1. In your own words, explain Sides’ concepts of heresthetics and issue trespassing. 2. Choose a candidate for the Democratic Party’s 2020 presidential nomination, and assume you were a strategic adv
B.J.Pol.S.36,407–436Copyright©2006 Cambridge University Press doi:10.1017/S0007123406000226Printed in the United Kingdom The Origins of Campaign Agendas JOHN SIDES* Why and how do candidates choose the issues on which their campaigns are based? Drawing on a large database of candidate advertisements from the 1998 House and Senate campaigns, extant theories of issue emphasis, which focus on factors such as party ownership and candidate record, are tested here and these theories are expanded by examining in more detail the role of constituency characteristics. Most notably, party ownership’s impact is demonstrated to be weak: candidates are more willing to ‘trespass’ or talk about the other party’s issues than previous literature has found. Also ‘trespassing’ is shown to be facilitated by framing the other party’s issues in certain ways. The results have implications for theories of candidate strategy and for normative questions, such as how much ‘dialogue’ occurs in campaigns. One chief function of electoral campaigns is to place issues before voters. Given that most voters are relatively inattentive to politics, campaigns provide a crucial arena for candidates to discuss their views on speci c issues. Commentators in turn refer to the agendas that candidates formulate as ‘what this election is about’. Campaign agendas also resonate beyond the campaign as elected of cials arrive in various capitals and draw on these agendas to guide policy making. Campaign agendas take on even greater import because there are electoral bene ts to agenda setting. Candidates can win votes by emphasizing issues where they perceive an advantage, thereby making these issues prominent in voters’ minds. This strategy is an appealing alternative to other, arguably more dif cult tactics, such as shifting one’s own position as a candidate to match the voters’ or persuading voters to adopt a speci c position on an issue. As candidates begin to formulate agendas, salient political events and public opinion are very much on their minds. For example, events such as a war or an economic recession often gure prominently in campaign agendas. Candidates also consult the electorate; answers to a polling question like ‘What do you think is the most important problem facing our country?’ may lead candidates to profess similar concerns. These kinds of factors undoubtedly contribute to variation in campaign agendas across election years. However, these factors may not fully explain variation among candidates within any given election. Salient events and the public’s agenda are often national in scope and should * Department of Political Science, George Washington University. The author wishes to thank Henry Brady, Jack Citrin, John Geer, Andy Karch, Stephen Medvic, Susan Rasky, Laura Stoker and Sean Theriault for comments on previous versions of this article, and the reviewers and Editors for their suggestions. The 1998 advertising data come from the Brennan Centre for Justice at New York University School of Law, which used data obtained from the Campaign Media Analysis Group in Washington, D.C. Ken Goldstein deserves particular thanks for allowing access to copies of the 1998 advertisements. The 2000 data are based on work supported by The Pew Charitable Trusts under a grant to the Brennan Centre and a subsequent subcontract to the Department of Political Science at the University of Wisconsin-Madison. The 2002 data were obtained from a project of the Wisconsin Advertising Project, under Professor Kenneth Goldstein and Joel Rivlin of the University of Wisconsin-Madison, and include media-tracking data from the Campaign Media Analysis Group. The Wisconsin Advertising Project was sponsored by a grant from The Pew Charitable Trusts. The opinions expressed in this article are the author’s and do not necessarily re ect the views of the Brennan Centre, the Campaign Media Analysis Group, the Wisconsin Advertising Project, Professor Goldstein, Joel Rivlin or The Pew Charitable Trusts. 408 SIDES thus in uence all candidates. Any cross-sectional variation must derive from other factors, especially the decisions of individual candidates. Unfortunately, relative to research on voter behaviour in campaigns and elections, there is much less research on candidate behaviour. As Franklin writes: The political nature of elections lies in the choices candidates make about strategy. How to present oneself and one’s opponent to the voters is the critical electoral heresthetic. At present, the spatial model is the closest thing we have to an explanation for candidate behavior. Yet we have very little empirical work on campaign behavior with which to test the predictions from this or any other model. As we have become adept at studying voters, it is ironic that we have virtually ignored the study of candidates. Yet it is in candidate behavior that politics intrudes into voting behavior. Without the candidates, there is only the psychology of vote choice and none of the politics. 1 The chief question examined here is, what are the bases of these heresthetical decisions? Or, put differently, what explains the content of campaign agendas? The analysis that follows focuses on the major theory of campaign agendas, party ‘ownership’ of issues, which predicts that candidates should advertise only on issues historically associated with their party. I compare this theory to an alternative, ‘riding the wave’, which predicts that candidates from both parties will tend to advertise on the same issues, namely those salient in the public’s mind. My ndings differ from past studies in important ways. Most notably, I nd that party ownership is not consistently important. Candidates of both parties tend to pursue similar agendas based on the public’s priorities. Consequently, candidates often ‘trespass’, or talk about issues the other party ostensibly ‘owns’. I then show how parties trespass by framing issues in particular ways. For example, candidates often talk about the other party’s issues by highlighting dimensions of that issue that are consonant with their party’s traditional philosophy. The empirical analysis is based on an extensive dataset of advertisements from 1998 House and Senate races that allows for a more detailed examination of campaign agendas. I also analyse data from the 2000 and 2002 races to determine whether the results are generalizable. The ndings have implications for normative questions about ‘dialogue’ in campaigns and for theories of candidate strategy. HERESTHETICS AS A THEORY OF CANDIDATE STRATEGY Heresthetics represents a departure from two other extant theories of campaign strategy: convergence and persuasion. Convergence, whereby candidates assume the position of the median voter, has proven theoretically fragile and empirically lacking. 2Persuasion, or 1Charles H. Franklin, ‘Eschewing Obfuscation? Campaigns and the Perception of US Senate Incumbents’, American Political Science Review, 85 (1991), 1193–1214, p. 1201. 2The strategy of convergence is explicated in Anthony Downs,An Economic Theory of Democracy(New York: Harper and Row, 1957). See also Harold Hoteling, ‘Stability in Competition’,Economic Journal, 39 (1929), 41–57; and Arthur Smithies, ‘Optimum Location in Spatial Competition’,Journal of Political Economy, 49 (1941), 423–9. Various factors complicate the spatial model’s prediction of convergence. One is risks to the candidate’s credibility – see Alberto Alesina, ‘Credibility and Policy Convergence in a Two-Party System with Rational Voters’,American Economic Review, 78 (1988), 796–805; Randall L. Calvert, ‘Robustness of the Multidimen- sional Voting Model: Candidate Motivations, Uncertainty, and Convergence’,American Journal of Political Science, 29 (1985), 69–95; William Riker, ‘Heresthetic and Rhetoric in the Spatial Model’, in James M. Enelow and Melvin J. Hinich, eds,Advances in the Spatial Theory of Voting(Cambridge: Cambridge University Press, 1990), pp. 46–65; and Donald Wittman, ‘Candidate Motivation: A Synthesis of Alternatives’,American Political The Origins of Campaign Agendas409 ‘rhetoric’ in Riker’s terminology, means that candidates attempt to convince the median voter to support their position. 3The dif culty with this strategy is that individuals often process information in ways that merely con rm pre-existing attitudes and thus remain immune to attempts at persuasion. 4 In contrast, heresthetics focuses attention on campaign agendas. 5Heresthetics is the art of agenda manipulation: ‘The point of an heresthetical act is to structure the situation so that the actor wins, regardless of whether or not the other participants are persuaded’. 6If (F’note continued)Science Review, 77 (1985), 142–57. A second is a multi-dimensional issue space – see Scott Feld, Bernard Grofman and Nicholas Miller, ‘Centripetal Forces in Spatial Voting: On the Size of the Yolk’,Public Choice, 59 (1988), 37–50; Richard McKelvey, ‘Covering, Dominance, and Institution-Free Properties of Social Choice’,America Journal of Political Science, 30 (1986), 282–314; and Charles R. Plott, ‘A Notion of Equilibrium and Its Possibility under Majority Rule’,American Economic Review, 57 (1967), 787–806. A third is elections that occur in stages – see John H. Aldrich,Before the Convention: Strategies and Choices in Presidential Nomination Campaigns (Chicago: University of Chicago Press, 1980). A fourth is imperfect information about voters’ preferences – see Randall L. Calvert, ‘Robustness of the Multidimensional Voting Model: Candidate Motivations, Uncertainty, and Convergence’,American Journal of Political Science, 29 (1985), 69–95; and Rebecca B. Morton, ‘Incomplete Information and Ideological Explanations of Platform Divergence’,American Political Science Review, 87 (1993), 382–92. A fth is the in uence of party activists – see John H. Aldrich, ‘A Downsian Spatial Model with Party Activists’,American Political Science Review, 77 (1983), 974–90; and Gary Miller and Norman Scho eld, ‘Activists and Partisan Realignments in the United States’,American Political Science Review, 97 (2003), 245–60. A sixth is ‘valence’ advantages, such as incumbency – see Stephen Ansolabehere and James M. Snyder Jr, ‘Valence Politics and Equilibrium in Spatial Election Models,’Public Choice, 103 (2000), 327–36; Tim Groseclose, ‘A Model of Candidate Location When One Candidate Has a Valence Advantage’,American Journal of Political Science, 45 (2001), 862–86; and Stuart Elaine MacDonald and George Rabinowitz, ‘Solving the Paradox of Nonconvergence: Valence, Position, and Direction in Democratic Politics’,Electoral Studies, 17 (1998), 281–300. A nal factor is directional voting – see George Rabinowitz and Stuart Elaine MacDonald, ‘A Directional Theory of Issue Voting’,American Political Science Review, 83 (1989), 93–121. Thus, it is perhaps no surprise that politicians routinely fail to seek the position of the median voter – see James Adams and Samuel Merrill, III, ‘Modeling Party Strategies and Policy Representation in Multiparty Elections: Why Are Strategies So Extreme?’American Journal of Political Science, 43 (1999), 765–91; Stephen Ansolabehere, James M. Snyder Jr and Charles Stewart III, ‘Candidate Positioning in US House Elections’, American Journal of Political Science, 45 (2001), 136–59; Barry C. Burden, ‘Candidate Positioning in US Congressional Elections’,British Journal of Political Science, 34 (2004), 211–27; Robert S. Erikson and Gerald C. Wright, ‘Voters, Candidates, and Issues in Congressional Elections’, in Lawrence C. Dodd and Bruce I. Oppenheimer, eds,Congress Reconsidered(Washington, D.C.: Congressional Quarterly Press, 2001), pp. 67–96; Bernard Grofman, Robert Grif n and Amihai Glazer, ‘Identical Geography, Different Party: A Natural Experiment on the Magnitude of Party Differences in the U.S. Senate, 1960–84’, in R. J. Johnston, F. M. Shelley and P. J. Taylor, eds,Developments in Electoral Geography(New York: Routledge, 1990), pp. 207–17; Norman Scho eld, Andrew D. Martin, Kevin M. Quinn and Andrew B. Whitford, ‘Multiparty Electoral Competition in the Netherlands and Germany: A Model Based on Multinomial Probit’,Public Choice, 97 (1998), 257–93; and Gerald C. Wright Jr and Michael B. Berkman’, ‘Candidates and Policy in United States Senate Elections’,American Political Science Review, 80 (1986), 567–88. 3William Riker, ‘Political Theory: The Art of Heresthetics’, in Ada Finifter, ed.,Political Science: The State of the Discipline(Washington, D.C.: American Political Science Association, 1983). 4Psychologically uncomfortable information can be explained away to reduce cognitive dissonance – see Leon Festinger,Theory of Cognitive Dissonance(Evanston, Ill.: Row, Peterson, 1957). Theories of ‘schematic’ processing describe how existing beliefs in uence perception and memory, such that we fail to remember information that would discon rm or qualify those beliefs – see Shelley E. Taylor and Jennifer Crocker, ‘Schematic Bases of Social Information Processing’, in E. Higginset al., eds,Social Cognition: The Ontario Symposium (Hillsdale, N.J.: Erlbaum, 1981). 5Riker, ‘Political Theory’.6Riker, ‘Political Theory’, p. 60. For examples, see William Riker,The Art of Political Manipulation(New Haven: Yale University Press, 1986); and William Riker,The Strategy of Rhetoric: Campaigning for the American Constitution(New Haven: Yale University Press, 1996). 410 SIDES candidates can structure the election’s agenda so that the issues where their positions are popular come to the fore in voters’ minds, then a larger number of voters will support them. No convergence or persuasion is necessary. Hammond and Humes apply this notion more formally to campaign strategy. They write, ‘Instead of the candidates trying to gure out what positions to take, then, political campaigns are turned into contests about what the issue dimensions of the campaign will be’. 7 This emphasis on heresthetics as a strategy coincides nicely with a literature that emphasizes how campaigns actually affect voters’ agendas and the predictors of vote choice. After arguing that Truman rallied undecided Democrats in the 1948 campaign by emphasizing working-class economic concerns, Berelsonet al. write, ‘It is dif cult to change people’s preferences; it is easier to affect the priorities and weights they give to subpreferences bearing on the central decision’. 8 In sum, campaign agendas have both theoretical and empirical relevance. Heresthetics can win votes not only in the rari ed climes of the spatial model but also on the ground, in the day-to-day hurly-burly of an actual campaign. However, we have no systematic understanding of this strategy. As Riker states, ‘It is hard to de ne heresthetics other than by examples because the purpose of the concept is to provide an open-ended set of categories for events that have not heretofore been systematically characterized’. 9Given 7Thomas H. Hammond and Brian D. Humes, ‘The Spatial Model and Elections,’ in Bernard Grofman, ed., Information, Participation, and Choice(Ann Arbor: University of Michigan Press, 1993), pp. 141–59, at p. 142. See also Holly Brasher, ‘Capitalizing on Contention: Issue Agendas in U.S. Senate Campaigns’,Political Communication, 20 (2003), 453–71; Thomas M. Carsey,Campaign Dynamics: The Race for Governor(Ann Arbor: University of Michigan Press, 2000); Amihai Glazer and Susanne Lohmann, ‘Setting the Agenda: Electoral Competition, Commitment of Policy, and Issue Salience’,Public Choice, 99 (1999), 377–94; John G. Geer, ‘Campaigns, Party Competition, and Political Advertising’, in John G. Geer, ed.,Politicians and Party Politics (Baltimore, Md.: Johns Hopkins University Press, 1998), pp. 186–217; Melvin J. Hinich, Michael C. Munger and Scott De Marchi, ‘Ideology and the Construction of Nationality: The Canadian Elections of 1993’,Public Choice, 97 (1998), 401–28; Stephen K. Medvic,Political Consultants in U.S. Congressional Elections(Columbus: Ohio State University Press, 2002); and Adam Simon,The Winning Message: Candidate Behavior, Campaign Discourse, and Democracy(Cambridge: Cambridge University Press, 2002). That a heresthetic logic matters to candidates is evident anecdotally. For example, after the 1994 California governor’s race, Dick Dresner, the pollster for then-incumbent governor Pete Wilson, described a strategy of agenda control: ‘If I can control the agenda and expand the number of people who are concerned about immigration or crime [Wilson’s focus in this campaign], then I can change their focus from something else, whether it’s the environment or education or whatever. We developed techniques to expand our audience. They worked, and the number of people concerned about these issues just kept growing’ (cited in Gerald C. Lubenow, ed.,California Votes – The 1994 Governor’s Race: An Inside Look at the Candidates and Their Campaigns by the People who Managed Them(Berkeley, Calif.: Institute of Governmental Studies Press, 1995), p. 79). 8Bernard R. Berelson, Paul F. Lazarsfeld and William N. McPhee,Voting: A Study of Opinion Formation in a Presidential Campaign(Chicago: University of Chicago Press, 1954), p. 206. See also Carsey,Campaign Dynamics; Richard Johnston, Andre´ Blais, Henry E. Brady and Jean Creˆ te,Letting the People Decide: Dynamics of a Canadian Election(Stanford, Calif.: Stanford University Press, 1992); Kim Fridkin Kahn and Patrick J. Kenney,The Spectacle of US Senate Campaigns(Princeton, N.J.: Princeton University Press, 1999); Tali Mendelberg,The Race Card: Campaign Strategy, Implicit Messages, and the Norm of Equality(Princeton, N.J.: Princeton University Press, 2001); Matthew Mendelsohn, ‘The Media and Interpersonal Communications: The Priming of Issues, Leaders, and Party Identi cation’,Journal of Politics, 58 (1996), 112–25; Simon,The Winning Message; Nicholas A. Valentino, Vincent L. Hutchings and Ismail K. White, ‘Cues that Matter: How Political Ads Prime Racial Attitudes during Campaigns’,American Political Science Review, 96 (2002), 75–90. These studies dovetail with others that show how the news media can have a similar effect – e.g., Shanto Iyengar and Donald Kinder,News That Matters: Television and American Opinion(Chicago: University of Chicago Press, 1987). 9Riker, ‘Political Theory: The Art of Heresthetics’, p. 56. The Origins of Campaign Agendas411 this lack of understanding about the bases of heresthetics, and indeed about the origins of campaign strategy generally, it is all the more important to identify how candidates come to choose the issues that dominate their campaigns. WHENCE COME CAMPAIGN AGENDAS? Candidates desire to win voters, and strategic choices about campaign agendas can help accomplish that goal. However, not all choices are equal. Certain areas of emphasis may bear more fruit than others. What factors should guide heresthetical choices? Below I explicate further Riker’s theory of heresthetics and connect it with the most prominent factor in the extant literature, party ownership of issues. I generate speci c hypotheses about what the theory of party ownership predicts and contrast these predictions with those derived from an alternative theory, known colloquially as ‘riding the wave’. 10 Dominance, Issue Ownership and ‘Riding the Wave’ In Riker’s explication of heresthetics, he derives two important principles that anticipate the ndings of subsequent empirical work. 11The Dominance Principle states that ‘when one side has an advantage on an issue, the other side ignores it’. The Dispersion Principle states that ‘when neither side has an advantage, both seek new and advantageous issues’. 12 Thus, parties and candidates will focus their campaign rhetoric on issues that advantage them and ignore all others. Focusing on issues that are disadvantageous is obviously irrational, and focusing on those on which neither party has a clear advantage is likely to be a waste of resources. The empirical prediction is that competing candidates or parties will diverge rhetorically during the campaign and emphasize quite different issues – an empirical pattern that Riker observes in his study of the campaign for the rati cation of the US Constitution. This formulation of the Dominance Principle begs the question of what provides parties with an advantage on any given issue. The theory of party ownership states that the advantage arises from reputations the parties have developed for effective policy making on certain issues. The parties have what Petrocik describes as a ‘a history of attention, initiative and innovation toward these problems, which leads voters to believe that one of the parties (and its candidates) is more sincere and committed to doing something about them’. 13Thus, ‘issue ownership’ appears to provide parties the kind of advantage that Riker would describe as ‘dominance’. The crucial mechanism behind this advantage is credibility: candidates cannot simply appropriate any issue they please and claim it as their own. Instead, heresthetical manœuvers will work best when candidates are perceived as credible – that is, as concerned about and committed to an issue. Party reputations, with 10Stephen Ansolabehere and Shanto Iyengar, ‘Riding the Wave and Claiming Ownership over Issues: The Joint Effects of Advertising and News Coverage in Campaigns’,Public Opinion Quarterly, 58 (1994), 335–57. 11Riker,The Strategy of Rhetoric. It should be noted, however, that this work was published posthumously and the editors believed Riker had not fully re ned the chapter in which these principles were elaborated (fn.1, p. 99). 12Riker,The Strategy of Rhetoric, p. 106.13John R. Petrocik, ‘Issue Ownership in Presidential Elections, with a 1980 Case Study’,American Journal of Political Science, 40 (1996), 825–50, p. 826. As these reputations accrue over time, voters may actually develop partisan stereotypes that mitigate their ability even to notice when candidates ‘trespass’ on the other party’s territory – see Helmut Norporth and Bruce Buchanan, ‘Wanted: The Education President: Issue Trespassing by Political Candidates’,Public Opinion Quarterly, 56 (1992), 87–99. 412 SIDES their accumulated historical evidence, provide this credibility. Thus, candidates should campaign on the issues that their party is perceived to ‘own’, and numerous studies of campaign content provide evidence that they do so. 14 Candidates from different parties appear to ‘talk past’ each other during campaigns, rarely speaking to the same issues – just as the Dominance Principle would predict. Relatively few scholars have argued that issue ownership’s effects could be limited. However, there are theoretical reasons to expect limitations. Instead of focusing on different issues based on their respective reputations, the parties may instead focus on a common set of issues that are highly salient to the public. Ansolabehere and Iyengar refer to this strategy as ‘riding the wave’. 15The theoretical motivation for candidates is to appear responsive to the public’s concerns, regardless of an historical reputation or claim to ‘ownership’. Salient issues are likely to motivate voter decision making, and a party that has not made an effort to speak to that issue may be perceived as indifferent. This strategy may account for apparent recent exceptions to the pattern of divergence predicted by the theory of issue ownership, such as George W. Bush’s focus on education in the 2000 election (see below). 16 Parties may thus have an incentive to ‘trespass’ on the opposing party’s territory. In sum, the theory of issue ownership predicts that opposing candidates will emphasize very different issues based on the historical record of their party, while the idea of ‘riding the wave’ predicts that opposing candidates will emphasize a similar set of highly salient issues. To formulate testable hypotheses for each theory, we must identify the issues that the Democratic and Republican parties own as well as the issues salient to voters. To measure issue ownership, Petrocik examines poll questions about which party is better able to ‘handle’ certain issues. 17Because the analysis of candidate strategy focuses on the 1998 election, I extend Petrocik’s analysis and evaluate issue ownership using polling data from 1990–98. 18The advantage of over-time data is that one can evaluate not 14Brasher, ‘Capitalizing on Contention’; Ian Budge and Dennis J. Farlie,Explaining and Predicting Elections: Issue Effects and Party Strategies in Twenty-Three Democracies(London: George Allen & Unwin, 1983); Kahn and Kenney,The Spectacle of US Senate Campaigns; Benjamin I. Page,Choices and Echoes in Presidential Campaigns(Chicago: University of Chicago Press, 1978); Petrocik, ‘Issue Ownership in Presidential Elections’; Simon,The Winning Message; Constantine J. Spiliotes and Lynn Vavreck, ‘Campaign Advertising: Partisan Convergence or Divergence?’Journal of Politics, 64 (2002), 249–61. In their mathematical model of candidate strategy, Hammond and Humes (‘The Spatial Model and Elections’) also nd it unlikely that two opposing candidates would pursue exactly the same strategy. 15Ansolabehere and Iyengar, ‘Riding the Wave and Claiming Ownership over Issues’.16For examples of trespassing, see John G. Geer, ‘Campaigns, Party Competition, and Political Advertising,’ in John G. Geer, ed.,Politicians and Party Politics(Baltimore: Johns Hopkins University Press, 1998); David B. Holian, ‘He’s Stealing My Issues! Clinton’s Crime Rhetoric and the Dynamics of Issue Ownership’,Journal of Politics, 26 (2004), 95–124; John R. Petrocik, William L. Benoit and Glenn J. Hansen, ‘Issue Ownership and Presidential Campaigning, 1952–2000’,Political Science Quarterly, 118 (2003), 599–626; and Lee Sigelman and Emmett H. Buell Jr, ‘Avoidance or Engagement? Issue Convergence in U.S. Presidential Campaigns, 1960–2000’, American Journal of Political Science, 48 (2004), 650–61. Brasher also nds that election-year salience drives campaign agendas (‘Capitalizing on Contention’). 17Petrocik, ‘Issue Ownership in Presidential Elections’.18The Lexis-Nexis search engine – which searches mainly polls conducted by the major television networks and newspapers – turned up 122 instances of questions evaluating issue ownership, excluding a general question about which party could be trusted ‘to do a better job in coping with the main problems the nation faces over the next few years’. These questions covered twenty different issues. (The precise wording of these questions is available from the author.) The Origins of Campaign Agendas413 only which, if any, party appears to ‘own’ each issue, but also whether ownership is a stable attribute of parties. In Figure 1, I present graphs for each of twelve issues that were also central in candidate advertising in 1998. 19 To measure ownership, I simply subtracted the percentage of respondents favouring the Republicans from the percentage favouring the Democrats. Thus, positive numbers indicate Democratic ownership and negative numbers indicate Republican ownership. Several of the results in Figure 1 conform to conventional wisdom and to Petrocik’s ndings. For example, from 1990 to 1998 Democrats had approximately a twenty-point advantage on social programmes, including Social Security and Medicare. It also appears that during this period Democrats ‘owned’ issues like education, the environment and health care. 20 Perhaps most interesting in these graphs is that a number of traditionally ‘Republican’ issues manifest little or no Republican advantage. The Democratic party had a slight advantage when it comes to ‘handling the economy’. There was near-parity Fig. 1. Trends in issue ownership 19About 80 per cent of the House and Senate candidate advertisements in 1998 mentioned at least one of these issues. 20It is interesting that during the Clinton years Democrats’ advantage on health care declined. A February 1994 ABC/Washington Postpoll found that 58 per cent of respondents trusted the Democrats to ‘do a better job on providing affordable health care’, whereas 22 per cent trusted the Republicans. But by October 1994, 46 per cent trusted Democrats and 29 per cent trusted Republicans. The ‘gap’ between the two parties shrank from 36 points to 17 points. The apparent reason – the failure of the Clinton health plan – demonstrates another of Petrocik’s points: ownership depends on party performance and can be a short-term ‘lease’ (‘Issue Ownership in Presidential Elections’, p. 827). This trend also nicely illustrates Geer’s observation that ‘the reputations of the parties are not xed’ (‘Campaigns, Party Competition, and Political Advertising’, p. 189). 414 SIDES between Republicans and Democrats in terms of ‘reducing the federal budget de cit’. 21 These ndings may not be surprising in that during the Clinton administration the economy was healthy and the de cit shrank considerably. Similarly, by 1998 there was no Republican advantage on crime. 22 There were also Democratic gains in ‘holding taxes down’, ‘maintaining a strong national defence’ and ‘handling foreign affairs’, though if one averages responses over this entire period, there was still a Republican advantage on these issues. Several hypotheses can be derived from these results. In its strongest formulation, the theory of issue ownership would predict that in 1998 Democratic candidates should focus exclusively on these issues: education, the environment, health care, Social Security and Medicare. Republican candidates should ignore these issues and focus instead on national defence, foreign affairs and taxes. The parties will diverge completely in their campaign agendas and there will be no instances of ‘issue trespassing’. 23Issues on which neither party has an apparent advantage, such as the budget and crime, should never appear in advertising, as predicted by Riker’s Dispersion Principle. In its weaker formulation, the theory of issue ownership would not necessarily predict complete divergence. Instead, one would expect a probabilistic, rather than deterministic, relationship between party ownership and campaign emphasis. This hypothesis may correspond better to the nature of ‘ownership’ suggested by the public opinion data, in which any party’s advantage is only relative (i.e., more voters trust that party than the other) rather than absolute (i.e., all voters trust that party). If this probabilistic relationship emerges empirically, then Democrats should advertise more on ‘Democratic’ issues than Republicans do, and conversely for Republicans. Consequently, there may be instances of issue trespassing, though these should still be rare relative to advertising on ‘owned’ issues. In addition, issues on which neither party has a clear advantage should be less prominent in campaign discourse relative to other issues, rather than completely absent. These probabilistic hypotheses are more in spirit with extant empirical evidence about candidate strategy. 24 The idea of ‘riding the wave’ generates very different predictions. According to this idea, both parties should emphasize issues that the public considers important priorities. What was the public’s agenda at this time? Early in the 1998 campaign, a July Harris Poll asked, ‘What do you think are the two most important issues for the government to address?’ The top ve were education (16 per cent mentioned this issue), health care (15 per cent), taxes (13 per cent), crime (12 per cent) and Social Security (10 per cent). Thus, if candidates were to ‘ride the wave’ in this election, both parties would focus primarily on these issues. Given that this list includes several issues that Democrats appear to ‘own’, there should be a notable degree of Republican trespassing, for example, on education, health care and Social Security. And given that for most of the 1990s the Republicans appeared to ‘own’ taxes, there should be some Democratic trespassing on this issue. Moreover, although neither party had a clear advantage on crime for most of the 1990s (see Figure 1), both 21For further discussion of party ownership and the federal budget, see Barry C. Burden and Joseph Neal Rice Sanberg, ‘Budget Rhetoric in Presidential Campaigns from 1952 to 2000’,Political Behavior, 25 (2003), 97–118. 22For an account of how Democrats mitigated Republicans’ traditional advantage on crime, see Holian, ‘He’s Stealing My Issues!’ 23This hypothesis also follows from the formal model in Simon,The Winning Message, which predicts that candidates should never discuss an issue on which they lack an advantage. 24See, e.g., Riker,The Strategy of Rhetoric, and Petrocik, ‘Issue Ownership in Presidential Elections’. The Origins of Campaign Agendas415 parties should advertise on this as well, despite what the Dispersion Principle suggests. In sum, the theories of issue ownership and ‘riding the wave’ generate clear and in many cases opposite predictions to test empirically. Other Sources of Campaign Agendas It is important to note that campaign agendas may derive from factors besides party ownership. 25One is a candidate’s individual record. Sellers argues that ‘candidates tend to emphasize issues on which they have built a record that appears favourable to voters’. 26 Credibility is again the relevant mechanism. Candidates will typically have experience in political of ce (such as bills written, co-sponsored or passed) or in their non-political careers (for example as educators, lawyers, soldiers, etc.) that is relevant to the campaign, and it is likely that voters will nd candidate appeals more compelling when backed up by evidence of a record. Sellers’s examination of Senate candidates shows that they tend to emphasize positive messages about themselves when they have an extensive background on an issue. Other factors that may in uence campaign agendas are the race and gender of candidates. Scholarship about female candidates suggests that their campaign agendas centre more on social issues and social programmes, and less on foreign affairs or economics issues. 27 Although similar research into racial representation has produced more mixed results, it seems reasonable to investigate whether white candidates and minority candidates emphasize different issues. 28 In particular, the expectation is that minority candidates, African-Americans in particular, will focus on issues salient to minority voters, such as social spending. 29 As candidates structure their agendas, they may also take account of their constituencies, choosing to emphasize issues that are particularly salient in their respective districts or states. How can we measure the preferences of constituencies? One possibility is to use demographic attributes as a proxy. Given that issues like education, Social Security and Medicare gure so prominently in the 1998 campaign (see below), we might expect candidate agendas to depend on two attributes in particular: rst, the number of constituents who have school-age children, which should be positively associated with the candidates’ attention to education; and, secondly, the number aged 65 and over, which should be 25A candidate’s decisions about issue emphasis might depend on what his or her opponent is emphasizing (see Hammond and Humes, ‘The Spatial Model and Elections’). Because the data examined here include numerous races where both candidates advertised, an investigation of strategic interaction is possible. However, doing so entails analysing the candidates’ advertising over-time, and such an analysis is beyond the scope of this article. An initial cross-sectional analysis reveals only a few signi cant associations between opposing candidates’ advertising, suggesting that strategic interaction is not a major factor in the races examined here. 26Patrick Sellers, ‘Strategy and Background in Congressional Campaigns,’American Political Science Review, 92 (1998), 159–71, p. 159. 27Kim Fridkin Kahn,The Political Consequences of Being a Woman(New York: Columbia University Press, 1996). Other research on women’s behaviour as legislators nds that their agendas frequently differ from men’s – see Michele Swers,The Difference Women Make(Chicago: Chicago University Press, 2002). 28See David Canon,Race, Redistricting, and Representation: The Unintended Consequences of Black Majority Districts(Chicago: Chicago University Press, 1999); Carol Swain,Black Faces, Black Interests: The Representation of African Americans in Congress(Cambridge, Mass.: Harvard University Press, 1993); and Katherine Tate,Black Faces in the Mirror: African Americans and Their Representatives in the US Congress (Princeton, N.J.: Princeton University Press, 2003). 29See Donald R. Kinder and Lynn M. Sanders,Divided by Color: Racial Politics and Democratic Ideals (Chicago: University of Chicago Press, 1996). 416 SIDES positively associated with attention to issues important to seniors, including Social Security, Medicare and health care. Previous research has demonstrated that these sorts of ‘demand-side’ characteristics in uence legislative behaviour. 30 Though conventional wisdom about campaign strategy routinely emphasizes such characteristics – in that candidates are seen as tailoring their remarks to the concerns of their audience – there has been little investigation of these characteristics in the academic literature on campaign strategy. Another aspect of constituency is its ideological disposition. Spiliotes and Vavreck argue that ‘constituency leaning’ drives candidates in liberal environments to emphasize ‘bread-and-butter’ liberal issues and candidates in conservative environments to emphasize issues that appeal similarly to conservatives. For example, they nd that candidates running in more conservative districts or states spend more time discussing the economy and tax cuts. 31However, it is dif cult to specifya prioriwhich issues are uniquely attractive to voters of different ideological persuasions. Even a stereotypically ‘liberal’ issue, such as the environment, could be emphasized successfully in conservative constituencies if the message – e.g., the risk environmental regulations pose to economic growth – is amenable to conservatives. Thus, the demographic factors enumerated above should have a more important impact than ideologyper se. Besides characteristics of candidates and constituencies, the circumstances of particular races may in uence campaign communication. Perhaps the most important attribute of a race is its competitiveness. Kahn and Kenney nd that candidates are more likely to mention issues in their campaign appeals when the race is closely contested. 32They argue that a crucial mechanism behind the effect of competitiveness is uncertainty. If candidates are uncertain of the outcome then they need both to mobilize their supporters and to appeal to swing voters. Furthermore, in competitive races, candidates may be besieged by larger numbers of interested and engaged activists. Candidates will feel the need to emphasize a wider array of issues to win the support of these activists. Finally, competitive campaigns will induce candidates to focus on issues because they must respond to their opponents’ claims. None of these incentives predicts that candidates will focus on any particular issue. Instead, these incentives will simply combine to produce an issue agenda that is broader, in that more diverse issues are discussed as candidates try to appeal to different types of voters, and that is also ‘deeper’, in that candidates will devote more time to discussing issues. ADVERTISING IN THE 1998 CONGRESSIONAL ELECTIONS To investigate the effect of these factors on candidate strategy, I draw on an extensive database of candidate advertisements from the 1998 House and Senate elections. Television advertising is not the only source of information about the decisions candidates 30E. Scott Adler,Why Congressional Reforms Fail: Reelection and the House Committee System(Chicago: University of Chicago Press, 2002); and E. Scott Adler and John Lapinski, ‘Demand-side Theory and Congressional Committee Composition: A Constituency Characteristics Approach’,American Journal of Political Science, 41 (1997), 895–918. For an analysis of legislators’ response to seniors citizens, see Andrea Louise Campbell,How Policies Make Citizens: Senior Political Activism and the American Welfare State(Princeton, N.J.: Princeton University Press, 2003). 31Spiliotes and Vavreck, ‘Campaign Advertising’.32Kahn and Kenney,The Spectacle of US Senate Campaigns. The Origins of Campaign Agendas417 make. One could, for example, examine newspaper coverage of campaigns. 33 However, newspapers and other media outlets employ their own ‘ lters’ on campaign speech and may therefore over- or under-emphasize certain aspects of the candidates’ agendas. Advertising is preferable because it is unmediated. Secondly, advertising should reveal a candidate’s priorities because it is costly; candidates presumably cannot afford to expend resources discussing issues they do not care about. Finally, advertising provides a very precise measure of emphasis because one can examine the content of advertising together with the volume of advertising to determine which advertisements discussed which issues and how often these advertisements were broadcast. The advertising data were originally collected by the Campaign Media Analysis Group (CMAG), a private company based in Arlington, Virginia. Their technology tracks satellite transmissions from the major national networks and recognizes the digital ‘ ngerprint’ of various television programmes and advertisements. With that ngerprint, CMAG then records when and where each advertisement ran and who its sponsor was. 34After the 1998 elections, the Brennan Centre for Justice at New York University purchased and compiled the House and Senate advertisements into a publicly available database. CMAG did not capture every single advertisement because it tracked only those that aired in the top seventy- ve media markets. However, the vast majority of the US population (about 80 per cent) lives in one of those markets. Thus, this dataset represents the closest thing to a universe of advertisements that exists for these races. In 1998, there were thirty-four contested Senate races. The CMAG data contain advertising from twenty-nine of them. 35 Of the 393 contested House races, CMAG identi ed advertising in 134 (the CMAG data contain no advertisements for any of the forty-two uncontested races). In 103 races (eighty-three House races and twenty Senate races), there was advertising from both of the candidates. 36For this analysis, I eliminated advertisements sponsored by the two parties and by interest groups and those that occurred before the primary date in each state. 37 This was done for theoretical reasons. I am interested in explaining the decision making of candidates, not that of parties or interest groups, which may have different motives for advertising. Secondly, I am concerned with candidate strategy in general elections, not in primary elections, where the dynamics are 33See Brasher, ‘Capitalizing on Contention’; Simon, ‘The Winning Message’; and Sigelman and Buell, ‘Avoidance or Engagement?’ 34Freedman and Goldstein (‘Measuring Media Exposure and the Effects of Negative Campaign Ads’,American Journal of Political Science, 43 (1999), 1189–208) were the rst to demonstrate the power of this data source for political science, using it to argue that negative advertising does not depress turnout. 35CMAG did not capture any advertisements for Senate races in ve states – Alaska, Hawaii, North Dakota, South Dakota and Vermont – because there is no media market amongst the top seventy- ve in any of these states. 36It may seem surprising that a Senate candidate would not air advertisements. Eight of the nine Senate races with only one candidate advertising featured incumbents, such as John McCain, who won by large margins – an average of 35 points, as compared to 15 points in the other contested races. The remaining race was between Republican Michael Crapo and Democrat Bill Mauk in Idaho. Crapo won 71 per cent of the two-party vote and spent over $1.5 million, while Mauk spent only about $240,000, according to the Federal Elections Commission. Thus the Senate candidates who failed to advertise were simply not competitive in comparison to their well-established and better-funded opposition. 37In the general election period, about 77 per cent of all advertising airings were from candidates. There were, of course, races in which advertising from parties and interest groups constituted a signi cant fraction of all the advertising, and my focus on candidates should not be construed as an attempt to downplay the signi cance of party or issue advertising. It is also possible that candidates might structure their agendas based on what they anticipate that other groups will do. This possibility could be investigated in future research. 418 SIDES obviously very different. 38 The dataset I employ contains 1,258 unique advertisements aired by the major-party candidates after the primary. These advertisements aired a combined total of 188,256 times during the general election period. The Brennan Centre rst coded these advertisements for a wide range of attributes related to content and tone. I obtained original copies of nearly all of these advertisements and augmented the original coding in two ways. First, I re-coded the issues that the advertisements mentioned, using the same categories as the Brennan Centre (see Appendix A). While the original dataset coded up to four separate issues in each advertisement, I coded as many as were mentioned (a maximum of eight in these data, though the vast majority of advertisements mentioned no more than four). Secondly, within a given issue I also coded the speci c positions the candidate took. This two-fold scheme provides information necessary to investigate both what issues candidates talk about and how exactly they talk about these issues. The resulting dataset has several assets. First, it includes a very large number of advertisements. Most other published analyses of advertisements rely on those publicly archived or on those that a source such as theNational Journalgathers. There is, however, no guarantee that either of these other sources actually gathers a representative sample of advertisements. 39The CMAG data, though certainly not a census, come close. Secondly, the CMAG data track not only which advertisements were aired but also when and how often they aired. Knowing how often a candidate aired an advertisement is crucial. As Prior shows, to characterize the information environment accurately, one must account for advertising volume; the characteristics of the advertisements produced may differ signi cantly from those of the advertisements that were aired. 40 ANALYSIS What were the most prominent issues in the 1998 election? Table 1 presents the total volume of advertising – that is, the number of airings – that mentioned an issue. 41The rst 38Spiliotes and Vavreck (‘Campaign Advertising’) analyse a dataset that combines primary and general election advertisements. However, I feel that these two sets of advertisements are best considered separately. If candidates do move towards the ‘extremes’ during primaries and towards the ‘centre’ during the general election, then their issue agendas could change accordingly. For example, one might suspect that in the primary candidates will stick to the issues their party owns, since these will appeal to more ideological primary voters, but in the general election candidates will draw on a more varied set of issues, including even those ‘owned’ by the other party. This is to say, the effect of issue ownership may be stronger in primaries than in the general election. Because Louisiana does not have separate primary and general elections, I included only those races in Louisiana which mimicked general election races in that, rst, they either were contested by only one major-party candidate, or, secondly, they featured a race contested by both parties where a Democratic candidate and a Republican candidate were the top vote-getters (i.e., where the main lines of competition were inter-party rather than intra-party). In this latter case, I do not examine any other Democratic or Republican candidates. This has little practical consequence, as the front-runners in each party typically aired the vast majority of advertisements in that race. 39TheNational Journal’s data utilized by Spiliotes and Vavreck (‘Campaign Advertising’) came only from competitive races. Thus those data cannot illuminate the effect of competitiveness itself. 40Markus Prior, ‘Weighted Content Analysis of Political Advertisements’,Political Communication,18 (2001), 335–45. 41The issues listed in Table 1 typically correspond to a single code in Appendix A. In a few cases, I combined several codes under a single issue heading. ‘Crime’ includes the codes for crime, drugs, the death penalty and any other reference to law and order. ‘Defence’ includes all codes under ‘defence’ and ‘missile defence’. Any advertisements coded as containing themes related to ‘other defence, foreign policy issue’ were investigated further; based on their content, they were coded as ‘defence.’ ‘Foreign affairs’ therefore includes ‘foreign policy’, The Origins of Campaign Agendas419 TABLE 1Advertising Volume on Various Issues, by Party All candidates (N 268) Republicans (N 139) Democrats (N 129) Taxes 59,318 Taxes 42,883 Education 30,383 Education 54,314 Social Security 28,662 Social Security 21,717 Social Security 50,379 Education 23,931 Health care 18,411 Health care 31,375 Health care 12,964 Taxes 16,435 Medicare 24,999 Crime 11,387 Medicare 13,964 Crime 24,780 Medicare 11,035 Crime 13,393 De cit etc. 20,042 De cit etc. 10,623 De cit etc. 9,419 Other 11,423 Welfare 8,879 Environment 7,484 Environment 10,879 Gov. spending 7,010 Gun control 5,594 Welfare 9,790 Other 6,859 Other 4,564 Gov. spending 8,618 Jobs 4,449 Jobs 3,803 Jobs 8,252 Clinton 3,575 Child care 3,659 Gun control 5,945 Environment 3,395 Poverty 2,976 Clinton 5,549 Other social 3,139 Abortion 2,834 Abortion 5,393 Veterans 2,880 Veterans 2,413 Veterans 5,293 Abortion 2,559 Other economic 2,200 Child care 4,998 Other economic 2,032 No issue mention 2,029 Poverty 4,735 Defence 1,928 Clinton 1,974 Other social 4,397 Poverty 1,759 Roads etc. 1,791 Other economic 4,232 Roads etc. 1,649 Gov. spending 1,608 No issue mention 3,591 No issue mention 1,562 Other social 1,258 Roads etc. 3,440 Child care 1,339 Campaign nance 1,059 Defence 2,216 Gun control 351 Welfare 911 Campaign nance 1,149 Foreign affairs 140 Defence 288 Civil rights 223 Campaign nance 90 Civil rights 179 Foreign affairs 140 Civil rights 44 Immigration 77 Immigration 77 Immigration 0 Foreign affairs 0 Note:Table entries are the number of advertisement airings. column of Table 1 presents the number of advertisements aired by all House and Senate candidates in the dataset. The most prominent issues are the mainstays of domestic politics in the United States: taxes, education, Social Security, health care, Medicare and crime. Advertising about the de cit, surplus and budget was also prominent, mostly because many advertisements referred to the successfully balanced budget and to using the surplus to shore up Social Security. The second and third columns present this same distribution separately for Republican and Democratic candidates. This breakdown reveals some differences in priorities between the two parties. For example, taxes were clearly the dominant issue in Republican advertising: 42,883 airings were devoted at least in part to this issue. The next most popular issue, Social Security, netted about two-thirds as much attention (28,662 airings). (F’note continued)‘Bosnia’ and ‘China’. ‘Clinton’ includes all references to Clinton, Ken Starr, Whitewater, impeachment or Paula Jones. ‘Child care’ includes explicit references to child care as well as any other reference to children. ‘Other economic’ includes any economic issue not given its own category. ‘Other social’ includes any social issue not given its own category. Table 1 does not include advertising related to ‘personal’ themes, such as a candidate’s background, experience, values etc. Most extant theories of campaign agendas, such as party ownership, speak only to advertising about policy-related issues. The vast majority (82 per cent) of advertisements aired contained at least one reference to a policy-related issue. 420 SIDES Education, health care, crime and Medicare were also prominent. The Democratic agenda prioritized education, Social Security and health care. But perhaps most striking about the results in Table 1 is that, while differences in the two parties’ priorities did exist, their issue agendas were in general quite similar. Candidates of both parties devoted most of their attention to taxes, Social Security, education, health care, crime and Medicare. These issues were also those that were most prominent on the public’s agenda, suggesting that the parties were ‘riding the wave’ and focusing on this same set of salient issues rather than focusing only on issues that their party ‘owns’, as the Dominance Principle and ownership theory suggest. Consequently, there is a notable degree of issue trespassing evident in this table, particularly among Republicans, who talked frequently about such ‘Democratic’ issues as Social Security, education and health care. Thus, the ‘strong’ version of the ownership hypothesis – which predicts a complete focus on ‘owned’ issues – does not hold true in these races. The results in Table 1 do not imply, however, that party ownership had no in uence on candidate strategy and, in particular, whether there were signi cant differences between Democrats and Republicans, as the ‘weak’ version of the ownership hypothesis implies. To determine more thoroughly the role of party ownership, as well as other factors, I investigate the decision making of the 268 candidates in this dataset, treating the candidate as the unit of analysis. 42To gauge issue emphasis, I constructed a measure of what I will call ‘advertising time’. This measure takes into account the number of issues any individual advertisement mentions and the number of times that advertisement aired to produce a proportion ranging between 0 and 1. This proportion captures the amount of time a candidate devoted to a speci c issue. Across all issues, including both policy-related issues and issues related to the candidate’s experience or character, this measure sums to 1. (Details of this measure are in Appendix B.) Having created this measure, the question to be asked is: do candidates spend more ‘advertising time’ on issues that their party owns? Table 2 presents the ndings for eleven of issues included in Figure 1. 43 For each issue, I present the extent of ‘ownership advantage’ averaged over the 1990–98 period. Six issues manifested a Democratic advantage, three a Republican advantage and two no clear advantage for either party. I present two quantities that capture emphasis: the percentage of Democratic and Republican candidates who aired at least some advertising on each issue; and the average percentage of Democratic and Republican ‘advertising time’ devoted to each issue (the quantity described above). The cells shaded grey indicate a statistically signi cant difference between the parties (p 0.05 or better). If candidates base their agendas on issue ownership, signi cant differences should predominate. 42These data obviously offer other modes of empirical purchase. One could, for example, make the individual advertisement the unit of analysis. However, given that the goal of this analysis is to explore candidate strategy, candidates seem the appropriate place to start. See Spiliotes and Vavreck (‘Campaign Advertising’) for a similar approach. 43I do not include ‘the economy’ in this analysis. The relevant survey question refers generally to stewardship of the economy. However, in the 1998 advertising data, it is dif cult to identify appeals that stressed the general idea of improving the economy. Typically, this goal is referenced only with regard to speci c issues such as taxes, government spending, the minimum wage, jobs, poverty and trade, which each have different expectations vis-a` -visownership theory. The catch-all ‘other economic reference’ category (see Appendix A) is comprised of only a handful of advertisements whose content (e.g., local economic development, protecting seniors from bankruptcy) does not centre on any one economic theme. The Origins of Campaign Agendas421 TABLE 2Issue Emphasis and Party Ownership Average percentage Percentage of candidates of advertising advertising on issue time about issue Ownership advantage Republicans Democrats Republicans Democrats Democratic advantage Education 16.6 58% 72% 9.8% 15.6% Environment 17.8 13 19 2.3 2.9 Health care 22.3 30 57 4.5 9.9 Jobs 9.8 18 16 3.5 2.9 Medicare 18.5 30 40 4.3 5.6 Social Security 17.5 64 69 11.6 12.2 No advantage Crime 1.1 36 37 5.4 5.8 De cit, budget, surplus 0.1 36 36 4.8 4.0 Republican advantage Defence 32.0 10 4 0.8 0.7 Foreign Affairs 15.3 2 0 0.5 0.0 Taxes 5.9 76 42 16.4 6.6 Note:‘Ownership advantage’ is equal to the percentage who trust the Democrat to handle each issue, minus the percentage who trust the Republican, averaged over 1990–98. Shaded cells indicate that the difference between Democrats and Republicans is statistically signi cant at p 0.05. Republicans,N 139, Democrats,N 129. Table 2 shows rst that the two issues which neither party ‘owned’, crime and the de cit, manifested no partisan differences, as expected. However, the mere existence of substantial amounts of advertising about these issues – in that slightly more than a third of candidates devoted at least some time to these issues – is evidence against the Dispersion Principle, which would predict little or no attention to issues that manifest no real advantage for either party. In regards to the other issues, on which the parties are hypothesized to differ, few signi cant differences emerge. Of the six ‘Democratic’ issues, Democratic emphasis outpaces Republican emphasis in only two cases, education and health care. For example, 72 per cent of Democratic candidates aired advertisements mentioning education, whereas 58 per cent of Republicans did so. The measure of advertising time generates a similar difference. However, Republicans were as likely as Democrats to emphasize issues such as the environment, jobs, Medicare and Social Security, despite the Republican party’s apparent disadvantage. On the three ‘Republican’ issues, only taxes earned different amounts of emphasis from Republicans and Democrats, with Republicans more likely to emphasize this issue. While more Republicans than Democrats aired advertisements on defence, the corresponding difference in the amount of advertising time is not signi cant. Even though Republicans seemingly held an advantage on defence and foreign affairs (although their advantage on the latter had weakened during the 1990s), very few advertisements focused on these issues. 422 SIDES Thus, even the weak version of the ownership hypothesis is not con rmed here. 44There are occasional relative differences between Democrats and Republicans, but more often than not the parties seem willing to ‘trespass’ on the other party’s territory. This is particularly true for Republicans, who were competing at a time when the public’s agenda centred on numerous ‘Democratic’ issues. Indeed, if one sums up the proportion of advertising time devoted to the other party’s issues, Republicans devoted twice as much advertising time to ‘Democratic’ issues (36 per cent) as to ‘Republican’ issues (18 per cent). Thus Republicans and Democrats do not pursue a ‘pure strategy’ akin to the Dominance Principle, whereby an attribute like issue ownership wholly guides their campaign agendas. They instead pursue what appears to be a ‘mixed strategy’, drawing on a varied repertoire of issues in constructing their campaign appeals. Issue ownership appears in uential in a few cases but its impact is neither robust nor deterministic. A Model of Issue Emphasis The next empirical task is to specify not only the impact of party but also that of record, constituency and competitiveness, as well as candidate race and gender. To do so, I estimated multivariate models of ‘advertising time’, with candidate as the unit of analysis. 45The measure of party ownership is a dichotomy, where 1 signi es a Democratic candidate. I coded candidate records in a manner similar to Sellers, drawing on the description of each race found in theAlmanac of American Politics. 46 I constructed a dichotomous indicator of whether the candidate had a record on a given issue, based on theAlmanac’s discussion of the candidate’s accomplishments. 47This strategy is undoubtedly imperfect, and it is very likely that it underestimates the accomplishments of candidates, simply because theAlmanacis not intended to be comprehensive in this respect. But it is nevertheless a reasonable rst cut at measuring this concept. The expectation is that candidates will devote more advertising time to issues on which they have a record. To measure constituency preferences, I rely rst on Clinton’s percentage of the two-party vote in 1996 as a proxy of each district or state’s ideological leaning. This measure is inexact but it is readily available for all states and congressional districts. I also employ two other measures of constituency characteristics: the percentage over the age 44In a footnote, Spiliotes and Vavreck (‘Campaign Advertising’, p. 254) say that they nd effects of issue ownership in forty issues (they focus on ve in the text). The results in this analysis do not suggest comparable robustness. 45Given that the dependent variable is a proportion, one possible strategy is to employ a simple logit transformation to place bounds of 0 and 1 on predicted values. However, as pointed out by one of the referees of this article, this transformation has the unfortunate consequence of accentuating differences at low levels of emphasis – e.g., the difference between proportions of 0.01 and 0.02 is 0.70 logits, while the difference between proportions of 0.20 and 0.21 is 0.06 logits. Thus, I do not transform the dependent variables in any fashion. Secondly, because these data contain opposing candidates within the same race as separate observations, these models were estimated with robust standard errors that allowed for this sort of clustering. Finally, the results presented below are also robust in seemingly unrelated regression models that assume some degree of correlation among the errors terms of these models. 46Sellers, ‘Strategy and Background in Congressional Campaigns’; Michael Barone and Grant Uji sa, eds, Almanac of American Politics(New York: E. P. Dutton, 2000). 47Initially, following Sellers, I compiled a trichotomous indicator that also captured whether the candidate’s record was relatively meagre or extensive. There were, in total, a paucity of candidates with only a meagre record and so I combined these two categories. The Origins of Campaign Agendas423 of 65 and the percentage enrolled in a public primary or secondary school. 48I expect that candidates who represent districts with larger proportions of senior citizens will tend to emphasize issues such as Medicare, Social Security and health care, while de-emphasizing issues of less relevance to seniors, in particular education. By contrast, candidates who represent districts or states with larger proportions of students (and therefore parents) should emphasize education more strongly. To measure competitiveness, I drew on theCook Political Report’s rating of each race and created a dichotomous measure, coding races as safe or unsafe seats. The primary expectation is that candidates in competitive races will be more likely to mention particular issues in their advertising, regardless of the nature of the issues or the candidates. The models also include a dummy variable for candidate gender (where 1 indicates a female) and for candidate race (where 1 indicates a non-white candidate). I also include a dummy variable for whether the race was a House or Senate race (coded 1 for Senate). 49Finally, I include incumbent status. Riker argues that it is the ‘losers’ in politics who often inject new issues into political debates. 50This may imply that challengers will tend to emphasize issues more than do incumbents. I include a dummy variable for incumbents and a dummy for open-seat candidates, with challengers the excluded category. 51 Table 3 presents models of ‘advertising time’ devoted to the ten issues presented in Table 2, which include the most salient issues in the 1998 election. 52 These models tend to recapitulate the bivariate results about party ownership presented in Table 2. Party ownership is associated with emphasis of education, health care and taxes, but no other issue. The magnitude of its effect in these cases is roughly equivalent to half the standard deviation of the dependent variable. Candidate record is positively and signi cantly associated with emphasis of three issues (crime, education and health care), though, somewhat oddly, negatively associated with emphasis of the de cit and Social Security. It is also close to signi cance in the case of the environment (p 0.14) and Medicare (p 0.10). The magnitude of its effect is variable, ranging in absolute value from less than 1 percentage point (Social Security) to over 8 points (crime). Record is thus in uential but by no means determinative. Candidates do not exclusively advertise on issues where they have a record of accomplishment. 48The measure of Clinton’s vote is the number reported in theAlmanac of American Politics. The measure of the age distribution comes from the US Census Bureau (see Table 33 of theStatistical Abstract of the United States 1999). In this sample the measure ranges from 5 to 19 per cent, with a mean of 12 per cent. The measure of the number of students enrolled comes from the National Center for Education Statistics. The measure is from the fall of 1997 – as close to the election of 1998 as could be found on the NCES website – and ranges from 8 to 23 per cent, with a mean of 16 per cent. 49Spiliotes and Vavreck (‘Campaign Advertising’) nd less attention to issues in House races.50Riker, ‘Political Theory’.51A few words about the distribution of these variables are in order. ‘Unsafe seats’ combines two of Cook’s classi cations, ‘leaning’ and ‘toss-up’. Empirically, these two categories were not very distinct: the average margin of victory in leaning races was 12 points and in toss-up races 8 points. By contrast, the average margin of victory in safe seats was 27 points. By this measure 167 candidates ran in ‘safe’ races and 101 ran in ‘unsafe’ races. In terms of gender and race, there were 42 female candidates out of 268 in this dataset, and there were 22 non-white candidates. Finally, 133 of these candidates were incumbents, 46 were open-seat candidates, and 89 were challengers. 52I do not model advertisements related to foreign affairs since only 140 of them aired in 1998. 424 SIDES TABLE 3Models of Issue Emphasis Health Social Crime Defence De cit Educ. Env. care Jobs Medicare Security Taxes Democrat 0.003 0.001 0.005 0.051** 0.010 0.056** 0.005 0.007 0.001 0.097** (0.013) (0.007) (0.008) (0.016) (0.011) (0.014) (0.014) (0.013) (0.004) (0.015) Record 0.084** 0.028 0.031* 0.054 # 0.026 0.067* 0.038 0.078 0.006 # 0.026 (0.030) (0.021) (0.015) (0.028) (0.018) (0.027) (0.057) (0.048) (0.004) (0.025) Female 0.013 0.002 0.007 0.002 0.001 0.010 0.004 0.003 0.001 0.001 (0.012) (0.005) (0.013) (0.025) (0.015) (0.023) (0.012) (0.014) (0.006) (0.020) Non-white 0.021 0.007 0.003 0.020 0.005 0.022 0.007 0.007 0.008* 0.014 (0.021) (0.004) (0.016) (0.032) (0.014) (0.026) (0.014) (0.019) (0.004) (0.028) Clinton % ’96 0.001 0.001 0.001* 0.003 0.002 # 0.001 0.001 0.001 0.001 0.001 (0.001) (0.001) (0.001) (0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) % Age 65 0.012** 0.003 0.004 # 0.001 (0.004) (0.003) (0.002) (0.001) % Student 0.005 (0.004) Unsafe seat 0.031* 0.008 0.004 0.001 0.013 0.011 0.031** 0.021 0.005 0.011 (0.013) (0.006) (0.009) (0.017) (0.012) (0.013) (0.011) (0.015) (0.006) (0.014) Senate race 0.027* 0.003 0.007 0.030 # 0.002 0.015 0.008 0.001 0.005 0.004 (0.012) (0.004) (0.011) (0.016) (0.011) (0.020) (0.010) (0.014) (0.007) (0.016) Incumbent 0.008 0.008 0.043** 0.001 0.008 0.023 0.003 0.014 0.008 0.018 (0.016) (0.005) (0.009) (0.018) (0.012) (0.017) (0.019) (0.014) (0.006) (0.017) Open seat 0.009 0.008 0.013 0.044 0.025 # 0.038* 0.006 0.004 0.011 0.028 (0.019) (0.012) (0.012) (0.029) (0.015) (0.017) (0.016) (0.018) (0.007) (0.021) Constant 0.022 0.051 # 0.092** 0.165 0.108 # 0.127 # 0.056 0.100 # 0.020 0.149** (0.044) (0.030) (0.036) (0.101) (0.057) (0.069) (0.040) (0.053) (0.016) (0.053) R 2 0.08 0.05 0.08 0.14 0.04 0.12 0.02 0.06 0.03 0.17 Note:Table entries are unstandardized OLS regression coef cients, with robust standard errors in parentheses. The dependent variable is the measure of ‘advertising time’ described in the text.N 268. #p 0.10; *p 0.05; **p 0.01 (two-tailed). The Origins of Campaign Agendas425 Other attributes of the candidate – gender and race – have very little effect. Female candidates are not signi cantly different from male candidates on these issues. Race has a signi cant effect in only one case: non-white candidates are less likely to emphasize Social Security. 53 In terms of constituency characteristics, the ideological disposition of these districts and states matters only occasionally, and not always in hypothesized ways. In more Democratic areas, candidates were less likely to emphasize the de cit/surplus and health care – even though health care is a ‘Democratic’ issue and lowering the de cit was a signature achievement of the Clinton administration. The age distribution within each district or state appears more relevant. Candidates who represented higher proportions of senior citizens emphasized Medicare more strongly, while de-emphasizing education. This is to say, they try harder to speak to at least some concerns of senior citizens. Student enrolment is not signi cantly related to education advertising. Competitiveness is also largely unrelated to the volume of advertising time on these issues. 54 It is signi cantly and positively related to emphasis of crime and negatively related to emphasis of jobs. Otherwise, it has no real impact on issue emphasis. This is surprising given the expectation that competitiveness would lead to more issue discussion. However, further analysis shows that the true effect of competitiveness is only to broaden the issue discussion, not to deepen it. While candidates in competitive campaigns are more likely to mention a variety of issues in their advertising, they do not necessarily devote more of their advertising time to issues. If one sums up the number of the eleven issues modelled in Table 3 that are mentioned by each candidate in their advertising, the median number of issues mentioned by candidates in uncompetitive races is three, while the median in competitive races is ve, suggesting that the issue agenda in competitive races is indeed more extensive. However, the total amount of advertising time devoted to these issues is not signi cantly higher in competitive races (the average total is about 53 per cent in uncompetitive races and 55 per cent in competitive races). The variety of pressures that candidates in these races face appears to encourage them to discuss a larger number of issues, but not to do so more frequently. Put differently, the ‘issue-ness’ of competitive campaigns is broader but not deeper. 55 Finally, there is no conclusive evidence that Senate and House candidates differ signi cantly, or that incumbent and open-seat candidates differ from challengers. In terms of advertising on these eleven issues as a whole, challengers tend to focus on fewer issues (a median of three, versus four for open-seat candidates and four for incumbents) and devote slightly less advertising time to issues as a whole (51 per cent versus 54 per cent for open-seat candidates and 55 per cent for incumbents). Thus, challengers’ campaigns conform in part to Riker’s expectations: they pursue fewer issues but do not necessarily devote more time to issuesper se. 53Dummy variables for black candidates and for black or Latino candidates generated similar ndings.54Substituting the eventual margin of victory as a measure of competitiveness generates substantively similar results. 55InThe Spectacle of US Senate Campaigns(Table 3.3, p. 63), Kahn and Kenney analyse only whether an advertisement mentions an issue, and, given their data, cannot measure the amount of time the candidates spent discussing issues. 426 SIDES Taken as a whole, these models produce several conclusions about campaign agendas. 56 First, it is striking that virtually none of the independent variables has a consistent relationship to issue emphasis. Campaign agendas do not depend systematically on party ownership, candidate record, candidate race or gender, the competitiveness of the race, or the ideological leaning of the district. There is some evidence that the demographic composition of a district, particularly the size of the senior population, may matter. Nevertheless, the amount of variation explained in these models, as represented by the r 2statistic, is quite low. This provides some further evidence of ‘riding the wave’. If factors like party, record and constituency are not dependably related to issue emphasis, then it appears that candidates tend to pursue similar agendas despite their diverse experiences, constituencies and so on. It is particularly striking that the effect of issue ownership is so anaemic. Though a great deal of literature suggests that the credibility that ownership provides is critical to candidate strategy, ownership’s effect in 1998 was limited to only a few issues, and even then, the probabilistic relationship captured by the coef cients in Table 3 conceals a large number of candidates who are apparently acting in an ostensibly irrational fashion. Many candidates violate the ‘strong’ prediction of a formal model like Simon’s – that candidates should never talk about an issue on which they do not possess an advantage over their opponent. 57Neither does candidate behaviour consistently conform to the weaker version of the ownership hypothesis, which expects only signi cant differences between Republicans and Democrats. Issue trespassing is a common strategy in this election. This nding is investigated further below. THE NATURE OF ‘ISSUE TRESPASSING’ Why do candidates out the dictates of issue ownership? I argue that candidates trespass byframingthe other party’s issues in ways favourable to them. 58 Framing is the act of emphasizing certain aspects or dimensions of an issue. The framing literature in the study 56I also tested for a variety of interaction effects in these models: between party and record; between party and the ideological orientation of the candidate’s constituency (see Spiliotes and Vavreck, ‘Campaign Advertising’); and between competitiveness and both party and candidate record. These results turned up few signi cant effects. I also investigated whether region of the country played any role by including a dummy variable for the South. It was signi cantly related to issue emphasis only once (candidates in the South gave less emphasis to the environment). I examined the role of seniority, coded as the number of Congresses in which member had served (if any). Its signi cant effects were few and inconsistent in sign: candidates with more seniority emphasized the de cit and health care less but emphasized crime more. This pattern of results has no obvious interpretation. Finally, I included a measure of candidate spending. It had no signi cant effects. In general, these alternative speci cations did not reveal any new systematic in uences on issue emphasis. 57Simon, ‘The Winning Message’, p. 64.58I also investigated whether factors like candidate record and competitiveness encourage trespassing. I estimated models identical to those in Table 3, except separately for Democrats and Republicans. Models of education, the environment, health care, jobs, Medicare and Social Security in particular revealed no factor consistently associated with Republican trespassing. Models of defence and taxes revealed no factor consistently associated with Democratic trespassing. I also investigated whether candidates trespass in reaction to their opponent’s advertising – e.g., if a Democratic candidate begins advertising on education, does his or her Republican opponent follow suit? Looking at the 103 races where there is advertising data for both candidates, I found little evidence that trespassing derives from ‘copycat’ behaviour. For example, 52 of 103 Republicans advertised about education, but only 19 of 52 did so after their Democratic opponent did. The others either advertised rst (29) or were the only candidate advertising on education (14). Similar ndings emerged for Democrats regarding taxes and for Republicans regarding health care, Medicare and Social Security. The Origins of Campaign Agendas427 of political communication typically focuses on how issues are presented in elite debate or in the media and on what consequences framing has for public opinion. 59In the context of campaign agendas, we are concerned with the strategic bene ts of framing for candidates. Scholars have suggested that framing is a particularly effective strategy for candidates when they need to discuss issues that may appear to advantage the other party. Petrocik articulates this idea: ‘When the opponent’s issues are unavoidable, they can be interpreted in a way to highlight some feature of the issue on which they are likely to be regarded as more competent’. 60Similarly, Holian notes that issues can be co-opted by a party ‘given the correct “spin” ’.61 Framing is thus a means by which candidates can trespass on the opposing party’s issues. 62 This begs the question, what kinds of frames facilitate trespassing? If a party does not have the record of attention to an issue suggested by ownership, how can it discuss that issue credibly? There has been little theoretical or empirical examination of these questions. I argue that candidates do so by pursuing two strategies. First, they trespass by talking about the issue in only the vaguest of ways. Scholars have suggested that candidates have an incentive to be ambiguous in their issue positions. 63Ambiguity may be especially valuable when trespassing. An ambiguous trespasser merely endorses a goal that is widely shared – a familiar strategy in elections generally. 64 If trespassing candidates were to suggest a more speci c position on that issue, such as their ‘plan’ once elected, they would open themselves up to criticism about the feasibility of this plan. In particular, given that it is the other party that has a history of developing and implementing policy related to that issue, trespassing candidates would be especially vulnerable to this kind of critique. It is a safer rhetorical strategy to make only vague and unobjectionable claims. A second framing strategy that facilitates trespassing is taking a position on an issue that comports with one’s party’s ideology and policy priorities. Candidates can take such a position safely and credibly because most political issues are broad enough to encompass a wide variety of secondary ‘sub-issues’ and policy debates. Nearly every issue thus contains sub-issues on which each of the parties has an advantage. Holian’s discussion of crime provides one example. 65Political debates about crime generally centre on two ideas, punishment and prevention. Holian demonstrates that during the 1990s Democrats, and in particular Bill Clinton, appeared to believe that their credibility in advocating punishment was weaker than that of Republicans, in part because segments of the Democratic party 59For example, William A. Gamson and Andre Modigliani examine how the framing of af rmative action has changed over time – see ‘The Changing Culture of Af rmative Action’, in Richard D. Braungart, ed.,Research in Political Sociology, Vol. 3 (Greenwich, Conn.: JAI Press, 1987), pp. 137–77. Shanto Iyengar examines how different frames in news stories about poverty affect the causes to which viewers attribute poverty and their support for government action to help the poor (see ‘Framing Responsibility for Political Issues: The Case of Poverty’, Political Behavior, 12 (1990), 19–40). 60Petrocik, ‘A Theory of Issue Ownership’, p. 829.61Holian, ‘He’s Stealing My Issues!’ p. 99.62One might ask whether issue trespassing is merely a manifestation of convergence to the median voter. While it is true that a Republican who is talking about education could be expressing the preference of the median voter, the logic of convergence itself cannot explain why the Republican would emphasize education as opposed to some other issue on which his or her position is both close to the median voter’s and on which the Republican party has an ownership advantage. 63Page,Choices and Echoes; and Kenneth A. Shepsle, ‘The Strategy of Ambiguity: Uncertainty and Electoral Competition’,American Political Science Review, 66 (1972), 555–68. 64Page, Choices and Echoes; and Donald P. Stokes, ‘Valence Politics’, in Dennis Kavanagh, ed.,Electoral Politics(New York: Oxford University Press, 1992), pp. 141–64. 65Holian, ‘He’s Stealing My Issues!’ 428 SIDES do not support measures such as capital punishment. Clinton instead advocated prevention, such as a proposal to put more police of cers on the street. This rhetorical strategy simultaneously allowed Democratic candidates to demonstrate concern about crime – an issue conventionally considered a ‘Republican’ issue – while advocating a policy in line with their party’s ideology and goals. 66 Thus, candidates can trespass by articulating positions that are credible given their party’s traditional strengths. What are the empirical implications of these two strategies in the 1998 elections? I generate expectations regarding Democrats’ trespassing on the issue of taxes and Republicans’ trespassing on education, health care and Social Security. 67For Democratic candidates to talk about taxes, a strategy of ambiguity would simply entail support for lower taxes. Candidates pursuing this strategy would argue in favour of ‘tax cuts’ or ‘tax relief’ without specifying which taxes, which taxpayers, etc. Were Democrats to take more speci c positions on taxes that accord with Democratic values and priorities, the major theme that should emerge is support for a progressive tax structure. The Democratic party has a historical commitment to a progressive taxation and its party platform in 1996 reiterated this commitment, accusing congressional Republicans of giving ‘a massive tax break to the wealthiest Americans, and pay[ing] for it by raising taxes on ordinary Americans’. 68 The crux of Democratic philosophy about taxation is to ensure that lower-income individuals do not bear too great a tax burden relative to the wealthy. Democratic candidates who discuss taxes would therefore be expected to emphasize this theme. In its most ambiguous form, Republican rhetoric about education, health care and Social Security should endorse only agreeable goals such as ‘improving education’, ‘better schools’, ‘improving health care’, ‘protecting Social Security’, and so on. There would be no elaboration of means by which those goals might be obtained. More speci c positions on these issues should comport with Republican ideology. In the case of education, the Republican party’s traditional position advocates the devolution of authority away from federal government to local authorities and to families. In their 1996 platform, the party pledged to ensure that each family ‘has control over the education of its children’. 69In the case of health care, the party’s philosophy is similar: empowering individuals and relying on markets rather than designing programmes administered by the government. An illustrative passage of the 1996 platform advocated tax-free ‘medical savings accounts’ that allow individuals to ‘plan for their own medical needs instead of relying on government’. In the case of Social Security, the Republican party has typically sought to connect this programme with other ‘owned’ issues, particularly taxes. A ready example also comes from the 1996 platform: ‘The Republican Party has always opposed the earnings limitation for Social Security bene ts, a con scatory tax that discourages older Americans from active engagement in all walks of life’. The platform continues by accusing then-President Bill Clinton of supporting this tax. Republicans are thus defending Social Security by advocating another long-standing position, reducing taxes. Ascertaining whether these rhetorical strategies were prominent in 1998 necessitates more detailed information about the claims candidates made. To generate this information, 66As Figure 1 demonstrates, Clinton’s efforts appear to have helped negate Republicans’ advantage on crime.67I do not present an analysis of Medicare. In 1998, both party’s advertising was entirely focused on vague promises to protect Medicare. 68See George Thomas Kurian,Supplement to the Encyclopedia of the Democratic Party(Armonk, N.Y.: M. E. Sharpe, 2001). 69See George Thomas Kurian,Supplement to the Encyclopedia of the Republican Party(Armonk, N.Y.: M. E. Sharpe, 2001). The Origins of Campaign Agendas429 I coded each of these 1,258 advertisements in more detail, documenting the positions candidates took on each issue. The coding scheme erred on the side of inclusiveness, adding a separate code for even esoteric positions. However, advertisements that talked about an issue in nothing but the vaguest terms – merely stating support for a consensual policy goal – were coded only as ‘valence’ since they endorsed a widely shared goal without proposing a means to achieve it. 70 Advertisements were thus coded either valence or non-valence. If the former, I coded them as ‘valence’ and nothing else. If the latter, I coded the speci c proposals made. I also recorded in separate codes candidates’ claims about their opponent, since it is one thing to state support for Social Security and another thing to accuse the opponent of inadequate support. 71 I compared the incidence of certain positions in both Democrats’ and Republicans’ advertising about education, Social Security, health care and taxes to identify where the parties may have differed in their rhetoric and in particular how the trespassing party handled these issues. For each issue, I present the ve most prominent positions taken by each party’s candidates, measured by considering only those candidates who actually aired advertising on that issue, and then calculating the average percentage of those Democrats’ and Republicans’ advertising on that issue that stated a particular position. Table 4 presents the results. Looking rst at taxes, it is clear that, in trespassing on this issue, Democratic candidates employ both types of frames. They employ the strategy of ambiguity in 23.6 per cent of their advertising about taxes, endorsing tax cuts without offering any further detail. This is the most prominent theme in Republican advertising as well, suggesting the more general appeal of ambiguity to candidates. Even more prominent in Democratic advertising about taxes is a theme directly related to the Democratic party’s commitment to progressive taxation: accusing their opponent of supporting tax cuts for the rich. For example, an advertisement issued by Arizona Democrat Steve Owens said of his Republican opponent, ‘J. D. Hayworth voted for a tax loophole that allows billionaires to avoid paying taxes’. This theme is completely absent in Republican advertising, which, when it is more speci c, focuses on such ideas as reforming the Internal Revenue Service, repealing the marriage penalty and revising the tax code. Democrats were also more likely to discuss speci c tax cuts and in particular a tax cut for college tuition – a proposal that gels nicely with Democrats’ traditional concern for education. Thus, Democratic candidates trespassed by linking taxes to their established ideology and policy priorities, drawing on ambiguous ‘valence’ appeals as well. What kinds of themes did Republicans mention in trespassing on ‘Democratic’ issue like education, health care and Social Security? In their advertising on education, Republicans also employed both ambiguous messages and traditionally ideological messages. Among Republicans, the ‘valence’ position, i.e., ‘improving education’, was the most prominent of them in their advertising. On average, it was the only ‘position’ taken in over 33 per cent of Republican candidates’ advertising on this subject (versus 23 per cent of Democrats’ advertising). 72 The next two most prominent Republican themes centred on the idea of opposing national authority: support for local control of schools, and putting education funding into the classroom as opposed to the federal bureaucracy. For example, 70Geer, ‘Assessing Attack Advertising’; Stokes, ‘Valence Politics’.71Valence advertisements were also coded for any references made to the opponent’s position.72An example of a ‘valence’ advertisement about education comes from Congressman Bob Riley (R-AL). Sitting among young children in a classroom, he said, ‘I want to see an educational system in Alabama second to none. These kids deserve the best we can give them.’ 430 SIDES TABLE 4Prevalence of Issue Positions, by Party Republicans Democrats Taxes Taxes Valence (‘supports tax cuts’) 46.2% Opponent wants tax cuts for 35.1% Opponent supports tax increases 24.9 wealthy IRS reform 12.9 Valence (‘supports tax cuts’) 23.6 Repeal marriage penalty 10.0 Targeted tax cut for tuition 14.8 Reform tax code 7.8 Opponent supports tax increases 7.2 IRS reform 5.8 Education Education Valence (‘improve education’) 33.7 Valence (‘improve education’) 22.9 Local control of schools 17.9 Smaller classes 22.9 Money in classrooms, not 16.7 Safer schools 21.5 bureaucracy Higher standards 11.7 Safer schools 15.0 More teachers 10.1 Smaller classes 9.1 Health Care Health Care Patients’ bill of rights 21.4 Doctors and patients make 38.6 Valence (‘improve health care’) 21.0 decisions Doctors and patients make 13.7 Patients’ bill of rights 26.9 decisions HMO reform or accountability 22.3 Increase research funding 12.7 Valence (‘improve health care’) 10.8 Expand coverage or insurance 7.2 Increase research funding 3.5 Social Security Social Security Valence (‘protect Social 59.3 Valence (‘protect Social 46.1 Security’) Security’) Opponent supports higher SS 14.3 Use surplus for SS 25.8 taxes Opponent hurt SS 10.1 Opponent hurt SS 10.4 Opponent supports 9.5 Use surplus for SS 9.0 privatization Opponent raided trust fund 1.9 Opponent raided trust fund 9.1 Note:Table entries are the average percentage of the candidates’ advertisement airings on each particular issue that mentioned a given position. Republican J. D. Hayworth of Arizona said in one advertisement, ‘I want 95 cents of every education dollar spent in our classrooms.’ 73 Democratic advertisements almost never touched on similar ideas, focusing instead on proposals such as smaller class size. Thus, Republican advertising about education was at times simply vague but it also drew on themes that resonated with their long-established goal of limiting the power of the federal government. In Republican advertising on health care, ‘valence’ rhetoric was again prominent. Just over a fth (21 per cent) of Republican advertisements only endorsed the generic goal of improving health care (versus 10.8 per cent of Democratic advertisements). The rest of Republican rhetoric, however, was not particularly ideological and indeed both parties emphasized the same basic positions, such as support for a patients’ bill of rights and an endorsement of doctor and patient decision making (as opposed to that of insurance 73Somewhat surprisingly, very few advertisements from either party mentioned school vouchers or even the broader notion of parent choice. The Origins of Campaign Agendas431 companies). Democratic advertising did advocate reform of the health management organization, a theme virtually absent in Republican advertising, but generally the differences between the parties are mostly matters of degree. ‘Valence’ rhetoric appears the most useful frame for Republican trespassing. In advertising about Social Security, the most prevalent position taken by both parties was the least speci c: a general promise to protect or preserve Social Security. 74However, this theme was more prominent in Republican advertising (59.3 per cent) than in Democratic advertising (46.1 per cent), suggesting the particular value of a ‘valence’ frame for the trespassing party. Republican rhetoric about Social Security, when it delved into speci cs, tended to link Social Security to other issues and in particular to taxes, as expected. About 14 per cent of Republican advertisements accused the Democratic opponent of supporting higher taxes on Social Security income. Democratic rhetoric was very different, focusing on applying the budget surplus to ‘save’ Social Security. In sum, a more nuanced reading of candidate advertising turns up interesting differences in the rhetoric of the two parties. These differences suggest parties can trespass on issues both by taking ambiguous positions on issues and by espousing positions that are consonant with their party’s traditional concerns or reputation. On these issues, the trespassing party is almost always more likely to discuss them only in generic terms. This is particularly true for Republicans when trespassing on ‘Democratic’ issues. It may be more plausible for candidates to talk about the opposing party’s issues by treating them as valence issues. This kind of commitment may seem more credible since candidates do not have to articulate how a proposed goal might be attained. 75 Trespassing parties also do so by drawing on their customary values and beliefs. Thus, Republicans can talk about education demanding that ‘95 cents of every dollar’ of federal education spending go to the classroom. Republicans can talk about Social Security by accusing their opponents of supporting higher taxes on Social Security income. Similarly, Democrats talk about taxes in part by opposing tax cuts for the wealthy. The consequence of this strategy is that the parties may ‘diverge’ in their rhetoric on issues. Even though parties often centre their agendas on similar issues – thereby ‘riding the wave’ – they employ somewhat different frames in discussing these issues. Thus, while the parties’ tendency to ‘ride the wave’ contradicts the Dominance Principle and ownership theory, their differing frames actually comport with the Dominance Principle, which holds that opposing sides will pursue contrasting rhetorical strategies. This suggests Riker’s theory better predicts how candidates discuss issues, rather than which issues candidates discuss. DISCUSSION: GENERALIZABILITY OF FINDINGS This analysis has focused in some detail on a single American election, and thus an important question is whether the results are generalizable. Are the strategies employed during the 1998 electionsui generisor indicative of a broader empirical pattern? Speci cally, should we expect comparable evidence of trespassing in other elections? More recent American elections suggest that the answer is yes. In Table 5, I present 74Typically, this would entail statements like these in an advertisement for Senate candidate Jim Bunning (R-KY): ‘A national leader ghting for Social Security …“What we ought to do is secure our Social Security system” ’. 75In general, the prominence of valence themes in advertising corresponds to the ndings of Geer (‘Assessing Attack Advertising’), Page (Choices and Echoes), and Spiliotes and Vavreck (‘Campaign Advertising’). 432 SIDES TABLE 5Issue Emphasis and Party Ownership in 2000 and 2002 Percentage of advertising time devoted to issue Ownership Republicans Democrats Republicans Democrats advantage (2000) (2000) (2002) (2002) Democratic advantage Education 10.5 10.4% 11.1% 6.9% 7.4% Environment 22.7 2.2 2.1 1.4 1.2 Health care 19.4 7.9 13.6 8.5 12.0 Jobs 9.8 2.0 1.1 6.0 5.7 Medicare 19.6 4.3 9.7 2.0 3.5 Social Security 15.9 8.4 6.9 4.7 7.3 No advantage Crime 2.4 2.2 2.2 0.9 0.8 De cit/budget 0.2 3.0 3.0 0.5 0.7 Republican advantage Defence 29.9 0.7 0.4 3.5 0.9 Foreign Affairs 13.0 0.4 0.8 0.2 0.1 Taxes 5.0 10.1 5.5 9.4 5.0 Terrorism 29.6 2.1 0.7 N129 128 147 129 Note:‘Ownership advantage’ is equal to the percentage who trust the Democrat to handle each issue, minus the percentage who trust the Republican, averaged over 1990–2002. Shaded cells indicate that the difference between Democrats and Republicans is statistically signi cant at p 0.05. the breakdown of ‘advertising time’ by party for the 2000 and 2002 elections, along with the average party ownership advantage for the 1990–2002 period. 76This table’s structure mirrors that of Table 2, with shaded cells indicating statistically signi cant differences between the parties. During the 2000 election, the public agenda was much the same as in 1998, and candidate behaviour was also quite similar. 77With education, Social Security and health care at the top of the agenda, Republican House and Senate candidates spent the vast majority of their time discussing these and other ‘Democratic’ issues. Summing up the advertising time devoted to the six ‘Democratic’ issues in Table 5, Republican candidates spent 35 per cent of their advertising time discussing these issues while only 11 per cent of their time discussing the three ‘Republican’ issues (defence, foreign affairs and taxes). Democratic candidates, not surprisingly, also devoted nearly half (45 per cent) of their advertising time to these ‘Democratic’ issues, which were conveniently salient to the public as well. 76Kenneth Goldstein, Michael Franz and Travis Ridout, ‘Political Advertising in 2000’, Combined le [dataset], nal release (Madison: University of Wisconsin-Madison and the Brennan Centre for Justice at New York University, 2002); and Kenneth Goldstein and Joel Rivlin, ‘Political Advertising in 2002’, Combined le [dataset], nal release (Madison: The Wisconsin Advertising Project, The Department of Political Science at The University of Wisconsin-Madison, 2005). 77In a Harris Poll of July 2000, the public’s ‘most important issues’ were virtually the same as in 1998: education (18 per cent), health care (14 per cent), Social Security (13 per cent), crime (11 per cent), taxes (10 per cent). The Origins of Campaign Agendas433 In 2000, the behaviour of the presidential candidates was quite similar. In terms of the volume of advertisement airings that mentioned the issue, Bush devoted the most attention to education, Social Security, the de cit/surplus, health care and taxes (data not shown). Gore’s agenda was almost identical: health care, taxes, education, the environment and Social Security. Thus, ‘riding the wave’ appears to guide presidential candidate strategy in this election as well. By the time of the 2002 election, the issue agenda had changed notably because of the terrorist attacks of 11 September and the concomitant economic downturn. In a June 2002 Harris Poll, respondents listed terrorism and the economy as their most important priorities. 78Thus, Table 5 includes data about the degree of ownership advantage and the volume of advertising time on terrorism. Table 5 provides some evidence that the parties responded to shifts in the public’s agenda. 79 Both parties spent less time on issues such as education, the environment and Medicare than they did in 2000. They spent more time discussing jobs, defence and terrorism – suggesting that they were, in part, ‘riding the wave’. The 2002 election is different, however, in that the public’s agenda does not neatly coincide with traditionally ‘Democratic’ issues. In particular, as Table 5 suggests, issues like terrorism and defence advantage Republicans. ‘Riding the wave’ would therefore entail, for Republicans, attention both to ‘Democratic’ issues, especially jobs, and to ‘Republican’ issues like defence. It is thus quite remarkable that Republicans continued to devote so much time to ‘Democratic’ issues unrelated to the economy. Republicans spent nearly twice as much time discussing these six ‘Democratic issues’ issues (29.5 per cent) than defence, foreign affairs, taxes and terrorism (15 per cent), despite their large ownership advantage on defence and terrorism. This seems to contradict what ownership theory would have them do. Given an election-year agenda that favoured defence and terrorism, Republicans still devoted more time to ‘trespassing’ on education, health care and Social Security. This strategy is particularly curious given that by discussing defence and terrorism Republicans could simultaneously ‘ride the wave’ and capitalize on their ownership advantage However, the behaviour of Democratic candidates better accords with ownership theory. Democrats largely ignored terrorism, foreign affairs and defence. They focused about 37 per cent of their advertising time on these six ‘Democratic’ issues and only about 7 per cent on these four ‘Republican’ issues. Quite strikingly, they devoted a paltry amount of time (1.6 per cent) to defence and terrorism combined. Thus, though Democratic attention to jobs suggests that they were attentive to the public’s concerns about the economy, they seemed to be ignoring the public’s concern about terrorism. In 2002 Democratic candidates appeared loathe to discuss traditionally ‘Republican’ issues such as defence. Candidate behaviour in the 2000 and 2002 elections con rms the previous results in important ways. In 2000, both parties continued to ‘ride the wave’, focusing on the issues most salient to the public. As a consequence, Republican candidates continued to trespass 78When asked ‘What do you think are the two most important issues for the government to address?’ 37 per cent of the public listed the economy or jobs and 45 per cent listed terrorism, national security or homeland security. Issues salient in 1998–2000 were less salient in 2002 – e.g., education (8 per cent), health care (10 per cent), taxes (5 per cent) and Social Security (4 per cent). 79In Table 5, the category ‘health care’ includes references both to health care and to prescription drugs. ‘Foreign affairs’ includes references to foreign policy, foreign aid, the Middle East, China and Afghanistan. 434 SIDES quite freely since many of these salient issues were ‘owned’ by the Democratic party. In 2002, when the public’s agenda was quite different, candidates from both parties shifted their attention accordingly, but not as much as one might expect. There was less evidence of ‘riding the wave’ and more evidence of continued attention to issues like education and health care, even though those ranked lower on the public’s agenda. Republican trespassing continued apace, despite their ostensible incentive to focus on defence and terrorism. Democrats continued to focus on ‘owned’ issues at the expense of defence and terrorism. Their behaviour in 2002 is the clearest evidence of ownership theory in any of these three elections. CONCLUSION AND IMPLICATIONS This analysis of a comprehensive dataset of candidate advertising from House and Senate races demonstrates that heresthetical choices in campaigns derive in important ways from the public’s agenda. Candidates from both parties tend to advertise on the same set of salient issues. As a consequence, heresthetical choices do not depend consistently on the credibility that the candidate’s or party’s record is presumed to convey. Candidates do not apparently feel that they need personal experience with an issue, or an af liation with the party traditionally linked to an issue, to talk credibly about that issue. Indeed, the results here seem to con rm what Downs wrote decades ago: ‘Each party casts some policies to the other’s territory in order to convince voters that its net position is near them’. 80This tendency to roam in ‘the other’s territory’ means that candidates routinely trespass and discuss the other party’s issues, often more frequently than they do their own party’s issues. Certain rhetorical strategies facilitate trespassing – notably, framing the issue in terms of their party’s traditional concerns and articulating only vague endorsement of an uncontroversial goal. If issue ownership’s effects are not entirely powerful or robust, then it may also be that candidates are not simply ‘talking past’ each other, as the Dominance Principle and ownership theory would predict, though their speci c rhetorical themes will not necessarily be identical. This nding has particular implications for the presence of dialogue in campaigns. Scholars have noted that dialogue is normatively bene cial, in that it more closely resembles the ideal of an ‘enlightening conversation’, but have also lamented its relative absence in campaigns. 81 In these elections, the fairly high level of trespassing suggests that candidates may be engaging in dialogue. In 1998, among the 103 races where these data included advertising from both candidates, there were many instances where both candidates talked about a given issue. For example, forty-eight races featured both Republican and Democratic advertising about education (versus forty-four where only one candidate advertised). And fty-six races featured both Republican and Democratic advertising about Social Security (versus thirty-four where only one candidate advertised). Dialogue may not be the norm on every issue, and certainly the mere appearance of 80Downs,An Economic Theory of Democracy, p. 135.81Simon,The Winning Message, p. 25. One such lament is that of Berelson, Lazarsfeld and McPhee (Voting, p. 236): ‘The opposing candidates tended to “talk past each other” almost as if they were participating in two different elections’. The Origins of Campaign Agendas435 dialogue says nothing about the quality of that dialogue, but neither is dialogue completely absent. 82 There is another, perhaps ironic, implication in these results. Heresthetics depends on structuring the agenda in advantageous ways. However, the inconsistent effect of factors like party ownership and candidate record – both of which are perhaps the most obvious sources of advantage – begs the question of what exactly ‘advantage’ is based on. Perhaps advantage comes from simply having an issue position closer to the median voter’s than is your opponent’s position, as Hammond and Humes hypothesize. 83But this kind of spatial imagery seems irrelevant when so many advertisements feature only the most unobjectionable positions, which was precisely the point of Stokes’s original argument. 84 If ‘valence politics’ is the norm, and if both parties appear willing to endorse similar goals regardless of who ‘owns’ the issue, then how do candidates differentiate themselves? And how do voters differentiate the candidates? It may be that heresthetics in campaigns depends not so much on policy issues but on other factors, such as traits of the candidates, including their personalities. It may also be that voters tend to assume that candidates differ in stereotypical partisan ways, even when the candidates’ messages to not support this assumption. 85To understand the science as well as the ‘art’ of heresthetics, we must make better sense of this complex reality. APPENDIX A: CODES FOR ISSUES IN 1998 HOUSE AND SENATE CANDIDATE ADVERTISING Personal Characteristics of Candidate(s) 1. Background 5. Personal values 2. Political record 6. Honesty, integrity 3. Attendance record 7. Special interests 4. Ideology Policy Issues Economy 10. Taxes 11. De cit, surplus, budget, debt 12. Government spending 13. Minimum wage 14. Farming (e.g., friend of) 15. Business (e.g., friend of) 16. Employment, jobs 17. Poverty 18. Trade, NAFTA 19. Other economic referenceForeign Policy and Defence 50. Defence 51. Missile defence, Star Wars 52. Veterans 53. Foreign policy 54. Bosnia 55. China 59. Other defence, foreign policy issue 82See Brasher (‘Capitalizing on Contention’) and Sigelman and Buell (‘Avoidance or Engagement?’) for similar ndings in Senate and presidential elections, respectively. 83Hammond and Humes, ‘The Spatial Model and Elections’.84Donald E. Stokes, ‘Spatial Models of Party Competition’,American Political Science Review, 57 (1963), 368–77. 85Helmut Norporth and Bruce Buchanan, ‘Wanted: The Education President: Issue Trespassing by Political Candidates’,Public Opinion Quarterly, 56 (1992), 87–99; and Wendy Rahn, ‘The Role of Partisan Stereotypes in Information Processing about Political Candidates’,American Journal of Political Science, 37 (1993), 472–96. 436 SIDES Social Issues 20. Abortion 21. Homosexuality 22. Moral values 23. Tobacco 24. Af rmative action 25. Assisted suicide 26. Gun control 27. Other reference to social issues Law and Order 30. Crime 31. Drugs 32. Death penalty 33. Other reference to law and order Children 40. Education 41. Lottery for education 42. Child care 43. Other child-related issue 44. School prayerClinton 60. Clinton 61. Ken Starr 62. Whitewater 63. Impeachment 64. Sexual harassment, Paula Jones Other 70. Environment 71. Immigration 72. Health care 73. Social security 74. Medicare 75. Welfare 76. Civil rights, race relations 77. Campaign nance reform 78. Government ethics 79. Roads, transportation 95. Other 99. None APPENDIX B: MEASUREMENT OF ADVERTISING TIME The measurement of advertising time takes into account both the number of issues mentioned in any one advertisement and the number of times this advertisement was aired. The former is necessary because advertisements that focus on one issue indicate more emphasis than do those mentioning multiple issues. The latter is important because certain advertisements will air more frequently than others. An example may best illustrate how the measure was constructed. Imagine a candidate produced two different advertisements (XandY) and airedX80 times andY20 times. The two themes inXwere education and taxes and the only theme inYwas taxes. To compute the amount of ‘advertising time’ the candidate devoted to each theme, divide the number of advertisements about that theme by the number of issues in the advertisement and by the total number of advertisements. So the amount of advertising time devoted to education is 80 divided by 2 (the number of issues inX) and then divided by 100 (the total number of advertisements, 80 20). So the proportion of time devoted to education is 0.4. A comparable calculation for taxes summarizes the attention in bothX(80 divided by 2 and by 100 0.4) andY(20 divided by 1 and by 100 0.2). So the total advertising time devoted to taxes is 0.4 0.2 0.6. The measure is thus a proportion that speci es the amount of advertising time that a candidate devoted to a speci c issue. Across all issues, this measure sums to 1. (In the example above, the time spent on education plus that spent on taxes is 0.4 0.6 1). I calculate a proportion because the raw number of advertisements varies wildly across races and thus some kind of standardization is in order. It is worth noting that dividing by the number of issues in the advertisement is inexact in that it assumes equal time for each issue within the advertisement. Nevertheless, given the dif culties inherent in ascertaining time precisely and developing a ner-grained measure, the version I employ here seems sensible. Finally, though the example above involves only policy-related issues, the measure of advertising time also takes into account themes related to personal attributes of the candidates (their experiences, background, etc.) in this measure so as not to exaggerate the amount of time candidates spend discussing policy-related issues.