journal of Economics 104(4), 1989. Tests the outside-option ('deal-me-out') principle and finds the data favour it decisively over split-the-difference: an option constrains the division only when it exceeds the holder's equilibrium share, and otherwise leaves the split untouched. registry ↩

[10] Williamson, Oliver E. The Economic Institutions of Capitalism: Firms, Markets, Relational Contracting. Free Press, 1985. Describes the fundamental transformation by which relationship-specific investment converts a competitive supply situation into bilateral dependence at renewal. registry ↩