The Paradox of Voting

In brief

Voting is nearly free, and yet from a strict cost-benefit view even that small cost looks like too much. The reason is the probability term. Your vote produces the outcome you want only in the rare event that it breaks an exact tie or creates one. In any large electorate that chance is minute, often smaller than the chance of being killed in a car accident on the way to the polls. Multiply a tiny probability by even a large benefit and the expected instrumental payoff shrinks below the cost of showing up. Anthony Downs drew this out in An Economic Theory of Democracy (1957): a rational, self-interested citizen appears to have no instrumental reason to vote. Turnout in real democracies routinely runs from a third to over ninety percent of eligible voters. The gap between what the model predicts and what people do is the paradox of voting, sometimes called the paradox of not voting.

The full treatment

The problem it answers

Economics assumes people act to advance their interests at acceptable cost. Applied to politics by Downs, that assumption should predict who votes and who stays home. Instead it predicts almost universal abstention, which is plainly false. The paradox is therefore a stress test for rational-choice theory itself. If the theory cannot explain an act as ordinary as voting, its claim to explain political behavior in general is in doubt. Every serious resolution is an attempt to save the theory, replace it, or mark its limits.

The calculus of voting

William H. Riker and Peter C. Ordeshook gave the decision its canonical form in "A Theory of the Calculus of Voting" (1968). A citizen votes if the reward R is positive, where R equals the instrumental payoff, minus the cost C, plus a duty term D.

B is the benefit the voter would gain if the preferred candidate won rather than lost. P is the probability that the voter's own ballot is decisive, that it makes or breaks a tie. C is the cost of voting: registering, informing oneself, traveling, queuing, the value of the time. The instrumental payoff is P multiplied by B, the expected benefit that flows through actually changing the result.

The trouble is P. In an electorate of millions the probability that one vote is pivotal is on the order of one in millions, and Gordon Tullock pressed exactly this in Toward a Mathematics of Politics (1967). Even if B is enormous, P times B is tiny, and any real cost C swamps it. On the payoff and cost terms alone, R is negative and the rational citizen abstains.

Riker and Ordeshook's move was to add D: a term for the satisfaction of voting itself, independent of whether the vote changes anything. D bundles the sense of civic duty, of complying with an ethic of participation, of affirming allegiance to the political system, of expressing a preference. If D exceeds C, a person votes even when the instrumental payoff is negligible. The equation is saved, but at a price the next sections will examine.

Rational ignorance

Downs paired the paradox with a second idea that follows from the same logic: rational ignorance. If the chance your vote matters is minuscule, the chance that being well informed changes what your vote does is smaller still. So the effort of studying candidates and policies has almost no instrumental payoff. A rational voter, on this account, chooses to stay largely uninformed, not from laziness but because acquiring political knowledge is a bad investment for the individual even though widespread ignorance may be bad for the collective. Rational ignorance is why the paradox is not merely academic: it predicts a democracy of thin, cheap information, and it explains why campaigns lean on slogans and cues rather than argument. It is the demand-side counterpart to the median voter theorem, which describes how office-seeking parties converge on the center of a poorly informed electorate.

Distinctions that matter

Three lines are worth keeping straight. First, the paradox is about instrumental self-interest, not about all reasons to vote. It says nothing follows from narrow cost-benefit calculation, and that is compatible with there being excellent non-instrumental reasons. Second, the D term is a genuine amendment, not a derivation. Nothing in rational-choice theory tells you how large D is; it is read off the fact that people vote. Third, the paradox scales with electorate size. It barely bites in a committee of ten, where one vote often is decisive, and bites hardest in a national election, which is why it is a paradox of mass democracy in particular.

Lineage

The paradox is a child of the marriage between economics and politics that produced public choice theory in the mid twentieth century. Its immediate parent is Downs's 1957 book, itself indebted to Joseph Schumpeter's redescription of democracy in Capitalism, Socialism and Democracy (1942) as a competitive market for votes rather than the expression of a general will. Behind Downs stands the wider apparatus of game theory and expected-utility decision theory, from which the payoff calculation is drawn directly. The paradox belongs to the same family as Mancur Olson's account of collective action in The Logic of Collective Action (1965), which showed why rational individuals often fail to contribute to a shared good whose benefit they receive regardless: voting is, on this reading, a collective-action problem in which the individual's contribution is nearly invisible in the aggregate.

The strongest case for it

The paradox is not a mistake to be explained away; it is a real and durable finding. Its first virtue is that it is true where it claims to be true. The probability of a decisive vote in a large election genuinely is negligible, and the instrumental payoff genuinely is dominated by cost. Any honest theory of voting has to start from that fact, and the paradox states it exactly.

Its second virtue is that it is fertile. Rational ignorance, retrospective rather than prospective voting, the reliance on cues and party labels, the vulnerability of turnout to the weather and to small changes in cost such as voter-identification laws or polling-place closures: all of these fall out of taking the calculus seriously. The finding that turnout is sensitive to cost, which the C term predicts, is one of the better-confirmed regularities in the study of elections.

Its third virtue is diagnostic honesty. By forcing the D term into the open, the model makes explicit that mass democracy runs on something other than individual instrumental payoff. It cannot itself say what that something is, but it identifies precisely the load-bearing element, duty and expression, that a fuller theory must explain. A theory that tells you exactly where it stops is more useful than one that pretends to cover everything.

The strongest case against it

The sharpest objection is that the D term is an admission of defeat dressed as a solution. Critics note that once you are allowed to add a variable large enough to explain any behavior, the model explains nothing: it can rationalize voting and abstention alike by adjusting D after the fact. Brian Barry, in Sociologists, Economists and Democracy (1970), argued that the rational-choice account of voting collapses into the sociological account it meant to replace, because everything doing the real work, duty and identity, comes from norms the economic model cannot generate.

A second line reframes the act entirely. On the expressive theory, developed by Geoffrey Brennan and Loren Lomasky in Democracy and Decision (1993), voting is more like cheering at a match than like buying a car. Because a single vote almost never changes the outcome, the low probability that so troubles the instrumental account also lowers the cost of voting your conscience, so people vote to express who they are and what they value. There is no paradox because there was never an instrumental decision to begin with.

A third line questions how the voter reasons about P. Some argue that citizens do not compute the true pivotal probability but reason as members of a group, asking what would happen if people like them all stayed home, which restores a large effective stake. Others, following the ethical-agent view associated with Derek Parfit's discussion of rational altruism, note that a policy benefit B is enjoyed by millions, so even a tiny probability multiplied by the benefit to everyone, not just to oneself, can outweigh the personal cost: the paradox dissolves once the voter is credited with caring about others.

Finally, empirical scholars including André Blais, in To Vote or Not to Vote (2000), find that the calculus predicts turnout patterns only weakly and that the strongest predictor of voting is simply whether a person believes voting is a duty, a belief the model treats as an unexplained constant. The complaint is not that the paradox is false but that, as a theory of why people vote, it fails to earn its keep.

Where it stands now

The paradox remains a fixed landmark. No one seriously disputes that the pivotal probability is tiny, so the instrumental account of turnout in large elections is regarded as inadequate, and the live question is what replaces it. The consensus has drifted toward the D term made respectable: turnout is driven mainly by civic duty, group identity, social pressure, habit, and expressive motivation, with instrumental cost still shaping the margins. Field experiments in the tradition of Alan Gerber and Donald Green have shown that social pressure, such as telling people their turnout is on public record, measurably raises participation, which fits the duty-and-social-norm picture far better than the pure calculus. Rational ignorance, meanwhile, has aged well and underwrites much of the current worry about misinformation and low-information voting. The paradox endures less as a theory of behavior than as the clarifying question every theory of turnout must answer: given that your vote will almost certainly not decide anything, why is it still worth casting?

Test yourself

Think back to the last time you voted, or chose not to. If you voted, try to name honestly what moved you: the belief that your single ballot would change the result, or something else, duty, identity, the wish to be counted. If it was something else, you have located your own D term. Now ask whether that reason is one the person next to you in line shares, and whether a democracy could run without it.