WikiSummary, the Social Science Summary Database

Riker and Ordeshook: A theory of the calculus of voting

From WikiSummary, the Free Social Science Summary Database

 

Riker and Ordeshook. 1970. A theory of the calculus of voting. American Political Science Review 1970:25-41.

In Brief: The Paradox of No Turnout

The authors begin where Downs (1957) left off: Rational choice appears to predict unrealistically low levels of turnout. They express this in equation 1, where R = the reward in utils of voting; B = the benefits of having your candidate win (compared to benefits of opponent); P = probability that your vote matters; and C = costs of voting:

Equation (1): R = (B*P) - C

In a population with 100 million voters, Downs assumed P would always be 1/100,000,000. Thus, any positive C would make R negative. The surprising result: Nobody will ever vote. Rather than ask why so many people don't vote, then, rational choice leads us to ask why so many people do.

The authors suggest a couple possible solutions to this paradox, although the "paradox of non-voting" has been the subjected of repeated scholarly inquiry. For more, see turnout.

Main Arguments

Reinventing the D Term

The authors insert D into the equation. D represents your sense of civic duty, your satisfaction from voting, your desire to affirm your partisanship or efficacy, etc (see pg 28). Thus, we get a new equation, in which C and D become the most important figures (since B*P is effectively zero, at least so far):

Equation (7): R = B*P - C + D

Downs had also proposed a D term, although his D represented a voter's desire to support democracy as a system. For Downs, a voter might turn out merely to prevent the democratic collapse that would follow zero turnout. (Note that this argument collapses on itself; just as no vote has much chance of deciding the election result, no vote has much chance of deciding the fate of democracy.)

P Can Matter

P can be much larger than Downs suggested. As the race gets closer to a tie, voters perceive that their vote has a much higher probability of affecting the outcome. Moreover, we tend to overestimate P (for more on the latter point, see Quattrone and Tversky 1988). Thus:

Equation (21): P = a bunch of calculus. The point: as your candidate's vote share approaches 50%, P increases.

Data and Testing

The authors test their ideas using survey data from 1952, 1956, and 1960, showing that P, B, C, and D (at the individual level) could predict Y, the probability that a particular respondent (claims that he) voted.

  • Y: Respondents declared in post-election interviews whether they recalled having voted.
  • P: Respondents were asked in pre-election interviews how close they thought the presidential election would be. This was converted to a dichotomous measure (high P vs low P) for unclear reasons.
  • B: Respondents were asked how much they care about the outcome (in pre-election interviews). This was also dichotomized.
  • D: A bunch of questions about citizen duty, trichotomized.
  • C: Not really included. Assumed to be constant for all voters, and if C is constant, then there is no need to control for it. (Note that later models use socioeconomic status and other variables to note that the costs of voting do vary; Downs had also speculated that income would matter.)

Findings

See Table 3 (p 38). The main hypotheses are validated.

Comments and Criticism

Note that a model driven by D and C doesn't explain why voters are strategic. ("Strategic voting" means that voters consider not only a candidate's positions, but also whether he is electable--thus, as an example, voters abandon the Green (or Libertarian) Party and vote for the Democrats (or Republicans) instead.) Strategic behavior implies that B (and therefore P) must matter.

An arcane and technical point that later critics have made: This analysis shows only that P, B, C, and D have marginal effects on Y (the probability of turnout). That is, it shows that increasing B, P, or D (or decreasing C) can change the probability of turnout, but it does not show that there is a particular (observed) combination of P, B, C, and D that will consistently lead us to predict that a particular voter will turn out.