The Dynamics of Issue Ownership

7/28/2019 The Dynamics of Issue Ownership

1/8

The Dynamics of Issue Ownership in Presidential Campaigns

David F. Damore

Political Research Quarterly, Vol. 57, No. 3. (Sep., 2004), pp. 391-397.

Stable URL:

http://links.jstor.org/sici?sici=1065-9129%28200409%2957%3A3%3C391%3ATDOIOI%3E2.0.CO%3B2-D

Political Research Quarterly is currently published by University of Utah.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtainedprior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content inthe JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/journals/utah.html.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. Formore information regarding JSTOR, please contact [email protected].

http://www.jstor.orgFri Jun 1 15:12:41 2007
http://links.jstor.org/sici?sici=1065-9129%28200409%2957%3A3%3C391%3ATDOIOI%3E2.0.CO%3B2-Dhttp://www.jstor.org/about/terms.htmlhttp://www.jstor.org/journals/utah.htmlhttp://www.jstor.org/journals/utah.htmlhttp://www.jstor.org/about/terms.htmlhttp://links.jstor.org/sici?sici=1065-9129%28200409%2957%3A3%3C391%3ATDOIOI%3E2.0.CO%3B2-D


2/8

- The Dynamics of Issue Ownership- in Presidential CampaignsDAVID E DAMORE, UNIVERSITY LASVEGASF NEVADA,This effort examines the dynam ics of the agenda-setting process in presiden tial campaigns by assess ing theconditions that motivate candidates to discuss issues associated with their opponen tk party, The article's argu-ment c onten ds that occurrences of issue trespassing are a function of the context in wh ich a campaign is occur-ring and factors stemming from the campaign process. The hypotheses are tested against data collected fromall available campaign advertisements produced by major party candidates competing in the 1976 through19 96 presidential elections. The results of the logit analysis indicate that candidates' decisions to address issuesowned by th eir opponent's party are a function of their competitive stand ing, their partisanship, the imp or-tance of an issue to the electorate, and the tone of their cam paign messages.

I ncreasingly, the field of campaign politics is movingaway from an emphasis on voters ( e.g .,voting behavior,information processing, and attitude formation) andtoward analyses of candida te behavior ( e.g., Damo re 20 02 ;Jacobs and Shapiro 199 4; Kahn and Kenney 1 999 ; Shaw199 9a, 1999 b; Simon 20 02) . This shift has resulted in theemergence of a body of knowledge that has improved ourunderstanding of the sup ply side of the camp aign process,particularly in terms of the study of campaign strategy.

I seek to further this research by building on p rior workexamining the theory of issue ownership (Petrocik 1991,19 96) to assess how presidential candidates compete todefine and control campaign agendas. Briefly, the theory ofissue ownership posits that the parties hold reputations fortheir ability to handle certain issu es. These reputa tions , inturn, provide candidates with credibility over issues associ-ated with their party By increasing the salience of party-owned issues, candidates can stack the campaign agendawith issues that accent their strengths and highlight theiropponents' weaknesses.

While the advantages that candidates may reap fromdiscussing party-owned issues is apparent, what is lessobvious, and which is the focus of this effort, is why can-didates would choose to trespass on issues associated withthe opposition's party. Indeed, evidence from recent presi-dential campaigns indicates that such behavior is comm on,if not always successful (i .e., Dukak is in the t ank ). Investi-gation of this behavior and the study of campaign agendasetting more generally have implications for our under-standing of campaign strategy and political communica-tion, as well as the linkage between campaign processesand election outcomes.

NOTE: An earller version of thls research was presented at the AnnualMeetlng of the Western Polltical Science Association, Denver,March 27-29, 2003. 1 am thankful for the comments and sugges-uon s of Scott Gartner, Bob Jackma n, Tom Hansford, Steve Nichol-son, and Jlm Spriggs.

Po l~i ~tu lew rth Qua~ier(v,Vol. 5 7, No . 3 (September 2004): pp. 391-397

The research begins by assessing the strategic considera-tions that underlie candidates' decisions about which issuesto discuss during a campaign and which issues to avoid.Next, I draw on prior research to develop a set of hyp othe-ses that capture the conditions under which presidentialcandidates will address issues associated with their oppo-nent's party In particular, I argue that these decisions areshaped by the con text in which a campaign is occurring andthe dynamics of the campaign process. These hypothesesare then tested with data collected from all available cam-paign ad vertisements produced by major party candidatescompeting in the 19 76 throug h 1 996 presidential elections.After presenting the results of the analysis, the paper con-cludes by discussing its implications for the literature.

As Riker (1993: 1) notes: "Agendas foreshadow out-comes: the shape of an agen da influences the choices m adefrom it." As applied to the stu dy of cam paign s, Riker's axiomsuggests that candidates who are able to define campaignagendas hold a distinct advantage over candidates operatingunder agendas established by their opponents. ' Given thislink between agendas and outcomes, understanding howcampaign agendas emerge is a critical component of cam-paign strategy.

Investigation of this process, however, necessitates exante specification of candidates' agenda setting strategies.To this end, Riker (1993) argues that candidates shouldfocus their attention on issues that either play to theirstrengths or their opponents ' weaknesses while avoidingissues that either accent the oppo sition's streng ths or high-light their own weaknesses (e. g.,the Dom inance Principle).Execution of this strategy may allow candidates to increase

In contrast to agendas in legislative settings, campaign agendas are morefluid in nature. Whereas legislative agendas are fixed sets of proposalsthat are voted on sequentially, campalgn agendas emerge over time andresult from the joint cholces candidates make a bout voter receptivity toth e~ rssue packages .


3/8

the saliency of issues over which they are perceived as cre d-ib le in the campaign envi r~nment .~his, in tu rn, increasesthe likelihood that candidates will be evaluated in terms oftheir strengths and the opposition candidate in terms oftheir weaknesses.

Clearly, the D ominanc e Principle provides a useful tem-plate for assessing agenda competition in the campaignarena. For the construct to be tractable, however, indicatorsof credibility need to be observable. To wit, past researchsuggests that policy reputations (Carmines and Kuklinski1990), incumbency (Kahn and Kenney 1999), and candi-dates' backgrounds (i.e ., military experience) can serve assources for credibility. However, incorporation of these cri-teria into a general model of agenda competition is prob-lematic because they are largely idiosyncratic and he nce d onot lend themselves to systematic evaluation. Because ofthis limitation, a more generalizable construct for exam iningthe agenda com petition process in presidential cam paigns isused here: candidates' partisanship and the theory of issueownership (Petrocik 199 1, 1996).

The theory of issue ownership rests on the propositionthat because of long standing party reputation, candidates'are more likely to be perceived as credible over issuesowned b y their party (also see Camp bell et al. 19 60 ; Nor-poth and Buchanan 1992 ). For instance, voters see Repub-licans as being better able to handle foreign policy and gov-ernment management issues, while the Democrats areperceived as mo re capable on civil rights an d social welfareissues. Prior research suggests that these partisan issue asso-ciations are consistent an d well defined (e .g. , Campbell etal. 1960; Hamill, Lodge, and Blake 1985).3

Given the basic thrust of the issue ownership argument,it might be expected that as cand idates compete for leverageover campaign age ndas they will focus their campaign m es-sages on issues owne d by their party and ignore issues asso-ciated with the oppo sition party to maximize pe rceptions oftheir credibility over the campaign agenda. The data pre-sente d in Table 1 , howev er, tell a different story. These datapresent the distribution of appeals taken from all availablecampaign advertisements produced by major party candi-dates competing for the presidency between 197 6 and199 6. The data are divided into five categories: appealsfocusing on candidate traits and backgrounds, appealsaddressing neutral an d performance issues, and appeals thathighlight either party-owned or opposition owned issues.

Credibility is a necessary antecedent for persuasion ( e.g., Lupia andMcCubbins 1998).Consistent with political psychological approaches to the study of votingbehavior (e.g ., Conover and Feldm an 1989; Jacobs and Shapiro 19 94;Krosnick and Kinder 199 0; Rahn 19931, I assume that voters are morelikely to rely on inform ation that is salient in the campaign env ironm entwhen formulating their vote choice. Thus, the purpose of campaign mes-sages is not to change voters' att itudes per se, but to prime the informa-tion that voters draw on when evaluating candidates. Consistent withstrategic actor m odels of agenda setting, 1 assume th at, it is the inter-can-didate competition for agenda control that affects what information issalient to voters.

Table 1 yields a number of insights into how major partypresidential candidates present their campaign messages; spe-cific to my purposes, two considerations merit discussion.First, to some de gree, it appears that all candidates issue tres-pass. This ranges from a high of 34 percent for Walter Mon-dale in 1988 to a low of 4 percent by George Bush in 1988.Second, issue trespassing is more co mm on am ong Democratsthan among Republicans. Of the total instances of issue tres-passing in the data, Democratic candidates account for 74percent. It appears that in some cases presidential candidatesabandon the efficacy of the Dominance Principle and cam-paign in their oppone nt's issue territories.

What accounts for the discrepancy between the theoret-ical predictions outlined above and these data? The work ofDowns (1957) provides a useful starting point for explain-ing this behavior. Specifically, Dow ns (135 ) argues that toappeal to the largest segment of voters, candidates in a two-party system sh ould cast "some policies in to the other's ter-ritory in order to convince voters that their net position isnear them ." Th us, some level of issue trespassing shou ld beexpected as a function of normal campaign ~trategy.~riorresearch also suggests additional factors, most notably thecontext in which a c ampaign is occurring an d factors stem-ming from the campaign process that may affect occur-rences of issue trespa ssing. Below, I more fully elucidate thehypothesized relationship between these factors and candi-dates' decisions to issue trespass.Environmental Factors

As the work of Holbrook (1996) and others suggests,macro-level forces (e.g. , the state of the econom y) sha pe thecontext in which campaigns transpire. This, in turn, mayconstrain candidates' selection or avoidance of specificissues. In particular, environmental factors may affect thedecision to issue trespass at two levels. First, candidates mayadapt the content of their campaign messages in response to

'Although this strategy is rational from the perspective of candidate s, itsimplications for voters may not be so satisfactory In addition to spend-ing a good deal of time discussing moderate policies, candidates whoemploy this strategy are likely to heighten ambiguity about what theystand for. Moreover, evidence from the voting behavior literature indi-cates that the payoffs for this strategy may be less than expected . In theiranalysis of voter recall of Dukakis and Bush's discussion of oppositionparty issues during the 198 8 presidential campaigns, Norpoth andBuchanan (1992) find that because of voters adherence to long-stand-ing partisan stereotypes, voters failed to associate the candidates withthe trespassed issues. With this said, it is important to remember thatcampaigns do not target all voters equally, but rather direct theirresources towards voters at the margins whose decisionmaking ulti-mately determine election outco mes (Shaw 1 999 a, 1999b ). As a conse-quence , seeking evidence of campaign effects by analyzing the behaviorof all voters is likely to miss subtle effects that have large implicationsin the aggregate.


4/8

= TABLE DI S TR I B UTI ONF CANDIDATEPPEALS C AM PAIG NS PA RTYN THE 1976-1996 PRESIDENTIAL (MAJOR CANDIDATES)

Traits and Neutral Party Opp osition PerformanceYear Can didate Background Issues Issues Issues Issues1976 Ford 66% (141) 1% (2) 16% (35) 8 % ( 1 7 ) 9 % ( 2 0 )Carter 44% (44) 1% (2) 16% (16) 13% (13) 26% (26)1980 Reagan 45% (94) 1% (3) 13% (28) 5% (10) 36% (76)Carter 53% (65) 1% (1) 16% (20) 16% (19) 14% (17)1984 Reagan 34% (45) 3 % (4 ) 2 6 % (3 3 ) 5% (6) 32% (41)

Mondale 30% (27) 2 % (2 ) 2 2 % (2 0 ) 34% (31) 12% (11)1988 Bush 52% (54) 0 31% (32) 3 % (3 ) 1 4 % (1 5 )Dukakis 29% (59) lo/, (2) 30% (60) 26% (52) 14% (27)1992 Bush 4 4 % (3 5 ) 2% (2) 24% (19) 16% (12) 14% (11)Clinton 36% (44) 1% (1) 23% (28) 17% (22) 23% (28)1996 Dole 34% (44) 0 50% (66) 11% (14) 5% (6)Clinton 8 % (1 6 ) 6 % ( 1 2 ) 42% (89) 30% (63) 14% (32)1976-1996 All Candidates 39% (668) 2% (31) 26% (446) 15% (262) 18% (310)Note: Data collected from all ava~lable ampalgn ad\zert~sements roduced by major party ca nd~ dat es Values In parentheses are number of unique issueappeals gleaned from the advert~sements De ter m~ na t~ on for each party5 candldates was taken from Petrocik (1996 ) and Petrocik et al.f I s s ue o~vner s h~p(2003-2004)

the mood of the coun try Stims on, MacKuen, and Erikson(1995) suggest that politicians are informed about move-men ts in the natlon's ideological tenor and adapt th eir behav-ior In response to changes In the publlc's mood, a strategythey refer to as rational anticipation. Thus, when the publicmood does not provide a context amenable to issues ownedby a candidates' party, candidates may opt to trespass inorder to app ear in line with the deslres of the electorate.

At a more expliclt level, this same loglc may result In agreater likelihood of candidates discussing issues that areimportant to voters even if the opposition party owns thoseIssues. In thls case, trespassing may allow candldates toassociate themselves wlth concerns that are already sallentin the mlnds of voters, as opposed to simply abdicatingthese issues to their o ppon ents. Aldrich an d Griffin (2003:247) succinctly capture this point by noting that "lf votersidentify certain issues as priorities, we would expect thecandidates to speak more often and show more advertise-ments about these issues."

Occu rrences of issue trespassing also may be affected byvoters' expectations abo ut the role of the pre siden t. Jacob -son (1990) , for instance, suggests that the electorate expectspresidents to pursue diffuse collective goods such as bal-ancing the b udge t, lowering taxes, and greater g overnmen-tal efficiency and evaluates presidential candidates in termsof their competence in dealing with national problems.Given that voters' expectations about the role of the presi-dent positively correlate with issues owned by the Republi-can Party these institutional expectations may work to theadvantage of Republican candidates. As a consequence,Democratic candidates may feel obliged to discuss Republi-can issues in order to satisfy voters' wants. Indeed, thisdynamic may account for much of the partisan differencesin the data presented in Table 1

The Campaign ProcessIn addition to the context in which a campaign 1s occur-

ring, factors stemming from the campaign process mayaffect the decision to issue trespass. Foremost here is candi-dates' level of sup po rt, whlc h past research has sho wn to bea strong predictor of variation in candldates' behavior (e.g.,Damore 2002; Kahn and Kenny 1999; Shaw 1999a). Asapplied to the context of Interest here, I expect that traillngcandidates will be more likely to issue trespass in order toreconnect with voters. That is, given that the opposition'slead may stem from voter receptl\,ity to its campaign mes-sages, issue trespassing may pro v~ de he means by w hichtrailing candidates can attach themselves to issues that mayloom la rge in vote rs ' e \~a lua t i~ns .~ore generally, becauseof their disadvantaged stat us, trailing candidate s may bemore willing to accept the risks associated with issue tres-passing. In contrast, candidates who are ahead in the pollshave little incentive to alter their messages

A second campaign attribute that may affect candidates'decisions to trespass is the tone of their campaign mes-sages. Here, I anticip ate that to increase the veracity of their

j However, 1 expect that when candldates Issue trespass they will alter theframes through whlch they discuss opposltlon owned lssues (Petroc~k1996) L'v'hereas a Republican may choose to discuss crlme reduction Interms of strengthening law enforcement, a Democrat may focus on theIssue by hlghllghtlng Investments In ed uc at~ on nd job tralnlng. By uslngan alternative frame to present an Issue candldates may be able toreshape voters' understanding of an Issue In man ner that hlghllghts t h e ~ rperce ~ved trengths More generally. Chong (19 96) and others argue thatthe frames through whlch elltes present lssues can have strong effects onhow voters understand what lssues are and are not about. However. todate, the study of how candlda tes choose to frame Issues IS a \old In thepolltlcal comm unlcatlon l~ teratur e Slmon 2002)


5/8

attacks, candidates' party-owned issues will form the basisof candida tes' negative appeals. In so do ing, candidates canexplicitly highlight their oppon ents' weaknesses over issuesthat accent their credibility. Moreover and as is argued else-where (Damore 2002), the returns for discussing party-owned issues positively may be limited given that votersmay be inclined to associate candidates with such issuesfavorably. However, when candidates issue trespass theywill do so positively in o rder to assert their credibility overissues that might otherwise be perceived as weaknesses. Ifcandidates were to use opposition-owned issues as thebasis of their negative appeals, these attacks may faltergiven that candidates may not be perceived as credible inthis regard.

Data gathered from all available campaign advertise-ments produced by major party candidates competing inthe 1976-1996 presidential elections are used to test thehypotheses specified above. An overview of the data collec-tion, coding, reliability, and validity is provided by Damore(2002, 2004). Due to the dichotomous nature of thedependent variable, logistic regression is used to estimatethe model. Robust standard errors (clustering on each can-didate) are used to control for lack of independence amongobservations. The model's unit of analysis is the appeal,which represents a unique mention of an issue within anad.6 The d etermination of issues follows Smith (1 985 ). Thenum ber of observations in the model is 708 taken from 476ads. Excluded from the d ataset are appeals focusing on ne u-tral and performance issues, as well as appeals addressingtraits and candidates' backgrounds.Variables and Measures

The de pen dent variable is a dichotom ous measure that iscoded 1 for a trespassed issue and 0 for a party-ow ned issue.Data used to create this measure are taken from Petrocik(199 6) and Petrocik et al. (2003-20 04) an d are summ arizedin Damore (2003).Four independent variables are specified to assess theaffect that the larger context in which a campaign is occur-ring has on candidates' likelihood of issue trespassing. Adum my variable, Democrat, is include d to test if Dem ocraticcandidates are systematically more likely to issue trespassthan their Republican counterparts. Salient is included toassess the affect that issue specific public opinion exerts oncandidates' decisionmaking. The variable is measured as thepercent of respondents identifying an issue as being themost important problem in pre-campaign Gallup polls.

Using the appeal as the u nit of analysis provides a finer measure of can-didate behamor glven that candidates discuss multiple Issues in anadvertisement. often focus on the same Issues or trans multiple timeswthin an advertisement, and typically include posltlve and negatlveappeals in the same advertisement.

Both variables should have a strong and positive effect onthe likelihood of issue trespassing.Two interaction terms, Democrat X Mood Change an d

Republican X Mood Change, are included to evaluate theinfluence of public mood on candidates' decisionmaking.The partisan co mpo nents of the terms are dumm y variablesand the mood component is measured as the change indomestic policy mood between the third quarter in the yearprior to the election and the third quarter of the electionyear using data taken from Stimson (19 99). Negative valuesfor the mood term suggest that the country is moving in aconservative direction a nd p ositive values are indicative of aliberal shift. Thu s, the sign for the Democratic term shouldbe negative and the sign for the Republican term should bepositive.

Support Dqjerence and Tone are included to assess theinfluence that the campaign process has on candidates' deci-sionmaking. Support Dij'Jerence is measured as the differencein candidates' poll standings (moving over time). With theexception of 1 99 2, the data for this measure are taken fromGallup. For 19 92, data presented by Goldman et al. (199 4)are used." anticipate that the variable's coefficient will benegative, indicating that candidates who are behind aremore likely to issue trespass. Tone is a dum my variable thattakes on the value of 1 if the tone of an appeal is negative,and 0 if positive. An appeal is considered negative if itfocuses on a candidate's opp one nt and positive if it focuseson the cand idate producing the ad (e.g., Damore 200 2). Thesign for the coefficient should be negative suggesting thatwhen candidates go negative party-owned issues will formthe basis of their attacks.Three control variables also are included in the model.Opposition Emphasis is used to control for discussion ofissues that may result from the dialogue that emerges overthe course of a campaign. Specifically, past research (e.g.,Damore 2003; Simon 2002) finds that as the opposition'sattention to an issue increases so does the likelihood thatcandidates will discuss the same issue. The variable is meas-ured as the percent (moving over time) of the oppositioncandidates' total appeals devoted to an issue. CandidateEmphasis is include d to control for the affect that candidates'decisionmaking at earlier points in a campaign may have o ntheir subsequent decisionmaking. Again, the variable ismeasured as the percent (moving over time) of candidates'total appeals devoted to an issue. Finally, because many ofthe model's variables are temporally dependent, I includeDays Prior to Election to control for time. The variable ismeasured as the number of days prior to Election Day onwhich an appeal was generated.

' ue to coll~nearityssues, Mood Change IS not ~n clu ded s a main effect.1use the data presented by Goldm an et al. (1994) for 1992 because theypromde tracking data for the campaign. Gallup does the same for 1996.However. for the other four campaigns, Gallup reports poll results inter-mittently. For dates when candidates' poll numbers were not available.these data are estimated assuming monotonic linearity between dateswhen Gallup did report polling data.


6/8

Table 2 presents the results of the logit analysis for themodel examlnlng the factors that affect candidates' deci-sions to discuss issues owned by their opponentk party. Thefit of the m odel a ppears to be stron g. The significance of thechi-square value suggests that the inclusion of the inde-pend ent variables provide a significant improvement in themodels explanatory power. The PRE statistic indicates thatthe model makes a modest improvement in reducing pre-diction error. Inspection of the paramete r estimates suggestssupport for the hypotheses specified above as the coeffi-cients for all six variables of substantive interest are in thepredicted and all but two obtain conventional levels of sta-tistical significance.'

The performance of the variables assessing the Influencethat the context In which a campaign IS occurring exerts oncandidates ' l ike l~h ood of issue trespassing suggests anumber of insights into the process by which campaignagendas are formed. First, consistent with the descriptivedata prese nted in Table 1 , the sign an d significance forDemocrat suggests that Democratic candidates are morelikely to discuss issues owned by the Republican Party ascompared to the propensity of Republican candidates to dis-cuss Democratic issues. This behavior is likely driven byvoters' expectations about the role of the presidency That is,because voters' expectations about the type of issues thatpresidential candidates should address correlate with issuesassociated w ~ t hhe GOP, i t appears that D emocratic candi-dates strategically adapt the content of the~rmessages Inresponse to these expectations.

Consistent with Aldrich and Gr~ffi n 200 3), the resultsalso indicate that candi dates are more likely to issue trespassif the electorate has iden tified an issue as a priority. Specifi-cally, the sign a nd significance for Salient suggests that as theimp ortanc e of an issue to the electorate increases so does thelikelihood that the candidates w11l discuss the issue even ~fit is associated w ith th e opp ositio n party. Lastly, althou gh inthe predicted direction, the insignificance of the publicmood interaction terms indicates that change in the coun-try's mood does not systematically affect candidates' deci-sions to issue trespass.'"

The performances of the two variables assessing theinfluence that the campaign process has o n occurrences ofissue trespassing are consistent with theoretical expecta-tions. Th e sign an d statistical significance of Support Differ-ence indicates that candidates who are trailing in pre-elec-tlon polls are more likely to discuss issues associated withtheir opponent's party, presumably as a means to jump-startth e~ r ampaigns. The performance of Tone suggests that

" Because Opposttton Emphasts, Candtdate Emphasts, an d Days Pnor to Elei-tton Day are Included as controls. 1do not d~sc usshelr substantive effects"' \%'hen the mode l IS run separately for each partyk cand~dateshe moodterm remalns In the predicted dlrectlon and IS statlstlcally slgnlf~ cant orDemocrats at the .03 level and approaches slgn~flc ance or Republicans(p = 08)

CoefficientsIndep enden t Variables (robust standard errors)Democrat 1 , 0 3 * *

( .358)SalientDemocrat * Mood Change -.007

( .037)Republican * Mood ChangeSupp ort DifferenceTone

Opposition Emphasis - .22 3( .257)Candidate EmphasisDays Prior to Election Day ,0 02

( . 005 )Constant

Num ber of Observations 708Wald chi-square 61 44Log Likelihood -415 64PRE (tau ) 7Oh* * * p< ,001, * * p < .01, * p < 05 (one tailed lest), robust slandard errorsclusler on each candidate.

when candidates attack they are more likely to do so byaddressing issues owned by their own party and when can-didates present a positive message they tend to use opposi-tion owned issues as the basis for these appeals. As dis-cussed above, this behavior is designed to maximize theveracity of candidates' appeals.Due to the nonlinear nature of logit analysis, the coeffi-cients presented in Table 2 do not indicate the linear effectsof the ind epend ent variables o n the d epend ent variable. Toassess more clearly the substantive effects of the analysis,Table 3 reports predicted probabilities for values of the sta-tistically significant variables. For these predictions, thevalues of the independent variables of interest are variedwhile all other variables in the model are held at theirmeans. Thus, these values capture the net affect of the Inde-pend ent variables on the depen dent variable.

Examin ation of the values presented in Table 3 Indicates thatthe model's statistically significant variables also have strongsubstantive effects. This IS most apparent in the difference in


7/8

Value of Probab ility ofInde pend ent Variable Issue TrespassingCandidate Party

Democrat .45Republican .22TonePositive

NegativeSalient

010 %2 0%

Sup port Difference+1 3 (-1 SD) .280 .35 -13 (+1 SD) .44 -26 (+2 SD) .52

Note: Values are net effects, all other v ar~ able sn the model are held at th e~ rmeans.

the probabilities of issue trespassing for Republican andDemocratic ca ndidates. All else equa l, Democratic candidatesare 23 percent more likely to issue trespass. The predictedprobabilities for Tone and Salient also suggest comparableeffects. If, for example, can didates pre sent a negative message,the probability that they will use opposition issues as the basisof these appeals decreases by 1 5 percent as comp ared towhen presenting a positive appeal. The change in the proba-bility of issue trespassing also appears to be strongly influ-enced by the saliency of an issue in the electorate. Ind eed , if10 percent of the electorate identifies an issue as a priority, thelikelihood that candidates will discuss the issue even if it isassociated with the o pposition party increases by 1 4 percent.This value increases an additional 13 percent as an issue'ssaliency increases to 20 pe rcent. Finally, changes in the p rob-ability of issue trespassing for candidates who are trailing inthe polls suggest similar effects.

As mentioned at the outset, the study of candidatebehavior and campaign strategy has received much lessattention in the literature, particularly as compared to analy-ses of voters. This effort seeks to help fill this lacuna byextending our understanding of the agenda-formationprocess in the campaign arena by assessing the conditionsund er wh ich presidential cand idates will discuss issues thatare associated with their op pone nts' party. The result of theanalysis indicate that candidates' decisions to issue trespassare shaped by their partisanship, the tone of their campaignmessage, their competitive standing, and the importance of

an issue to the electorate. The model failed to detect sys-tematic evidence that changes in the public moo d affect can-didates' decisionmaking (although see note 10).

The implications these findings have for our unde rstan d-ing of campaign strategy in presidential elections are at leastthreefold. First, consistent with prior research, t he analysissuggests that candidates' level of support accounts for agood deal of variation in candidate behavior. Here, I findthat cand idates who are trailing in the polls are more likelyto alter their messages in a manner that distance their can-didacies from the issues traditionally championed by theirparty In so doin g, these cand idates are willing to accept thepotential risks that may result from issue trespassing inhopes of reenergizing their campaigns. In contrast, candi-dates who are ahead in the polls have little incentive toadjust their messages, and as a consequence , they app ear tobehave in a more risk adverse manner.

Second, the evidence presented here lends credence toAldrich and Griffin's (2003) argument that the priorities ofvoters should exert a strong influence on the issues that can-dida tes choose to discuss. Specifically, the re sults indicatethat issues that are perceived as important to the electorateare more likely to be discussed by presidential candidatesregardless of the issues' partisan ownership. Although thisfinding could be construed as suppor t for Aldrich and Grif-fin's (2003) c ontention that the issue ownership appro ach toagenda setting is problematic (because it presupposes thatcandidates' selection of issues are made irrespective of thewants of voters), I am unw illing to go that far." Given thepreponderance of evidence indicating that party-ownedissues are a staple of candidate rhetoric in presidential cam-paigns (e .g., Petrocik et al. 2003 -200 4), a more fitting con-clusion on this point might be that the discussion of party-owned issues is tempered by pragmatic assessments of theissue environment in which a campaign takes place.

Finally, these findings offer insight into the difficultiesfacing Democratic presidential candidates as they competeto define campaign agenda s on terms favorable to their can-didacies. As Jacobson (1990) suggests, traditional Democra-tic themes run counter to the expectations that voters haveabout the type of policies and goals that presidents shouldarticulate. More to the po int, as the data in Table 1 and themultivariate analysis suggest, Democratic candidates areaware of this dynamic and have generally sought to over-come it by discussing Republican issues. However, with theexception of Bill Clinton, these candidates have been unsuc-cessful in convincing the electorate that they are credibleover suc h issues. More broadly, the agenda setting strategy

'I Aldnch and Gr~ffin 2003) base t he ~r nalys~s pon campalgn adver-tlsements and speeches made by A1 Gore and George Bush du nng the200 0 presidentla1 campalgn Interest~ ngly n th e ~ r naly s~sof Issue ou n-e r s h ~ pn the 1952-2000 pres ~den t~alampaigns Petroc~ k t al (2003-2004) hnd that the 2000 campalgn IS an outl~erhat does not comportw t h e ~t he r he ~r r m y theoretical expectations Specifically, they findthat In 2000, Republ~canGeorge Bush not Democrat A1 Gore, was thecan d ~d a temost likely to Issue trespass


8/8

employed by recent Democratic presidential candidatesraises a host of questions. For instance, pragmatically, onemight speculate if it is in the Democratic Party's best Inter-est to nominate cand idates who have credibility in Republi-can turf. Normatively, we might questlon if candidateswhose m essages emulate those of the opposition are capableof engaging voters an d promding a meaningful choice.

While the above insights are Important a nd for the mostpart novel, I would be remlss to ignore the limitations ofthis effort. Most notably, the theory and analysis advancedhere is llmited to the study of presidential candidates.Clearly, these con tests are atypical of m ost cam palgnsbecause of the medla attention they co mm and , the voterinterest they receive, and the ability of presidential candi-dates to disseminate their messages. As a consequence,determining the generalizability of the analysis presentedhere IS difficult. At th e very least, these findings are consls-tent with Petrocik et a1.k (2003-2004) analysis examiningpatterns of candidate attention to party-owned issues in the1952-2000 presidentlal elections.

With this sald, I conclude by notlng that this effort hasprovided a unlque examination of an important, albeltunder-studied, component of the campaign process.Whereas past research has tended to concentrate on theeffects that information presented durin g a campaign has onvoters' decislonmaking, this effort has sought to increaseour knowledge about the process that shapes the informa-tion that candidates choose to present to the electorate, andin so doing, extend ou r understanding of the su pply side ofthe campalgn process

Al dr~ ch , John . . and John D . Griffin 2003 . . 'The Pres~dency ndt he C ampa ~ gn: r ea tl ng Voter P r~ o r~ t ~ es "n the 2000 Elect~onIn M~chaelNelso n, e d . , The Presidency and thc Po lit~ia l ystem,7th e d. Washington DC: Congressional Quarterly.

Campbell . Angus. Phll~pConverse. Warren Miller, and DonaldStokes. 1 96 0. The Amcncan Voter. New York: W ~l ey

Carmines, Edward G.. and James H. Kuklmski 1990 "lnc ent~ ves ,Opportunit ies, and the Log~ c f Public Op lnlon In Amer~c anPo l~t ~ca lepresentat~on." n John A Ferejohn and James H.K u k l ~ n s k ~ . d s . Inform at~on and Democratic Processes, edChicago. University of Chicago Press.

Cho ng, Dennis. 1 996 . "Creatlng Comm on F rames of Reference onP o l ~ t ~ c a lssues." In D~anaC. Mutz, Paul M. Sn~der ma n, ndR~chardA. Brody, eds , Polit~calPersuasion and Attitude Change.Ann Arbor: Un ~v ers ~ty Press.f M ~ c h ~ g a n

Conover, Pamela Johns ton, and Stanley Feldman 1989 . "Ca nd~ datePerception In an Amb ~gu ousWorld. Campaigns, Cues, and Infer-ence Processes." Amencan Journal of Political Science 3 3: 91 2-40.

Damore, David F 2 002. "C and ~da te trategy and the Dec~s ion oGo Negat~ve." olltical Research Quarterly 55: 669-86.

2003. "Issue Response in Pres~dentialCam pa~gn s." niver-sity of Nevada, Las \'egas. Typescript.

2004 "Us~ ngCampaign Advertisements to Assess Cam-palgn Processes."Journal of Pol~ticalMarketing 3: 39-59 .

Do wn s, An tho ny 19 57 . An Economic Theory of Democracy. NewYork: Harper.

Goldman, Peter, Thomas M. DeFrank, Mark Miller, Andrew Murr,and Tom M athew s. 1 99 4. Questfor the Presidency. College Sta-tlo n: Texas A&M Uni\gersit):

Hamill, Ruth C ., Milton Lod ge, and Frederick Blake. 198 5. "TheBreadth, Depth, and Util~ty f Class, Part isan, and ldeolog ~calSchem ata." AmeticanJournal of Political Science 29 85 0-7 0.

Jacob s, Lawrence, and Robert Y. Shapiro. 1994 . "Issues, CandidateImage, and Priming: The Use of Pr~vat e olls In Kennedyk 196 0Presidential Campa~gn."American Political Science Review 8527-40.

Jac obs on, Gary C. 1 99 0 The Electoral Origins of Divided Govern-ment. Boulder,CO: W estview Press.

Kah n, Kim F r~ d ki n , nd Patrick J. Kenney. 1 999 . The Spectacle ofU.S. Senate Campaigns. Princeton, NJ: Princeton Un~versityPress

Krosnick, Jon A , and Donald R. Klnder. 19 90. "Altering the Fo un-dations of Support for the President Through Prlmlng " Amerl-can Polltical Science Review 8 4: 49 7-5 12 .

L u p ~ a ,Arthur. and Mathew D . McC ubb ~ns 1998 The DemocraticDilemma. New York: Cambridge Un ~v ers ~tyress

Norpoth. H elmut, and Bruce Buchanan. 19 92. "Wanted: The Edu-cation Pre s~ de nt, ssue Trespassing by Political Cam paigns ."Public Opinion Quarterly 56: 87-99Petro c~k ,ohn R. 19 91. "Divided Governm ent: Is ~tAll In the Cam-paigns?" In Gary W. Cox an d Samuel Kernell, eds. The Politicsof Divided Government Boulder, CO: Westv~ew ress.

1996 . "Issue Ownership In Pre s~d en t~a ll ec t ~ ons ,with a1980 Case Stud):" AmericanJournal of Polit~calScience 40: 82 5-50

Petrocik, Joh n R.. Willham L Ben o~t, nd Glenn J Hansen 2003 -2004 "Issue Ow ners h~p nd Pres~den t~alampa i gn~ ng , 952-2000 " P ol ~ t ~ t a lcicnce Quarterly 118 599-626

Rahn. Wendy M 19 93 . "The Role of Pa rt~ san tereotyp es In Infor-mauon Processing about Pollt~calCand~ da tes Amrn~ unournalof Polit~ccil ~ien ie 7 472-9 6

Riker, Willlam H . 199 3. "The Speclhcatlon of Issues " In WllllrimH Riker. ed . Agenda Formailon Ann Arbo r: Unlverslty ofMich~gan ress.

Shaw, Daren R. 1 999a . "The Methods beh lnd the M adness: Presl-dential Electoral College Strateg~es, 988-1996." ournal of Pol-itics 61 . 893 -91 3.Shaw, Daron R 199 9b. "The Effect of TV Ads and C an d~ da teAppearances on Statew~de res~dentialVotes, 1988-96."Amer-ican Polit~calScience Rev~ew 3 . 34 5-6 2.

S ~ m o n ,Adam F. 20 02 The Winning Messuge. Cand idate Beha v~or;Campaign Discourse, and Democracy. New York. CambridgeUniversity Press.

S m ~ t h ,Tom W. 198 5. "The Polls: Am er~cakMost lmportant Prob-lems, Part I: Nat~onal nd International " Public Opinion Quar-terly 49: 264-74

St ~m so n,am es A. 19 99 . Public Opinion in Amcnca Moods, Cycles,and Swings, 2nd ed . Boulder, CO Westv~ew.

S t ~ mso n ,ames A , , M ~ chae lB. MacKuen, and Robert S. Er~kson.1995 . "Dy nam ~c Representation." Am er~an Pol~tical ScienceReview 89 543-65.

Received: October 20 , 2003Accepted for Publication: December 5 , [email protected]

Documents

The Dynamics of Issue Ownership