Majority rule feels like an obvious way to turn many preferences into one decision. It is not nearly so well-behaved as it feels. This module assembles the central results of social-choice theory, which together show that no voting rule can do everything we want, that majority rule can cycle endlessly, and that whoever controls the agenda often controls the outcome.
The Condorcet paradox
The Marquis de Condorcet noticed in 1785 that majority preferences can be intransitive even when every individual's are perfectly rational. Three voters rank three options:
A majority cycle
Voter 1: A > B > C Voter 2: B > C > A Voter 3: C > A > B • A beats B (voters 1 and 3 prefer A to B) • B beats C (voters 1 and 2 prefer B to C) • C beats A (voters 2 and 3 prefer C to A) So the majority prefers A to B to C to A — a cycle with no Condorcet winner. There is no 'will of the majority' to discover; the outcome depends entirely on which pair gets voted on last. Majority rule has produced not a decision but a merry-go-round.
Arrow's impossibility theorem
Kenneth Arrow (1951) asked whether any rule for aggregating individual rankings into a social ranking could satisfy a short list of minimal, reasonable conditions. The answer, for three or more options, is no.
- Unrestricted domain — the rule must handle any logically possible profile of individual preferences
- Pareto — if everyone prefers X to Y, the social ranking must prefer X to Y
- Independence of irrelevant alternatives — the social ranking of X vs Y depends only on individuals' rankings of X vs Y, not on where some third option Z sits
- Non-dictatorship — no single individual's ranking is imposed as the social ranking regardless of everyone else
Arrow's theorem
No social-welfare function can satisfy unrestricted domain, Pareto, independence of irrelevant alternatives, and non-dictatorship simultaneously. Any rule that always produces a consistent social ranking must violate at least one — in the limit, it must be a dictatorship. There is no perfect democratic aggregation rule. This is not a flaw in a particular system; it is a logical impossibility, and it won Arrow the Nobel in 1972.
The median-voter theorem — order out of chaos, conditionally
If the chaos were total, politics would be unpredictable, yet much of it is not. The median-voter theorem (Duncan Black, 1948; popularised by Anthony Downs, 1957) explains the regularity. If choices lie on a single dimension (say, low to high public spending) and every voter has single-peaked preferences (each has an ideal point and likes options less the further they are from it), then the median voter's ideal point is a Condorcet winner — it beats every alternative in a pairwise majority vote.
Why parties converge to the centre
The median-voter theorem is why two-party competition tends to push both parties toward the centre: the party that captures the median voter wins. It predicts policy moderation, the importance of swing voters, and why manifestos often look similar. But it holds only under its conditions — one dimension and single-peaked preferences — and those conditions are exactly what fails next.
Two dimensions and the chaos theorem
Add a second dimension — say, spend more, but also redistribute toward one region — and the tidy median result collapses. McKelvey (1976) and Schofield proved a startling chaos theorem: with multidimensional choices and majority rule, there is generically no equilibrium, and an agenda-setter who controls the order of votes can lead the assembly, by a sequence of majority-winning steps, from any starting point to any outcome whatsoever, including ones a majority would reject head-to-head. Real budgets are massively multidimensional, so this is the relevant case.
Agenda control is power
Once majority rule can cycle, the power shifts to whoever sets the agenda — who decides which options are voted on, in what order, and what the default is if nothing passes. The committee chair, the Speaker, the finance ministry that drafts the bill. Social-choice theory's deepest practical lesson is that procedure is not neutral: control of the agenda is control of the outcome, which is why fights over rules of order are never really about order.
Exercise
A county assembly must allocate a fixed development budget across three wards. Each ward's representative most wants the money in their own ward, second-most wants it split, and least wants it entirely in a rival ward. (1) Show that this can generate a Condorcet cycle and explain what that means for the 'will of the assembly'. (2) The Speaker controls the order in which proposals are put to a vote. Using the chaos result, explain how the Speaker can engineer a preferred outcome. (3) The median-voter theorem predicted stable, centrist outcomes — why does it not rescue us here? (4) Propose one institutional rule that would tame the instability, and identify which Arrow axiom or social-choice problem it addresses.