New Perspectives and Evidence on Political Communication and Campaign Effects

S. Iyengar and A. Simon

Annual Review of Psychology

These are heady times for students of political communication. To an extent not previously seen, the use -- even manipulation -- of the mass media to promote political objectives is now not only standard practice, but in fact essential to survival. Recent months have seen political practitioners engage in high-stakes media games, from the apparently deliberate leaking of secret grand jury testimony to reporters, the national telecast of presidential deposition testimony and the forced resignation of Speaker Gingrich in the immediate aftermath of a failed advertising campaign.

It is true that public relations and media advocacy have penetrated virtually all governmental arenas, even extending into the traditionally invisible judicial process. However, it is in the context of political campaigns that these strategies are most extensive and likely to have the greatest real-world impact. Electoral reforms and the influence of the broadcast media have gradually rendered grass roots organizations and party infrastructure less relevant, thereby transforming political campaigns from labor-intensive clashes between disciplined party organizations to capital-intensive, choreographed media spectacles. The ever-increasing level of investment made by the candidates is truly stunning -- over one billion dollars for the 1998 cycle alone.

One question that defies analysis is why the participants in the political marketplace continue to invest at these levels when decades of academic research into the effects of media-based political campaigns purports to demonstrate that exposure to campaigns mainly reinforces voters= preexisting partisan loyalties. Political scientists routinely attribute electoral outcomes to structural variables -- most notably, the state of the national economy and the level of the incumbent president=s popularity -- giving short shrift to the specifics of day-to-day campaign events, which are viewed as having but Aminimal consequences@ (see Abramowitz, 1996; Gelman and King, 1993; Campbell and Mann, 1992).

Our objective in this chapter is to demonstrate that the conventional academic wisdom is mistaken. Campaigns do matter and can be pivotal. Our approach will be to identify the existing roadblocks -- both intellectual and methodological -- to a full-fledged understanding of the effects of campaigns; then to describe several cases in which scholars have documented that campaigns do matter; and, finally, to propose theoretical and methodological vantage points for reviving the study of modern, media-based campaigns.

The Conceptual and Methodological Roadblocks

Perhaps the most fundamental obstacle to understanding the real-world role of political campaigns is definitional; traditional research has looked mainly at persuasion (i.e., the effect of a campaign on voter preference). Within this definition, the law of minimal consequences has some validity because the evidence does suggest that exposure to campaigns merely activates voters' prevailing partisan sentiments (for a recent review of the evidence, see Iyengar and Petrocik, 1998). Limiting the search to persuasion effects necessarily ignores a variety of other highly relevant campaign effects, the most significant of which is turnout. Changes in the size and composition of the electorate can, of course, alter the distribution of candidate support. The single-minded quest for persuasion effects has also ignored altogether the transmission of information, the setting of campaign agendas, and alteration of the criteria by which candidates are judged. Moreover, even if one were to accept persuasion as the benchmark for campaign effects, identifiable traces of persuasion are bound to be minimal since most campaigns feature offsetting messages. Observable effects should thus be limited to campaign in which one candidate has a significant resource or skills advantage. This condition rarely occurs in presidential campaigns.

Turning to the ever-contentious issue of methodology, researchers have been (and still are) limited by their dependence on survey methods. The founding fathers of the field (including Lazarsfeld, Berelson, and their successors at the University of Michigan) were all trained in survey research and accepted the logic of treating respondents' self-reported exposure to campaign communication as a reliable surrogate for actual exposure. In this tradition, the standard test of a campaign effect is the scale of differences in some criterion variable (typically vote choice) between respondents who self report high or low levels of exposure to the campaign.

The assumption that self-reported exposure converges with actual exposure is problematic on several grounds. Most importantly, people have notoriously weak memories for past events, especially when the "event" in question concerns an encounter with a particular political campaign (see, for instance, Bradburn, Rips and Shevell, 1987; Pierce and Lovrich, 1982). In the experiments conducted by Ansolabehere and Iyengar concerning the effects of campaign advertising, for example, over fifty percent of the participants who were exposed to an advertisement were unable, some thirty minutes later, to recall having seen the advertisement (Ansolabehere and Iyengar, 1999). Errors in the opposite direction are also likely as many respondents may feel compelled to respond affirmatively to questions monitoring their media exposure on the grounds that such responses speak well of their civic virtue. In sum, the considerable measurement error in self reports will necessarily attenuate estimates of the effects of political campaigns (see Bartels, 1993, 1997).

Adding even further damage to survey studies, self-reported exposure to campaign messages is typically endogenous to political attitudes, including candidate preference. That is, those who choose to tune in to the campaign differ systematically (in ways that matter to their vote choice) from those who do not. To take the case of campaign advertising, people who can and cannot recall an advertisement are likely to differ in innumerable ways. In addition to having better memories, the former are likely to be more interested in politics, more attached to the candidates, more informed about the issues, and more likely to vote. The presence of feedback between vote intention and recall of advertising seriously undermines claims about the impact of exposure to advertising on participation. Respondents who recalled having seen a Anegative@ campaign advertisement in the 1992 National Election Study survey, for instance, were more likely to intend to vote (Wattenberg and Brians, 1999) than those who did not, leading the researchers to suggest that negative campaigning stimulated turnout. But was it exposure to negative advertising that prompted turnout, or was the greater interest in campaigns among likely voters responsible for their higher level of recall? When recall of advertising in the same survey was treated as endogenous to vote intention and the effects re-estimated using appropriate two-stage methods, the sign of the coefficient for recall was reversed: those who recalled negative advertisements were less likely to intend to vote (see Ansolabehere et al., 1999). Unfortunately, most survey-based analyses fail to disentangle the reciprocal effects of self-reported exposure to the campaign and partisan attitudes/behaviors.

In recent years, survey researchers have turned to longitudinal designs for monitoring the effects of campaigns. This is understandable in light of the fact that modern presidential campaigns are year-long affairs (if not longer). The Iowa caucuses occur in January, the primary season ends in June, the nominating conventions are held in August, and Acampaigning@ in serious begins after Labor Day. In this environment, most voters are likely to arrive at their presidential candidate choice well before September 1 (see Iyengar and Petrocik, 1998; Bartels, 1997a). Despite a handful of studies which track campaigns over their entire life span, most of the longitudinal evidence is derived from surveys administered after September 1. Not surprisingly, this evidence has hardly made a dent in the conventional wisdom. Finkel's thorough analysis of the 1980 NES panel survey typifies the literature: 80% of the variance in vote choice could be accounted for by pre-campaign attitudes and beliefs (Finkel, 1993). When attitudes did change during the campaign, they invariably did so in a manner consistent with voters' partisanship.

Aggregate-level longitudinal studies, by tracking changes in national sentiment, provide estimates of campaign-based "interventions" in public opinion. Given the importance of contextual factors such as the state of the economy, these studies are especially valuable in their ability to compare the effects of short-term campaign events on voter preference with the effects of the campaign context. Holbrook's studies of trends in candidate support during recent presidential campaigns (1994, 1995) found that voters' initial attitudes are paramount. Although events (such as conventions and debates) contributed modestly to net changes in the level of each candidate's support, these effects were dwarfed by the effects of the economy and the incumbent president's popularity (Holbrook, 1994).

While aggregate time series and panels can yield estimates for the effects of specific events, like the cross-sectional survey, neither can address the daily variation in campaign communication. In the case of panels, this would require that each respondent be matched to a particular set of media outlets for which the messages be scored for political content and slant -- a near impossibility. In the case of the few time series studies that do incorporate measures of media content, because they necessarily aggregate to the national level, all voters in any given campaign are assumed to have received the identical message; the only variation in campaign communication occurs over time (elections). The problem with this aggregate approach, of course, is that it masks the considerable cross-sectional variation in the volume and tone of advertising. Bill Clinton's advertisements were nowhere to be seen in California during the 1992 campaign simply because George Bush had conceded the state. Viewers in the "battleground" states would have been exposed to much higher doses of advertising than those living elsewhere (see Shaw, 1997; Goldstein, 1998). Even if one assumes that the candidates choose to campaign everywhere, they are hardly likely to deliver uniform messages to farmers in South Dakota and high-technology workers in California. All this contemporaneous diversity in the content and tone of campaigning is ignored in the aggregated, time series method (see, for instance, Finkel and Geer, 1998 in which it is asserted that all voters encounter the same level and type of campaign advertising).

The survey design offering the greatest potential for the study of campaigns is a hybrid of the panel and time series. This is the repeated cross section in which interviews are carried out sufficiently frequently (on a daily basis) to yield reliable weekly or biweekly national samples that span the entire life of the campaign. This design (dubbed the "rolling cross section") was implemented by the NES group for the 1984 presidential campaign and featured a January-November time frame (see Brady and Johnston, 1988). More recently, a similar design was implemented to study the Canadian election of 1988 (Johnston et al., 1992). Because respondents are interviewed more-or-less continuously from the earliest stages of the campaign, it is relatively easy to aggregate groups of respondents into precise temporal groupings corresponding to "pretest" and "posttest." Not surprisingly, the published reports based on these data sets (some of which are described in the next section) reveal several instances of campaigns that "work."

In summary, the study of campaigns has been impeded by a "drunkard's search" syndrome of sorts in which researchers congregate at the persuasion "lamppost," and by methodological "correctness." By excluding effects other than persuasion and by remaining committed to the flawed logic of survey design, the state of campaign effects research remains mired in the uninteresting rut of minimal consequences.

Bypassing the Roadblocks

Multiple Effects

The diminishing marginal returns from research into the persuasive effects of presidential and other campaigns naturally encouraged scholars to explore other facets of voter engagement, none of which required the abandonment of long-held partisan loyalties. Based on this research, we can include voter "learning" -- the acquisition of information about the candidates and issues -- and "agenda control" -- the use of campaign rhetoric to set the public's political agenda -- in the catalogue of robust campaign effects.

Learning

Since voters tend to be risk averse and are loathe to support unknown candidates (see Alvarez and Franklin, 1994; Bartels, 1986), information is a precious campaign commodity. To be known is to be liked, and campaigns generate large quantities of "knowns" on subjects as diverse as family background, military service, and the details of policy proposals. As campaigns have evolved into media affairs, however, the information-function of campaigns has fallen into disrepute. The candidates' media presentations, especially their reliance on the thirty second televised advertisement, appear all too superficial and it is suspected that superficial campaigns breed superficial voters.

The fears that campaigns are non-substantive are based on several observed regularities. First, the news media tend to treat the campaign first and foremost as a horse race, with the candidates the entries, the consultants the jockeys, and the electorate the betters (for evidence of the predominance of horse-race coverage, see Robinson and Sheehan, 1980; Lichter, 1988; Task Force on Campaign Reform, 1998). Given this framework, daily events are accorded coverage in proportion to their bearing on the "odds." The Dukakis campaign's detailed position paper on education policy received no coverage in 1988, but the Massachusetts Governor's brief answer (during a televised debate) to a question about the hypothetical rape of his wife attracted headlines and endless commentary about the candidate's diminishing chances to win.

In addition to the media's preoccupation with the contest, market-based journalism is bound to keep voters interested in news about the campaign (see Kalb, 1998). What better enticement than sex, sleaze and scandal? In recent months, Americans have become all too familiar with the details of President Clinton=s relationship with Monica Lewinsky, as well as the moral quotient of some of his Republican antagonists (such as Congressman Hyde). In what might be called Gary Hart's Law, news about character drives out news about public policy (see Sabato, 1991; Reeves, 1998).

The final piece of evidence cited by critics of modern campaigns is that even when the candidates and the press do turn their attention to the issues or questions of performance and ideology, the rhetoric takes the form of truncated soundbites rather than well-developed and detailed arguments (Hallin, 1994). Moreover, the press has increasingly taken to interpretive forms of reporting in which policy proposals are typically seen as cynical electoral ploys rather than sincere political commitments (Patterson, 1994).

Given the rampant cynicism about the information value of media campaigns, it comes as a considerable surprise that exposure to these campaigns -- either in the form of paid advertising or news coverage -- boosts citizens' political information allowing them (if they so choose) to cast votes on the basis of their relative proximity to the candidates on the issues. The evidence comes in several forms. In a series of experiments carried out by Ansolabehere and Iyengar (see Ansolabehere and Iyengar, 1996), participants (ordinary citizens of southern California) exposed to a single thirty second advertisement were more able (by a significant margin) than those not exposed to the advertisement to identify the sponsoring candidate's position on the "target" issue. Exposure to issue-oriented advertising also increased the likelihood that participants would cite the candidates' positions on issues as a basis for supporting the sponsoring candidate. Similar results were obtained by Brians and Wattenberg (1996) and Patterson and McClure (1976) using national and regional surveys respectively. Interestingly, both these studies found that exposure to campaign advertising (measured in terms of recall of campaign ads) contributed more to issue information than exposure to newspaper or television coverage of the campaign.

The question of source factors in voter learning is still a matter of controversy. Contrary to the studies cited above, which suggest that advertisements provide greater information value than news coverage, a series of statewide and local surveys by Chaffee and his colleagues has uncovered the opposite pattern of evidence favoring news coverage over advertising as a source of information (Chaffee, Zhao, and Leshner, 1994; Zhao and Chaffee, 1996). While it is difficult to compare results from studies utilizing different designs and indicators, one possible explanation of the divergence is that presidential advertisements (found to be informative by Ansolabehere and Iyengar, Brians and Wattenberg, and Patterson and McClure) are more likely to feature issue appeals, while advertising in statewide or congressional races is more evenly divided between "issue" and "image" appeals (see West, 1994; Kern, 1989).

The conclusion that campaigns provide voters with substantive information is buttressed by studies that rely on alternative indicators of information. Franklin's study of U.S. Senate campaigns found that voters were more precise in their perceptions of their incumbent senator's position on a liberal-conservative scale when exposed to a senatorial campaign featuring the incumbent (Franklin, 1993). Using a somewhat looser definition of uncertainty, Bartels (1988) also found a significant pattern of uncertainty reduction or information gain during the presidential primaries of 1984. As expected, the rate of gain was accelerated among respondents most attuned to the media. (Adam: describe Bartels in AJPS; Alvarez book)

While the conclusion that voters gain substantive information from campaigns may seem counterintuitive given what we know about the behavior of reporters, it should come as no surprise that information about the candidates' personal traits and electoral chances is there for the taking. Popkin's work (1992, 1996) has shown that voters, like good psychologists, make inferences about the candidates' personalities based on what they see and read about the campaign. The sight of President Ford attempting to eat an unshucked tamale sent Hispanic voters a clear signal about Ford's responsiveness to their concerns; the Republicans= unrelenting focus on impeachment in 1998 was understood by voters as a sign that Democrats were more concerned about education and social security. Popkin's work is especially revealing because it shows that voters' information about the candidates is highly sensitive to the ebb and flow of news. During the 1992 presidential campaign, Bill Clinton was saddled with the "Slick Willie" image in the aftermath of news coverage of his marital difficulties and avoidance of military service. Following his triumphant acceptance speech at the convention and the Clinton campaign's masterful use of "alternative" media outlets, voters began to learn about Clinton's economic plan and his track record on welfare reform and other issues. As information about Clinton's policy expertise came to the forefront, impressions of Clinton's character began to be couched more in terms of "the man from Hope" than "draft-dodger" (Popkin, 1996).

Finally, there is ample evidence that the barrage of horse- race stories raises the public's consciousness of the candidates' prospects (Mutz, 1997). In addition, beliefs about the candidates' chances influence preferences, thus setting in motion campaign "bandwagons," especially during the initial round of primaries (for evidence from the 1984 primaries see Bartels, 1989; experimental studies demonstrating that favorable poll results lead to increased voter support can be found in Brady, 1984 and Ansolabehere and Iyengar, 1995).

In sum, campaigns are information-rich events. Contrary to the prevailing wisdom, the information they yield is multifaceted encompassing the candidates' chances of winning, their personal traits and mannerisms, and, most importantly, their policy and ideological bearings. Media campaigns may appear superficial, but they do educate citizens.

Agenda-Control

If there is one thing political scientists and communications specialists all agree upon, it is that ordinary citizens remain at arms length from the world of public affairs. As casual observers of the political scene, individuals do not monitor the entire political universe; instead, they attend selectively to a few issues that appear important at the moment. Of course, the appearance of importance is very much a matter of what editors and journalists choose to cover or ignore; the more prominent some issue in the news media, the higher the level of importance people accord that issue (see Iyengar and Kinder, 1987; Rogers, Hart and Dearing, 1996; McCombs and Estrada, 1996). Since candidates are the principal sources of news during campaigns, they are in an advantageous position to simultaneously influence the media and public agendas (see Semetko et al., 1991; Dalton et al., 1995)

The core implication of agenda-setting for the study of campaign effects is that issues deemed significant by the electorate become the principal yardsticks for evaluating the candidates. This pattern (what the social psychologist would call "priming") of weighting issues in accordance with their perceived salience has been documented in a series of experimental and non-experimental studies (for a recent review of priming research, see Krosnick and Miller, 1996). Kinder and Iyengar (1987), for instance, found that the news media's sudden preoccupation with the Iranian hostage issue in the closing days of the 1980 presidential campaign caused voters to think about the candidates' ability to control terrorism when choosing between Carter and Reagan. Naturally, this logic proved disadvantageous to President Carter.

Given the limited number of policy domains that voters take into account, and the attitudinal consequences of priming (which are tantamount to indirect persuasion in that altering the criteria by which one chooses can alter the choice), candidates are motivated to introduce and pursue issues on which they enjoy a comparative advantage. The candidate closer to the median voter on tax reform might be some distance away on the issue of free trade. Accordingly, a great deal of campaign rhetoric and strategy (about which we will have more to say in the next section) is designed to capitalize on this "disequilibrium of tastes" (Riker, 1980; Iyengar, 1993). The fact that the news coverage of the economy drowned out news about the Gulf War at the time of the 1992 election cost President Bush dearly; on the economic dimension voters preferred Clinton by a wide margin. Had the media played up military or security issues, of course, the tables could have been turned given Bush's considerable edge on matters of national defense (see Iyengar and Simon, 1993; Krosnick and Brannon, 1993).

The most compelling evidence of priming effects in the course of campaigns comes from the Canadian election study of 1988. The authors (Johnston et al., 1992) show how the free trade agreement between Canada and the U.S., as a result of the candidates' and parties' rhetorical posturing, came to the forefront of the public issue agenda. The candidates' efforts to position themselves on this issue had clear consequences; as the campaign progressed, voters' preferences on the issue increasingly came to influence their vote choice. Particularly striking in this case is the fact that a rival issue of equal, if not greater relevance to Canadian politics -- the Meech Lake Accords concerning Anglo-French relations -- remained dormant. As the authors of this model study conclude:

Rhetoric, then, does play an important role in campaigns, but not just by persuading people. Rhetoric also plays a role -- possibly its biggest role -- by directing voters towards a specific agenda and considerations surrounding that agenda. (Johnston et al., p. 249)

As the examples of learning and agenda-control suggest, campaigns can influence voters in more ways than one. Confronted with partisan messages, most voters are loathe to roll over and declare themselves converted; instead, they resist and rebut messages at odds with their prior preference. The acquisition of information and agenda-control are far more subtle and modest forms of influence, often occurring "automatically" without the voters' awareness. Their net impact on the "bottom line," however, can be just as significant electorally.

New Perspectives on Campaign Effects

The earliest and still predominant way of thinking about the effects of campaigns has been to assume that candidates persuade voters simply by "injecting" them with appropriately designed messages. Assuming that the message is perfected (by carefully crafting the content and form of presentation), adequate exposure is all that candidates need; the more people they reach, the more likely they are to win. The all-absorbing task of the campaign consultant is to predict which variety of messages will be most advantageous and to design the campaign accordingly. In effect, campaigns are as good as the resources and talents behind them.

The hypodermic model presumes that an effective campaign will be effective regardless of the office at stake, the political leanings of the electorate, the nature of the opposing candidate, or the circumstances of the moment. In contrast, more promising theorizing about campaigns has emphasized the importance of the "environmental surrounding" and strategic logic of campaigns.

The Resonance Model

While accepting the premise that voters can be persuaded, the distinct premise of the "resonance" model is that campaign messages -- whatever form they take -- work their influence in concert with voters' prevailing predispositions and sentiments. In contrast to the hypodermic model, which assumes that effects are due entirely to particular characteristics of each campaign, the resonance model anticipates that effects are contingent upon the degree of fit between campaign messages and prevailing attitudes. New information intermingles with the old and, depending on the chemistry, voters' choices will or will not be affected.

The most "charged" ingredient of the electorate's prior predispositions is, of course, party identification. Acquired during early childhood, this psychological anchor is known to withstand the vicissitudes of events and the passage of time (Jennings and Markus, 1984; Niemi and Jennings, 1991). While much has been made of recent increases in the percentage of Americans who reject partisan labels, a good many of these "independents" are, in fact, closet partisans (Keith et al., 1994). Party identification remains the salient feature of the American electoral landscape.

The importance of partisanship suggests an obvious prediction about the impact of campaigns. In "high volume" races such as presidential contests where virtually everyone is likely to encounter snippets of the campaign and where the contestants attain approximately the same decibel level, the principal impact of the campaign will be to push partisans into their respective corners. This prediction has been substantiated in nearly every systematic study of presidential elections since the 1940s: campaigns reinforce voters' partisanship.

From the perspective of the resonance framework, the reinforcing effects of campaigns can be attributed to the interaction between the content of campaign messages (their slant) and voters' prior preferences. As part of one's attitudinal "endowment," party identification is unlikely to be relinquished because of a particularly attractive advertisement or nasty news report. Messages that are counter-attitudinal will be actively resisted, while those that are consonant are accepted (for the classic discussion of acceptance factors in the persuasion process, see McGuire, 1968; John Zaller has extended the McGuire model to political campaigns in several recent publications; see Zaller, 1995, 1992). It is this interaction between message content and prior attitudes that governs the reinforcement or "polarization" effect. Since it may be expected that the most intense partisans are least in need of reinforcement, it is the less-intense Democrats and Republicans who will move the most during campaigns. In Ansolabehere and Iyengar's experimental studies, partisan reinforcement was significantly higher among partisans with lower levels of political interest (Ansolabehere and Iyengar, 1996). Similarly, Iyengar and Petrocik (1998) found that as campaigns progress, weak partisans and younger voters become especially likely to express candidate preferences that were in accord with their partisanship.

Presidential campaigns mask a further interaction between campaign messages and voter predispositions. As we noted earlier, the level of exposure to the campaign is bound to fluctuate with individual-differences in political engagement and involvement. Some people are more likely than others to tune in. These systematic differences in exposure produce, in Zaller's (1995) terminology, "reception gaps" whereby some voters are likely to encounter a message from one candidate, but not the other. These voters tend to be drawn disproportionately from the middle level of the political involvement strata; as a result, the relationship between involvement and support for the louder candidate is curvilinear (for several tests of this model involving House elections where incumbents are presumptively the louder candidates, see Zaller 1992). In short, when studying campaigns of differential loudness, the interactions that define campaign effects become more complex; voters' preferences become a function of the volume/intensity of the contestants' appeals, voters' partisan preferences, and their level of exposure to the campaign.

The relevance of voters' partisanship extends well beyond the mere fact that Democrats will be more responsive to the Democratic candidate and vice-versa. Not only do most voters acquire a partisan identity, they also acquire beliefs about the groups served by the political parties and, by inference, the issues or problems on which they will deliver (see Petrocik, 1995). For example, the public generally considers Democrats more able than Republicans to deal with the problem of education. Conversely, Republicans are seen as better than Democrats on taxation.

These stereotypes about the differential policy responsiveness of the parties influence campaign strategy. Campaigns that take advantage of (or resonate with) voters' expectations are considered most likely to be effective; a Democrat should be better off using appeals that emphasize his intent to strengthen public education, while a Republican should promote his support for lower taxes.

Ansolabehere and Iyengar (1996) have tested this hypothesis experimentally by examining differences in the persuasiveness of Democratic and Republican campaign advertisements on the issues of crime and unemployment. Using the identical message that was broadcast on behalf of either the Democratic or Republican candidate for U.S. Senate in California (1992), they found that the gains from advertising on unemployment tended to be greater for the Democrats than the Republicans, with the opposite pattern holding for crime leading the researchers to suggest that candidates have differing degrees of credibility, depending upon the issues they address. In a later study, this hypothesis was tested; Iyengar and Valentino (1998) asked experimental participants to rate campaign ads aired by presidential candidates Dole and Clinton during the 1996 presidential campaign. They found that Republicans were more likely to rate Dole=s ads as informative (and less likely to rate them as misleading) when the ads addressed ARepublican@ issues (drug abuse, crime, and illegal immigration). Conversely, Democrats were more impressed by Clinton=s ads when they dwelled on ADemocratic@ issues (social security, welfare reform, health care).

The logic of "issue ownership" extends easily to attributes of the candidates other than their party affiliation. Gender is an especially visible attribute and the popular culture provides several cues about the traits of male and females, cues that are amply reinforced by the media's depiction of women candidates (Kahn, 1994; Kahn, 1994a). Given the availability of gender stereotypes, it might be anticipated that "masculine" issues such as defence or crime will be especially persuasive as campaign material when the candidate is a male war hero, while child care and matters of educational policy will resonate well with voters' beliefs about a female candidate who happens to be a mother.

Ansolabehere and Iyengar (1995), Valentino et al. (1996) and Iyengar and Valentino (1998) have designed experimental tests of the predictions of the resonance model as it applies to the issue appeals of male and female candidates. Their research compared the success of U.S. senatorial candidates Barbara Boxer and Dianne Feinstein with presidential candidate Bill Clinton when all three candidates broadcast a campaign advertisement that dealt with the issue of sexual harassment. The results showed that advertising on gender-related issues yielded significant gains for Boxer and Feinstein, but virtually no advantages for Clinton. (Insert discussion of credibility results from 1998 paper) Given the presence of built-in differences between candidates in their ability to gain from specific issue appeals, these results suggest quite strongly that candidates should emphasize issues on which they enjoy comparatively favorable stereotypes. As we describe in the next section, this is an important ingredient of campaign strategy.

In sum, the insight offered by the resonance approach is that campaigns do not occur in vacuums, but instead blend in with voters' partisan motives and attitudes. As such, the effects of interest are inherently interactive -- either involving interactions between the content or source of campaign messages and voters' partisanship, or higher-order interactions that also capture individual differences in exposure to campaign messages. The clear implication of the resonance model is that efforts to study campaigns as "main effects" (e.g. by entering indicators of media exposure into a regression equation) are doomed to fail.

The Strategic Model

This model also specifies campaign effects as interactions, but focuses on strategic interactions between the competing candidates and between the candidates and the press. In the hypodermic and resonance frameworks, the candidates act to maximize their position vis-a-vis the electorate with little regard for the actions of their opponents. The competitive model, on the other hand, sees the effects of any particular message as conditioned by the effects of competitors' messages; that is, candidates are interdependent rather than autonomous actors.

The recognition that campaigns are interdependent is especially important in the political world. Unlike product manufacturers and retailers, candidates for public office feel free to air advertisements that feature their opponents. Campaigns have increasingly turned to "attack advertising," in which candidates or their surrogates directly attack or seek to discredit their opponents (see Jamieson, 1992; Ansolabehere and Iyengar, 1996). Although the impact of attack messages is thought to depend upon certain qualities of the sponsoring candidates (such as their popularity), practitioners generally acknowledge that it is the response of the attacked candidate that is more important. Generally, the attacked candidate is thought to suffer if he or she fails to rebut or otherwise discredit the attack. According to Roger Ailes' "First Law" of political advertising: "once you get punched, you punch back."

The question of advertising tone (i.e., attack vs. self-promotion) is only one element of the strategic equation. Candidates must also choose from any number of advertising themes. A candidate whose opponent is considered especially qualified on a particular issue might prefer to organize his advertising around alternative issues or, instead, attempt to neutralize his opponent's advantage by highlighting his own credentials on the issue in question.

Game theory provides an elementary yet powerful tool for studying campaign strategy and the behavior of candidates. Concept of equilibrium. Adam: need to insert discussion of your model here.

 

.Example of 1994 California gubernatorial campaign study.

As this example reveals, game theory provides both descriptive and prescriptive insights into campaigns. More generally, the theory of games provides a well-developed analytical tool which can lead to an enriched and often counter-intuitive understanding of campaigns. Candidates for elective office would seem to meet the definition of rational actors -- they have clearly defined interests (gaining election), they invest considerable time, money and effort in order to maximize their own "utility," namely, their vote share, and they are less than certain about the motives and actions of their competitor(s). As such, campaigns can be modeled as simple games of incomplete information (Austen-Smith, 1992; Banks, 1990).

In structuring an advertising campaign, a candidate must anticipate not only his opponent's probable strategy, but also the activities of the news media. In the aftermath of the 1988 presidential campaign, reporters have shown a penchant for examining campaign advertisements. In 1992, all major news outlets regularly offered "ad watch" reports which scrutinized particular campaign advertisements for their accuracy and fairness. The intent behind this new genre of campaign journalism is, of course, to deter candidates from using false, distorted, or exaggerated claims. However, this form of news coverage also has the potential to provide the candidates with considerable free exposure by recirculating the campaign message to a vast audience. Some studies of ad watch reports have concluded that they actually have the effect of strengthening the impact of the scrutinized advertisement; others have shown that ad watch reporting works as intended (for a review of the evidence, see Task Force on Campaign Reform, 1998).

The emergence of ad watch journalism provides candidates with incentives to design advertisements that are especially likely to attract news coverage. By making their advertisements newsworthy, candidates obtain additional (and free) exposure on subjects of their choosing. In 1996, the Republican National Committee produced an advertisement criticizing President Clinton for attempting to use his status as Commander-in-Chief of the U.S. military to evade prosecution on sexual harassment charges. This controversial attack attracted front-page and prime-time coverage for several days, but not once did it air as an advertisement.

In sum, a second layer of strategic interactions involves candidates and the press. with the former attempting to recirculate their appeals in the guise of news reports. From the perspective of the candidate, accordingly, the goal is to design advertisements that not only counter the opponent's strategy, but succeed in attracting extensive news coverage.

Beyond Surveys: Experimentation and Content Analysis

As outlined earlier, the field's commitment to survey research has impeded researchers' ability to detect the traces of campaigns in public opinion. The advantages of studying campaigns via experiments are obvious and follow directly from the fact that experiments enable researchers to systematically assess the specific components of campaign messages. A thirty second advertisement, for instance, is a concatenated sequence of images and text. What was it that moved viewers in the infamous "Revolving Door" advertisement of 1988?. Was it, as is widely alleged, Mr. Horton's race? Or was it the violent and brutal nature of his behavior, the fact that he was a convict, the race of his victim, or what? Modern, digital-based techniques of audiovisual editing make it possible to zero in on the explanation, whether it be text-based, or in the form of audio or visual cues. In the case of the Horton advertisement, for instance, we might construct identical facial composites of Mr. Horton, but then vary his skin color so as to estimate the effects of race. Alternatively, we might present the convict's face with and without a beard. In effect, it is possible to dissect a message into core visual components and input these components into experimental manipulations.

Of course, experiments are not without their limitations. Most experiments are administered upon "captive" populations -- college students who must participate in order to gain course credit. As Carl Hovland warned many years ago, college sophomores are a convenient subject population, but they are not comparable to "real people." Experiments also require a somewhat sterile, laboratory-like environment which bears little resemblance to the "blooming, buzzing confusion" of election campaigns. In general, a considerable leap of faith is required to generalize experimental results to the real world.

Experiments can be made more naturalistic by relying on participants who represent a fair cross-section of the electorate and by resorting to experimental procedures and settings that closely reflect the typical citizen's media experiences. In the case of campaign advertising, realism can be enhanced by administering the experiments during ongoing campaigns characterized by heavy levels of advertising (see Ansolabehere and Iyengar, 1996).

No matter how realistic their designs, experimenters must strive to replicate their results using alternative sources of evidence. Content analysis merged with surveys provides one such opportunity. Traditionally, content analysis has been used as a descriptive tool to identify characteristics of messages. This descriptive function enables researchers to identify the relevant experimental manipulations. In the past, this process was sufficiently labor-intensive to deter most researchers. Attempts to increase the rate of return by using computers to interpret and categorize language have yet to bear fruit. Nonetheless, technological developments have introduced dramatic economies of scale. These developments will accelerate the ability to simultaneously identify the distribution of campaign messages and incorporate these messages into experimental manipulations.

Until recently, researchers attempting to content analyze news coverage would have to either subscribe to a representative set of newspapers from across the country or rely on one or two "prestige" outlets. Either method would require vast amounts of labor, storage, and organization. The process was tedious, expensive, and subject to considerable human error. Today, most newspapers can be accessed through electronic databases offering full-text retrieval of articles that can be digitally searched and transferred to digital storage in a matter of minutes. On-line services such as Nexis and Westlaw include major newspapers as far back as 1980. By using Nexis and other archives (such as Datatimes), the researcher has access to virtually every newspaper published in the U.S. What would take months and years to collect and compile can be accomplished in a week. For instance, for research on candidates strategy in U.S. Senate campaigns between 1988 and 1992, we compiled over 14,000 newspaper articles bearing on fifty races. The entire data set was constructed within a period of six months.

Using this approach, experimental findings have been replicated with relative ease. One of the more consistent results from the Ansolabehere and Iyengar experiments concerning the effects of campaign advertising is that exposure to negative (rather than positive) advertising reduces voter turnout (see Ansolabehere and Iyengar, 1996). This result was replicated by content analyzing newspaper coverage of all 34 U.S. Senate campaigns in 1992. Based on the news reports of the candidates' advertising, the races were classified as either positive or negative in tone. The content-based measure of advertising tone was then used to predict voter turnout in the various senate elections.

Conclusion

As this review essay has suggested, recent research has begun to take a toll on the long-dominant "minimalist" view of campaigns. A major stimulus to this progression has been the increasing volume of traffic between political science, communications, and allied disciplines. The work of the eminent social psychologists Carl Hovland and William McGuire has served as an invaluable beacon for the current generation of campaign researchers. Hovland's calls for methodological pluralism led gradually to the current flourishing of experimental and quasi-experimental work which, as we have seen, provides the most unequivocal evidence of campaign effects. McGuire's theorizing about the exposure-acceptance matrix has been no less fertile, contributing to the current interest in interactionist specifications of campaigns.

We may anticipate a further growth spurt in campaign research from exchanges on other cross-disciplinary fronts, particularly the synthesizing of empirical work on campaign effects with rational choice, game-theoretic models of communication. These models provide a well-developed set of analytic and mathematical tools from which to derive testable propositions. With its emphasis on strategic interactions and the dependency of outcomes on individual choices, game theory is an especially appropriate to the study of campaign strategy and decision making.

In sum, the study of campaign effects stands poised to make significant theoretical and methodological advances. As ongoing inter-disciplinary efforts mature, we may expect an outpouring of evidence that campaigns contribute to the selection of political leaders and the formulation of public policy.

References

Alvarez, R. M., and C. H. Franklin. 1994. "Uncertainty and political perceptions." Journal of Politics, 56, 671-88.

Alvarez, M. 1997.

Ansolabehere, S. L., and S. Iyengar. 1996. Going Negative: How Political Advertisements Shrink and Polarize the Electorate. New York: Free Press.

Ansolabehere, S., S. Iyengar, and A. Simon. 1998.

Ansolabehere, S., S. Iyengar, A. Simon, and N. Valentino. 1997. ARunning as a woman: gender stereotyping in women=s campaigns,@ in P. Norris ed. Women, Media, and Politics. New York: Oxford University Press.

___. 1996a "The craft of political advertising: a progress report," in D. C. Mutz, P. M. Sniderman, and R. A. Brody eds., Political Persuasion and Attitude Change. Ann Arbor: University of Michigan Press.

___. 1998. Messages forgotten: misreporting in surveys and the bias towards minimal effects." Unpublished paper,Massachusetts Institute of Technology.

___. 1995. "Of horseshoes and horse races: experimental studies of the impact of poll results on electoral behavior," Political Communication, 5, 413-29.

Austen-Smith, D. 1992. "Strategic models of talk in political decision making." International Political Science Review, 3, 45-58.

Banks, J. 1990. "A model of electoral competition with incomplete information." Journal of Economic Theory, 50, 309-25.

Bartels, L. M. 1986. "Issue voting under uncertainty." American Journal of Political Science, 30, 709-28.

___. 1988. Presidential Primaries and the Dynamics of Public Choice. Princeton: Princeton University Press.

___. 1993. "Messages received: the political impact of media exposure." American Political Science Review, 87, 267-85.

___. 1997. AThree virtues of panel data for the analysis of campaign effects,@ Presented at the Coference on Campaign Effects, Vancouver, British Columbia.

Bradburn, N. M., L. J. Rips, and S. K. Shevell. 1987. "Answering autobiographical questions: the impact of memory and inference in surveys." Science, 236, 157-61.

Brady, H. E. 1984. "Chances, utilities and voting in presidential primaries," Paper Presented at the Annual Meeting of the Public Choice Society.

Brady, H. E., and R. Johnston. 1987. "What's the primary message: horse race or issue journalism?" in G. Orren and N. Polsby eds., Media and Momentum. Chatham, NJ: Chatham House.

Brians, C. L., and M. P. Wattenberg. 1996. "Campaign issue knowledge and salience: Comparing reception from TV commercials, TV news, and newspapers." American Journal of Political Science, 40, 172-93.

Chaffee, S. H., Zhao, X., and G. Leshner. 1994. "Political knowledge and the campaign media of 1992." Communication Research, 21, 305-24.

Dalton, R. J., P. A. Beck, R. Huckfeldt, and W. Koetzle. 1995. "Agenda-setting in the 1992 campaign: the flow of campaign information," Paper Presented at the Annual Meeting of the Midwest Political Science Association.

Finkel, S. E. 1993. "Reexamining the 'minimal effects' model in recent presidential campaigns." Journal of Politics, 55, 1-21.

Franklin, Charles H. 1991. "Eschewing obfuscation? Campaigns and the perception of U.S. Senate incumbents." American Political Science Review, 85, 1193-1214.

Hallin, D. 1994. We Keep America on Top of the World: Television Journalism and the Public Sphere. London: Routledge.

Huddy, L. and N. Terkildsen. 1993. "Gender stereotypes and the perception of male and female candidates." American Journal of Political Science, 37, 119-147.

Holbrook, T. M. 1994. "Campaigns, national conditions, and U.S. presidential elections." American Journal of Political Science, 38, 973-98.

___. 1996. Do Campaigns Matter? Thousand Oaks, CA: Sage Publications.

Iyengar, S. 1993. AAgenda-setting and beyond: television news and the strength of political issues,@ in W. Riker ed., Agenda Formation. Ann Arbor: University of Michigan Press.

Iyengar, S., and D. R. Kinder. 1987. News That Matters: Television and American Opinion. Chicago: University of Chicago Press.

Iyengar, S., and A. Simon. 1993. "News coverage of the Gulf War and public opinion: a study of agenda-setting, priming, and framing." Communication Research, 365-383.

Iyengar, S., and N. Valentino. 1998. AWho says what: source credibility as a mediator of campaign advertising,@ in A. Lupia, M. McCubbins and S. Popkin eds. Elements of Reason. New York: Cambridge University Press, forthcoming.

Jamieson, K. H. 1992. Dirty Politics. New York: Oxford University Press.

Jennings, M. K., and R. G. Niemi. 1981. Generations and Politics: A Panel Study of Young Adults and Their Parents. Princeton: Princeton University Press.

Johnston, R., A. Blais, H. E. Brady, and J. Crete. 1992. Letting the People Decide: Dynamics of a Canadian Election. Stanford: Stanford University Press.

Kahn, K. F. 1994. "Does gender make a difference? An experimental examination of sex stereotypes and press patterns in statewide campaigns." American Journal of Political Science, 38, 162-95.

___. 1994a. "The distorted mirror: press coverage of women candidates for statewide office." Journal of Politics, 56, 154-73.

Keith, B. E., D. B. Magleby, C. J. Nelson, E. Orr, M. C. Westlye, R. E. Wolfinger. 1992. The Myth of the Independent Voter. Berkeley: University of California Press.

Kern, M. 1989. Thirty-Second Politics: Political Advertising in the 1980s. New York: Praeger.

Krosnick, J. A., and L A. Brannon. 1993. "The impact of the Gulf War on the ingredients of presidential evaluations: multidimensional effects of political involvement." American Political Science Review, 87, 963-78.

Lichter, S. R. 1988. The Video Campaign: Network Coverage of the 1988 Presidential Primaries. Washington, D.C.: American Enterprise Institute.

McCombs, M. C., and G. Estrada. 1996. "The news media and the pictures in our heads,"

in S. Iyengar and R. Reeves eds., Do the Media Govern?. Thousand Oaks, CA: Sage Publications.

McGuire, W. J. 1968. "Personality and susceptibility to social influence," in E. F. Borgatta and

W. W. Lambert eds., Handbook of Personality Theory and Research. Chicago: Rand McNally.

Mutz, D. C. 1997. AMechanisms of momentum: does thinking make it so?.@ Journal of Politics, 59: 104-124.

Miller, J. M., and J. A. Krosnick. 1996. "The anatomy of news media priming," in S. Iyengar and R. Reeves eds. Do the Media Govern?. Thousand Oaks: Sage Publications.

Niemi, R. G., and M. K. Jennings. 1991. "Issues and inheritance in the formation of party identification." American Journal of Political Science, 35, 970-89.

Patterson, T. E., and R. D. McClure. 1976. The Unseeing Eye: The Myth of Television Power in National Elections. New York: G. P. Putnam's Sons.

Petrocik, J. R. 1996. "Issue ownership in presidential elections with a 1980 case study.@ American Journal of Political Science, 40, 825-50.

Pierce, J. C., and N. P. Lovrich. 1982. "Survey measurement of political participation." Social Science Quarterly, 63, 164-71.

Popkin, S. L. 1992. The Reasoning Voter. Chicago: University of Chicago Press.

___. 1996. "Voter learning in the 1992 presidential election," in S. Iyengar and R. Reeves eds. Do the Media Govern?. Thousand Oaks, CA: Sage Publications.

Riker, W. H. 1980. "Implications from the disequilibrium of majority rule for the study of institutions." American Political Science Review, 74, 432-46.

Robinson, M., and M. Sheehan. 1983. Over the Wire and on TV. N.Y.: Russell Sage Foundation.

Rogers, E. M., W. B. Hart, and J. W. Dearing. 1996. "A paradigmatic history of agenda-setting research, in S. Iyengar and R. Reeves eds. Do the Media Govern?. Thousand Oaks, CA: Sage Publications.

Sabato, L. 1991. Feeding Frenzy: How Attack Journalism has Transformed American Politics. New York: Free Press.

Semetko, H., J. G. Blumler, M. Gurevitch, and D. H. Weaver. 1991. The Formation of Campaign Agendas: A Comparative Analysis of Party and Media Roles in Recent U.S. and British Election Campaigns. Hillsdale, N.J.: Lawrence Erlbaum Associates.

Shaw, D. R. 1996. "Blaming the messenger: media coverage and issue agendas in the 1992 election." Unpublished manuscript, University of Texas, Austin.

Shaw, D. R. 1997. "Presidential campaigning and electoral college strategies, 1988-1996.@ Presented at the Annual Meeting of the Midwestern Political Science Association.

Simon, A. F. 1998. The Winning Message: A Game-Theoretic Study of Campaigns and Candidate Strategy. Ph.D. Dissertation, University of California, Los Angeles.

Task Force on Campaign Reform. 1998. Campaign Reform: Insights and Evidence. Princeton, N.J.: Woodrow Wilson School of Public and International Affairs.

Wattenberg, M. P. and C. Brians. 1998. "Do negative ads . . . American Political Science Review, in press.

West, D. M. 1994. Air Wars: Television Advertising in Election Campaigns, 1952-92. Washington: Congressional Quarterly Press.

___. 1994. "Political advertising and news coverage in the 1992 California U.S. Senate campaigns." Journal of Politics, 56, 1053-75.

Zaller, J. 1992. The Nature and Origins of Mass Opinion. N.Y.: Cambridge University Press.

___. 1996. "The myth of massive media impact revived: new support for a discredited idea," in D. C. Mutz, P. M. Sniderman, and R. A. Brody eds., Political Persuasion and Attitude Change. Ann Arbor: University of Michigan Press.

Zhao, X., and S. H. Chaffee. 1996. "Campaign ads versus television news as information sources." Public Opinion Quarterly, 59, 41-65.