Party bias in political-advertising processing: results from an experiment involving the 1998...

The processing of political advertising is inherently complex. Variations in the environment in which political advertising operates make it difficult to detect the effects of advertising, and differences in voters make examinations even more complicated. Findings regarding the influence of

political advertising generated from past studies are inconsistent. The inconsistency in findings is partly due to the exclusion of important variables. Specifically, one of the important variables that needed to be taken into account was voters' party orientations. This study proposes that voters' party orientations will bias the way they process political-advertising messages and moderate the impacts of political advertising. Taiwan was considered a suitable location for conducting the study because 68% of voters in Taiwan identify with a specific party (Xie 1996).

The reviewed evidence pertaining to party-bias effects in political-advertising literature suggests that subjects' attitudes toward an affiliated candidate may alter the way they respond to political-advertising messages (e.g., An 2002). Biased processing of political-advertising information results in the acceptance of proattitudinal information and the denigration of counterattitudinal information. Most important, past research has indicated that existing party alignment leads to polarized effects, generating strengthened support for affiliated candidates and reduced support for opposing candidates (Ansolabehere and Iyengar 1995; Kaid 1997; Kaid and Tedesco 1999). Drawing on past research, this study reasons that the influence of exposure to political advertising will not be robust across all voter classifications in the 1998 Taipei mayoral election. Specifically, for voters with a specific party identification, exposure to positive ads will help the candidate of the same party but will have a negative impa ct on the competing candidates of other parties. In a similar way, the damage incurred by initiating attacks on opponents and by being attacked will not be observed in voters with affiliated party orientations. In addition, past literature has suggested that independent voters are less influenced by campaign advertising (e.g., Ansolabehere and Iyengar 1995; Kaid 1997; Kaid and Tedesco 1999). Thus, this study argues that only partisans, who are more emotionally involved with either candidate, will show a bias in their processing of political-advertising messages.

Most studies of political advertising published in international academic journals are concerned with how political advertising works in Western cultures. Given that the effects of ad campaigns should be context-specific (West 1994), findings generated from Western settings may not apply to elections in an Asian country. However, party bias should be an inherently important variable when examining campaign effects across cultures or nations with a party-system government. Therefore, this study attempts to determine whether findings and theories regarding party bias in processing ad messages developed in the United States can be applied to explain or anticipate what effects political advertising may have in other cultures.

This study investigates the influence of political advertising in an election in Taiwan. Past studies conducted in Taiwan have primarily analyzed advertising strategies employed by candidates through the use of content analysis (e.g., Wang and Chen 1995), with little research devoted to exploring how political advertising affects candidate evaluations or voting preferences. Therefore, it is still not clear whether exposure to political advertising has any political consequences. In contrast to past research in Taiwan, the present study focuses on the possible influences that political advertising may exert. This exploration will concern three effectiveness measures pertaining to voters' responses toward the ads, their evaluations of the candidates, and their intention to vote for the candidates. Since 66.1% of the population of Taiwan reads newspapers regularly, and more than half of the campaign money was spent on newspaper advertising in this election, this study will focus solely on newspaper advertisement s. Due to a lack of literature regarding newspaper advertisements, however, the studies reviewed are concerned mainly with televised political spots.

This investigation includes two basic forms of political advertising: positive advertising and negative advertising. The use of negative political advertising is on the rise in Taiwan. Specifically, 19% of the ads aired in the 1996 presidential election in Taiwan were negative (Chang 2000), whereas 32% of the ads in the 1998 Taipei mayoral election contained degrading information about opponents (Chang 1999a). Therefore, it is also important to examine the influence of negative ads in Taiwan, a state whose culture shuns conflict and is more oriented toward harmony among its people.

BACKGROUND TO THE ELECTION

The 1998 Taipei City Mayoral Election

Taipei is the center of commerce and the hub of transportation in Taiwan. It is the most populous city in Taiwan, with 2,263,000 inhabitants, accounting for 13% of the total population of Taiwan. The central government is also located in Taipei. The election for Taipei's mayor is held once every four years and is highly visible in the mass media nationwide. The competitors in the 1998 election were the incumbent, Chen Shui-bian, representing the Democratic Progressive Party (DPP), and two challengers, Ma Ying-jeou and Wang Chienshien, who were nominated by the Kuomintang (KMT) and the New Party (NP), respectively. It was a close race between Chen and Ma, with Wang running a distant third. Both leading candidates allocated a significant amount of campaign money to media advertising. Chen spent U.S.$ 1.3 million on newspaper ads and U.S.$580,000 on television ads. Ma spent U.S.$1.7 million on newspaper ads and U.S.$1.5 million on television ads. Eighty-one percent of eligible voters cast their votes, and Ma won the election by a margin of 5%.

The Party System in Taiwan

Taiwan has a multiparty system of government. There are three leading political parties: the KMT, the DPP, and the NP. A new party, the People First Party (PFP), was formed after the 2000 presidential election, but this new change has no bearing on the 1998 Taipei mayoral election. The KMT was the ruling party from 1949 until the DPP took power in 2000, but this new change does not pertain to the political situation when the 1998 Taipei mayoral election took place. In 1986, antigovernment forces formed the DPP, whose mission was to fight for Taiwan's independence. It was not until later in that same year, however, that martial law was lifted and people were legally allowed to form political parties. Eighty-four political parties have registered since then, but among these parties, only the DPP, the PFP, and the NP have had significant political influence.

POSITIVE ADVERTISING VERSUS NEGATIVE ADVERTISING

Political advertising can be generally categorized into positive advertising and negative advertising. It has been commonly agreed that positive advertising focuses on the sponsoring candidate by promoting his or her merits and showcasing his or her accomplishments or issue positions (e.g., Ansolabehere et al. 1994; Shapiro and Rieger 1992; Tinkham and Weaver-Lariscy 1993). In contrast, there seems to be no agreement among researchers regarding exactly what constitutes negative advertising or how many different subtypes it has. It has been suggested that, in general, negative political advertising assails the sponsor's opponent by pointing out his or her personality flaws or by criticizing his or her poor job performance or broken promises, thereby creating an unattractive image of the opponent (e.g., Gronbeck 1992; Johnson-Cartee and Copeland 1991; Pinkleton 1997, 1998). Although a review of extant literature does indicate that different subtypes of negative advertising have been identified (Gronbeck 1992; J ohnson-Cartee and Copeland 1991), the term "negative advertising" was used loosely in past literature, with little discussion about which subtypes were specifically being explored. The most common approach seemed to be to distinguish negative ads from positive ads on the basis of whether the candidate criticized the opponent's weakness or advanced the candidate's own strength.

Therefore, based on this review of past research, negative advertising will be viewed in this broad sense and no specific subtypes will be identified when discussing past findings. In addition, in this study, positive advertising and negative advertising will also be operationalized in this broad sense, with the former promoting candidates' issue policies, thereby highlighting the candidates' capability, and the latter attacking opponents' issue policies, thereby revealing the opponents incompetence.

INFLUENCE OF POLITICAL ADVERTISING: INCONSISTENCY IN FINDINGS

Because attitudes toward candidates are important determinants of voting intention (Miller, Wattenberg, and Malanchuk 1986), researchers have attempted to establish the relationship between ad exposure and change of attitude toward the candidates. Survey studies show that ad exposure is positively related to respondents' evaluative responses toward candidates (Atkin and Heald 1976; Mulder 1979). However, West (1993) shows that although ad viewing may yield a positive influence on viewers' assessments of a candidate's personality traits--one of the basic elements of a candidate's image--the influence is nor strong across all presidential elections examined during a 40-year period. Similarly, West (1994) finds that ad exposure does nor always contribute to a significant increase in how much viewers like a given candidate. Experimental studies have also been conducted to explore the impacts of political advertising. For example, Cundy's 1986 quasi-experimental study indicates that exposure to advertising has a s ignificant positive impact on subjects' evaluations of the candidates. Kaid's 1997 experimental study has documented contradictory findings, however. Therefore, conclusions regarding the influence of political advertising on candidate evaluations cannot be easily reached due to inconsistent findings generated from past studies.

Inconsistent findings suggest that it is important to explore the moderating effects of other variables in establishing the effects of political advertising on candidate evaluations. Indeed, past studies have demonstrated the moderating impacts of ad attribute differences and individual voter differences. Ad attribute variations include the use of image/issue ads (e.g., Kaid, Chanslor, and Hovind 1992; Kaid and Sanders 1978), the use of music and settings (e.g., Thorson, Christ, and Caywood 1991), the credibility of the ad sponsor (e.g., Garramone and Smith 1984), and the length of the ad message (e.g., Kaid and Sanders 1978). Individual differences pertain to viewers' motivations (Garramone 1983), involvement (Garramone 1984), information-processing orientation (e.g., Garramone, Steele, and Pinkleton 1991), and party orientation (Kaid 1997).

Similarly, negative advertising has been shown to have impacts on candidate evaluations, yet the patterns of its influence appear to be complex and inconsistent. On the one hand, evidence for the boomerang effect, generating more negative effects toward the sponsor, rather than the target, has been documented by survey studies (Garramone 1984), experimental studies (Hill 1989; Thorson, Christ, and Caywood 1991), and time-series analyses (Jasperson and Fan 2002). On the other hand, negative ads have been shown to generate double-impairment effects by evoking negative responses toward both the sponsor and the opponent in survey studies (e.g., Merritt 1984) and experimental explorations (e.g., Basil, Schooler, and Reeves 1991; Shapiro and Rieger 1992).

Because past findings on the impacts of negative political advertising appear to be inconsistent--suggesting that the influence of negative political advertising sometimes hurts the opponent, whereas at other times it backfires--it is problematic to examine the influence of political advertising without taking important moderators into account. For example, the effectiveness of negative political advertising has been shown to be a function of variables such as ad attributes (i.e., type of ad--image ad versus issue ad, the perceived truthfulness of the ad, and the sponsor of the ad) (e.g., Roddy and Garramone 1988; Shapiro and Rieger 1992) and individual differences (i.e., existing preferences, partisanship, involvement levels, and general attitudes toward negative ads) (e.g., Christ, Thorson, and Caywood 1994; Faber, Tims, and Schmitt 1993; Garramone 1984). This study will thus focus on whether individual differences, such as partisanship, will have an influence on voter response to political advertising.

THEROLE OF PARTISANSHIP IN PROCESSING POLITICAL-ADVERTISING MESSAGES

One important individual difference well identified in political-behavior literature, but not as extensively examined in political-advertising research, is party identification. Party identification is a strong determinant of voting preference in Taiwan (e.g., Chen 1996), as well as in the United States (e.g., Campbell et al. 1960). Moreover, voters' party orientations have been shown to affect voters' processing of campaign information, including debates (e.g., Bothwell and Brigham 1983) and campaign ads (e.g., Ansolabehere and Iyengar 1995).

Past studies have demonstrated that voters with opposing party orientations respond to advertising messages in different ways. Kaid (1997; see also Kaid and Tedesco 1999) demonstrates that after being exposed to the positive and negative advertising of competing candidates, partisans polarize their image perceptions of competing candidates by strengthening their preferences for candidates of affiliated parties and lessening their evaluations of opposing candidates. Independents, however, were less susceptible to the influence of ad exposure, showing no significant evaluation shifts in any condition except for a negative evaluation change toward one candidate in an experimental session. Other studies focusing on the effects of negative political advertising suggest a double-impairment effect on candidate evaluations for partisans. In a survey study, Merritt (1984) demonstrates that exposure to negative political ads generates unfavorable evaluations of both the sponsor and the opponent. Similarly, Kaid and Boy dston's (1987) experimental investigation shows that Democrats are swayed less by a negative ad targeting a Democratic politician compared with Republicans, who react significantly more negatively to a Democratic politician. However, it is important to note that although attitude shifts are not as serious for Democrats as they are for Republicans, Democrats do express significantly more negative evaluations of the affiliated politician after ad exposure than before. Independents, however, did nor show significant attitude change.

In terms of voting support, Ansolabehere and Iyengar (1995) show that exposure to campaign advertising produces vote-preference changes in line with voters' party orientations. Partisans express striking vote-preference reinforcement for the affiliated candidate, even after one ad exposure. Independents, however, being distrustful of and alienated from politics, are less receptive to campaign advertising and show much less vote-preference change after ad exposure. Faber, Tims, and Schmitt (1990) obtained similar results on voting support, showing that the intended effect of an attack ad is more likely to occur among partisans of the sponsor, whereas backlash is mote likely to occur among partisans of the target. Findings of their study demonstrate that partisans' voting support for two competing candidates polarizes in accordance with their party affiliation.

In general, regardless of the use of negative attacks or positive appeals, consistent patterns of results point to the conclusion that a voter's existing preference biases his or her responses to political-advertising messages. What can explain the mechanism underlying parry-bias effects? The obtained evidence seems to be consistent with selective-processing literature, which argues that processing bias is a function of message perceivers' existing attitudes. In social psychology, it has been well established that when individuals have readily established preferences for one of a set of alternatives, they are motivated to process incoming information in a selective way, which produces additional reasons for the preference (for a review, see Frey 1986). This line of literature suggests that individuals process information in a biased way so as to confirm their existing attitudes and maintain their cognitive consistency. As a result, individuals pay more attention to evaluative consistent information and avoid evaluative inconsistent information. In political literature, for example, Sweeney and Gruber (1984) found that due to selective processing, Nixon supporters reported paying less attention to Watergate-related matters, and An (2002) has shown that Clinton's supporters paid more attention to Clinton's ads than did Dole's supporters.

In addition, psychology literature indicates that new information is distorted perceptually so as to favor an existing preference (Russo, Medvec, and Meloy 1996), Most important, confirming messages, in general, are perceived to be more positive and convincing, whereas discrepant information has been shown to be evaluated as more negative and less convincing (Lord, Ross, and Lepper 1979; Vallone, Ross, and Lepper 1985; Vidmar and Rokeach 1974). In political literature, Lodge and Milton (1999) show that when evaluating political issues, individuals' prior attitudes bias their responses, with attitude-congruent arguments being perceived as inherently stronger than attitude-incongruent arguments. In exploring the effects of the 1980 Reagan-Carter debate, Bothwell and Brigham (1983) demonstrate that judgments of who won the debate were biased in a way that was consistent with viewers' party orientations. Political advertising has also been shown to be evaluated in ways that suggest voters are nor fair-minded (e.g ., An 2002).

In line with evidence in selective-processing literature and political-communication literature, this study suggests that subjects are more likely to perceive political advertising by candidates nominated by their parties as more likable and persuasive than political advertising by candidates nominated by other parties. Driven by existing preference, this perceptual bias will apply to situations in which either positive or negative advertising is processed. Moreover, selective-processing literature suggests that selective exposure to information is more likely to emerge when individuals are committed (Brock and Balloun 1967) and when the choice is up to them (Frey and Wicklund 1978)--a situation somewhat similar to elections, in which voters are free to make their own choices. In a similar vein, Houston and Fazio (1989) demonstrate that biased processing is a function of attitude accessibility, which is determined by attitude extremity. Therefore, it is plausible to argue that independents, who do not have st rong existing preferences to start with, should have low accessibility of existing attitudes toward candidates when processing advertising messages, and thus will not respond to candidates' ad messages in a biased manner. Specifically, this study proposes the following hypotheses:

H1: There is a significant interaction between party identification and candidate difference on liking of the positive ads (H1a) and negative ads (H1b). Specifically, partisans perceive positive political ads (negative political ads) sponsored by candidates nominated by their affiliated parties to be more likable than positive political ads (negative political ads) sponsored by candidates of competing parties, whereas independents do not evaluate ads by different candidates in divergent ways.

H2: There is a significant interaction between party identification and candidate difference on the persuasiveness ratings of the positive ads (H2a) and negative ads (H2b). Specifically, partisans perceive positive political ads (negative political ads) sponsored by candidates nominated by their affiliated parties to be more persuasive than positive political ads (negative political ads) sponsored by candidates of competing parties, whereas independents do not evaluate ads by different candidates in divergent ways.

Biased processing is not limited to subjects' judgments of advertising messages. How subjects perceive the persuasive communication has been shown to further determine how much, and in which directions, that persuasive information will exert an influence (see Eagly and Chaiken 1993 for a review). In Lord, Ross, and Lepper's (1979) study, after reading mixed sets of information containing confirming messages and disconfirming messages, subjects reported polarized shifts in their attitudes toward the proposed issue. Similarly, as reviewed earlier, political-advertising literature also indicates that after being exposed to the positive and negative advertising of competing candidates, partisans polarized their image evaluations of competing candidates (e.g., Kaid 1997; Kaid and Tedesco 1999).

In an actual campaign context, as well as in this study, voters are exposed to advertising messages supporting opposing views on the same candidates. Drawing upon selective-processing literature, it is reasonable to argue that in the 1998 Taipei mayoral election, due to selective processing, advertising exposure is likely to polarize partisans' attitudes toward the competing candidates. Specifically, this study hypothesizes that in comparison to partisans who have no ad exposure, partisans who are exposed to positive ad messages sponsored by an affiliated candidate that are perceived to be favorable and credible by partisans, are more likely to express more positive evaluations of the affiliated candidate. Similarly, the positive bias of partisans toward the affiliated candidate will emerge when subjects are exposed to negative political-advertising messages. In addition, this study proposes that, given that political advertising--positive or negative--initiated by the opposing candidate is perceived to be un acceptable and highly incongruent with their prior attitudes, the perceptual bias of partisans of the competing party will lead to a negative effect on the opposing candidate.

H3a: There is a significant interaction between ad difference (no ad exposure versus positive ad exposure,) , party identification, and candidate difference on candidate evaluations. Specifically, compared with partisans who are not exposed to any ads, partisans who are exposed to positive ads generate more positive attitudes toward candidates nominated by affiliated parties and more negative attitudes toward candidates nominated by competing parties. Compared with independents in the no ad exposure condition, independents in the positive ad condition will not have different attitudes toward either candidate.

H3b: There is a significant interaction between ad difference (no ad exposure versus negative ad exposure), party identification, and candidate difference on candidate evaluations. Specifically, compared with partisans who are not exposed to any ads, partisans who are exposed to negative ads generate more positive attitudes toward candidates nominated by affiliated parties and more negative attitudes toward candidates nominated by competing parties. Compared with independents in the no ad exposure condition, independents in the negative ad condition will not have different attitudes toward either candidate.

The main purpose of employing political advertising is to win voting support. Ansolabehere and Iyengar (1995) find that exposure to even an experimental advertising message increases support for the sponsoring candidate substantially. They also demonstrate that for partisans, the positive or negative direction of change of voting support for candidates is determined by party affiliation congruency between voters and candidates. Similarly, Faber, Tims, and Schmitt (1990) demonstrate polarization effects on partisans' voting support for affiliated and opposing candidates. This study thus suggests that voters' party orientations will moderate the influence of ad exposure on voters' voting intentions for the two candidates.

H4a: There is a significant interaction between ad difference (no ad exposure versus positive ad exposure) , party identification, and candidate difference on voting intention for candidates. Specifically, compared with partisans who are not exposed to any ads, partisans who are exposed to positive ads generate higher voting support for candidates nominated by affiliated parties and lower voting support for candidates nominated by competing parties. Compared with independents in the no ad exposure condition, independents in the positive ad condition will not express different voting support for either candidate.

H4b: There is a significant interaction between ad difference (no ad exposure versus negative ad exposure), party identification, and candidate difference on vote intention for candidates. Specifically, compared with partisans who are not exposed to any ads, partisans who are exposed to negative ads generate higher voting support for candidates nominated by affiliated parties and lower voting support for candidates nominated by competing parties. Compared with independents in the no ad exposure condition, independents in the negative ad condition will not express different voting support for either candidate.

METHOD

Design

This is an experimental design with three ad type conditions: no ad, positive ad, and negative ad. Even though there were three candidates running in this campaign, opinion polls throughout the campaign indicated that only two candidates, Ma (KMT) and Chen (DPP), were major runners. Therefore, in this study the only concern was with how people who identified with either the KMT or the DPP responded differently to political-advertising messages concerning these two candidates. Given that in a real campaign, subjects would be exposed to competing ad messages from each candidate, subjects in each positive ad and negative ad condition were exposed to competing ad messages for both candidates.

Subjects

This study recruited 754 subjects. Two hundred and sixty-eight of them were eligible voters from Taipei City, recruited from all walks of life. The other subjects were students from a national university located in Taipei City, who had residency in Taipei City and were eligible voters. They were randomly assigned to each of the following conditions: no ad, positive ad with Ma's ad being exposed first, positive ad with Chen's ad being exposed first, negative ad with Ma's ad being exposed first, negative ad with Chen's ad being exposed first. Approximately 150 subjects were assigned to each condition. None of the ad order effects on dependent measures were significant. Therefore, responses of subjects who were assigned to the two different order conditions for positive ads were later collapsed, as were those for the negative ads.

Stimuli

Stimuli ads were mainly about issues, and real campaign issues were featured in the ads. Specifically, the stimuli ads were created by editing actual campaign-brochure materials issued by Ma and Chen, which were only available for the press or upon citizens' requests. Therefore, prior exposure to the campaign materials was not likely. It is important to note that sometimes political-advertising issues used in previous experimental studies were designed to be presented nonspecifically to avoid producing either agreement or disagreement. This study, however, tried to create political advertising that was as close as possible to what voters might be exposed to in an actual campaign, but new wording, visuals, and layouts were specifically developed to reduce any familiarity effects. Advertisements were imbedded in two magazine articles and two filler ads to reduce obtrusiveness.

Procedures

The study was conducted during the first week of the official election campaign period, which was from November 20 to December 4. Folders containing different stimuli were sorted randomly and subjects were given a folder without following any specific order. A typical folder included instructions and two news articles, followed by one ad for candidate Chen/Ma, one filler ad, one ad for candidate Ma/Chen, and a second filler ad. Those who were assigned to the "no ad" group were provided with folders that only contained news articles and filler ads. The news articles were either in issue frames or strategy frames. The type of news frames had been shown to exert no influence on subjects' evaluations of ads and candidates or on their voting support for candidates. Therefore, collapsing responses of subjects who were assigned to the two news frames was justified.

Subjects began by reading the instructions which stated that this study involved research on how people processed news articles. They were then asked to read the articles and ads at their own pace. Finally, they answered questions that measured their evaluations of the news articles and the ads, their attitudes toward candidates Ma and Chen, their voting intentions for each candidate, and their party identification. Subjects also responded to questions designed to assess other types of responses that were not specifically hypothesized in this study. They included interests and concerns in politics, political knowledge, political attitudes, such as cynicism toward political candidates, and internal and external efficacy.

Independent Variables

Ad Type

Ad type was a between-subjects factor, with three levels: no ad, positive ad, and negative ad. A positive ad for a candidate addressed the candidate's issue policies, thereby showing the candidate's competence. A negative ad by a candidate (the attacker) criticized faults or weaknesses in opponents' issue policies, thereby indicating that the target candidate was not capable. A pretest of ads utilizing a seven-point scale indicated that subjects perceived that positive ads for both Ma (KMT) and Chen (DPP) used more positive appeals, F(1, 753) = 45.10, p < .01, and involved fewer attacks than did negative ads for both Ma and Chen, F(1, 753) = 42.73, p < .01. At the same time, the positive ad for Ma did not differ from the positive ad for Chen on the following important characteristics: like/dislike, persuasive/not persuasive, convincing/not convincing, involving attacks/not involving attacks, using positive appeal/negative appeal. The negative ad for Ma did nor differ from the negative ad for Chen on these fiv e items, either. Therefore, the manipulation was deemed successful.

Party Identification

Subjects were asked to indicate the party with which they identified. All major political parties in Taiwan were listed, along with an option to choose no specific party alignment. Three hundred and forty-five subjects, or 45.8% of the recruited subjects, indicated that they did not identify with any of the listed parties. Four hundred and one subjects, or 54.1% of the recruited subjects, indicated specific party alignment. Among them, 61 subjects who identified with political parties other than the KMT and DPP were dropped from the analysis. Moreover, eight subjects did not respond to this question and were also dropped from the analyses. The number of subjects that remained in each condition is shown in Table 1.

Further analyses indicated that independents were less interested in and concerned about politics than were partisans. In addition, independents were more cynical than partisans and they expressed lower levels of internal political efficacy and external political efficacy. Partisans and independents did not differ, however, in terms of political knowledge and self-assessed susceptibility to the influence of political advertising. Finally, partisans and independents did not differ in their prior exposure to Ma's and Chen's ads.

Dependent Measures

Ad Evaluations

Ad evaluations were assessed with eight 7-point bipolar adjective pairs. The items were adopted from Christ, Thorson, and Caywood (1994); Hitchon, Chang, and Harris (1997); and Pinkleton (1997). Factor analysis with varimax rotation produced two factors with eigenvalues larger than 1. Factor 1, with an eigenvalue of 4.88, was labeled "ad liking." Factor 2, with an eigenvalue of 1.40, was labeled "ad persuasion." The scale was then treated as two subscales. The first scale, ad liking, which was composed of five items, had a satisfactory Cronbach's reliability a value assessed at .92. The five items were like/dislike, good/bad, pleasant/unpleasant, believable/unbelievable, and fair/unfair. The second scale, ad persuasiveness, which was composed of three items, had a satisfactory Cronbach's reliability a value assessed at .88. The three items were persuasive/nor persuasive, ethical/not ethical, and accurate/inaccurate.

Candidate Evaluations

Candidate evaluations were assessed with eight 7-point bipolar adjective pairs. Three items were used to determine subjects' liking/dislike of the candidates: like/dislike, good/bad, and pleasant/not pleasant. These three items were adopted from Hitchon, Chang, and Harris (1997). The other five items were obtained from Chang (1999b), who employed an open-ended question to explore voters' perceptions of what attributes an ideal political candidate in Taiwan should possess. The five items were: being concerned about the general public, being capable, being responsible, being moral, and being zealous to serve the people. A factor analysis with varimax rotation produced only one factor with an eigenvalue larger than one; therefore, the scale was summed and averaged. Cronbach's a was estimated as satisfactory at .93.

Intention to Vote for the Candidates

Subjects were asked "how likely" and "how probable" it was that they would vote for each candidate. The correlation of the two items was .89 (p < .01) for candidate Ma and .91 (p < .01) for candidate Chen.

RESULTS AND ANALYSES

MANOVA and ANOVA involving responses toward both types of political advertising were first conducted. Subject status (general voter or college student voter) was included as a covariate in all the analyses. Results of MANOVA indicated that the interaction between party identification and candidate difference was significant, F(1, 565) = 48.37, p < .01. ANOVA further showed that the interaction between party identification and candidate difference was significant on ad liking, F(1, 565) = 102.37, p < .01, and ad persuasiveness, F(l, 565) 90.75,p < .01. These significant results justified further tests of the hypotheses.

It is proposed that there is a significant interaction between party identification and candidate difference on liking of the positive ads (H1a) and negative ads (H1b). As expected, when subjects' responses to positive ads were analyzed, repeated ANOVA generated a significant two-way interaction between party identification and candidate difference on ad liking, F(2, 279) = 49.57, p < .01, which qualified further analysis of contrasts. Similarly, when subjects' responses to negative ads were analyzed, ANOVA generated a significant two-way interaction between party identification and candidate difference on ad liking, F(2, 279) = 49.61, p < .01. As shown in Table 2, the results of all the contrast analyses in the positive ad and negative ad conditions were consistent with expectations. Therefore, H1a and H1b were fully supported.

Hypotheses 2a and 2b concern ad persuasiveness. It is predicted that there is a significant interaction between party identification and candidate difference on persuasiveness ratings of the positive ads (H2a) and negative ads (H2b). As expected, when subjects' responses to the positive ads were analyzed, ANOVA generated a significant two-way interaction between party identification and candidate difference on ad persuasiveness, F(2, 285) = 40.74, p < .01. Consistent with expectations, when subjects' responses to the negative ads were analyzed, ANOVA generated a significant two-way interaction between party identification and candidate difference on ad persuasiveness, F(2, 285) = 40.97, p < .01. As indicated in Table 2, the results of all the contrast analyses were consistent with expectations. Therefore, H2a and H2b were fully supported.

Hypothesis 3a proposes that there is a significant interaction between ad difference (no ad exposure versus positive ad exposure), party identification, and candidate difference on candidate evaluations. Results of repeated-measure ANOVA indicated that the three-way interaction was not significant, F(1, 401) = 2.06, p = .13. However, because some researchers have argued that for theory-based hypotheses (e.g., Winer, Brown, and Michels 1991), simple-level comparison or lower-level interactions are acceptable, further analytical tests were conducted.

As expected, when evaluations of Ma were considered, a significant ad difference (no ad exposure versus positive ad exposure) by party-identification two-way interaction emerged, F(1, 401) = 4.93, p < .01. As shown in Table 3, however, the result of the planned contrast was significant for KMT identifiers, but the means were not in the expected direction. For DPP identifiers, a planned contrast between the positive ad and no ad condition on evaluations of Ma was not significant, but the means were in the expected directions. Moreover, results indicated that exposure to positive ads had no significant impact on independents' evaluations of Ma, which was consistent with expectations.

Similarly, when evaluations of Chen were considered, the ad difference (no ad exposure versus positive ad exposure) by party-identification two-way interaction was not significant, F(1, 401) = .08, p = .93. As shown in Table 3, a planned contrast between the positive ad and no ad conditions on evaluations of Chen was significant for DPP identifiers, but the means were not in the expected directions. For KMT identifiers, a planned contrast between the positive ad and no ad condition on evaluations of Chen was not significant, but the means were in the expected directions. In addition, consistent with expectations, results indicated that exposure to positive ads had no significant impact on independents' evaluations of Chen.

Hypothesis 3b suggests that there is a significant interaction between ad difference (no ad exposure versus negative ad exposure), party identification, and candidate difference on candidate evaluations. Results of repeated-measure ANOVA indicated that the three-way interaction was nor significant, F(1, 407) = .21, p .81. When evaluations of Ma were considered, the ad difference (no ad exposure versus negative ad exposure) by party-identification two-way interaction was not significant, F(l, 407) = 1.47,p = .23. Further analyses showed that for KMT identifiers, a planned contrast between the negative ad and no ad conditions on evaluations of Ma was not significant and the means were not in the expected direction. Although a planned contrast between the negative ad and no ad condition on evaluations of Ma was not significant for DPP identifiers, the means were in the expected directions, Moreover, results indicated that, consistent with expectations, exposure to negative ads had no significant impact on indepe ndents' evaluations of Ma.

On the other hand, when evaluations of Chen were considered, the ad difference (no ad exposure versus negative ad exposure) by party-identification two-way interaction was not significant, F(1, 407) = .43, p = .65. As shown in Table 3, a planned contrast between the negative ad and no ad conditions on evaluations of Chen was significant for DPP identifiers, but the means were not in the expected directions. For KMT identifiers, a planned contrast between the negative ad and no ad condition on evaluations of Chen was significant and the means were in the expected directions. In addition, as expected, results indicated that exposure to negative ads had no significant impact on independents' evaluations of Chen.

In sum, inconsistent with expectations, partisans expressed more negative attitudes toward affiliated candidates, as well as competing candidates, after exposure to either positive or negative advertising. In clear contrast, independents did not change their evaluations of either candidate after exposure to negative advertising. Therefore, H3a and H3b were weakly supported.

Hypothesis 4a suggests that there is a significant interaction between ad difference (no ad exposure versus positive ad exposure), party identification, and candidate difference on voting intention for the candidates. Results of repeatedmeasure ANOVA indicated that the three-way interaction was significant, F(1, 401) = 7.78,p < .01. As expected, when voting support for Ma was considered, the interaction between ad difference (no ad exposure versus positive ad exposure) and party identification was significant, F(1, 401) = 7.89, p < .01. As shown in Table 3, the results of all the planned contrasts were as expected. Also, when voting support for Chen was considered, the interaction between ad difference (no ad exposure versus positive ad exposure) and party identification was significant, F(1, 401) = 4.69, p < .01. As indicated in Table 3, the results of the planned contrasts for DPP identifiers and independents were as expected. Although a planned contrast was not significant for KMT identifiers, the means we re in the expected direction.

Hypothesis 4b suggests that there is a significant interaction between ad difference (no ad exposure versus negative ad exposure), party identification, and candidate difference on voting intention for the candidates. Results of repeatedmeasure ANOVA indicated that the three-way interaction was significant, F(1, 407) = 7.5 5,p < .01. Consistent with expectations, when voting support for Ma was considered, a significant ad difference (no ad exposure versus negative ad exposure) by party-identification interaction emerged, F(1, 407) = 5.27, p < .01. As indicated in Table 3, the results of all the planned contrasts were as expected. Similarly, when voting support for Chen was considered, a significant ad difference (no ad exposure versus negative ad exposure) by party-identification interaction emerged, F(1, 407) 6.11, p < .01. As shown in Table 3, the results of the planned contrasts for DPP identifiers and independents were consistent with expectations. For KMT identifiers, however, the result of the planned c ontrast approached the significant level and the means were in the expected direction.

In sum, partisans' voting support for their affiliated candidates increased significantly after exposure to either positive ads or negative ads. On the other hand, DPP partisans' voting support for Ma, the competing candidate, declined significantly. KMT partisans' voting support for Chen also declined after ad exposure, although the decline was not significant. Moreover, independents did not change their voting support for either candidate after ad exposure. Therefore, H4a and H4b were mostly supported.

It is important to note that although student-voter subjects and general-voter subjects in this study differed in percentages of party identification--for student voters, 18% were KMT partisans, 28% were DPP partisans, and 54% were independents, whereas for general voters, 31% were KMT partisans, 25% were DPP partisans, and 44% were independents, [chi square] (2, N = 688) = 14.81, p < .01--when responses of student voters and general voters were analyzed separately, results of all the analyses revealed similar patterns, suggesting that selective bias in accordance with subjects' party orientations occurred regardless of whether they were student voters or general voters.

DISCUSSION

Findings from this study suggested that the selective-processing mechanism that was implicated by voters' party orientations determined their responses to campaign advertising in Taiwan. It is important to note that when responses of all subjects, regardless of their party identification, were analyzed, exposure to either positive or negative ads did not exert significant influence on candidate evaluations or on voting support for candidates. This did not mean, however, that exposure to political advertising exerted no influence on voters. Specifically, when responses of subjects who identified with either the KMT or DPP were analyzed separately', interesting and consistent results emerged. These results generally confirm the idea that voters respond to ad information in a selective way, such that it reinforces their existing preferences. Therefore, if party orientation had not been taken into account, a researcher would have concluded that political advertising--either positive or negative--did not generate any impacts in the 1998 Taipei mayoral election.

In terms of responses to candidates' advertising messages, evidence in this study indicated that a selective bias appears to have emerged in partisans' ratings of ad messages sponsored by different candidates, in that advertising by affiliated candidates was favored to advertising by opposing candidates. However, it is important to note that in candidate evaluations, double-impairment effects emerged for both positive and negative political advertising when partisans' responses were analyzed. This is consistent with Merritt's 1984 findings that exposure to negative political ads generates unfavorable evaluations of both the sponsor and the target.

In terms of voting intention for each candidate, ad exposure polarized voters' responses. The inconsistency of the influence of ad exposure on voting support and candidate evaluations points out the importance of examining more than one effect indicator. Yet how can the inconsistency be explained? In this study, subjects were exposed to either positive messages or negative messages for two candidates at the same time. It was likely that seeing two candidates either promoting themselves or attacking each other reminded subjects of the manipulation techniques that candidates employ. As a result, their evaluations of both candidates deteriorated. Indeed, as noted earlier, subjects in this study were asked to rate their agreement with the statement "What candidates say depends on who is listening" with their mean response assessed at 5.93 on a seven-point scale. This indicated that their general attitudes toward political candidates were somewhat negative and cynical to start with.

At the same time, however, the enhanced voting support generated by partisans seemed to indicate that voters with strong political orientations had an enhanced sense of competitiveness between candidates after ad exposure and were worried that others would be influenced by advertising messages initiated by competing candidates. To lessen their discomfort, partisans showed stronger voting support for their candidates. Specifically, analyses of subjects' agreement with the statement "How likely is it that the general public is subject to the influence of political advertising" showed that a significant main effect of ad type emerged, with subjects who were exposed to negative ads and subjects who were exposed to positive ads generating relatively higher ratings than subjects who did not have ad exposure, F(2, 682) = .07, p = .05, [M.sub.negative ad] = 4.95, [M.sub.positive ad] = 4.76, [M.sub.no ad] = 4.58. This indicated that exposure to advertising might enhance their concern that political advertising would i nfluence others. Their enhanced concern might further increase their willingness to support their affiliated candidates regardless of deteriorated evaluations. It is therefore necessary to include measures of candidate evaluations as well as measures of voting support.

In view of the findings, it seems that two types of processes might have been functioning simultaneously when subjects were exposed to ad messages. On the one hand, voters felt disappointment at seeing candidates try to manipulate voters through campaign advertising. On the other hand, they were worried that others would be affected by the opponents manipulation attempts, and thus expressed stronger voting support for the affiliated candidate. It is likely that the closeness of a race may determine which process will have a dominant influence. For example, when a race is not competitive, double-impairment effects may emerge for candidate evaluations as well as for voting support, whereas when a race is very close, such as this one was, double-impairment may occur only for candidate evaluations, not for voting support, due to voters' worries that their affiliated candidate will not get elected. However, more explorations across races are necessary to provide more support regarding this possible theory.

Most important, findings in this study highlighted the importance of contextual variation, such as the competitiveness of the race, in understanding the effectiveness of campaign ads. Indeed, the race context also serves as an important foundation from which researchers may make meaningful interpretations of findings. Findings from experimental studies, void of race contexts, may not clearly help predict or explain campaign ad effects in actual elections. The importance of race competitiveness has been pointed out by West (1994), who argues that whereas it is less likely for lopsided races to show substantial media effects, it is more likely for competitive races to demonstrate significant media influence. Therefore, the competitiveness of a race always warrants consideration as an important aspect of campaign context differences for future research.

Independents in this study were not receptive to advertising messages. After exposure to either positive ads or negative ads, they did not evaluate candidates differently or express significantly different voting support for the two candidates. As discussed earlier, although independents and partisans were similar in their political knowledge and level of exposure to campaign ads, independents held more cynical attitudes toward politics, and thus might have been less vulnerable to the influence of campaign ads. It is also important to note that among the independents, 70.7% expressed their intention to cast a vote on election day, which was significantly lower than KMT partisans (89%) and DPP partisans (94%), [chi square](4, N = 593) = 14.49, p < .01. Taken together, independents' lack of response to advertising and their low willingness to cast their vote can be attributed to their cynical attitudes toward politics.

Findings of this study should be considered within its limitations, one of which concerns the manipulation of ad stimuli. Given that the experiment was conducted during an actual election campaign, the issues discussed in this experiment had to be real issues, presented in such a way as to avoid evoking any suspicion or counterarguments from the subjects. It was known that Chen and Ma were concerned about different issues; therefore, different issues were discussed in the ads for or against Chen and Ma. Scholars have raised concerns about using fictitious issues in experimental studies (e.g., Thorson et al. 1991), but in this study, using real issues did not provide a clean control of ad messages, and thus issue differences might have confounded the results. The second limitation concerns the use of student subjects in this study. All the students recruited to participate in this study resided in Taipei and had the right to vote in the 1998 Taipei mayoral election, but it should be noted that young voters wer e known to favor Chen to a greater degree than Ma. Another limitation of this study is the use of print ads. Print ads, lacking vivid imagery, might have generated different responses than televised spots. Regardless of these limitations, this study presents important insights into information processing of advertising messages in a non-Western context and thus enriches the existing literature.

TABLE 1

Sample Sizes: Party Identification and Student Versus Nonstudent Voters

                       KMT identifiers               DPP identifiers

                                   Percentage
              Number   Percentage      of        Number   Percentage
                of     of student  non-student     of     of student
             subjects    voters      voters     subjects  voters

Positive ad     67        48%          52%         83     71%
Negative ad     65        58%          42%         74     62%
No ad           25        52%          48%         29     72%
Total          156        53%          47%        184     68%

                  DPP                  Independents                Total

              identifiers
             Percentage                         Percentage
                 of        Number   Percentage      of        Number
             non-student     of     of student  non-student     of
               voters     subjects   voters       voters     subjects

Positive ad      29%        130       68%           32%        280
Negative ad      38%        147       71%           29%        286
No ad            28%         68       71%           29%        122
Total            32%        345       70%           30%        688

                       Total

                         Percentage
             Percentage      of
             of student  non-student
               voters      voters

Positive ad     64%          36%
Negative ad     66%          34%
No ad           67%          33%
Total           66%          34%

Note: KMT = Kuomintang; DDP = Democratic Progressive Party.

TABLE 2

Results of Planned Contrasts for Testing H2a and H2b

                                       Positive


                       Ma      Chen

                     Means     Means     Results of planned
                                         contrasts

Ad liking

  KMT                 5.05     3.95      F(1, 66) = 28.24, p < .01
  DPP                 3.71     5.35      F(1, 82) = 66.06, p < .01
  Independents        4.46     4.24      F(1, 129) = 2.21, p = .14

Ad persuasiveness

  KMT                 4.54     3.80      F(1, 66) = 19.85, p < .01
  DPP                 3.55     4.93      F(1, 82) = 57.05, p < .01
  Independents        4.16     3.99      F(1, 129) = 1.66, p < .20

                                       Negative

                       Ma      Chen

                     Means     Means     Results of planned
                                         contrasts

Ad liking

  KMT                 4.40     2.81      F(1, 64) = 35.15, p < .01
  DPP                 3.07     4.66      F(1, 73) = 47.57, p < .01
  Independents        3.41     3.39      F(1, 146) = .02, p = .89

Ad persuasiveness

  KMT                 4.22     2.95      F(1, 64) = 29.38, p < .01
  DPP                 3.06     4.46      F(1, 73) = 49.86, p < .01
  Independents        3.56     3.53      F(1, 146) = .08, p = .78

Note: KMT = Kuomintang

DPP = Democratic Progressive Party.

TABLE 3

Results of Planned Contrasts for Testing H3a, H3b, H4a, and H4b

Candidate evaluations

                     Ma                                            Chen

              Positive  No ad                                   Positive
               Means    Means     Results of planned contrasts   Means

KMT             5.57    6.06      F(1, 91) = 6.36, p < .01        4.25
DPP             4.05    4.50      F(1, 111) = 3.01, p = .09       5.74
Independents    5.00    4.72      F(1, 197) = 2.73, p = .10       4.63

                 Chen

              No ad
              Means     Results of planned contrasts

KMT           4.52      F(1, 91) = .63, p = .43
DPP           6.09      F(1, 111) = 4.08, p = .05
Independents  4.90      F(1, 197) = 2.31, p = .13

Candidate evaluations

                     Ma                                            Chen

              Negative  No ad                                   Negative
               Means    Means     Results of planned contrasts   Means

KMT             5.70    6.07      F(1, 89) = 3.03, p = .09        3.99
DPP             4.23    4.50      F(1, 102) = 1.05, p = .31       5.59
Independents    4.80    4.72      F(1, 214) = .28, p = .60        4.56

                 Chen

              No ad
              Means     Results of planned contrasts

KMT           4.52      F(1, 89) = 4.29, p = .04
DPP           6.09      F(1, 102) = 7.07, p < .01
Independents  4.90      F(1, 214) = 2.60, p = .11

Voting support for candidates

                     Ma                                            Chen

              Positive  No ad                                   Positive
               Means    Means     Results of planned contrasts   Means

KMT             5.71    4.83      F(1, 91) = 6.92, p < .01        3.54
DPP             3.36    4.47      F(1, 111) = 7.92, p < .01       5.73
Independents    4.55    4.57      F(1, 197) = .02, p = .90        4.27

                 Chen

              No ad
              Means     Results of planned contrasts

KMT           3.93      F(1, 91) = .78, p = .38
DPP           4.64      F(1, 111) = 10.47, p < .01
Independents  4.35      F(1, 197) = .09, p = .77

Voting support for candidates

                     Ma                                            Chen

              Negative  No ad                                   Negative
               Means    Means     Results of planned contrasts   Means

KMT             5.53    4.83      F(1, 89) = 4.07, p = .05        3.21
DPP             3.58    4.47      F(1, 102) = 5.34, p = .02       5.61
Independents    4.43    4.57      F(1, 214) = .47, p = .49        4.21

                 Chen

              No ad
              Means     Results of planned contrasts

KMT           3.93      F(1, 89) = 2.88, p = .09
DPP           4.64      F(1, 102) = 8.83, p < .01
Independents  4.35      F(1, 214) = .39, p = .54

Note: KMT = Kuomintang

DPP = Democratic Progressive Party.

REFERENCES

An, Soontae (2002), Voters' Attitude Selectivity Toward Political Advertising: Pre-decisional and Post-decisional Information, paper presented at the annual conference of the American Academy of Advertising, Jacksonville, FL, March.

Ansolabehere, Stephen, and Shanto Iyengar (1995), Going Negative, New York: Free Press.

-----, -----, Adam Simon, and N. Valentino (1994), "Does Attack Advertising Demobilize the Electorate?" American Political Science Review, 88 (December), 829-838.

Atkin, Charles, and Gary Heald (1976), "Effects of Political Advertising," Public Opinion Quarterly, 40 (Summer), 216-228.

Basil, Michael, Caroline Schooler, and Byron Reeves (1991), "Positive and Negative Political Advertising: Effectiveness of Ads and Perceptions of Candidates," in Television and Political Advertising: Psychological Processes, vol. 1, F. Biocca, ed., Hillsdale, NJ: Lawrence Eribaum, 245-262.

Bothwell, Robert K., and John C. Brigham (1983), "Selective Evaluation and Recall During the 1980 Reagan-Carter Debate," Journal of Applied Social Psychology, 13(5), 427-442.

Brock, Timothy C., and Joseph C. Balloun (1967), "Behavioral Receptivity to Dissonant Information," Journal of Personality and Social Psychology, 6 (August), 413-428.

Campbell, Angus, Philip. E. Converse, Warren E. Miller, and Donald E. Stokes (1960), The American Voter, New York: Wiley.

Chang, Chingching (1999a), "The Impacts of News Frames and Ad Types on Candidate Perception and Political Cynicism During the 1998 Taipei Mayoral Election in Taiwan," Scholar-to-Scholar exhibition session presented at the meeting of the Association for Education in Journalism and Mass Communication, New Orleans, August.

-----, (1999b), "The Advantages and Disadvantages of Being a Female Candidate" (in Chinese), Hsiian-ju yen-chiu [Election Studies], 6 (May), 113-136.

----- (2000), "A Cross-Cultural Comparison of Political Advertising in the 1996 Presidential Election Campaign in Taiwan and the United States," Asian Journal of Communications, 10 (1), 1-17.

Chen, Yi-ian (1996), Voting Election and Democratization of Taiwan (in Chinese), National Science Council Research Report no. 84-2414-H004-053 B2, Taipei, Taiwan: National Chengchi University.

Christ, William G., Esther Thorson, and Clarke Caywood (1994), "Do Attitudes Toward Political Advertising Affect Information Processing of Televised Political Commercials?" Journal of Broadcasting and Electronic Media, 38 (Summer), 251-270.

Cundy, Donald T. (1986), "Political Commercials and Candidate Image," in New Perspectives on Political Advertising, L. L. Kaid, D. Nimmo, and K. R. Sanders, eds., Carbondale: Southern Illinois University Press, 210-234.

Eagly, Alice H., and Shelly Chaiken (1993), The Psychology of Attitudes, Orlando: Harcourt Brace Jovanovich.

Faber, Ronald J., Albert R. Tims, and Kay G. Schmitt (1990), "Accentuate the Negative? The Impact of Negative Political Appeals on Voting Intent," in Proceedings of the American Academy of Advertising, P. Stout, ed., Austin: American Academy of Advertising, 10-16.

-----, -----, and -----, 1993), "Negative Political Advertising and Voting Intent: The Role of Involvement and Alternative Information Sources," Journal of Advertising 12 (Winter), 67-76.

Frey, Dieter (1986), "Recent Research on Selective Exposure to Information," in Advances in Experimental Social Psychology, vol. 19, L. Berkowitz, ed., San Diego: Academic Press, 41-80.

-----, and Robert Wicklund (1978), "A Clarification of Selective Exposure: The Impact of Choice," Journal of Experimental Social Psychology, 14 (1), 132-139.

Garramone, Gina M. (1983), "Issues Versus Image Orientation and Effects of Political Advertising," Communication Research, 10 (January), 59-76.

----- (1984), "Voters' Responses to Negative Political Ads," Journalism Quarterly, 61 (Summer), 250-259.

-----, and Sandra J. Smith (1984), "Reactions to Political Advertising Clarifying Sponsor Effects," Journalism Quarterly, 61 (Winter), 771-775.

-----, Michael Steele, and Bruce Pinkleton (1991), "The Role of Cognitive Schemata in Determining Candidate Characteristic Effects," in Television and Political Advertising: Psychological Processes, vol. 1, F. Biocca, ed., Hillsdale, NJ: Lawrence Erlbaum, 311-328.

Gronbeck, Bruce (1992), "Negative Narratives in 1988 Presidential Campaign Ads," Quarterly Journal of Speech, 78 (August), 333-346.

Hill, Ronald P. (1989), "An Exploration of Voter Responses to Political Advertisements," Journal of Advertising, 18 (Winter), 14-22.

Hitchon, Jacqueline, Chingching Chang, and Rhonda Harris (1997), "Should Women Emote? Perceptual Bias and Opinion Change in Response to Political Ads for Candidates of Different Genders," Political Communication, 14 (January), 49-69.

Houston, David A., and Russell H. Fazio (1989), "Biased Processing as a Function of Attitude Accessibility: Making Objective Judgments Subjectively," Social Cognition, 7 (Spring), 51-66.

Jasperson, A. E., and D. P. Fan (2002), "An Aggregate Examination of the Backlash Effect in Political Advertising: The Case of the 1996 U.S. Senate Race in Minnesota," Journal of Advertising, 31 (Spring), 1-12.

Johnson-Cartee, Karen S., and Gary A. Copeland (1991), Negative Political Advertising: Coming of Age, Hillsdale, NJ: Lawrence Erlbaum.

Kaid, Lynda Lee (1997), "Effects of the Television Spots on Images of Dole and Clinton," American Behavioral Scientist, 40 (August), 1085-1094.

-----, and John Boydston (1987), "An Experimental Study of the Effectiveness of Negative Political Advertisements," Communication Quarterly, 35 (Spring), 193-201.

-----, Mike Chanslor, and Mark Hovind (1992), "The Influence of Program and Commercial Type on Political Advertising Effectiveness," Journal of Broadcasting and Electronic Media, 36 (Summer), 303-320.

-----, and K. R. Sanders (1978), "Political Television Commercials: An Experimental Study of Type and Length," Communication Research, 5 (January), 57-70.

-----, and John C. Tedesco (1999), "Tracking Voter Reactions to the Television Advertising," in The Electronic Election, Lynda Lee Kaid and Dianne G. Bystrom, eds., Mahwah, NJ: Lawrence Erlbaum, 233-245.

Lodge, Milton, and Charles Taber (1999), "Motivated Skepticism in the Evaluation of Political Beliefs," paper presented at the 57th annual meeting of the Midwest Political Science Association, Chicago, April.

Lord, Charles G., Lee Ross, and Mark R. Lepper (1979), "Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence," Journal of Personality and Social Psychology, 37 (November), 2098-2109.

Merritt, Sharyne (1984), "Negative Political Advertising: Some Empirical Findings," Journal of Advertising, 13 (Summer), 27-38.

Miller, Arthur H., Martin P. Wattenberg, and Okasana Malanchuk (1986), "Schematic Assessments of Presidential Candidates," American Political Science Review, 80 (June), 521-540.

Mulder, Ronald (1979), "The Effects of Televised Political Ads in the 1975 Chicago Mayoral Election," Journalism Quarterly, 56 (Summer), 336-349.

Pinkleton, Bruce (1997), "The Effects of Negative Comparative Political Advertising on Candidate Evaluations and Advertising Evaluations: An Exploration," Journal of Advertising, 26 (Spring), 19-29.

----- (1998), "Effects of Print Comparative Political Advertising on Political Decision-Making and Participation," Journal of Communication, 48 (Autumn), 24-36.

Roddy, Brian L., and Gina M. Garramone (1988), "Appeals and Strategies of Negative Political Advertising," Journal of Broadcasting and Electronic Media, 29 (Fall), 147-159.

Russo, Edward J., Victoria H. Medvec, and Margaret G. Meloy (1996), "The Distortion of Information During Decisions," Organizational Behavior and Human Decision Processes, 66 (April), 102-110.

Shapiro, Michael A., and Robert H. Reiger (1992), "Comparing Positive and Negative Advertising on Radio," Journalism Quarterly, 69 (Spring), 135-145.

Sweeney, Paul D., and Kathy L. Gruber (1984), "Selective Exposure: Voter Information Preferences and the Watergate Affair," Journal of Personality and Social Psychology, 46 (June), 1208-1221.

Thorson, Esther, William G. Christ, and Clarke Caywood (1991), "Effects of Issue-Image Strategies, Attack and Support Appeals, Music and Visual Content in Political Commercials," Journal of Broadcasting and Electronic Media, 35 (Fall), 465-486.

Tinkham, Spencer F., and Ruth Ann Weaver-Lariscy (1993), "A Diagnostic Approach to Assessing the Impact of Negative Political Television Commercials," Journal of Broadcasting and Electronic Media, 37 (Fall), 377-399.

Vallone, Robert P., Lee Ross, and Mark R. Lepper (1985), "The Hostile Media Phenomenon: Biased Perception and Perceptions of Media Bias in Coverage of the Beirut Massacre," Journal of Personality and Social Psychology, 49 (September), 577-785.

Vidmar, Neil, and Milton Rokeach (1974), "Archie Bunker's Bigotry: A Study in Selective Perception and Exposure," Journal of Communication, 24 (September), 36-47.

Wang, Song-In, and Ian-Min Chen (1995), "Analysis of Television Political Ads of the DPP" (in Chinese), Kuang-kao hsiieh yenchiu [Journal of Advertising Research], 6 (July), 137-161.

West, Darrell M. (1993), Air War, Washington, DC: Congressional Quarterly.

----- (1994), "Political Advertising and News Coverage in the 1992 California U.S. Senate Campaigns," Journal of Politics, 56 (November), 1053-1075.

Winer, B.J., Donald R. Brown, and Kenneth M. Michel (1991), Statistical Principles in Experimental Design, 3rd ed., New York: McGraw-Hill.

Xie, Fu-sheng (1996), Interdisciplinary Research on Voting Behavior in Presidential Elections (in Chinese), National Science Council Report no. 85-2414-H-004-017 Q3, Taipei, Taiwan: National Chengchi University.

The author thanks Julie Drive for making this paper possible, as well as Dr. Ron Faber and the three anonymous reviewers for their thorough comments on a previous version of this manuscript.

Chingching Chang (Ph.D., University of Wisconsin-Madison) is associate professor of advertising, National Chengchi University, Taipei, Taiwan.