In 1951, economist Kenneth Arrow proved something that should have changed how we think about democracy: it is mathematically impossible to design a voting system that satisfies all of the basic fairness conditions we'd want it to satisfy. Every possible voting system has to violate at least one of them.
The Four Conditions
Arrow started with conditions that seem so minimal they're almost not worth stating:
- Unrestricted domain: The system must handle any possible set of voter preferences.
- Non-dictatorship: The outcome shouldn't just reflect one voter's preferences regardless of everyone else.
- Pareto efficiency: If every single voter prefers A to B, the system should rank A above B.
- Independence of irrelevant alternatives: The ranking of A vs. B should depend only on voters' preferences between A and B — not on how they feel about a third option C.
These aren't ambitious ideals. They're the minimum you'd want from a system you'd call "fair." Arrow proved that no system can satisfy all four when there are three or more candidates. Any voting system violates at least one of them, always, necessarily.
What Each Violation Looks Like
Most common voting systems fail independence of irrelevant alternatives — the condition that a third candidate can't affect the outcome between two others.
Plurality voting (most votes wins) fails this constantly. In the 2000 US presidential election, Ralph Nader received 97,000 votes in Florida. Florida decided the election by 537 votes. If the independence condition held, Nader's presence shouldn't have mattered to the Gore/Bush outcome. But it did, because voters who preferred Nader might have chosen Gore as a second choice. Removing Nader from the ballot would likely have changed the outcome. This is the spoiler effect, and it's a direct consequence of Arrow's theorem.
Ranked-choice voting (instant runoff) fixes some problems but still fails independence of irrelevant alternatives in different ways. It's possible to construct scenarios where a candidate wins, but if some voters increase their ranking of that candidate, the candidate loses — a property called non-monotonicity that's even more counterintuitive than the spoiler effect.
The Condorcet Paradox
Before Arrow, the Marquis de Condorcet identified a related problem in 1785. Imagine three voters and three candidates: A, B, C.
- Voter 1: A > B > C
- Voter 2: B > C > A
- Voter 3: C > A > B
In a head-to-head vote: A beats B (voters 1 and 3 prefer A), B beats C (voters 1 and 2 prefer B), C beats A (voters 2 and 3 prefer C). Majority preferences form a cycle. There is no "correct" winner — the group preference is intransitive even though each individual preference is perfectly rational and consistent.
This isn't a failure of voters or candidates. It's a structural property of aggregating preferences. The group has no coherent ranking even when every individual does.
Why This Matters
Arrow's theorem isn't a technical curiosity for voting theorists. It has real implications for how we should think about democracy and collective decision-making.
First: any voting system you praise has a fatal flaw by necessity. The question is never "which system is fair" but "which unfairness is most tolerable for your specific context." Plurality voting's spoiler effect might be acceptable in an election where third parties are weak. Ranked-choice's complexity might be acceptable for local elections but confusing at scale. Strategic bloc voting in approval systems might be acceptable when voters have clear coalitions. These are value judgments, not technical solutions.
Second: the theorem shows that collective preferences are not simply the sum of individual preferences. Groups don't have preferences the way individuals do. When we talk about "what the people want," we're obscuring a mathematical reality: it depends on the system we use to measure it, and different legitimate systems give different answers.
The Alternatives
There are ways around the theorem. Arrow's conditions can be relaxed. If you allow cardinal preferences — voters expressing intensity of feeling, not just ranking — different results become possible. Score voting (rate each candidate 1-10) and approval voting (vote for all acceptable candidates) both escape some of Arrow's constraints, though they introduce other problems like strategic score inflation.
The theorem also only applies to deterministic systems. Randomized voting methods — literally selecting a winner by lottery weighted by votes — can satisfy all of Arrow's conditions, but the randomness introduces its own unfairness that most people find unacceptable.
What Arrow Was Actually Saying
Arrow himself was careful about what the theorem implied. It's not that democracy is incoherent or impossible — it's that the concept of a "group preference" that accurately aggregates individual preferences doesn't exist in general. We can build systems that work reasonably well in practice while violating theoretical fairness conditions in edge cases. That's what we've done.
But the theorem should make us humble about what voting systems can achieve and skeptical of anyone who claims their preferred system is simply better. Every system makes tradeoffs. The tradeoffs are unavoidable. The only choice is which tradeoffs you're making consciously.