With the 2016 Presidential Election just 19 days away, this seems an appropriate juncture to pause and consider what it means to cast an informed and rational vote. The results from the earliest studies of public opinion in the 1940’s and 1950’s caused serious doubts about the American voter’s ability and willingness to acquire and process the information necessary to cast an informed vote. In 1957, Anthony Downs articulated a set of rationality assumptions by which vote choice could be evaluated. Since then, scholars have learned a great deal about how attitudes are acquired and expressed and how information is processed. Meanwhile, the project of evaluating vote choice has continued. Here, I briefly summarize some research in the Downsian stream concerning voters’ limitations and how they try to overcome them. Ultimately, I conclude that evaluating vote choice for rationality may be an unfruitful enterprise, but there have been fruitful developments besides.
Downs’ An Economic Theory of Democracy (1957) was an attempt to formulate a positive theory of democracy by applying the methodology of economics—specifically, methodological individualism and the rationality assumption—to actors in the political sphere. This undertaking required, in part, an account of the individual voter analogous to the individual consumer, a homo politicus (political man) corresponding to the stylized type homo economicus (economic man). Indeed, rational choice scholars have long contended that both private and public sector decision making are amenable to evaluation by one unified method.
It is important to note what rationality is and is not. As Downs explains, rational action is action “reasonably directed toward the achievement of conscious goals” (Downs 1957, 4). A rational man is not “a man whose thought processes consist exclusively of logical propositions, or a man without prejudices, or a man whose emotions are inoperative” (Id. at 5). Rather, rationality implies simply “a man who moves toward his goals in a way which, to the best of his knowledge, uses the least possible input of scarce resources per unit of valued output” (Id.). So, a rational voter is one who votes in order to maximize his individual utility. His utility function (bundle of preferences) need not consist solely of selfish desires; it need only consist of those ends which are his own.
Measuring rationality can be tricky because it is difficult to observe an agent’s ends directly. One way of dealing with this problem is to assume that “every man’s behavior is always rational because (1) it is aimed at some end and (2) its returns must have outweighed its costs in his eyes or he would not have undertaken it” (Id. at 6). Of course, this assumption is tautological: every man’s behavior becomes rational qua definitione. Downs avoids this pitfall by another key assumption: “If a theorist knows the ends of some decision-maker, he can predict what actions will be taken to achieve them” (Id. at 4, emphasis mine). If the decision-maker’s end is given, the theorist can then measure rationality by assessing the degree to which an agent’s means are reasonably suited to his ends. This deft evasion of the tautological pitfall is not without cost, however, which shall become clear in later discussion.
Given the structure of the American two-party system, the American voter typically confronts a dichotomous choice between competing providers of public goods—the Democratic and Republican parties—neither of which fully aligns with his preferences. If the differential between the two parties is significant, the Downsian voter casts his ballot for the party which most closely aligns with his preferences. If the differential between the two parties is not significant, the rational voter may choose to adopt a decision rule based on an assessment of prior performance or forego voting altogether.
It is also important to be clear about the import of the model under discussion here. For vote choice scholars in the Downsian stream, the extent to which the typical voter approximates the rational voter has important normative implications for the quality and viability of American democracy. In a political order guided by the will of the people, the stability and health of that order may well depend upon the ways in which the people make decisions.
By definition, homo politicus votes to achieve desired policy goals. In the voter choice literature, the relationship between the rational voter and the issue voter is well-articulated by Carmine and Stimson:
The common—indeed, universal—view has been that voting choices based on policy concerns are superior to decisions based on party loyalty or candidate image. Only the former represent clearly sophisticated behavior. Indeed, the policy-oriented vote is a defining characteristic of that mythical specimen, the classic democratic citizen (1980, 79).
Carmines and Stimson, however, challenged the notion that issue voting is necessarily a sign of political sophistication. They argued that not all issues are created equal, and issue voters could be sub-divided into two types—easy and hard—based upon the characteristics of their chosen issue. An easy issue is defined as “symbolic rather than technical”, focused on “policy ends rather than means”, and is “long on the political agenda” rather than newly arising (Id. at 80). An easy issue “makes possible a gut response elicited equally from well-informed and ill-informed, from interested and uninterested, from active and apathetic voters” (Id.). While hard-issue voters do seem to resemble the Downsian type, employing a well-reasoned decision calculus across a range of issues, easy-issue voters do not significantly differ from their non-issue-oriented counterparts, the authors claim.
Carmines and Stimson support their argument about the bifurcated nature of issue voting with an examination of the 1972 presidential election, “widely described as issue-oriented, perhaps more than any other” (Id. at 81). This election is an apt study because “it is dominated by two issues, Vietnam and desegregation, each an exemplification of hard and easy issues, respectively” (Id.). Carmines and Stimson use survey interview responses to ascertain the policy preferences of voters. Then, having operationalized an issue voter as an individual for whom knowledge of the individual’s policy preference increases the predictability of the individual’s vote over and above that of his party identification, Carmines and Stimson use probit analysis to assess the probability of an individual voting for McGovern. By comparing the predicted vote to the actual vote, the authors establish a fourfold typology: the non-issue voter, the easy-issue voter, the hard-issue voter, and the constrained-issue voter. Only the first three types are necessary to consider here. The non-issue voter is one for whom no greater predictability can be obtained beyond party identification. The easy-issue and hard-issue voter are voters for whom a desegregation or Vietnam policy preference yielded greater predictability beyond party identification alone.
Finally, Carmine and Stimson compare the easy-issue and hard-issue voter and find that the two groups differ systematically and significantly in levels of formal education, factual political knowledge (measured by a short test of basic civics knowledge), and level of political activism. Across all three dimensions, each deemed closely related with voter sophistication, the hard-issue voter outperformed his easy-issue counterpart. Carmine and Stimson’s conclusion is that easy-issue voters are issue voters but that the issue which motivates their vote requires much less information. Thus, the policy-oriented nature of the vote aside, there is little to suggest that the easy-issue voter is distinguishable from the non-issue voter, they argue. The authors conclude that “The surge of easy-issue voting is not an encouraging phenomenon for those who would hinge the viability of democracy on the ability of citizens to choose rationally between alternative issue positions of parties and candidates. The surge in issue voting seems likely to occur on a large scale only when the choices are simplistic” (Id. at 88).
But does the hard versus easy-issue distinction hold up? The categorization seems arbitrary. The two examples Carmines and Stimson proffer, desegregation and Vietnam withdrawal, are defended almost entirely on intuitive grounds.
Although the policy conflicts involved in desegregation can be detailed in great complexity, we think it reasonable to assume that the typical voter sees in it a simple issue…there are virtually no technical or pragmatic issues in it…The Vietnam War was quite a different issue….War and peace are simple enough ends, but the candidates did not offer that choice. Instead, the electorate was presented with alternative plans (one of them “secret”) to end the war…While antiwar activists may have a seen a wide gulf separating the candidates’ positions, we believe most voters saw the issue in far more narrow terms, focusing mainly on speed and conditions of withdrawal (Id. at 80-81).
Ultimately, desegregation is easy not because voters might not perceive complex social consequences involved in the question of if, when, or how desegregation took place, but because Carmines and Stimson “think it reasonable to assume that the typical voter” did not see it that way. Likewise, the Vietnam conflict is hard not because voters might not harbor simple affective preferences about if, when, or how the conflict was concluded, but because Carmines and Stimson “believe most voters saw the issue” in a more nuanced light. This hardly seems an adequate defense for a categorical distinction on which the entire study hangs.
Leaving intact, arguendo, the hard versus easy-issue distinction, Carmines and Stimson may push the point too far. Recall that a rational voter is one whose vote is reasonably well-calculated to achieve his own policy ends. According to Downs, the agent’s ends are not the proper objects of evaluation but are to be taken as given. But Carmines and Stimson seem to smuggle back into the analysis at least some evaluation of the voters’ ends. Once homo politicus is put up on a pedestal, it is perhaps too tempting to resist ascribing to him the noblest ends as well as the most efficient means.
Not all scholars in the Downsian stream are as pessimistic about the lack of sophistication of some issue voters. Arthur Lupia, in “Shortcuts Versus Encyclopedias: Information and Voting Behavior in California Insurance Reform Election” (1994), presents evidence from exit interviews with California ballot initiative voters that low information voters tend to vote similarly to high information voters through the use of informational signals from information providers. In short, voters who, through lack of time or interest, admittedly knew very few details about the multiple and highly technical propositions on the ballot, were surprisingly likely to vote their putative interests through the use of informational shortcuts, such as knowing the source of support for a given proposition. Lupia’s work suggests that there may be hope for the unsophisticated issue voter after all.
Yet, these studies highlight an important weakness in the rational voter model: we can only evaluate the fit between means and end if we know an agent’s end. Carmines and Stimson rely on survey response data to determine voter preferences while Lupia makes his own assumptions about what is in a voter’s economic interest. In the Lupia study, voters were selecting between five different propositions, each supported by one of three information providers: a consumer group, an insurance company, and the Trial Lawyers Association. So-called sophisticated voters were those who, in an exit interview, demonstrated a general grasp of the content of the propositions, and unsophisticated voters were those who did not. All voters who cast their ballot in favor of the consumer group-backed proposition were deemed by Lupia to have been acting in their own economic self-interest. Voters who voted for one of the propositions not supported by the consumer group were deemed to be acting contrary to their own economic self-interest.
Each of the complex propositions, however, were designed to deal with two separate considerations—auto insurance rates and access to court for auto accident claims—in different ways. Each proposition constituted a basket of trade-offs designed to balance the desire for lower insurance rates with the desire to protect access to court for meritorious auto accident claims. Lupia assumes that a vote cast for any of the non-consumer group-backed propositions is an irrational vote because he presumes to know what balance the individual voter is attempting to strike. However, we cannot be sure whether a given vote was reasonably calculated to achieve the desired end unless we can reliably determine what the desired end is. Lupia’s mere assertion of a voter’s interest lacks epistemic warrant. Even if we could determine with reasonable certainty the voter’s aim, it still may not be clear what the most effective means are and reasonable persons may disagree. Indeed, policy experts are often uncertain about the expected benefits of a given policy, especially when important trade-offs are involved. For this very reason policy makers often pursue incremental reforms in order to isolate the cause of error and minimize its effects (See Lindblom’s “The Science of ‘Muddling Through’” (1959)). Therefore, assuming a voter’s end may avoid the tautological pitfall yet still prove untenable.
Public Opinion Surveys: What Are They Good For?
Evaluating vote choice by assuming a voter’s end is problematic, but so is reliance on survey data as Carmines and Stimson do. John Zaller, in The Nature and Origins of Mass Opinion (1992) demonstrates both theoretically and empirically what a weak reed public opinion survey data can be and provides a more complex way of understanding the way voters acquire and process information. After a broad literature review demonstrating the general ignorance of members of the American public on political matters and the inexplicable variability of survey responses, Zaller lays out four axioms which form the foundation of his theory and from which he draws several testable deductions. Testing these deductions occupy the latter half of the book. I will briefly discuss each of these axioms and Zaller’s most significant deduction before offering concluding thoughts.
Axiom 1, the Reception Axiom, holds that “the greater a person’s cognitive engagement with an issue, the more likely he or she is to be exposed to and comprehend—in a word, to receive—political messages concerning that issue” (Zaller 1992, 42). Zaller chooses cognitive engagement over political interest because the former focuses on individual acquisition of political information, while the latter less helpfully focuses on affective involvement. “For example,” Zaller adds, “people who score higher on tests of political knowledge are substantially more stable in their attitude reports than people who score low in political awareness; however, people who describe themselves as highly interested in politics…are not significantly more stable than persons who express little political interest” (Id. at 43). Thus, Zaller’s measure of cognitive engagement consists of a neutral test of political knowledge involving questions on public affairs, which he finds to be more reliable gauge of an individual’s level of attentiveness to politics.
Axiom 2, the Resistance Axiom, holds that “people tend to resist arguments that are inconsistent with their political predispositions, but they do so only to the extent that they possess the contextual information necessary to perceive a relationship between the message and their predispositions” (Id. at 44). In other words, individuals are resistant to arguments from “the other side” of the political spectrum but only to the extent that they are aware that those arguments are from the other side. Here, Zaller relies on psychological literature indicating that individuals more readily embrace ideas from trusted sources and are highly skeptical of ideas from rival sources.
Axiom 3, the Accessibility Axiom, holds that “the more recently a consideration has been called to mind or thought about, the less time it takes to receive that consideration or related considerations from memory and bring them to the top of the head for use” (Id. at 48). This insight, which Zaller draws from cognitive psychology, means that individuals are not super computers capable of holding all the information they have previously consumed readily available at once for application to a given question; rather, they most easily recall ideas which they have thought about most recently.
Axiom 4, the Response Axiom, is closely related to Axiom 3. It holds that “individuals answer survey questions by averaging across the considerations that are immediately salient or accessible to them” (Id. at 49). Simply, individuals respond to survey questions by sampling only from those ideas that most readily spring to mind, and their opinion statement constitutes a mere composite of perhaps diverse and even conflicting ideas.
Zaller’s most theoretically significant deduction, the Ambivalence Deduction, follows most directly from Axiom 2 and combines with Axiom 4 in interesting ways. If, as Axiom 2 states, individuals can adequately resist arguments inconsistent with their attitudes only to the extent that those individuals are cued in to the directional implications of those arguments, and members of the mass public are as generally ignorant of politics as the literature suggests, it follows that members of the mass public might unwittingly hold internally inconsistent political attitudes. Thus, relatively politically inattentive people can be expected to demonstrate a high degree of ambivalence.
Understanding the variability in survey responses then requires only the combination of the Ambivalence Deduction and Axiom 4, the Response Axiom. If most members of the mass public are walking repositories of inconsistent political ideas, and answer survey queries by a haphazard sampling of those ideas which are most cognitively available, survey responses may tell us more about the most recent news coverage to which a given respondent has been exposed or the framing of the survey question than the respondents’ fixed true attitudes. Zaller recounts a host of survey results which demonstrate this point rather neatly. Indeed, Zaller asserts, “individuals do not typically possess ‘true attitudes’ on issues, as conventional theorizing assumes, but a series of partially independent and often inconsistent ones…[T]rue attitude seems a grievous simplification” (Id. at 93).
Most of the heavy lifting in Zaller’s theory is done by the Ambivalence Deduction and the Response Axiom working in tandem. One example of this is found in chapter six, “The mainstream and polarization effects.” This chapter opens with an examination of public reaction to Nixon’s 1971 speech announcing wage and price controls and the reversal of support for the war in Vietnam in the 1960’s. Zaller’s aim is to account for the effects of elite communications on mass attitudes by examining a case “in which elites achieve a consensus or near consensus on a value or policy” (Nixonian controls) and another “in which elites disagree along partisan or ideological lines” (Vietnam in the late 1960’s) (Id. at 97).
In the case of Nixonian wage and price controls, public support grew after Nixon’s speech on the subject and remained high among members of the public as conservative and liberal elites embraced the policy. Likewise, public support for the war in Vietnam was high during the early 1960’s while elites provided unified support. Zaller calls this the mainstream effect—when members of the mass public receive the same cues from partisan elites, mass liberals and conservatives can be expected to be very close in their attitudes. But, by the late 1960’s, elite support had divided, with conservative elites maintaining support for the war and liberal elites turning in opposition. Zaller notes that, when elites divided over the war, “increases in awareness [were] associated with greater polarization of the attitudes of mass liberals and conservatives” (Id. at 102). Simply, the more cognitively engaged members of the mass public received more cues from their respective elites and were more resistant to cues from ‘the other side.’ This is the polarization effect and shows the power of Zaller’s model.
Though Zaller presents a wide range of empirical support for the mainstream and polarization effects, the implied causation—from elite preference formation, to elite cues, to mass preference formation—does not seem fully substantiated. Could there be some factor driving elite preference change and mass preference change? For example, educated mass liberals began to change their mind about the war around the same time educated elite liberals began to change their mind about the war. Elite influence is one plausible explanation, but could it not be that individuals who are similarly disposed, educated, and socialized will tend to react similarly to political events? Moreover, once one considers that similarly educated elite liberals react similarly to educated mass liberals, the partition erected by terms like “elite” and “mass” begins to appear rather thin.
Nevertheless, several important implications follow from Zaller’s model. One implication is that polarization may be here to stay. Recall from the Resistance Axiom that “people tend to resist arguments that are inconsistent with their political predispositions, but they do so only to the extent that they possess the contextual information necessary to perceive a relationship between the message and their predispositions” (Id. at 44). Members of the mass public are predicted to be ambivalent to the extent that they are inundated with messages from both sides of the American ideological divide and are unable to discern the source of each message. Thus, an inattentive liberal (or conservative) member of the mass public is likely to be receptive to and sympathetic with rival arguments confronted in the public sphere. But some social scientists warn that Americans are becoming more atomized in their lifestyles (Putnam 1995) while the proliferation of niche media outlets can effectively insulate the consumer of public information from rival views.
Most importantly, Zaller’s work demonstrates that public opinion survey data is unreliable. If survey data is unreliable, then it is not available to save the Downsian rational voter analysis from the tautological pitfall and the blind assignment of preferences to voters by researchers. Zaller’s work shifts attention away from evaluating voters for rationality and focuses more fruitfully on the question of how public opinion is influenced in the context of public life. In the Zaller research program, voters who are limited in time and cognitive capacity respond to the complexities of political life by acquiring ideology which then permits the selective acquisition, processing, and application of information gleaned from public discourse. This process is messier than the pristine Downsian model, but also appears to capture more faithfully the realities of American political life.