Politics, Public Policy
Research by
Jonathan Bendor
Walter and Elise Haas Professor of Political Economics and Organizations
Stanford Graduate School of Business
Daniel Diermeier
Northwestern University
Michael Ting
Columbia University and Harvard University
FOR FURTHER INFORMATION: Helen K. Chang, 650-723-3358, Fax: 650-725-6750
Citizens Get Satisfaction from Voting
October 2004
STANFORD GRADUATE SCHOOL OF BUSINESS—When voters go to the polls this November they will be demonstrating in the flesh a phenomenon that has long troubled political scientists.
Called the "paradox of voting," it refers to the strange behavior that occurs in large elections: Voters actually show up.
Observers of the U.S. electoral process have long bemoaned that only 50 to 60 percent of registered voters turn up at the polls, even for presidential elections. But based on most "rational choice" theoretical models of human behavior, citizens would not be expected to vote even in numbers as large as these. It is simply irrational—or so the theory goes—given that it is exceedingly unlikely that any single individual will influence the course of the election.
Rational choice theory is a way of looking at deliberations among a number of potential courses of action and deciding which one is best. In the case of large elections, with millions of potential voters, the rational assumption would be that a single person would be extremely unlikely to influence the outcome. Hence: Why bother?
"The chance of being hit by lightning as you go to the polls is higher than the chance that you will be pivotal in a large election," said Jonathan Bendor, the Walter and Elise Haas Professor of Political Economics and Organizations at the Business School. "If people were perfectly rational and found voting costly, they'd say, 'If I'm not going to be pivotal, I'll just stay home with my kids.' Yet many people do vote. There's something fundamentally askew with the theories."
At least, there was until last May, when Bendor and two colleagues published a paper that proposed a new theory: that people learned—and adapted their behavior—based on whether the results of voting were satisfactory or unsatisfactory to them. The theory, called bounded rationality, is an idea expounded by the Nobel Prize-winning social scientist Herbert Simon, who pointed out that most people are only partly rational: Their reasoning is constrained by cognitive limitations and by emotions.
Called a model of "adaptive rationality," Bendor's theory marries two ideas: 1950s-style psychological learning reinforcement theory and Simon's notion of "aspiration levels." "It's basically very simply: An agent tries an action—in this case, voting—and if the payoff is satisfactory, then he or she is more likely to try that action again," said Bendor.
At the heart of the theory is the learning behavior of individual voters, who adjust their propensities to vote with each new election. The idea of "aspiration levels" is that every voter has a threshold that distinguishes a satisfactory outcome from an unsatisfactory one. If voting is associated with a payoff that is higher than that person's aspiration (or hoped-for outcome), then he or she is satisfied and will become more likely to vote next time. If the payoff is unsatisfactory, he or she will become less likely to vote.
"Williams James, the great American psychologist, once said that 'happiness equals get over want,'" said Bendor, adding that a voter's satisfaction with the results is not simply a reaction to winning or losing the election. Rather, it depends on how the winning party behaves after the election, and how satisfactory or dissatisfactory that behavior ultimately is to the voter.
And it's equally important to look at the reverse side of the theory: If someone doesn't vote and is unhappy with the results, then he or she will subsequently become more likely to change behavior and vote. This helps to buoy the voting statistics: Non-voters who are dissatisfied with the way the elections turn out may then come back to the polls and try again. Depending on their satisfaction with those results, their inclination to vote the next time may rise or fall in the next election. And thus the cycle starts again.
"I was not at all optimistic that we would be able to solve this problem," Bendor said, adding that he was surprised when he got a very excited email from one of his coauthors saying, "They [the voters in the computational model] are turning out!"
This model of voter turnout is just one of three that Bendor and his coauthors will use to write a book about elections. Titled Behavioral Models of Elections, the book will include a model of how voters choose among candidates—which will attempt to explain what happens when citizens actually get to the polls—and a model of candidate behavior, which examines how they position themselves along the political spectrum of left to right.
The book is designed to be "at least somewhat explanatory" of why elections turn out the way they do. It's emphatically not just a predictive model. "There are at least a dozen predictive voting models, and most of them turn on the state of the economy in one way or another," said Bendor. "But they don't provide explanations as to why voters and political parties do what they do and hence why we get the outcomes that we do."
Bendor's primary MBA teaching is on negotiation, which unites the strategic perspective of classical game theory with some appreciation of how people who are "boundedly rational" make decisions. Indeed, Bendor has been working on theories of bounded rationality for some time; what led him into studying elections was his work on building an aspiration-based model of legislative behavior. A colleague suggested using this technique on "a really difficult problem"—voter turnout—and, as Bendor said, "we took the bait."
—Alice LaPlante
Related Information
A Behavioral Model of Turnout
Jonathan Bendor, Daniel Diermeier, Michael Ting
American Political Science Review, May, 2003
The Empirical Content of Adaptive Models
Jonathan Bendor; Daniel Diermeier; Michael M. Ting
GSB Research Paper Series, 2005
Reinforcement Learning in Repeated Interaction Games
Jonathan Bendor, Dilip Mookherjee, and Debraj Ray
Advances in Theoretical Economics, March 17, 2003
A Theory of Participation in Elections
Timothy Feddersen and Alvaro Sandroni
MEDS Department, Kellogg School of Management, Northwestern University, mimeo, 2003
The Paradox of Not Voting
John Ferejohn and Morris Fiorina
American Political Science Review, 1974
A New Solution to the Collective Action Problem: The Paradox of Voter Turnout
Satoshi Kanazawa
American Sociological Review, 2000
Research News:
Concentrating Minority Voters Builds Liberal Strength in the South
Voting at the Extremes Costs U.S. Representatives Their Seats