General · EdgepediaOctober 10, 20269 min read

Nash bargaining solution

The Nash bargaining solution is the unique outcome of a two-person bargaining problem that maximizes the product of the players' utility gains over the disagreement point, the payoff each receives if negotiation fails. John F. Nash introduced it in "The Bargaining Problem" (Econometrica, volume 18, pages 155–162, 1950) and proved that it is the only solution satisfying four axioms: Pareto optimality, symmetry, scale invariance, and independence of irrelevant alternatives.¹^•² It remains the foundational solution concept in bargaining theory, alongside the Kalai–Smorodinsky and egalitarian solutions axiomatized in the 1970s.³

Key fact	Detail
Formula	Maximize $(u_1 - d_1)(u_2 - d_2)$ over feasible utility pairs $(u_1, u_2) \ge (d_1, d_2)$ , where $d$ is the disagreement (threat) point¹^•⁴
Axioms	Pareto optimality, symmetry, scale invariance (invariance to affine utility transformations), and independence of irrelevant alternatives uniquely characterize it²^•⁵
Risk aversion	All else equal, the solution awards a lower share of the surplus to a more risk-averse bargainer, who concedes more fearing breakdown²
Asymmetric version	Maximize $(u_1 - d_1)^{\alpha} (u_2 - d_2)^{1-\alpha}$ with bargaining power $\alpha \in (0,1)$ ; $\alpha = 1/2$ is the symmetric case⁶
Rubinstein limit	In alternating-offers bargaining with discount factor $\delta$ , the unique subgame-perfect division is $(1/(1+\delta), \delta/(1+\delta))$ , converging to the Nash solution as $\delta \to 1$ ⁷
Laboratory record	Experiments support efficiency, symmetry, IIA, and monotonicity but reject scale invariance; the Nash and Kalai–Smorodinsky solutions perform poorly in the lab⁸

The bargaining problem and Nash's axioms

A bargaining problem is a pair $(S, d)$ : a feasible set $S$ of utility pairs and a disagreement point $d$ in utility space. Nash imposed four requirements on any rule $F$ selecting an outcome for every such problem.¹

Pareto optimality. A solution $f(U, d)$ is Pareto efficient if there is no feasible $v$ with $v \ge f(U,d)$ and $v_i > f_i(U,d)$ for some player. An inefficient outcome is unlikely to persist, since it leaves room for renegotiation.⁹

Symmetry requires that if the players are interchangeable, the outcome treats them identically. Scale invariance (invariance to affine transformations of utility) says the solution should not depend on the arbitrary units in which each player's utility is measured. Independence of irrelevant alternatives (IIA) says that if the feasible set shrinks but the previously chosen outcome remains feasible, the choice does not change; formally, if $S \subseteq T$ , both problems share disagreement point $d$ , and $F(T,d) \in S$ , then $F(S,d) = F(T,d)$ .¹⁰ Nash justified this property by saying that there is "no action at a distance" in the determination of the solution.⁷

Nash's theorem states that exactly one solution satisfies all four axioms: the point of $S$ at which the product of the players' utility gains from $d$ is largest among all points dominating $d$ .¹^•¹¹ The result extends to arbitrarily many agents: in an $n$ -player problem the unique solution satisfying the axioms still maximizes the product of utility gains over the disagreement point, and Lensberg (1988) provides an alternative axiomatization.²

How the solution works

The solution solves

$\max_{v_1, v_2} \; (v_1 - d_1)(v_2 - d_2) \quad \text{subject to } (v_1, v_2) \in U, \; (v_1, v_2) \ge (d_1, d_2).$

Existence follows from compactness of the feasible set $U$ and continuity of the objective, and uniqueness from strict quasi-concavity.⁹ Geometrically, the solution lies at the point on the Pareto frontier where the hyperbola $(u_1 - d_1)(u_2 - d_2) = c$ is tangent to the frontier for the highest attainable $c$ .⁶ Equivalently, on a smooth frontier it picks the point where the absolute elasticity of utility gains, $\left| (du_2/(u_2-d_2))/(du_1/(u_1-d_1)) \right|$ , equals 1.²

Risk aversion matters. The Nash solution punishes risk aversion: all else equal, it awards a lower portion of the surplus to a risk-averse agent, who concedes more fearing bargaining breakdown.² Because of scale invariance, only the curvature of each utility function, not its units, affects the split.

The asymmetric (generalized) version assigns bargaining power $\alpha \in (0,1)$ to Player 1 and $1 - \alpha$ to Player 2, maximizing $(u_1 - d_1)^{\alpha} (u_2 - d_2)^{1-\alpha}$ ; $\alpha = 1/2$ recovers the symmetric solution, $\alpha \to 1$ gives Player 1 the entire surplus, and $\alpha \to 0$ gives Player 2 the entire surplus.⁶ In a search-theoretic labor-market application the generalized objective is $[u(q) - y]^{\theta_N} [y - c(q)]^{1-\theta_N}$ , where $\theta_N \in [0,1]$ is the buyer's bargaining power.¹² A 2026 result shows that any solution satisfying scale covariance and contraction independence on convex feasible sets, together with its sacrifice version, is an asymmetric Nash solution with weights $(\alpha, 1-\alpha)$ , and only the symmetric case $\alpha = 1/2$ satisfies perspective invariance.¹³

By the numbers

Rubinstein division. In the infinite-horizon alternating-offers game with discount factor $\delta$ , the unique subgame-perfect equilibrium prescribes immediate agreement on the division $(1/(1+\delta), \; \delta/(1+\delta))$ .⁷
Triangle example. In a truncated bargaining problem $T$ , the Kalai–Smorodinsky solution awards $(2/3, 1/3)$ , while the Nash solution, by IIA, still gives $(1/2, 1/2)$ .²
Estimated bargaining weight. In unconstrained laboratory bargaining in a New Monetarist search experiment, the estimated buyer bargaining weight equals 1/2.¹²
First offers and rejections. In a Nash demand game experiment, first participants offered 40% on average, although backward induction prescribes giving the second participant the least possible share; about 16% of offers were rejected.¹⁴
Dominant power. In experiments with earned disagreement payoffs, about one-fourth of the time one bargainer's disagreement payoff exceeded half the cake size, making equal splits not individually rational.¹⁵

How it compares with other bargaining solutions

Kalai–Smorodinsky. Ehud Kalai and Meir Smorodinsky (1975) retained Nash's first three axioms but dropped IIA, replacing it with an individual monotonicity axiom: if the bargaining set expands but the ideal points remain the same, the new solution should give at least as much payoff to both individuals.²^•¹⁶ Their theorem states there is exactly one solution satisfying monotonicity: the maximal element of the feasible set on the line joining the disagreement point to the ideal point $b(S)$ .¹⁷ The two solutions also differ locally versus globally: the Nash solution picks the point of the smooth Pareto frontier where the utility elasticity equals 1, while Kalai–Smorodinsky is driven by each bargainer's maximal utility.²

Egalitarian rules. The egalitarian solution is the unique weakly efficient point satisfying the equal-gains condition $\lambda_1 (x_1 - v_1) = \lambda_2 (x_2 - v_2)$ ; the $\lambda$ -egalitarian solution generally differs from the simple egalitarian solution and does not satisfy scale covariance.¹⁸ Kalai's (1977) proportional solution, a related monotonicity-based rule, replaces scale invariance with a strong monotonicity axiom; the choice between Nash and Kalai solutions has quantitative implications for the welfare cost of inflation in search-theoretic models of money.¹²

Noncooperative foundations: the Nash program

The Nash program asks which noncooperative games produce the Nash solution as an equilibrium outcome. The central result concerns alternating offers: as $\delta \to 1$ , the limit of the unique subgame-perfect-equilibrium payoff of the alternating-offers game coincides with the Nash bargaining solution payoff.⁷ A variant with exogenous probabilistic breakdown converges to the Nash solution as the breakdown probability $\alpha \to 0$ , with the limit payoff $1/2 + 1/2(d_1 - d_2)$ in the symmetric case.⁹

Nash himself contributed a strategic foundation in 1953: he proved that if a small amount of uncertainty is added in the specification of his demand game, the Nash equilibrium payoff vector is close to the outcome selected by the Nash bargaining solution.¹¹ Not every sequential model points the same way: Bossert and Tan (1995) constructed a sequential game whose equilibrium yields the egalitarian outcome instead.¹¹ The disagreement point $t$ is the payoff achieved if players fail to agree, and the product $(u_1 - t_1)(u_2 - t_2)$ is maximized over points $u \ge t$ .⁴

Applications and practice

The two-person bargaining problem has been applied in contexts including management–labor arbitration, where management negotiates contracts with labor unions.¹⁸ In wage determination, the Nash bargaining solution and its asymmetric form are standard tools of macroeconomic modeling.⁶ A 2024 applied paper uses the solution for patent reasonable-royalty estimation, integrating three cost functions, Absolute-Value, Uniform, and Square Error, representing risk-neutral, risk-seeking, and risk-averse negotiations respectively, combined with a Bayesian Cost approach.¹⁹

In multi-agent systems, distributed optimization algorithms compute the Nash bargaining solution, which maximizes $\prod_{i=1}^{N} (y_i - d_i)$ subject to feasibility and $y \ge d$ ; the solution satisfies Pareto optimality, strict gains from cooperation for all agents, and scale invariance, meaning it takes into account the scale each agent uses to represent its utility.²⁰ Autonomous AI agents are increasingly deployed in high-stakes resource allocation such as supply chain contracting and spectrum allocation, where a reliable system must propose allocations that are individually rational, Pareto-efficient, and axiomatically fair; one 2026 preprint treats NBS-consistency as a requirement for trustworthy autonomous negotiating agents.²¹^•²²

References

Topic: Encyclopedia › Society and history › Economics and business › Economics › Economic theory and methods › Microeconomics › Property rights, exchange, and institutional microfoundations

Initially written Oct 10, 2026 · Edited: Oct 11, 2026; Oct 15, 2026

Notice something wrong?