CHAPTER SIX The Third Way: The Theory of Affectiv igence and American Democracy Intell MICHAEL MACKUEN, CEORCE £ V. RUSSELL NEUMAN MARCUS. AND LUKE K E E L E jp the late 1940s and early 1950s political scientists began to make use 0( large national surveys to develop empirical theories of American po-Ijtitdl behavior and political judgment. From scholars at Columbia University and the University of Michigan came what has come to be called the psychological model: a now well-known and widely accepted portrait describing public ignorance of the major candidates and where they stood with respect to the predominant issues of the day. Moreover, the psychological model advanced the claim that partisan voting decisions were derived from a robust reliance on partisanship, whereas the voting decisions of independents resulted from responsiveness to "short-term" forces (hence the colloquial name "swing voters").1 The psychological model, more commonly called the "normal vote" model, best articulated -jn The American Voter (Campbell et al. i960), has often been taken as challenge to democracy. And although there have been many attempts to reCast these findings in a more positive light (Achen 1975; Key and Cum-mjngs 1966; Mueller 1999; Page and Shapiro 1992; Stimson 2004), the challenge to the competence of the voters remains unchecked. Another account of voting, that of rational choice, arrived shortly afterward from yet another sister social science, economics. In its initial foniU'lation, rational choice held that voters engaged in a rational consider^'00 of the alternatives presented to them, choosing that which best , _ Short-term forces are good times or bad, such as a good economy or a bad one, war 01 pCaCfi scandals or their absence, a particularly good or bad candidate or campaign—and, of coufsc-1 verY 'arSc Plrt °f lhp adult population not sufficiently interested to participate, no mattcr w',at situation, a fact that is of considerable concern and attention (Burn-ham i°So: LidtI '9*S: Schattschneider i960). heir interests (Downs 1957)- Rational choice posits an attentive *****1 j f0\ electorate that makes explicit comparisons and adjudi-them through rational evaluation of their respective costs c*teS g{J ^ jack of empirical support for rational choice (Quattrone *fd ^"^y jcj88) led to a more qualified theory of "bounded rationality." ""'^^PP1'03'-1"1 nas attcmP,ec*t0 Ciedl WIln tne problems of satisfying rive standards while dealing with seemingly intractable empiri-^"hallenge8- Scholars have wrestled with the dystrophic implications ^fthe "black and white" model advanced by Phil Converse (1964, 1966, ° - ) which asserts that most people have few, if any, organized political J^as If democracy requires an attentive and politically learned elcctor-and requires voters to give at least modest attention and thoughtful onsideration to the policy and leadership choices before them, then neither account suffices. We advance the claim that the theory of affective intelligence offers a comprehensive account that incorporates the insights of these two accounts of voter behavior. We argue that each of them identifies a special case of a more general range of outcomes. By this we mean that each has erroneously taken a special case of political judgment and treated it as if it were the general case. How can it be that the normal vote and rational choice accounts are special cases, that is, theoretical specifications that apply only in some rather than in all circumstances? The two established theories presume that voters have invariant patterns of judgment and behavior. In the case of the normal vote account, voters are either partisan or not, and these immutable qualities fully control what people do, for example, whether they will pay attention (partisans do, independents do not), when they decide for whom to vote (partisans early in campaigns and nonpartisans late), and so forth, Partisans have certain qualities and they consistently display them, just as nonpartisans display their characteristic qualities (as we shall see, a similar case can be made for ideology as a stable defining quality).2 In the case of rational choice theory (or its more recent variant, bounded rationality), voters think and act rationally all the time and in every circumstance so long as at least minimal stakes are in play. There is no logical barrier preventing the psychological and rational choice theorists from entertaining the possibility that voters shift between different decision strategies, but doing so requires formulating what the 2. John Zaller's (1992) work, justly celebrated, is the most sophisticated example of •hose that work from this premise. His Receive-Accept-Samp'.e (RAS) explains how people attend and respond to information. 126 MAfKl'fX, MARCUS, NIUMANi AND KF.UE alternative strategies might be and what would initiate shifting from one to another. The theory of affective intelligence offers an alternative account that specifies the alternative strategies, the factor that shifts voters from one strategy to another, and the consequences.3 Perhaps the most often noted feature of the theory of affective intelligence is that it makes the counterconventinnal claim that emotion's impact is largely functional and rational. Equally important, it explains how emotion controls the way in which voters make political judgments. We argue that the effect of anxiety is largely conditional. This conditionality generates a model that is far more dynamic than a conventional model. Conditionality enables an important "if then" component that allows the theory of affective intelligence to model rwo courses of action for each individual. It is this conditionality, operationalized as statistical interactions, that enables the theory of affective intelligence to subsume what are most often seen as contending and antagonistic theories, the psychological, or normal vote, and rational choice approaches. the political geography of affective intelligence The theory of affective intelligence holds that people have two basic decision strategies available and that they easily move from one to the other and back again. Why do people need more than one strategy? If people-have the capacity to be rational why do they not rely on that capacity in all situations given that rationality promises much and its opposite, irrationality, promises so little? The standard answer is that rationality, as a cognitive process, is very demanding (even for those who might be gifted in its practice) and so its demanding character prevents its universal display. The theory holds that rationality is appropriate only in some situations. More fundamentally, the theory holds that people have alternative decision strategies because different environments require them. The theory identifies two geographies, each of which demands a different strategy. In the first, the geography of familiar situations, it is efficient to swiftly and automatically (Bargh and Chartrand 1999) rely on previously learned routines. People in familiar and recurring choice situations can rely on the same decision-making strategies as they have in the past, for often. 3. Wc do not mean to imply that these are two discrete and thereby mutually exclusive strategies. Indeed, they may form ends of a continuum with a mixture of both available when both aie engaged. The Third Way 117 past success predicts future success. Moreover, the swift and deft management of social and behavioral interaction depends on unself-conscious niodes 0f articulating speech and action. It is costly and unnecessary to use the time and effort required to arrive at a decision via explicitly rational calculation when the same decision has to bo made again and again in the same environment. If all Democratic and all Republican candidates advance the same consistent policy stances, then why invest the time to learn about the newest candidate? If today's economic situation seems similar to yesterdays, why watch the business report? In such situations, what worked before, what becomes embedded in the heuristics of choice, operates to swiftly avail us of prior choices that will likely be as effective in the present as they were in the past.4 In such circumstances, voters display habitual choice as their decision strategy. But we do not always find ourselves in the domain of the familiar. Sometimes, we find ourselves in unexpected and novel situations. When we find ourselves in the political geography of uncertainty, we cannot safely or prudently rely on past lessons, especially lessons that are embedded in automatic judgments (Bargh et a). 1992). Practiced routines become unreliable guides and are likely to be ill suited to novel terrain. The theory of affective intelligence holds that in such circumstances we turn to the less often used mode of explicit consideration. Rationality, as a decision-making process, if not well suited to the familiar realm of habit—being too time-consuming and too costly—is critical to managing uncertain conditions (see table 6.1). Increased anxiety tells us when we are entering the geography of uncertainty.5 Absence of anxiety tells us we are in the realm of the safe and familiar and that we can rely on past actions that will, as they have before, successfully manage our lives. And in such circumstances people display habituated choice as their decision strategy. But there is more at stake than merely asserting that the central role of anxiety has been ignored. 4. Heuristic, as wx use the term, is synonymous with the following equivalent terms: preference, predisposition, conviction, standing decision, affective disposition, Milne, opinion. »nd attitude. Some of these, e.g., value, opinion, and attitude, are thought to be primarily semantic and consciously available, whereas othets, e.g.. preference and predisposition, ■nay be less accessible (Wilson 2002; Wilson, Kraft, and Dunn 1989: Wilson and Schooler •991). Notwithstanding that distinction, people often make curtent choices by relying on »otne previously learned standards, and we take all of the above terms to depict embedded choices, however they have been acquired or however they maybe expressed. 5. We do not claim infallibility for emotional preconscious appraisals. As with conscious perception, preconscious appraisals may well be erroneous and, further, share with Conscious assessments the prospect of elite manipulation. >l8 MACKUEN Mien v. v-ri... ... o . M*RCLS. NElMAJi, AND KFELC tabic 6.i: Responses to po't cal geography Po' tical geography familiar and rewarding context cOnteo t Erth^jiasm Habituated choice: Rely on routines (e.g., partisan a-d ideo'ogy convictions govern)—the situation depicted as the general case of the normal vote account Unfarn1 ar and uncertain An«iety Deliberative choir: Rely on expl cit 'ea'ningof a'ternatr.es ard thojghtful cons deration of available choices—the situation depicted as the general case by t*-e ra: o-nal choice approach We argue that the theory of affective intelligence significantly revises conventional thinking about electoral behavior in America. Spezio and Adolphs (chapter 4 in this volume) argue that affective processes and cognitive processes are simultaneously and continuously active, raising a question as to whether their careful critique of dual models of emotion and cognition applies to our work. Their view is one we concur with, though that might sound surprising given our adoption of a dual model of affect and cognition (with habit and hence the affective processes that sustain execution of habits being more dominant at some times and less so at others). As Spezio and Adolphs clearly state, neuroscience generally defines cognition as information processing, whether conscious or not. With such a definition affective processes become one variant of cognition (since affective processes are primarily but not exclusively information processing). We use the older, traditional meaning of cognition, "to cogitate," that is. to expressly think and reflect belore acting (and here also affective processes are active, though not the same as sustained execution of habits). And though the Tht Third Way uo, Ution of nonconscious roles for cognition is important (and likely ^become more so as neuroscience progresses), the role of expressed t0 sidcration has importance not only to our lives as humans but in Hollar f°r tne sPet'a' ro'e >l plays in liberal political democracy. Only by public sharing of our intentions, goals, and values can we resolve our differences and reach common purpose via democratic political mechanisms. THE THIRD WAY: RECONSIDERINC AMERICAN POLITICS The standard view of the American electorate, derived largely from the psychological school developed over the course of the past fifty years, articulates a set of expectations that have become conventional wisdom. We demonstrate how the theory of affective intelligence recasts many of these conventional assertions. CONVENTIONAL WISDOM li PARTISAN VOTERS DECIDE BY MEANS OF INTRANSIGENT RELIANCE ON DEEPLY HELD CONVICTIONS Normal vote assertion: Dispositions anchor partisans and ideologues. A representative and current example of this view is given by Stimson (2004, 183): "The committed partisans make their decisions long in advance, many before the campaign begins, before the candidates are known." Affective intelligence elaboration: Dispositions are used when appropriate and abandoned when inappropriate. Hence we expect that partisans will assert loyalty, but when they are anxious, we should observe a propensity for rejecting reliance on established partisan convictions. The theory of affective intelligence argues that voter competence is dynamically responsive to the strategic character of the political geography. For this purpose, being rational is using different strategies of choice depending on the political context. First, it makes rational sense that voters rely on heuristics in familiar, recurring situations because they yield a high probability of success. Second, it makes rational sense to abandon heuristics when situations are novel and or uncertain and instead rely on considered judgments based on contemporary information. Hence voters are sophisticated to the extent that they shift from reliance on heuristics to considered judgments when conditions change from certain and familiar to uncertain. And it is rational for voters to return to reliance on 13° UACXUBM, MARCOS, N E O M A N , AND KHLF heuristics when conditions return to normal. It, therefore, follows Uu, voter sophistication comprises the ability to rely on predispositions und* the appropriate circumstances as well as the ability to abandon them 3 other appropriate circumstances. 'n CONVENTIONAL WISDOM 21 SWING VOTERS DETERMINE ELECTION OUTCOMES Normal vote ussertion; Election outcomes depend on the dynamic of the swing vote and mobilization (turning out the partisan base). Partisans are intransigently loyal, but they may turn out at higher or lower levels in any particular election as a result of particular mobilization (and demobilization) efforts. But it is, as Stimson (2004,182) describes, the independent voter who is decisive in an election: "That leaves the horse race to be decided by those of middling interest and knowledge, but no commitment to one side, our score keepers. Attentive to outcomes, not party or ideology, they are not involved enough to care much about the early, primary, stage of the campaign. They sit on the sideline as judges, watching to see what the parties will do. They are detached, usually having no party and not wishing to involve themselves in producing a candidate."4 Affective Intelligence elaboration: Partisan defection is a more frequent and consequential part of every election than one might expect from the normal vote model. Election outcomes are largely the result of a party's securing a net advantage in partisan defection, not the movement of swing voters. The claim of partisan intransigence is central to the standard view although election narratives have long noted that partisans may support the candidates of the other party (for example, "Reagan Democrats" or "Clinton Republicans"). Hence our understanding of campaigns and how they are run largely misses an important feature of American elections: effective political campaigns often turn on their ability to recruit support from the hostile opposition. 6. Many American who could participate if the)- were sufficiently motivated ignore elections. Again, Stimson (2004. 181) offers a representative account: "The inattentive never tune in. They answer questions when pressed, but their answers signal neither conviction nor intent. Mostly they will not vote, constituting the great bulk of the eligible electorate which sits out even presidential elections." The Third Way 1 ji eMTlONAL WISDOM 3: AMERICAN ELECTIONS NATURALLY PERIODIC I vote assertion: Scholars have noted a cyclical pattern in American -s whereby periods of liberalism are followed by periods of conser-tisw andso on (Stimson 1991). The standard view has little to say about why such sinuous shifts take place. Affective intelligence elaboration: The public pays more attention than the conventional wisdom holds but not as a constant feature of all voters or of most elections. Attention levels rise and fall according to emotional signals about the strategic character of the political geography. Moreover, attentiveness does not automatically lead to partisan cheerleading (loyalty). Major events (for example, shifts in economic conditions) impact the electorate through the mechanisms of emotional appraisal. In addition, as administrations gain greater success they cause anxiety among their partisan base and generate the conditions of defection. Moreover, cyclical patterns (from conservative to liberal to conservative) may have their foundations in the emotional responses to governance. Hence the theory of affective intelligence has macro as well as micro implications. The theory of affective intelligences advances a counterintuitive explanation: the more successful governments are at enacting their policy proposals, the more likely they are to generate anxiety among their supporters. That is, support for a government will begin to crumble as the part)- leaders have success in enacting their programs. We expect that as governments achieve more of their political goals, thereby changing familiar terrain into the unfamiliar, their partisan supporters will become increasingly anxious, creating the conditions that make them open to withdrawing their support. This latter pattern may provide a critical element in accounting for the cyclical pattern of American politics. In sum, we expect that partisans play an active role in determining election outcomes by shifting from reliance on their established convictions to deliberation about the best options when unfamiliar conditions generate anxiety about their "normal" choice. We also expect that, as a result, the net defection among partisan camps advantaging one party or another will have a considerable impact on election outcomes. We now turn to the empirical evidence for the hypotheses we have advanced. At the center of these claims is that anxiety is a pivotal assessment of the nature of the immediate circumstances. The level of anxiety-should modify the mode of rationality to ensure that the appropriate form MACKUEN. MARCUS. N f. V M A N . AND KEELE of rationality is articulated: reliance on tested convictions in familiar and settled circumstances, on one hand, and attentive and deliberate conftfl eration of the available choices in novel and unsettled circumstances, on the other. In t his discussion we use the American National Election Studies (ANES) for the five presidential elections from 1980 to 1996. This group of data sets contains the measures that we have previously used (Marcus, Neuman and MacKuen 2000) and hence allows for comparable analyses." IDEOLOCY AS A POLITICAL DISPOSITION For much of the twentieth century Americans used the liberal-conservative ideological continuum to talk about the character of politics and policy. Despite the prevalence of ideological terminology in politics, however, it is clear that the terms have no precise meaning for many citizens and that specific understandings of ideology vary widely. Nevertheless, the general meaning of the terms liberal and conservative is largely consensual at any given time, particularly among politicians and elite commentators who have no trouble sorting out the "liberal" and "conservative" actors and policy options. The extent to which the electorate, on the other hand, uses ideology to structure political thinking has long been a matter of scientific interest. Clearly, there is a real difference between the intellectual frameworks of political elites and ordinary people, with the latter using ideological terms idiosyncratically and indifferently (Converse 1964,1975). During the past quarter-century, more and more Americans have started to use ideology to organize their political world. As a quick measure of ideological literacy consider the ability of citizens to (1) choose to identify themselves as liberal or conservative, (2) understand that political parties and candidates can be associated with one side or the other, and (3) correctly describe the Democratic Party or the Democratic candidate as more liberal than the Republican counterpart. This is a relatively easy task (Jacoby 1995) because it implies nothing about the citizens' conceptualization of liberalism and conservatism or their willingness to translate their own identification into a judgment about the parties or candidates. Nevertheless, it is a reasonable standard of a functional "ideological literacy" in that it reveals citizens' ability to get the terms straight. This is a major concern of ours because our view of predispositions is 7. In the 2000 ANUS the emotion measures were slightly modified (a change we approve of). In order to ensure that the data analyses are not contingent on using different measures, we restrict our analyses to the data from 1980 through 1096. flCe»F .6 .51 JTie Third Wav 133 6 ,. THE GROWTH OF IDEOLOGICAL LITERACY OVER TIME 1972 19SC Year 1968 that they are functional, that is to say, that they exist to provide a reliable guide to action. Hence, that people may label themselves "liberal" or "conservative" does not establish that they can and do competently use these terms of reference in making political choices. Figure 6.1 shows that ideological literacy has grown steadily. It plots the percentage of literates as identified by the ANES from 1972 through 1996—the portion of the public that chooses an ideological identification (including "moderate") and can correctly identify- the Democrats as more liberal than the Republicans.8 Only 35 percent of the public could do so in 1972, an election year in which Richard Nixon painted George McGovern as a liberal, But by the 1980s this portion increased to more than 40 percent, and by the end of the century, it reached 57 percent. By this measure, the public is catching up with the terms of elite political discussion. This is important because the ideologically literate public does use its ideological disposition as a guide for voting. During this period. 8. For this measure we use the standard seven-point scale prompts where the respondent is asked to choose a position from "Extremely Conservative" to "Moderate" to "Extremely Liberal" and then asked the positions of the two parties and their presidential candidates. Someone is ranked as literate if he or she chooses some position and gets the parties and presidential candidates straight. Below, when we use the full range of the ANES data (including the off-year election studies), we score as literate those who merely get the parties correct (in the off-year studies there are no presidential candidates). 134 MACKUINi MARCUS, NIVMAM, AND K E E L E ideological identification worked nearly as well as party identification as a predictor for how citizens will vote for candidates for the presidency and for the Senate (MacKuen, Erikson, and Stimson 1988)—a relation that appears to be relatively constant. The rise in literacy supports the conclusion that the electorate as a whole has voted increasingly along ideological lines. Given the importance of ideology as a political disposition, we want to understand how the systems of affecti%-e intelligence modulate ideology's operation. In the past forty years, the party system has evolved to produce ideologically consistent partisan alternatives across the country so that, much of the time, the Democratic candidate will consistently offer a more liberal alternative than the Republican. Understandingthesimilaritiesand differences between the two dispositions (party and ideology') and their reliance on affective intelligence will push our understanding further. affective intelligence and the modulation of political dispositions: ideolocy The ability of anxiety to modulate partisan dispositions has already been shown to be substantial and consistent with the predictions of the theory of affective intelligence (Marcus, Neuman, and MacKuen 2000), We examine the theory's effect on ideological dispositions as opposed to partisan dispositions. In the analyses that follow, we use the same analytic approach that we previously used to assess reliance on partisan identification as the habituated basis for political choices, but we add reliance on ideological convictions as another habituated predisposition to see whether it functions in much the same fashion.5 So in table 6.2 y. The candidate likes and dislikes are the volunteered comment* about the candidates, summed up for directional content. For the candidate policy proximity we use the seven point issue scales set out by the ANES staff in each election from 197a to 1996, relying on the investigators' judgment to get the right items. For each issue we calculate the simple distance from the individuals preferred policy choice and the individual's satyfCtM perception of the candidate's position oil the scale. Then we subtract the candidates' position from the respondent's placement and sum up straightforwardly across the different issues of the day. All variables are rescaled to a unit interval—bounded by the range of the variable's possible responses—to make the coefficients roughly comparable. We use linear regressions to make substantive interpretations relatively easy. Of course, the dichotomous nature of the dependent variable makes this a tradeoff between statistical probity and substantive feel. Anxiety is measured by the repeated ANES questions about whether the candidate had done anything to make the respondent feel angry or afraid. Although these items were not explicitly designed for the current theoretical test, they do well enough. TheTJiird Way 135 ta»l£ 6 - 2: A^ective intel' geice rrodu'atioi of disposition's impact on vote choice, i^j0_g6. idec!og:ca'!'/ iterate votes Partisanship ideokgt Complacent Anxious about Comp'acent Anxious about about candidate candidate about your your party's with same *»tfi some party's candidate candidate ideology ideolcg? Partisanship .62 .39 — — (.02) (.03) Ideology — — .80 .35 (.04) (.05) Pol»cy comparison .19 35 .21 .42 (.OS) (.04) (03) (05) Candidate qua ties .31 .46 .31 .47 (.02) (.03) (02) (03) Constant 18 .28 .10 33 (.01) (.02) (.02) (.03) N= 1995 1013 1610 808 Adjusted R: 80 .64 .79 .64 RMSE 22 .30 .22 .30 we include results for partisanship for comparative purposes, We might normally expect self-identified liberals to vote for the Democratic candidate and conservatives to vote for the Republican. And, given the ideological literacy of our target group, we can be sure that they understand the relative attractiveness of the two parties' candidates. So there is no surprise when the third column of table 6.2 shows that ideology dominates the choice of complacent voters—voters who feel no uneasiness about "their" candidate. On the other hand, when engaged by their emotional alert mechanisms, people do change their behavior. As was the case with partisanship, we see that the reliance on disposition diminishes: compare The scores are the averages of the two items normed to the unit interval. The astute reader will note that Keldman, Huddy, and Cassese (Chapter • 1 o w a. ■o B 0 ■5 01 w 0- CO 0 CD O O CM O O Ó - Lo.v Anxiety - - • Mean Anxiety - - High Anxiety 0.0 0.2 i 0.4 0.6 0.8 1.0 Issue Distance SWre; American National Election Surveys. 19S4. 19S8.1992. across three levels of anxiety, holding the other variables constant at their means. Figure 6.4 shows how the probabilit)- oť defection among Republican partisans as anxiety moves from its lowest levels, when Republican partisans are feeling complacent, to its highest, when they are most anxious. As figure 6.3 shows, when an issue comparison favors the Democrats there is little prospect of defection if anxiety is at a minimum, but as anxiety increases in the same circumstances, the probability of defection moves to well over 90 percent. Partisans are not nearly the intractably loyal group that conventional wisdom portrays them to be. Column 2 in table 6.3 shows that a focus on issues is not always the primary desiderata of rational, that is, anxious, voters. In 1980, when President Jimmy Carter was seeking reelection in the midst of the Iran hostage crisis, double-digit inflation, and unemployment, the key interaction term is not about the issue differences between Carter and his opponent. Ronald Reagan. Rather, the interaction of anxiety and candidate qualities is instead the key force behind defections. This suggests that Reagan attracted Democratic partisans, the so-called Reagan Democrats, not by 144 MACK VMM, MARCUS, NEC MAN, AND KEUt appealing to a more conservative array of issues but rather by suggesting he had more of the "right stuff" to be president. Anxiety was harnessed to Jimmy Carter's suitability for a second term, which led to a comparison of Carter's qualities to those of Reagan, and Democrats who found Carter wanting defected.17 Column 3 of table 6.3 shows the results for 1996, when President Ri]| Clinton was challenged by Republican Robert Dole. The results in this case are quite similar to those for Republican partisans. Anxiety initiates an explicit rational comparison of the parties' issue positions, with loyalty and defection the possible outcomes (and in 1980 and 1996 low levels of anxiety were sufficient, because of their interactive effects with candidate qualities and comparisons of the issues, respectively, to guarantee defection). So again we witness how the surveillance system of affect detaches partisans from efficient and automatic reliance on the dispositions system of affect (in this case, the expression of standard partisan and ideological positions). And when the surveillance system is active, what follows is a close reappraisal that has clear behavioral consequences in the voting booth. the macro consequences of anxiety The shifting of the American electorate from more liberal to more conservative predilections and back again is well known (Burnham 1970; Stimson 1991). The role of economics, in the form of business cycles, helps account for such sinuosity (Alesina, Londregan, and Rosenthal 1983). In addition, political movements, whether liberal or conservative, seem to have a natural life span, finding enthusiasm when young and exhaustion when older. If they are long enough, such cycles might reflect the passage of age cohorts whose political convictions are defined by the characteristics of their times (the Great Depression, World War II, Vietnam, or, most recently, 9/11). The theory of affective intelligence substantially revises the conventional wisdom about the periodicity of elections. That a downturn in the economy would cause anxiety and hence diminish support for an administration at election time is hardly a novel prediction, nor a prediction that could discriminate between this and any other theory. Administrations do more than simply manage the economy (with the limited tools available). Administrations also have policy goals that they seek to advance by way of creating laws. But as they do so, they change the envi- 17. As before with Republican partisans, anxiety had no direct impact on the odds of defection. Anxiety functioned indirectly, through its interaction with candidate qualities. The Third Way 145 jonrnent. New laws compel new behavior (either by applying sanctions to discourage current beha%iors or by offering inducements to engage in nev,. behaviors). The theory of affective intelligence advances a counterintuitive claim: that a regime's supporters should become increasingly anxious as the a(jministration becomes increasingly successful and that this dynamic should be more robust than the comparable increase in anxiety among the administration's detractors. We derive this prediction by noting that supporters of an administration are wedded to a world they already find familiar and congenial by virtue of the many predilections that offer reassuring guides to their everyday life. Having an administration in power that shares their commitments defends those predilections. Having an administration dynamically and effectively change the world will, we predict, make the regime's supporters increasingly nervous as the administration achieves its policy goals. In order to test the claim of affective intelligence theory, we need to develop a model that includes the features that would account lor the observed periodicity in American politics. Primary among the characteristics of governments is their management of public affairs—and it is surely in the political interests of presidents to be seen as competent rather than incompetent managers. In the contemporary United States, the most common standard of government competence (whether it makes sense or not) is management of the economy. Presidents who preside over economic booms are presumed to have done well, and those who encounter harsh times on their watch will find themselves having to defend their record. As an indicator of a sitting presidents economic fortunes, we use the University of Michigan's Index of Consumer Sentiment, a set of ongoing surveys that assess the public's views of the contemporary economy." Unremarkably, the theory of affective intelligence predicts that anxiety should rise and that the surveillance system should become active when the economy plummets. And, unremarkably, this prediction is sustained. When we model individual-level anxiety as a function of national economic conditions, we see a strong negative relation: the better the 18. The Index of Consumer Sentiment combines measures tapping retrospective and prospective assessments of family finances and "business conditions." These stem from national surveys, now conducted on a monthly basis, taken from 1952 through the present. We use the year's average sentiment (that is, for January to December of the election year) *» a proxy for contemporary economic conditions. In this sense we use information about both the recent past and the immediate future that might go into popular perceptions of the Incumbent's eennomic management. We have rescaled the measure to the unit interval to make it roughly comparable to our other measures. 146 MACKUEN. MARCUS. XEUMA.V, AND KEF. 1. E economy, the lower the anxiety, and the worse the economy, the greater the anxiety. The numbers are large and statistically solid. Performance matters for affective intelligence. Another feature is policy. We know that many citizens are only dimly aware of national policymaking and would be surprised to see that they react to legislative victories and defeats. To be sure, few citizens can enunciate anything specific about what the Congress and the president have done during the past year. And yet, from a normative point of view, we should at least hope that they are able to react to surges of liberal or conservative changes in national policy. It is surely the case that citizens do react in terms of their preferences for policy change. Evidence from the past half-century shows that when the national government passes liberal legislation, the public's demand for liberal policies drops discernibly, and when the government passes conservative policies the public begins to demand more activism (Erikson, MacKuen, and Stimson 2002). It this periodicity that we seek to illuminate. The question is whether people's affective intelligence operates to capture changes in national policymaking. By all accounts, we should expect that the passage of liberal legislation will lead to an increase in anxiety about liberal presidents and that the passage of a conservative program will lead to worries about conservative presidents. Success by presidents in pushing their policy agendas will typically exhaust their symbolic reservoir because that very success will impel changes that, even if welcome, generate a counterreaction based on the public feeling that things have gone far enough. We test this proposition by modeling incumbent anxiety as a function of the number of major (liberal-conservative) laws that were passed by Congress and signed into law by the president during the year preceding the election campaign.'-9 We expect a negative relation, and this expectation, too, is sustained. Putting the pieces together in the first column of table 6.4, we observe an elementary model of incumbent anxiety written as a function of our two macro-level political conditions: economic performance and partisan policymaking. Taken together, each is statistically and powerfully 19. We use a version of David Mayhew's (2991) major laws. These include pieces of legislation thought at the time to be significant (and described as so in the year-end reviews of the Sew York Times and the Washington Post). The sei ics has been updated by Jay Cieene using the same methodology. We have coded the particular laws as liberal or conservative, double-counting landmark legislation such as the 1996 welfare bill (Erikson. MacKuen, and Stimson 2002). . „Kumbent- and cha'lenger-elcited an-oety as a function of mac „, t spos tons and evaluations, 1980-96. ideology .Jb&** ________—--- The Third Way 147 to-level politics and iterate voters A1/ voten Incumbents support** CMSengeťt supporters ^ofconsurrcr j^tíment ^(„rt passed frrfory"') 5jtisansh'P Policy comparison Constant Mtusted R: USE J/Kurttbfit- Cna'lenger- Incumbers elicited thc'ttd elicited anxiety anxiety fly -.22 -.OS"' -.32 (OS) (OS) (07) .33 .03' 38 (.04) (.04) (OS) -.31 .22 -.24 (02) (.02) (OS) -.21 .27 -.12 (.03) (.03) (04) 1 -29 .20 -.12 (03) (.03) (.04) .94 .13 .89 (-04) (.04) (07) 4548 4S46 2235 .27 .18 .09 34 .35 32 Challenger-elicited anxiety Incumbent- Challenger elicited elicited anxiety anxiety -.16" (08) .02-(06) .14 (.06) .36 (OS) .26 (.05) .23 (OS) 2234 .06 ,39 -.12'' (.08) .24 (06) -.12 (.06) -.31 (04) -.46 (05) .83 (07) -.02'' (.06) .01"' (.05) .10 (-05) .14 (04) 15 (04) .IS (.05) 2052 2050 .13 .36 .02 .29 ' Not statistically discern ble f'om zero related to people's feeling complacent or uneasy about the incumbent. We can use these results to model the substantive consequences of the public's assessment of the economy and of an administration's translating its agenda into law. The difference between the best and worst years in consumer sentiment in our sample is enough to move anxiety downward by a little more than a quarter of its effective range. More striking, the difference in the most and least successful presidential policy years yields movement of about one-half of the range. Good times lead to a reduction •n anxiety; success in a legislative agenda leads to an increase in anxiety. The picture is different for challengers. The second column of the table shows an identical model, positing feelings about the challenger to °e a function of the political environment. It is fairly clear they are not. Tile coefficients for both of the macro political measures are decidedly small, and the peiformance measure is statistically insignificant. The ta8 M A C K U E N . MARCUS, N HUMAN, AND KEE1E surveillance mechanism's emotional cues about incumbents do ref] genuine shifts in the political world, but the cues given about the ch»j. lengers do not. And, of course, this is exactly what we would expect fr0q) affective intelligence: the surveillance emotions should reflect some tea. sonable approximation of reality and not represent a mere extrapolation from standing dispositions.20 Of greater interest is that, by dividing voters into supporters and op. ponents of the candidates, we can test whether the theory ol affective intelligence works as a signaling device for supporters to alert them \n the potent consequences of the success of the administration. It is conventional to presume that these factors work not for supporters but only for those in opposition, inasmuch as supporters should encourage the success of an administration in enshrining agreed-on policies in law. In this case, however, dividing voters into supporters of the incumbent administration and those opposed to it shows that anxiety is increased more among the former than among the latter. The signal that something is changing works for opponents and for supporters. This is absolutely critical because it is the supporters, after all, who will use the affective surveillance system to interrupt their routine reliance on partisan and ideological dispositions. And, as we have seen, when these dispositions are interrupted, defections may follow. CONCLUSION It is quite common to describe people as manifesting uniform characteristics: people are liberal or conservative, engaged partisans or inattentive independents. The established theories that we have discussed, the rational choice and the psychological, are exemplars of this approach." A core assertion of the theory of affective intelligence is the fundamental imperative of knowing at the earliest moment the character of the political geography in which one finds oneself. Preconscious appraisals identify our location as either in the realm of the familiar, where reliance on existing habits is efficient and productive, or the realm of the unexpected 20. For the macro political causal forces, it is anxiety about only the incumbent that carries a political signal. 21. Hence, when repeated studies show that people depart from rationality, rational choice theorists retreat to some lesser version, such as "bounded rationality," rather dun consider the alternative that wc advance: sometimes and in some circumstances people are quite rational whereas in perhaps most other circumstances they are not, and for good reason (Marcus aooz). The Third Way 149 nfainil>ar' where a different mode of decision making if required. un''°rm capabilities 'nolt^dc both the capacity to rely on heuristics 'rtJdtbe capacity to engage in explicit deliberation. The theory of affective i t ,]|igence provides an account of why we have both capabilities and "hen each is likely to be manifested. .. theory offers important insight into the dynamic character of litical dispositions. It holds that people use heuristics such as partisan and ideologic.il cum ictions when they are most likely to achieve conservation of limited cognitive resources. Reliance on heuristics enables one to secure rational, that is. near optimal results." Thus we expect that partisan loyalists and the ideologically driven are likely to most often rely on their dispositions to formulate their political judgments. But they are also likely to set aside those firmly held convictions and apply those limited cognitive resources in the service of deliberate judgment when appropriate to do so. As a result, the micro picture of intransigence in the face of contentious information is not a universal truth. Furthermore, the macro expectation that elections turn on how the least informed of likely voters decide how to vote is also a portrait that misses a crucial dynamic, namely, that the defection of partisans has far greater weight in determining who wins and who loses in American elections. The record of presidential elections in the past quarter-century amply shows that defection is the principal explanation for electoral outcomes and that defection arises from two sequential steps. First, the campaign that can induce anxiety in its opponent's partisan and ideological base creates an opening that can be used to encourage defection. But in order for defection to occur, that campaign must have an issue or a candidate that will attract voters who make a thoughtful comparison because of their anxiety. An older tradition, the cognitive appraisal theory of emotion, holds that feelings are informed by self-conscious considerations (Ortony, Clore, and Collins 1988; Roseman, Antoniou, and Jose 1996). As discussed by Crigler, Just and Belt (chapter 10 in this volume), the way in which people understand their feelings may have impacts above and beyond those that are impelled by the multiple preconscious systems that are the focus 22. By optimal we do not mean that some newly implemented analysis would not yield **tae marginal improvement over that resulting from reliance on a pertinent heuristic. Rather, we mean that reliance on the heuristic provides a highly certain and, if conditions are as they have been, favorable result. Furthermore, embedded w-.th the heuristic is the earned capacity for implementing said decision (something not generally available to support a newly derived solution). 15<3 MACKUEN. MARCUS, NEUMAN, AND KtElE of the theory of affective intelligence/1 But to properly explore the rolt of preconscious affective appraisals and post-awareness affective intro. spection requires a research design that can identify and specify each di$-tinct array of influences. Perhaps it is time to abandon the largely spatiaj metaphor that shapes our understanding of reason and passion, cognition and affect, a metaphor that construes these two states as concurrent bm distinct and often antagonistic (their "separateness" often articulated in such familiar tropes as heart and mind). The spatial metaphor misleads by distracting us from the temporal sequencing tasks that are the core responsibilities of preconscious appraisal systems, responsibilities that control and are followed by conscious awareness and thereafter by introspection and refection (Libet 2004; Marcus 2002). By abandoning this metaphor we would gain a fuller and more incisive understanding of the dynamic capabilities that people have to adapt to the different demands and challenges that democratic politics presents. Cassino and Lodge (Chapter 5 this volume) illustrate another point of contention. It has long been held that emotions provide an assessment that resolves the critical evaluative issue of approach to situations, objects, or people that we like or desire as against avoidance of situations, objects, or people that we dislike and avoid (Tooby and Cosmides 1990). This notion of emotion as valence has the advantage of parsimony. It makes the further presumption that emotion "tags" items in declarative memory (Fiske 1981; Fiske and Pavelchak 1986). We hold that this view, and the research it has spawned, miss essential and powerful roles for emotion. The first is that emotion is rarely experienced as a simple valence, like-dislike (Marcus 2000). The second is that, though the role of emotion is associative memory is ignored by this conception, what we do, as contrasted with what we think, is largely driven by associative memory, not declarative (or semantic) memory (especially when people choose automatic reliance on extant heuristics). Third, as a consequence, by ignoring the role of anxiety, the dynamic shifting from one decision strategy to the other (and back) is largely missed by the theory of emotion as valence?"1 Fourth, Brader and Valentino (chapter 8 in 23. And in this as well, neuroscicnce can be useful in applying its methodologic.il apparatus to understanding the discrete emotions that result from conscious consideration (Takahashi et al. 2004). 24. Furthermore, though Cassino and Lodge couch their findings as contradicting the normative implications of the theory of affective intelligence, their findings are quite consistent with the theory of affective intelligence. One of the major tasks of affective appraisals is to give the earliest assessment of the immediate context, classifying the environment as familiar and rewarding, familiar and punishing, or unfamiliar and uncertain. That an The Third Way 151 lume) find that anger is driven by prejudice, a relationship that is Ml* consistent with the theory of affective intelligence (Marcus 2002). - « we have labeled this class of emotion aversion, is driven by agger, a» / frontation with familiar punishing stimuli and by the commitment f psychic and physical resources to learned defensive and aggressive behavioral routines. Finally, though not likely to be the sole explanation, the theory of affective intelligence provides another counterintuitive insight. The cyclical character of American elections may have its roots not only in the way the "moderate middle" moves but also in the way that partisan supporters of a successful administration respond to its success. That partisan opponents become more determined to oppose successful administrations is hardly novel (on that score the theory of affective intelligence has no dispute with conventional wisdom). But that partisan supporters become more anxious than the administration's opponents, with attendant political consequences, is hardly conventional wisdom. As such the theory of affective intelligence provides a framework that has macro applications (see chapters 12 and 13 in this volume). Although the theory provides a reason to expect robust and comprehensive roles for preconscious affective appraisal systems, other paths are also worth exploring. By incorporating elements of the rational choice and the psychological theories and by and specifying the conditions when each is likely to be applicable, we resolve the long-lived conflict between an attractive normative macro theory—rational choice—and a seemingly more accurate but normatively disappointing micro theory: psychological (or normal vote) theory. The theory of affective intelligence offers empirical promise for new lines of inquiry that can reveal hitherto unknown dynamics of political behavior and judgment. It also provides a micro-account of a political psychology that sustains a normative portrait of democracy that is more encouraging than has previously been thought plausible. initial emotional appraisal, having identified which of these contexts we are in, would thereafter Impact subsequent information is precisely what the theory of affective intelligence says ought to happen.