INTRODUCTION

 

Is It Rat­ional to Vote?

 

This book addresses a simple question: What makes people decide to vote or not to vote?

As we shall see, the answer may not be as simple as the ques­tion. 

 

I start with a simple model, the rational choice model. According to it, a citizen makes up her mind to vote or not through a simple calculus. She de­cides to vote if, in her view, the benefits of voting are greater than the costs; if, on the contrary, the costs are greater than the benefits, she decides not to vote.

 

The "'calculus of voting" model was initially developed by Downs (1957) and extended by Riker and Ordeshook (1968). According to this model, the voter must first estimate the expected benefits of voting. In order to do this, she must first determine what she could potentially gain by voting rather than not voting. This means that, in the simple case of a plurality election with only two candidates, she must determine what difference it makes to her whether candidate A or B is elected. If it makes no difference, the potential benefit is nil. The greater the perceived difference, the greater the poten­tial benefit.

 

In the model, however, it is not the potential benefit but the expected benefit that matters. The rational individual must ask what the chances are that her vote will decide whether her preferred candidate wins or loses. If her preferred candidate is sure to win, the expected benefit of voting is nil, since her candidate will win whether she votes or not. If her preferred candidate is sure to lose, the same logic applies; the outcome of the election is the same whatever she does and the expected value of voting is nil.

 

The rational individual must thus determine the probability that her vote will be decisive, that is, that is, that it will determine whether her candidate wins or loses. Suppose a plurality election with 100,000 electors and two candidates. If the individual expects that 70,000 electors (besides herself) will vote, she must determine the probability that candidate A and candidate B will each receive 35,000 votes. In that case, it is her vote that will decide which candidate wins.'

 

How high is that probability? It depends on two factors. The first is the number of voters. The probability of a tie vote is much higher if there are only 1,000 voters than if there are 100,000. The second is the closeness of the election. The closer the race between the two candidates, the higher the probability of her casting the decisive vote. But however close the race, the probability of her vote being decisive is very small when the electorate is large. In the above example with 70,000 voters, even in a close race the chance that both candidates will get exactly the same number of votes is ex­tremely small.

 

In the rational choice model, the expected benefit of voting equals the benefit (B) she would gain from having her preferred candidate win rather than lose, multiplied by the probability (P) of her casting the decisive vote. This expected benefit is necessarily extremely low since P is bound to be extremely small.

 

Having estimated the expected benefits of voting, the rational individual calculates the costs (Q). These costs are essentially opportunity costs. They represent the time it takes to get registered, go to the poll, and mark the ballot, but also the time required to obtain and digest information about the candidates in order to determine which candidate the individual prefers. Since time is a scarce resource, not voting would allow her to do something else (including just relaxing), and it is that something else that she loses when she decides to vote.

 

The standard conclusion that is reached from the application of such a model is that in an election with a large number of voters the rational citizen decides not to vote. The cost of voting is small, but the expected benefit is bound to be smaller for just about everyone because of the tiny probability of casting a decisive vote.

 

The Paradox of Voting

 

Unfortunately for the theory, many people do vote. In fact, a clear majority vote in the most important elections, where the number of voters is extremely large and the probability of casting a decisive vote is minuscule. Hence the paradox of voting.

 

This paradox is the starting point of my inquiry. The calculus of voting model, in its simplest form, does not seem to work. Does this mean that the model is basically flawed and ought to be dismissed as completely inappropriate? Or does it need to be amended to provide a more compelling explanation of voting behavior? Or is the model, in its simple or more complex form, a valid though partial one that remains useful even though it cannot account for the frequency of voting?

 

In order to address these questions, I need to review the amendments to the model that have been proposed by rational choice theorists. Seven amendments have been proposed. Citizens may decide to vote: (1) to maintain democracy; (2) out of a sense of duty; (3) because they are risk‑averse and wish to avoid the regret of having not voted and seeing their preferred candidate lose by one vote; (4) because they reason that other citizens will not vote and that their own vote could become decisive; (5) because group leaders and politicians make it easy for them to vote; (6) because the cost of voting is practically nil; and (7) because they find it rational not to calculate benefits and costs when both are very small.

 

The first amendment was proposed by Downs (1957), who argues that the rational individual decides to vote in order to avoid the collapse of democracy. She will reason that everyone else will go through the same cost-benefit calculus and will come to the conclusion that it is rational to abstain. But when no one votes, democracy is threatened, and she benefits from the existence of democracy and has a long‑term interest in its maintenance. As a consequence, she is willing to bear the cost of voting in order to insure against the potential breakdown of democracy

 

As Barry (1978, 20) has pointed out, such an amendment cannot salvage the theory.2 The fact is that one ballot cannot make any difference to the maintenance or collapse of democracy any more than a single vote can determine the outcome of an election. The rational citizen should calculate expected ‑benefits, which entails multiplying long‑term benefits attached to the maintenance of democracy by the probability that her vote will be decisive in whether democracy survives or not. Since that latter probability is practically nil, expected benefits are minuscule and she decides not to vote.

 

This does not mean that sustaining democracy is not a major inducement for many people to vote. The point is rather that such a motivation is inconsistent with the rational choice approach, which assumes that the individual calculates the benefits and costs of a given action for herself, not for

 

the whole community. If we observe that people vote to preserve democracy, we will have to conclude that their behavior does not conform to the predictions of rational choice theory.

 

The second amendment comes from Riker and Ordeshook (1968). They argue that citizens can also derive psychic gratifications from the act of voting: the satisfaction of "complying with the ethic of voting," of "affirming allegiance to the political system," of "affirming a partisan preference," and of "affirming one's efficacy in the political system," as well as the enjoyment of going to the polls and deciding how to vote. Riker and Ordeshook call such satisfaction D, or citizen duty.

 

From this perspective, two kinds of benefits are associated with voting: investment (or instrumental) and consumption benefits. The investment benefits are those linked to the outcome of the election: they are the difference in utility the individual attaches to having candidate A win rather than candidate B. These investment benefits are contingent on her casting the decisive vote, and the expected benefit is bound to be exceedingly small. The consumption benefits‑the sense of satisfaction one derives from fulfilling her sense of duty‑are not contingent: she feels satisfied when she votes, whatever the outcome of the election. She consumes voting for its own sake.

 

It makes sense to believe that some people vote because they feel it is their duty to do so. The question is whether such an account fits with the rational choice model. It depends on whether one adopts a broad or a narrow notion of rationality.

 

There are those who adhere to a broad definition of rationality. From that perspective, rationality "'is nothing more than an optimal correspondence between ends and means" (Tsebelis 1990, 18). What counts as rational behavior has nothing to do with a person's objectives or motivations (the theory has nothing to say about them); all that is needed is that the individual behaves so as to maximize her utility, whatever that utility happens to be.

 

This broad perspective has one major advantage and one important drawback. The advantage is one of open‑mindedness. Those who use a broad definition of rationality do not have to impute specific goals to individuals; they do not even have to presume that they are egocentric. The drawback is the risk of tautology.  It can always be argued that a person chooses to do something because she believes the rewards for performing that act outweigh the costs. At this point, the theory cannot be falsified and loses any predictive power.'

 

Downs was keenly aware of this potential pitfall. He opted for a narrow definition of rationality. In his model, he made a specific assumption about the goal pursued by citizens: "The political function of elections in a democracy ... is to select a government. Therefore rational behavior in connection with elections is behavior oriented toward this end and no other" (Downs 1957, 7). He made an exception to that rule later on in the book, when he stated that citizens would want to vote in order to maintain democracy. He was consistent, though, in his unwillingness to resort to social or psychological gratifications. Otherwise, “all behavior whatsoever becomes rational be because every act is a means to some end the actor values. To avoid this sterile conclusion, we have regarded only actions leading to strictly political or economic ends as rational" (276).

 

In my view, the risk of tautology is a very serious one, and the theory cannot be saved by merely adding sense of duty as an explanatory variable. The same verdict is reached by most analysts (see, especially, Barry 1978, 16; Mueller 1989, 3 5 1; Green and Shapiro 1994, 52). It is more fruitful, I believe, to construe psychic gratifications such as citizen duty as factors that fall out- side the rational choice model. In short, the fact that some people vote out of a sense of duty does not support or invalidate the model. The core of rational choice has to be, as Downs has cogently argued, investment benefits, not consumption ones.

 

There are two possibilities to distinguish. The first is that all the story is in D and that the decision to vote is hardly affected by B, P, and C. Those who adhere to a broad definition of rationality would insist that the theory is confirmed. My response is that sense of duty can be incorporated in alternative schools of thought. It is dubious whether finding that people vote out of a sense of duty corroborates rational choice more than other interpretations. If all the story is in D, we should conclude that the rational choice model is not very helpful.

 

The second possibility is that people vote on the basis of each of B, P, C, and D. Such findings would confirm the validity of the rational choice model. In other words, rational choice does not rule out the presence of psychic gratifications, but these gratifications should not be used as positive testimony of the model. The crucial test is whether B, P, and C matter.

 

Ferejohn and Fiorina (1974) provided the third attempt to rescue the model. They argue that we ought to distinguish between decision‑making under risk and decision‑making under uncertainty. Under the former, probabilities of different outcomes are taken into account; under the latter, such probabilities are unknown or unknowable. Uncertainty calls for different sets of rules. Ferejohn and Fiorina examine one specific procedure, the minimax regret criterion, under which one calculates the loss associated with various outcomes without estimating the probabilities of these outcomes and chooses the option that minimizes regret.

 

The minimax regretter thus asks herself how much regret she would have if, on the one hand, she voted and her vote was not decisive, and, on the other hand, she did not vote and her preferred candidate lost by one vote. If the latter regret is greater than the former, she decides to vote. This procedure boils down to dropping the P from the calculus of voting. Since P cannot be known, the individual does not take it into account. She votes if B is greater than C and abstains if it is the reverse.

 

The great advantage of the minimax regret procedure is that it predicts more turnout than the rule of maximizing expected utility. Is it convincing? Two major criticisms have been made about this interpretation. The first concerns its assumption about P. Critics (see especially Beck 1975, 918; Aldrich 1993, 259) point out that even though it may be exceedingly difficult to calculate the exact value of P, it is easy for the rational citizen to figure out that in a large electorate it is minuscule. She is able to give a rough guess of what P is likely to be, and the expected utility model thus applies.

 

The second criticism is that the minimax regret strategy leads to bizarre behavior. Suppose ten candidates are running in a constituency. According to the minimax regret criterion, the individual should envisage the worst scenario, that the candidate she likes the least wins, without taking into account her probability of winning, and vote if this regret out‑weighs the cost of voting. Thus the presence of an extremist candidate should substantially increase turnout, whatever the degree of support she enjoys (see Mueller 1989,353).

 

The predictions of minimax, regret are even more problematic in other contexts. Minimax regretters should not cross the street to buy a newspaper because there is a possibility that they could be killed by a car while doing so (Beck 1975, 918). Ferejohn and Fiorina (1975, 92 1; 1974, 535) have responded that different decision rules may be used in different contexts (and by different types of individuals). But when it comes to deciding whether to vote or not, should not the minimax regretter consider the possibility that she could get killed by a car going to the polls? Is not the appropriate decision, then, not to vote?

 

Minimax regret does not appear to be a satisfactory solution to the paradox of voting. Ferejohn and Fiorina, however, raise an important question about what counts as a rational calculus. The standard assumption in the literature is that the rational citizen attempts to maximize expected utility, and thus multiplies B by P. Is it possible that people do not seek to maximize, that they merely attempt to satisfice (Simon 1955)? What should we think of an individual who calculates B and P and gives each some weight in her decision whether to vote or not without multiplying them, as the theory implies? Should such a person count as rational or irrational? Even though minimax regret does not solve the puzzle, perhaps other approaches could.

 

The fourth potential solution to the paradox of voting is of a game‑theoretical nature. The reasoning is the following: "If each rational voter were to decide not to vote because her vote has too small of a chance of affecting the outcome, and all voters were rational, no one would vote. But then any one voter could determine the outcome of the election by voting.... The greater the number of other voters I expect will rationally abstain, the more rational it is for me to vote" (Mueller 1989, 351‑52).

 

This raises the possibility that P may not be as minuscule as initially assumed once we take into account the fact that many other citizens will go through the same reasoning. Under some assumptions, some equilibrium solutions entail "substantial" turnout (Palfrey and Rosenthal 1983; Ledyard 1984).

 

This line of reasoning is not very convincing. As Palfrey and Rosenthal (1985, 73) have demonstrated, "'in the presence of a relatively small degree of strategic uncertainty . . . voters with positive net voting costs will abstain."' The bottom line is that the rational individual who is not certain that others will abstain, abstains.

 

Even though the argument is not compelling, it conveys an important message. We should not take for granted that P is perceived to be minuscule by all electors. We should not rule out the possibility that some people vote because they overestimate the probability that their vote could be decisive.

 

The fifth attempt to salvage rational choice theory is to shift the focus from the individual citizen to politicians and/or group leaders. Politicians attempt to mobilize voters and make it as easy as possible for their supporters to vote and by doing so reduce the cost of voting (see especially Aldrich 1993). A similar line of reasoning is pursued with respect to group leaders. Uhlaner (1986, 1989a, 1989b, 1993) provides the basic model. In the model, politicians and group leaders make bargains in which the latter promise increased support from the group and the former stake issue positions favorable to the group's interests.

 

The problem with such an approach is that it is not clear why it becomes rational for the individual citizen to vote. True, politicians and group leaders can reduce the cost of voting and can even increase the potential benefit. The point remains, however, that P is bound to be minuscule and that expected benefits do not outweigh the cost. These models must thus rely on other consumption benefits.

 

Aldrich argues for an expanded notion of D, which makes the model very close to being tautological, notwithstanding his claim to the contrary. Uhlaner resorts to what she calls "relational goods." In her model, group members identify with the group and wish to vote to receive approval from other group members or simply because voting makes them feel truly like members of the group. In this case as well, sense of duty plays a crucial part, though the driving force is a sense of duty toward the group, not toward the whole community. This approach leads to the same problems discussed with respect to the introduction of the D term. There is no inherent contradiction in including consumption benefits, but the real hard test of the rational choice model is whether investment benefits matter.

 

The sixth amendment boils down to arguing that the cost of voting is practically nil. This argument has been put forward most straightforwardly by Niemi (1976). Niemi makes two points. First, the cost of voting, at least in a national election, is extremely low. It takes very little time to vote, and because it is taken out of leisure time many people must perceive very little opportunity cost.

 

Though we have no hard evidence on this question, Niemi is probably right in asserting that the opportunity cost of voting, on election day, is very low. He fails to consider, however, the opportunity cost of getting the information to decide how to vote. And, most important, however low the costs are, they are unlikely to be small6r than expected benefits, which are bound to be minuscule. But what if people perceive no cost at all in deciding how to vote and in going to the polling station? It is thus crucial to determine whether citizens feel there are costs involved. I have alluded to the possibility that people vote because they overestimate P. Another possibility is suggested here, that they vote because they underestimate C.

 

The second point made by Niemi is that we should also consider the cost of not voting. This is "'the psychological cost of saying 'no' when asked whether or not you voted" (Niemi 1976, 117). But why should there be a psychological cost in admitting abstention? Perhaps because one feels a moral obligation to vote, in other words, a sense of duty. Niemi wants to distinguish satisfaction from duty, the D term, but the distinction is far from clear. If someone fails to fulfill her duty, should she not feel guilty, in the same way she should feel proud and satisfied if she has voted?

 

Niemi also raises the issue of social pressure. If someone wants to avoid embarrassment from admitting not voting, it is probably because she perceives voting to be a social norm, which relatives and friends expect her to obey. The presence of social pressure indicates that many people believe it is right to vote and wrong not to vote, and such beliefs presumably flow from the view that it is the duty of every citizen to vote. There is social pressure because most people feel voting is a duty.

 

It could be, of course, that the individual does not personally think that she has a duty to vote, but that she is aware that most people in her community believe that it is a citizen's duty to vote. She might then reason that her personal reputation would suffer if people knew she did not vote and decide that it is in her best interest to vote.

 

This is an intriguing hypothesis. Whether people actually engage in such complex reasoning remains to be shown. My point again is not that such psychic gratifications are inconsistent with a rational choice perspective but that they do not suffice to confirm the model. In other words, it must be shown that concerns for one's reputation overcome the initial propensity to abstain because of the expected investment benefits and costs. The voting decision must depend at least partly on B, P, and C.

 

This leaves one final amendment to the model, which was first proposed by Barry (1978) and is reiterated, in a different fashion, by Aldrich (1993). The argument is simple: "It may well be that both the costs and the (suitably discounted) benefits of voting are so low that it is simply not worth being I rational' about it" (Barry 1978, 23). Such a position acknowledges the limits of the model: it cannot explain why so many people vote. At the same time, it makes the point that this is no reason to conclude that the rational choice model is of no use; it might be extremely fruitful in accounting for other kinds of behavior, where costs and benefits are higher.

 

Green and Shapiro (1994, 58‑59) criticize this position on two grounds. First, they point to specific cases, such as those "100,000 African‑Americans who persevered through the intimidation and poll taxes of the Jim Crow South and voted in the national elections of the 1950s," where the cost of voting could not be construed as low. This is not a very compelling objection, as the phenomenon to be explained is the decision to vote or not to vote in an open democratic system.

 

The second objection is more serious. The argument is an arbitrary domain restriction because "there is nothing in rational choice theory that specifies the level at which costs or benefits are sufficiently small to render the theory inapplicable." The concern is that it is an easy way to salvage the theory when its predictions do not fit with reality, to decide, ex post, that it does not apply to the particular phenomenon being examined. Yet it is consistent with the basic postulates of the model to argue that, since costs are involved in calculating benefits and costs, it may be rational, when the stakes are low, not to calculate. In other words, we would expect rational choice theory to perform better when the stakes are high than when the stakes are low.

 

This last amendment salvages the rational choice model but asserts its inapplicability to the simplest political act, voting. A somewhat different position, put forward by Aldrich (1993), argues for differences of degrees. Rational choice would apply to all phenomena, including voting, but it would perform less well for low‑stakes behavior.

 

There have been several attempts, some of them quite imaginative, to solve the paradox of voting. I have discussed the merits but also the limitations of each of these attempts. This discussion leads me to make two general observations. First, most amendments have consisted of adding psychic gratifications to the core model. My position is clear and simple: there is nothing wrong in adding consumption benefits, but there is a great risk of tautology in doing so. The prudent solution is to focus on the core elements of the model and to determine whether they do affect the decision to vote. If people are inclined not to vote because they feel that the expected investment benefits are just too low but psychic gratifications such as sense of duty tend to overcome this inclination, I would conclude that rational choice offers a useful interpretation. If, however, the decision to vote is almost entirely accounted for by the sense that it is a moral obligation to vote, then it seems to me that we should conclude that the model should be discarded.  

 

Second, it must be kept in mind that from the perspective of rational choice, it is individuals' perceptions that matter. Most analysts take for granted that P is minuscule and that C is not.                                                                          

But we have precious little evi­dence on how people actually perceive the probability that their vote could be decisive and the costs of voting. If P is overestimated and/or if C is perceived to be minuscule, then it might be rational to vote. I present in chapters 3 and 4 new evidence on these perceptions.

 

Is It Marginally Rational to Vote?

 

Groftman, in a spirited defense of rational choice, makes two Points. First, we should not expect rational choice to explain everything. As he puts it,

 

the view that a rational choice model requires that the only element to be allowed in a voter's utility function is the expected short‑run instrumental benefit from voting is like saying that

the only way we can use rational choice models to predict eating behavior is to limit people's motivations for their food choices to a single parameter‑for example, an instrumentally rational choice of a consumption bundle that maximizes expected longevity. (1993, 94)

 

Rational choice provides partial explanations of human behavior, and the theory should be judged on its ability to account for some aspects of human behavior, neither more nor less.                                                                                                                                                    

 

Groftnan's second point is that the theory should be tested for its capacity to predict changes in turnout rather than turnout per se. This second point is related to the first. According to Groffinan, there are rational and nonrational reasons to vote (and to abstain). Because of this complexity of factors, it is impossible to predict the absolute level of turnout. It is possible,

however, to predict that if P or B increases in an election or C decreases, turnout will increase. Conversely, "turnout is lower when the weather is bad, when the barriers to registration are steep, and in elections whose outcomes few care about" (Grofman 1993, 102). As a consequence, Groffinan claims, the model is quite capable of doing what it is supposed to do: account for change at the margin.

 

I shall determine in later chapters whether this empirical claim is well established. For the time being, I am concerned with the epistemological relevance of the argument. According to Grofinan, we should test a theory according to what it tells us about dynamics. In the case at hand, the prediction of rational choice is that if P or B increases, turnout should increase, and that if C increases, turnout should decrease. The paradox of voting pertains to the level of turnout and is not really relevant as a test of the theory.

 

In my judgment, Grofman is partially (marginally?) right and partially wrong. It must be pointed out that the rational choice model itself is not couched in terms of marginal change. The model takes the individual as the basic unit of analysis and identifies the basic parameters of her decision to vote or not to vote. It is thus consistent with the model to look at the decision that is made at a given point in time by each individual, and by the aggregation of these individuals. I agree with Grofman, however, when he asserts that the mere fact that many individuals vote does not necessarily invalidate the model. It could be that for these individuals nonrational reasons for voting trump rational reasons for abstaining. I would add, though, that if we were to observe that an overwhelming majority of people vote, we would have to conclude that rational reasons for abstaining had a very "'marginal" impact. Evidence about the level of turnout may not be conclusive, but it is nevertheless suggestive.

 

Furthermore, it is not enough to show that changes in B, P, or C lead to changes in turnout or that individuals who are different with respect to these variables have a different propensity to vote. It is also imperative to take into account the strength of the relationship. If we observe that turnout is reduced when it rains but that it is reduced by only one percentage point, we may be inclined to conclude that the contribution of rational choice is indeed very marginal.

 

Grofman is right in insisting that we should not be looking solely or even primarily at the level of turnout. The fact that so many people vote does not require that the rational choice model be discarded. But this suggests that the model provides, at best, a partial explanation of voting. This opens the possibility that there are more powerful theories, and that, marginally speaking, we would be better off exploring alternative avenues.

 

What Are the Alternatives?

 

There are four major alternative explanations for why people vote. These are the resource and mobilization models, and what could be called the psychological and sociological interpretations. I will briefly review each of them and indicate how they converge on and diverge from rational choice. It should be kept in mind that these alternative models usually have as their explanandum political participation writ large, and not voting per se.

 

The first model, laid out most explicitly by Verba, Schlozman, and Brady (1995; see also Brady, Verba, and Schlozman 1995), focuses on resources.' According to this model, the decision to participate hinges on three main resources: time, money, and civic skills. The more free time, money, and civic skills one has, the more likely one ­is to participate. Of course, the impact of these various resources varies from one mode of participation to the other. Money, in particular, is less relevant to voting than to other forms of participation.

 

The fact that time is a crucial resource and that the more free time one has, the more likely one is to vote, is quite consistent with rational choice. Here, the two models clearly converge. The convergence is more ambiguous when it comes to skills. Brady, Verba, and Schlozman (1995, 273) assume that "'citizens who can speak or write well or who are comfortable organizing and taking part in meetings are likely to be more effective when they get involved in politics." It is possible to construe these skills as lowering transaction and information costs (290, n. 10), and to infer that the resource model emphasizes costs.

 

What distinguishes the two models? The resource model assumes that most of the action is in the cost section of the equation, whereas the rational model pays as much attention to expected benefits. In the resource model, people are inclined to vote if the cost is not too high; in the rational model, they weigh benefits and costs and vote only if they perceive the latter to be smaller than the former. The resource model seems to fit better with the facts (it predicts high turnout) but appears unduly partial: it focuses entirely on costs, taking benefits for granted!

 

The second model, which stresses the role of mobilization, has been put forward most forcefully by Rosenstone and Hansen (1993). Their starting point is that the only way to explain political participation is to "move beyond the worlds of individuals to include family, friends, neighbors, and coworkers, plus politicians, parties, activists, and interest groups" (23). These social networks exert pressures on people to behave as members of a group rather than as isolated individuals.

 

Politicians have an interest in getting out the vote. They mobilize individuals, especially through social networks, and induce them to vote. They do so by reducing the cost of voting: they drive people to the polls on election day and provide cheap information about the issues.

 

This second model, with its emphasis on social networks, overlaps to some extent with the sociological interpretation discussed below. It also has some obvious connections with the resource model, since both pay particular attention to how the cost of voting can be reduced. The two models seem complementary, as the resource model may better explain who participates, while mobilization may do a better job of explaining when (Rosenstone and Hansen 1993, 20).'

 

The problem is that, from the perspective of rational choice, the paradox of voting lies with the expected benefit, not with the cost. Even if a party drives an individual to the poll, it will take her some time to vote and it is difficult to imagine that this "waste" of time is worth less than the minuscule probability of casting a decisive vote.

 

Rosenstone and Hansen argue that social networks help to overcome the paradox of participation because they "create solidary rewards and bestow them, selectively, on those who act in the public interest." How do they do so? Basically through social pressure. Rosenstone and Hansen assume that "people want to be accepted, valued, and liked" and that they accept being mobilized (that is, participating) in order to be well esteemed. The mobilization model, then, pays particular attention to the impact of social pressure, a factor that has also been mentioned by some rational choice authors (1993, 23).

 

The third model, or explanatory mode, looks at psychological involvement. Bluntly put, it asserts that the more interested a person is in politics, the more likely she is to participate in general and to vote. The main problem with this line of explanation is that it appears trivial. The hypothesis is eminently plausible, but it is not clear what it tells us.

 

Despite its apparent superficiality, the psychological involvement interpretation should not be discarded too hastily. From the perspective of rational choice, the relationship between political interest and voting is not straightforward. If someone is highly interested in politics, she is likely to perceive the stakes of an election to be great: her B will be high. Her information costs might also be lower, but not her P. The bottom line is that from a rational choice perspective, the relationship between interest and voting should be only modest.

 

This indicates that when we want to explain why some people vote and others don't, it is imperative to take into account individuals' overall level of political interest. For the same reason that people are unlikely to watch basketball if they don't like the game, they are less likely to vote if they hardly follow politics.

 

The last approach to be considered is what I would call the sociological interpretation, which most directly challenges the rational choice model. It challenges the basic assumption that the individual acts in her own (narrowly defined) self‑interest. According to the sociological interpretation, individuals behave as members of a collectivity.

 

We should distinguish two strands within this school. The first strand stresses the role of social pressure: people vote because their friends or relatives tell them to vote. The second strand assumes that citizens are concerned with the well‑being of their community as much as with their own self‑interest. The concept of sense of duty encapsulates such a view. People vote because they feel it is a moral obligation for citizens in a democracy to vote. The reference point is not the self but the whole community. The individual feels she owes it to society to perform the act of voting, whatever the personal benefits and costs.

 

This review of other theoretical explanations for why people do or do not vote leads to three main conclusions. First, there is some overlap between the various theories. Rational choice, resource, and mobilization theories all insist on the costs of participation and they converge on the conceptualization of these costs. Second, the trademark of rational choice is in the P term. It is the sole theory to include it as a key variable. This is why the paradox of voting emerges only within a rational choice perspective. Consequently, determining whether P matters should be a priority when it comes to assessing the fruitfulness of the model. Third, the concept of sense of duty, which is often added in rational choice models, is also quite consistent with a sociological interpretation. Finding that people vote out of a sense of duty should not be construed as supporting a rational choice perspective.

 

Outline of the Book

 

The objective of my inquiry is to come to a comprehensive understanding of why people do or do not vote and to assess the capacity of the rational choice model to account for the simple act of voting. At first sight, rational choice seems to be on shaky ground in explaining voting behavior. As we have seen, the various attempts to solve the paradox of voting are not completely satisfactory. All these attempts, however, have been made at the theoretical level. We need to look more closely at the empirical evidence and determine whether there is at least partial support for the model.

 

I start with the presumption that rational choice is unlikely to provide a fully satisfactory theory of turnout. Perhaps, though, the theory provides some helpful insights into why some people are more likely to vote than others and why turnout is higher or lower in certain places and times.

 

The inquiry proceeds in the following fashion. In the first two chapters, I leave aside rational choice theory to focus on the "facts" of turnout. I address two questions. First, where and when is turnout higher and lower? I look at cross‑national variations in turnout and identify the factors that seem to explain why it is higher in certain countries than in others. Second, who votes and who does not? To answer this question, I look at the socioeconomic variables that are most closely associated with turnout at the individual level.

 

I start with empirical observations because rational choice theory has been sharply criticized for its lack of empirical rigor (Green and Shapiro 1994). It is a useful antidote to forget about the theory temporarily and to immerse ourselves in the data, to assemble as much information as possible about the patterns of voting at the micro and macro levels, and to digest that information before returning to the big theoretical questions.

 

In chapters 3 to 5, 1 proceed to a full empirical assessment of the rational choice model. As already indicated, the core elements of the model are B, P, and C. I also include D, sense of duty, that is often incorporated into rational choice models but whose status is ambiguous since it is consistent with a sociological interpretation, and level of political interest, an extraneous taste factor that should be taken into account as a control variable. The test of the theory is whether each of B, P, and C has an independent effect on the decision to vote.

 

I start with P, in chapter 3. As just mentioned, P represents the most interesting and problematic contribution of rational choice model. It is thus crucial to determine whether people believe that their vote could be decisive and whether perceptions of P affect the propensity to vote. I review the empirical literature on this question and provide new evidence.

 

I address the cost term in chapter 4. In the extreme, if the cost of voting is perceived to be nil, it could become rational to vote. Do citizens perceive any cost at all in the act of voting and/or in the process of finding out which candidate they should support? How much impact do these perceptions have on the decision to vote or not to vote?

 

Chapter 5 deals with sense of duty. Do people feel they have a moral obligation to vote? How strong is that feeling? Where does it come from? And are people who believe it is their duty to vote insensitive to expected benefits and costs?

 

Chapter 6 goes beyond the simple act of voting and considers the broad issue of rationality and participation. Not voting is part of a larger social phenomenon, the propensity to "free ride." While turnout and the capacity of rational choice to explain it are the central focus of this book, it is imperative, at the end of our journey, to adopt a broader perspective. We need, especially, to determine whether rational choice does a better job of explaining other kinds of behavior, especially those with higher stakes. When the costs of participation are higher, is the propensity to free ride higher? Could it be that, as Barry has argued, rational choice does not perform well for voting because people do not make cost‑benefit calculations when the stakes are too low?

 

In the conclusion, I render my verdict on why people do or do not vote, and on the capacity or incapacity of rational choice to provide a compelling explanation of turnout.

 

 

CONCLUSION

 

Rational Choice and Voting

 

I started my inquiry with the observation that the rational choice model of voting, in its simplest formulation, does not appear to work. An individual who takes into account the tiny probability that her vote could be decisive should rationally abstain. Yet most people vote in national elections, and most of them vote regularly. Voting seems to be a paradox.

 

One may adopt either of the following positions with regard to this paradox. The first is to argue that the rational choice model is basically flawed 'and should be discarded. The second is to assert that it is possible to come up with a more complex interpretation within the same framework.  The third is to propose that rational choice provides a valid, though very partial, explanation of turnout.

 

I support the third position. Rational choice can contribute to understanding why people vote or abstain but its contribution is quite limited. It is limited in three ways. First, for about half of the electorate, which possesses a strong sense of duty, the model has simply no explanatory power. These people vote out of a sense of moral obligation, because they believe it would be wrong not to vote; they do not calculate benefits and costs.

 

Second, even among those with a relatively weak sense of duty, rational choice's contribution is very partial. It is true that for these people, the propensity to vote increases with B and P and decreases with C, but these factors affect the vote much less than the model would lead us to believe.

 

Yes, turnout increases with the perceived importance of the election, but the fact remains that in countries where the president has very little power, such as Austria, Iceland, and Ireland, median turnout in presidential elections is only two points lower than in the more crucial legislative elections. Yes, turnout is higher when the race is closer, but a gap of ten points between the leading and the second parties seems to reduce electoral partici­pation by only one point. Yes, people are less likely to vote when it is more costly, but rain has no impact at all on those with a strong sense of duty and a very small effect among others (Knack 1994).

 

Third, rational choice is only partly right with respect to its most original component, the P term. It is true that people are more likely to vote when they feel that their vote can make a difference, and turnout increases when the race is close. According to the model, however, people should think about the probability that their vote will be decisive, that it will decide who wins the election. People's reasoning appears to be much more fuzzy. They do not really think about the possibility that the election could be decided by one single vote. What seems to matter (at the margin, for those with a weak sense of duty) is the vague impression that the outcome is not a foregone conclusion, that there is a race of some sort, that their vote may count."'

 

Moreover, people do not multiply B by P, they simply add them. The fact that the model is additive rather than multiplicative may appear to be a trivial technical consideration, but it has important substantive implications. In the standard rational choice model it is expected benefits that matter, that is, the utility a citizen derives from a party being elected times the probability that her vote will decide whether that party will be elected or not. Because P is bound to be infinitesimal, B*P is practically nil, and the rational citizen decides to abstain.

 

Minimax regret tried to solve the problem by getting rid of P entirely I have shown that minimax regret is neither theoretically satisfactory nor empirically supported. The analyses presented in this book suggest that P, not in the strict sense defined by the model but the vague feeling that one's vote may count, matters (a little, among a fraction of the electorate).

 

In a nutshell, people do not calculate expected benefits. They grossly estimate the benefits derived from their party or candidate being elected and they take into account the closeness of the race. But these are separate, not intertwined, considerations.

 

The upshot is that the expected utility model does not fit with the facts. The limits of expected utility have been known for a long time (see, especially, Allais 1953), and my study has illustrated them. We have seen that many voters overestimate P. And, perhaps even more important, what seems to matter is the perception not that one single vote might be decisive, as it should be in theory, but rather that the outcome may be close.

 

This suggests that there is some illusion at work. As Quattrone and Tversky (1988, 733) argue, many people fail to distinguish between causal and diagnostic contingencies: "some people may reason that if they decide to vote, that decision would imply that others with similar political attitudes will also abstain ... that is each citizen may regard his or her single vote as diagnostic of millions of votes, which would substantially inflate the subjective probability of one's vote making a difference."

 

Such reasoning is irrational. Quattrone and Tversky present experimental evidence that demonstrates that many people think in those terms. And I have shown in chapter 3 that responses to the statement that "my vote may not count much, but if all people who think like me vote, it could make a big difference" are related to turnout, even controlling for B, C, and D.

 

People, then, don't have it quite right. The half of the electorate who thinks about benefits and costs makes very rough estimates and fails to focus on the causal impact of their own personal behavior. Their (loose) decision rule seems to be that since the cost of voting is low, they should vote unless they perceive practically no difference between the parties and candidates. They are also slightly more likely to vote if they think the race is close.

 

P matters, but the way it matters is different from what rational choice would predict. On the one hand, P is just one marginal consideration that does not interact with B. As a consequence, perceiving P to be nil slightly decreases the propensity to vote; it does not, as in the standard rational choice model, necessarily lead to abstention. On the other hand, what affects the voting decision is not, as the model assumes, the strict probability that the individual could cast a decisive vote, but rather the probability that the outcome could be decided by a relatively small number of votes.

 

On empirical grounds, then, the rational choice model is only very partially confirmed. The model has considerable limitations. It would be tempting to conclude that the model should be discarded entirely. This is not the conclusion I have reached. I believe the model has some utility.

 

The fact remains that rational choice is more theoretically satisfactory than any of the alternative explanations. Take the resource and mobilization models. They tell us that people vote if th‑ey have the resources to overcome the cost of voting or if they are asked to. The problem is that these models take benefits for granted; they assume that everyone would like to vote if it were easy. Rational choice proposes a more complete picture, in which people consider both the advantages and disadvantages of voting.

 

It does not suffice, of course, for a model to be theoretically elegant. It must fit with the facts. But the rational choice model does fit partially with the facts. I have shown in chapter 5 that among those with a weak sense of civic duty, B, P, and C each has an independent impact on the propensity to vote. The elements of the calculus of voting play only at the margin, and only among a fraction of the electorate, but they do have an effect.

 

Furthermore, there are good theoretical reasons why rational choice does not perform as well In the case of voting. The reason is simply that whether the individual makes the "right"' or the "wrong" decision when she votes or abstains has minimal consequences in her life. Because voting is a low‑stake decision, people can easily afford to make "'irrational" decisions; they pay hardly any price for making the "wrong" choice. We should expect rational choice to offer less satisfactory explanations for these kinds of decisions, and this is precisely what we find. As I have indicated in chapter 6, rational choice faces some of the same problems when it comes to understanding why people decide to join or not to join groups, but the problems are less severe and the model does a better job. Voting is a tough test for rational choice. This has to be taken into account in assessing the model.

 

I have argued that the major motivation that leads most people to vote is the feeling that if one truly believes in democracy one has a moral obligation to vote. Oddly enough, rational choice may have facilitated the due recognition of this quintessentially irrational factor.

 

Rational choice did not invent the notion of duty. It has always been part of the voting literature. The American Voter (Campbell, Converse, Miller, and Stokes, 1960), for instance, identifies five factors affecting turnout, and sense of citizen duty is one of these five. The four others are partisan preference, interest in the campaign, concern over the election outcome, and sense of political efficacy Likewise, Milbrath and Goel (1977) list a whole litany of attitudes and beliefs that are related to electoral participation, among which sense of civic obligation appears.

 

Even though sense of duty is mentioned by these authors, it does not have any predominant status. In contrast, when it appears in rational choice models, it is usually acknowledged as a key factor. The first rational choice authors to do so were, of course, Riker and Ordeshook (1968). In their model, D is one of the three basic components of the calculus of voting, alongside B*P and C.

 

Indeed, once it is acknowledged that it is irrational, in the strict sense of the term, to vote, it becomes imperative to look for other factors that make most people vote. It is in part because I was interested in the rational choice model, both in its merits and its limits, that I felt the need to explore other kinds of.motivations such as sense of duty. And I would argue that the import of such feelings has been underestimated in the literature because there have been too few empirically minded researchers who were taking rational choice seriously.

 

The bottom line is that I cannot personally make sense of why people vote without taking into account the rational choice model. Those who are not fascinated by the fact that most people vote even if it is not in their best personal interest to do so miss something important, and only rational choice can nourish such fascination.

 

A few years ago, Green and Shapiro (1994) published a devastating critique of rational choice theory How does my verdict square with theirs? Our evaluations converge when it comes to assessing the empirical foundations of the model. Green and Shapiro argue that "despite its enormous and growing prestige in the discipline, rational choice theory has yet to deliver on its promise to advance the empirical study of politics" (7), and conclude their chapter on the paradox of voter turnout by saying that "readers interested in the determinants of voter turnout, in sum, derive little insight from the empirical work in the rational choice tradition" (68).

 

My own evaluation is slightly more sanguine. In my view, the model makes a real contribution to understanding of why people vote though that contribution is quite limited.

 

I would argue, contrary to Green and Shapiro, that the model provides a small but useful contribution to the explanation of voting for the following two reasons. First, rational choice has forced us to give as much thought to reasons for voting as to reasons for not voting. Before rational choice, the dominant paradigm was the alienationalist school (see Dennis 199 1), according to which nonvoting represents the absence of integration into the political. system. From such a perspective, voting is the norm‑the kind of thing people who have learned to deal with the system do‑and what needs to be explained is why some do not adhere to this norm. Rational choice, of course, reverses the perspective, as voting appears irrational. It is useful for political scientists to be reminded that there may be very good reasons not to vote. And, as I have indicated, it is easier to appreciate the crucial importance of motivations such as sense of duty once it is recognized that voting, from a pure self‑interest perspective, is irrational.

 

I disagree with Green and Shapiro on another important matter. Green and Shapiro dismiss too easily, in my view, the argument that rational choice is particularly ill‑suited to explaining voter turnout because turnout is associated with low costs and low benefits, and that its poor performance in accounting for why people vote is not a sufficient reason for questioning its overall validity.

 

Green and Shapiro call this arbitrary domain restriction. They question whether turnout is indeed a low‑cost activity by referring to "'Latin Arnerican elections, in which voters spend hours in polling lines, sometimes amid threat of violences" and to "'the more than 100,000 Aftican‑Arnericans who persevered through the intimidation and poll taxes of the Jim Crow South" (58). This is a cheap argument. What the rational choice literature has been dealing with is why people vote in a democracy; the theory is not supposed to explain behavior in situations where the democratic process is not functioning properly.

 

Green and Shapiro add that "there is nothing in rational choice theory that specifies the level at which costs or benefits are sufficiently small to render the theory applicable." According to them, rational choice advocates have been looking at turnout for a long time, thus implicitly acknowledging that the theory could be applied to this kind of behavior, and it is just an easy, post‑hoc excuse that is used "once it became apparent that no satisfying solution could be worked out" (58).

 

It is indeed not entirely clear at what precise level of costs and benefits the theory is or is not applicable. But should we expect such a degree of precision in a theory? Is it not sufficient to assume that the greater the stakes, the more relevant the theory, and, conversely, the lower the stakes, the less useful it is? This is at least what I would claim, and this does not strike me as a more imprecise proposition than those provided by other theories.

 

Green and Shapiro say that it is too easy to argue that a theory does not apply to a domain in which it does not seem to work. And, indeed, rational choice people would have been more convincing if they had made that point up front, before undertaking to solve the paradox of voting, rather than ex post. Still, the point is not whether it is an excuse or not but rather whether the argument makes sense. It does. It does because it is entirely consistent with the logic of rational choice theory. The rational person should allow more time, resources, and attention to those decisions that have greater implications on her wel.1‑being than to those with marginal consequences. While the argument may be suspect, given the context in which it has been advanced, it is nevertheless a valid one.

 

In short, Green and Shapiro seem to believe that the contribution of rational choice to an understanding of why people vote is tiny and close to nil. My own verdict is that its contribution is quite small but still useful, and that there are good reasons why the model performs poorly in low‑stakes situations.

 

I feel compelled to end this book on a personal note. Over the last few years, I have been reflecting on the reasons that lead people to vote or not to vote. I have looked at turnout rates around the world for all kinds of elections, I have done surveys, semi‑structured interviews, and an experiment with university students. I have also talked informally with many people, obsessed as I was to grasp their B, P, C, D, or other motivations.

 

During the same period, I have had to decide whether I should vote or not. This was no easy decision, especially as I like to think of myself as a rational person and as, like most of my fellow citizens, I have tended to vote in most elections in the past. Is it rational for me, or for anyone for that matter, to vote?

 

I have been and I am still struggling with this question. I have tried to formulate a general guideline that I am willing to recommend. At this stage, the reader will not be surprised to learn that my decision to vote or to abstain depends on my perceptions of B, P, C, and D.

 

The very first question to address concerns D: Do I have a moral obligation to vote? My answer is: Yes ... to a certain extent. My reasoning is quite similar to that of those I have described as voting out of a sense of duty. I strongly believe in democracy; I am convinced that the democratic form of government, while not a panacea for all human problems, contributes to the quality of life in a country, and I am looking forward to an era where all the governments of the world will be selected through free and fair elections.

 

This, of course, is not a sufficient reason to vote. Should not the rational individual reason that whether she votes or not will not salvage or jeopardize democracy in her country? My only response is that, like many of my fellow citizens, I feel that I must act in accordance with the principles I believe in. As I think of myself as a democrat, it would be incongruous not to vote. I vote, then, because I want to be consistent with my principles. Yes, I would feel somewhat guilty not to vote.

 

The obligation to vote is not absolute, however. Most important, I do not feel compelled to vote if I have reached the conclusion that it does not matter who wins the election (B is nil) or if I think that none of the parties or candidates deserves to be supported. In other words, I do not feel obliged to vote if I have no good reasons to vote for anyone.

 

What about P? Is the obligation to vote less compelling if the race is not close, or if my preferred candidate has no chance of winning? Here again much depends on B. If I have strong views about which party is the best, I feel I should vote. If I have only slight preferences, then I might decide that it is not worth voting if the outcome is a foregone conclusion.

 

Am I not deluding myself when I decide to vote partly because the race is close? I hope not. I start with the initial position that I should normally vote but I am willing to reconsider this position if B is nil or if B is small and P is nil.

 

What about C? In my view, the cost of voting is extremely small and does not vary much from one election to another. As a consequence, C does not have much impact on my decision, except in an indirect fashion; the fact that C is low makes it easier to vote.

 

According to this account, I vote most of the time but not necessarily all the time. My most important reason for voting has nothing to do with rational considerations, although I do pay some attention to these, first and foremost to B but also to P and C. Even though I would not claim to be a typical citizen (especially with respect to my level amount of time I devote to whether I should of Political interest and to the vote or not), the same message applies: rational choice provides a small but useful insight into my reasons for voting.

 

My last word will be a plea for more research. In my view, further research is needed along three lines. First, we need more qualitative studies to deepen our understanding of the sense of civic duty. I hope to have thrown some light on what goes through people's minds when they feel they have a moral obligation to vote. But how strong and stable are such feelings. How are they correlated to other attitudes and ideologies? When are they formed? And how do people define the limits of obligation, that is under which circumstances does the citizen feel free not to vote? These are crucial questions that deserve in‑depth analyses.

 

Second, we need cross‑national comparative analyses of why people vote. Rational choice advocates have traditionally relied on secondary data that were often of aggregate nature or that had been assembled for different purposes. It is time to design studies with specific measures of B, P, and C. Robert Young and I have done precisely that in a number of studies, the findings of which have been reported here. This is, of course, not enough. We need similar surveys in different countries to determine whether the patterns we have established can be generalized‑ A particularly intriguing question is whether P is perceived and plays differently in PR systems. Strictly speaking, the probability of casting the decisive vote is the same in all kinds of electoral systems but I would formulate the hypothesis that the perceived probability is greater in PR systems.

 

Third, we need panel studies to grasp how individuals approach the decision to vote over time. As I have shown in chapter 2, age is closely associated with voting: people are more likely to vote as they grow older, Why? I have demonstrated in chapter 5 that this is partly because sense of duty increases with age, but we need to understand what people learn over time that makes them feel a greater obligation to vote. Panel studies should also help us to determine whether habit is an important factor in the decision to vote. My own analyses have little to say about this.

 

Further research may lead me to revisit my verdict. For the time being, I conclude that voting is mostly irrational but that rational choice contributes slightly to understanding why so many people decide to vote in most elections.