Arrow's Impossibility Theorem — What No Voting System Can Satisfy

In 1951, Kenneth Arrow published a proof showing that no ranked voting system can satisfy a set of four seemingly reasonable fairness conditions. The result — now called Arrow’s impossibility theorem — is one of the most influential findings in social choice theory. It demonstrates that collective decision-making through ranked voting faces a structural limitation: you cannot have it all.

The theorem does not say voting is useless or that democracy is flawed. It says something more precise: any system that aggregates individual preferences into a group ranking must give up at least one property that we intuitively expect a fair system to have.

The Condorcet paradox

Arrow’s work built on an earlier discovery by the French mathematician and philosopher Nicolas de Condorcet. In his 1785 Essay on the Application of Mathematical Analysis to the Probability of Decisions by Majority, Condorcet showed that majority rule can produce cycles.

Consider three voters and three candidates — A, B, and C:

• Voter 1 ranks: A > B > C
• Voter 2 ranks: B > C > A
• Voter 3 ranks: C > A > B

In a head-to-head contest, A beats B (two voters prefer A), B beats C (two voters prefer B), and C beats A (two voters prefer C). The group preference is cyclic: A beats B, B beats C, but C beats A. There is no Condorcet winner — no candidate who defeats every other candidate in pairwise comparison.

Condorcet recognized that majority rule, the most natural method for collective choice, can produce outcomes that are logically inconsistent. The paradox does not require irrational voters. Each voter has a perfectly consistent ranking. The inconsistency emerges from the aggregation process itself.

Arrow’s four conditions

Arrow formalized the problem by asking what properties a social welfare function — a rule that converts individual rankings into a group ranking — should satisfy. He identified four conditions that any reasonable voting system ought to meet:

Unrestricted domain: The rule should work for any set of individual preference orderings. Voters can rank candidates however they like, and the system should produce a result. No preference pattern should be excluded in advance.

Pareto efficiency (Pareto criterion): If every voter prefers candidate A to candidate B, then the group ranking should also place A above B. When everyone agrees on a preference, the collective outcome should respect that agreement.

Independence of irrelevant alternatives (IIA): The group’s relative ranking of any two candidates should depend only on how individual voters rank those two candidates. The presence or absence of a third candidate — one who is “irrelevant” to the comparison between A and B — should not change whether the group ranks A above B or vice versa.

Non-dictatorship: No single voter should determine the group’s ranking regardless of how everyone else votes. The outcome should reflect the preferences of the group, not one individual.

Each condition is difficult to reject on its own terms. Unrestricted domain means the system works for any electorate. Pareto efficiency means unanimous preferences are respected. Independence of irrelevant alternatives means the relative ranking of two options does not depend on a third option that neither party cares about. Non-dictatorship means the system is genuinely collective.

The theorem

Arrow proved that no social welfare function satisfying all four conditions exists when there are three or more candidates. Any ranked voting system must violate at least one of them. If the system satisfies unrestricted domain, Pareto efficiency, and independence of irrelevant alternatives, then it is a dictatorship — one voter’s preferences always become the group’s preferences.

The proof proceeds by showing that any system satisfying the first three conditions has a “decisive” set of voters whose preferences determine the outcome for at least one pair of candidates. Arrow then shows that this decisive set can be shrunk until it contains a single voter — a dictator. The argument does not assume malice or coordination. It follows from the logical structure of preference aggregation.

Arrow developed this proof as a 22-year-old graduate student at Columbia University. His 1951 book, Social Choice and Individual Values, published by John Wiley & Sons, laid out the formal framework that became modern social choice theory. He was awarded the Nobel Memorial Prize in Economic Sciences in 1972 for this body of work.

What independence of irrelevant alternatives means in practice

The condition that is most commonly violated — and hardest to defend intuitively — is independence of irrelevant alternatives. To see why, consider a real-world voting scenario.

Three candidates run for office: A, B, and C. In a pairwise comparison, the group prefers A to B. Now suppose candidate C drops out of the race. Under IIA, the group should still prefer A to B, because C’s presence or absence is irrelevant to the comparison between A and B.

Most ranked voting systems do not satisfy this. In instant-runoff voting (also called ranked-choice voting), the elimination of a candidate can change how votes are redistributed, which can flip the relative ranking of the remaining candidates. The “irrelevant” candidate was not irrelevant after all — their presence shaped how ballots were counted.

This phenomenon is commonly called the spoiler effect. A candidate who cannot win changes the outcome between two other candidates simply by running. Arrow’s theorem shows that the spoiler effect is not a bug in one particular system. It is a structural feature of ranked voting more broadly.

The Gibbard–Satterthwaite theorem

A related result strengthens the picture. The Gibbard–Satterthwaite theorem, proved independently by Allan Gibbard (1973) and Mark Satterthwaite (1975), shows that any deterministic ranked voting system with three or more outcomes is either dictatorial or manipulable.

Manipulable means there are situations in which a voter can achieve a better outcome by misrepresenting their true preferences. No ranked voting system eliminates the incentive to strategically vote. If you care about the outcome, you sometimes have an incentive to rank candidates differently from how you actually feel.

The Gibbard–Satterthwaite theorem and Arrow’s impossibility theorem address different questions — one about the logical consistency of aggregation, the other about strategic incentives — but together they paint a picture of ranked voting as fundamentally constrained.

What the theorem does not rule out

Arrow’s impossibility theorem applies to ranked voting systems that produce a complete social ordering. It does not apply to every method of collective decision-making. Several approaches sidestep the result by relaxing one of the four conditions:

Rated voting systems: Approval voting, score voting, and range voting ask voters to rate or score candidates rather than rank them. Because the input is not a strict ordering, the theorem’s assumptions do not hold. These systems can satisfy analogues of Arrow’s conditions that ranked systems cannot.

Randomized rules: Lotteries and other probabilistic methods escape the theorem because they do not produce deterministic outcomes. Random selection satisfies fairness criteria that deterministic rules cannot.

Restricting domain: If voters’ preferences fall into a restricted class — for example, single-peaked preferences along a one-dimensional spectrum — then consistent aggregation becomes possible. The median voter theorem shows that majority rule produces transitive outcomes when preferences are single-peaked.

Positional systems: Borda count and other point-based ranking methods trade independence of irrelevant alternatives for freedom from Condorcet cycles. They produce consistent rankings but are sensitive to the presence of candidates who would not win.

Each workaround has trade-offs. Rated voting requires voters to calibrate scores, which introduces its own strategic considerations. Randomized rules sacrifice predictability. Domain restrictions do not hold in every election. Positional systems remain vulnerable to the spoiler effect that Arrow identified.

Why it matters

Arrow’s theorem changed how economists, political scientists, and mathematicians think about collective choice. Before the theorem, the search for a better voting system was largely practical — find the rule that minimizes manipulation or maximizes representation. After the theorem, the question became more fundamental: what properties are worth trading off against each other, and which trade-offs are acceptable?

The result also influenced mechanism design — the study of how to structure rules so that individual incentives produce desirable collective outcomes. The 2007 Nobel Memorial Prize in Economic Sciences went to Leonid Hurwicz, Eric Maskin, and Roger Myerson for building on Arrow’s framework to develop mechanism design theory. Their work shows how to design systems that align individual incentives with social objectives, even when full consistency is unattainable.

In practice, the theorem reminds us that no voting system is neutral. Every rule encodes values about what fairness means. Instant-runoff voting prioritizes broad support over intensity of preference. Plurality voting prioritizes simplicity over representation. Borda count prioritizes consistency over independence. The choice among them is not a technical decision. It is a normative one.

What remains open

Social choice theory has matured considerably since Arrow’s original proof. Researchers have explored conditions under which approximate versions of the four criteria can be satisfied simultaneously. The field of computational social choice studies how complexity constraints affect voting outcomes and manipulation. Empirical work examines how different voting rules perform in real elections with actual voter preferences.

One persistent question is how closely real-world preferences resemble the unrestricted domain that Arrow assumed. If voters’ preferences are structured — clustered along ideological dimensions, for example — then the impossibility result may be less binding in practice. The theorem describes what is logically possible across all preference profiles, not what typically occurs in a given electorate.

Another open direction is the design of voting systems for specific contexts. Arrow’s theorem addresses general-purpose aggregation, but specialized settings — committee selection, resource allocation, multi-winner elections — have different constraints and may admit rules that avoid the worst trade-offs.

The core insight endures: collective decision-making through ranked voting involves irresolvable trade-offs. The task is not to find a system that satisfies every fairness criterion. It is to choose which criteria matter most for the decisions at hand.

Sources

• Arrow, K. J. (1951). Social Choice and Individual Values. John Wiley & Sons. Yale University press reprint

• Condorcet, N. (1785). Essay on the Application of Mathematical Analysis to the Probability of Decisions by Majority. Original text (French)

• Gibbard, A. (1973). “Manipulation of Voting Schemes: A General Result.” Econometrica, 41(4), 757–772. Manipulability proof

• Satterthwaite, M. A. (1975). “Strategy-Proofness and Arrow’s Conditions: Existence and Correspondence Theorems for Voting Procedures and Social Welfare Functions.” Journal of Economic Theory, 10(2), 187–217. Strategy-proofness theorem

• Nobel Prize in Economic Sciences 1972. Kenneth J. Arrow. Nobel Prize announcement