The Patterns of Partisan Defection in the 2011 Mayoral Election
2010年臺灣直轄市市長選舉選 民投票背離其黨性的行為模式分析
Paper to be presented in the TEDS2011 Conference.
This is an earlier draft. Please notify the authors before citing it.
Frank Cheng-Shan Liu Associate Professor
Institute of Political Science, National Sun Yet-San University csliu@mail.nsysu.edu.tw
Albert Shih-Yi Chiu Assistant Professor
Department of Political Science, Tunghai University s12357@ms24.hinet.net
Alexander C. Tan Associate Professor
School of Social and Political Sciences University of Canterbury, NEW ZEALAND
alex.tan@canterbury.ac.nz Abstract
How well are the models of partisan defection that finds support from national election data applied to sub-national elections? This paper employs Paul Beck’s (2002a) social-support theory of partisan defection and examines the models using Taiwan's most recent mayoral election. The theory suggests that a partisan voter becomes likely to vote against his or her partisanship when he or she is exposed to heterogeneous communication network in which family or close friends supporting for a candidate of the opposite political party. This study finds the limit of the theory and tries to provide explanation. Based on three independent face-to-face survey data sets collected in Taiwan's top three cities—Taipei, Taichung, and Kaohsiung—our discussion about the findings on the patterns of partisan defection across the cities shed light on the on-going debates about the area difference regarding voter preferences and voting behavior.
Keywords: partisan defection, network heterogeneity, social support theory, Taiwan mayoral elections
It is reasonable to expect that partisan voters in a democracy seldom change their party choices when they vote. But in reality some partisans do “defect” and a small proportion of voters who do change their partisan orientation and preferences during an election can result in a shift in an election, particularly a close one. Guided by Michigan School, political scientists continue to believe that partisan strength explain the stability of partisan votes (Paul Allen Beck, Dalton, Greene, & Huckfeldt, 2002a;
Krosnick, 1991; Schickler & Green, 1997). Despite this, researchers have found pieces of empirical evidence that Columbia School left to us: communication networks also play a role in explaining this instability.
According to Beck’s (2002a) theory of social support a voter who changes his partisan choices, i.e., not voting for a candidate of his party but vote for somebody else, finds cognitive support from individuals of his communication network of political discussion. “Support from personal discussants helps the voter to surmount these [cognitive] difficulties—in short, to encourage political defection.” (p.313)
This causal effect has been elaborated in the literature on network heterogeneity, suggesting that heterogeneous communication networks of political discussion are where individuals find such cognitive support for the changes of their vote choices (Eveland, William P. & Hively, 2009; R. Huckfeldt, Mendez, & Osborn, 2004; McClurg, 2006; Scheufele, Nisbet, & Brossard, 2004).
The question is that such network variables are not proven as robust as the partisan strength variable, suggesting that the effect of social networks may be contingent upon some circumstances. Whereas most supportive evidence is found at the national level (F. C.-S. Liu, 2011; J. H. Liu, Ikeda, & Wilson, 1998; J. H. Liu et al., 1998; Pattie & Johnston, 2001), it has not been confirmed if this causal effects will “step down” to a more local level election and hold true. If not, how could we explain it?
We pay our attention to sub-national cases by utilizing data collected simultaneous in the top three major cities in Taiwan’s 2012 mayoral election. We ask, given the three datasets with sufficient sample size, will network heterogeneity variable be significant while controlling for partisan strength. Below we will briefly review how current studies on network heterogeneity suggest its potential and limits.
Next we introduce datasets by which we test the hypothesis that individuals exposed to heterogeneous networks, other things being equal, are likely to defect from their party orientation and change their vote choices.
Note that the three cities, although they all well represent Taiwan’s metropolitan citizens, may differ from each other due to their specific electoral contexts. For example, Taipei had two major candidates from KMT and DPP respectively but had three other independent candidates; Taichung had no other
independent candidates but two from the major parties; and Kaohsiung had two candidates from the major parties but plus one deviated from DPP. While we will focus our attention to the extent to which the causal effect finds support from the three datasets, we are also interested in how the stories lying behind the causal relationship affect the patterns of defection we try to depict.
Network Heterogeneity and Partisan Defection
Ordinary voters theoretically votes for candidates whose party is consistent with those voters’ party identification. The debate regarding whether party identification is one of the dominant factors of voting choice has been settled since the rise of Michigan School in the 1960s (Campbell, Converse, Miller, & Stokes, 1960).
Socio-psychological factors, such as party identification, empirically matter to vote choice more so than sociological factors (Berelson, Lazarsfeld, & McPhee, 1954; Paul Felix Lazarsfeld, Berelson, & Gaudet, 1944), such as personal discussion network.
Yet, even proponents of the Michigan school did not rule out that other factors are also likely to alter the vote choice of party loyalists such as voters’ evaluation of candidates and inconsistent issue stances of the voters and the voters’ party to name a few (Campbell et al. 1952, 141-142). These factors include voters’ evaluation of candidates and voters’ issue stance that is inconsistent with the issue stance of voters’
party. In fact, the explanatory power of cross-cutting pressure is increasing when it comes to understanding American elections in recent years (Paul A. Beck, 2002a;
Paul Allen Beck, Dalton, Greene, & Huckfeldt, 2002b).
Berelson and his colleagues (1954) first introduced the term “cross-pressure”
to imply a communication network where individuals encounter different political opinions. Because we emphasize the influence of network and avoid paying excessive attention to individual voters’ perspective, we replace cross-pressure with the term network heterogeneity in this paper. In the 1960s, the “hows and whys” of network heterogeneity remains unclear. Neither “status inconsistency” nor “inter-generational status mobility” and “geographical mobility” have been shown to lead to network heterogeneity (Boyd, 1969, p. 63). That is, (1) network heterogeneity does not form as a result of individuals residing or working in an environment where others have different ethnicity, income and education levels, or occupations; (2) differences between fathers’ occupational prestige and his son’s social class does not produce network heterogeneity; (3) heterogeneity does not serve as function as residentially moving around. A heterogeneous network is where a voter encounters personal discussants who support a candidate of different political party.
Earlier works on the relationship between network heterogeneity and participation did not find much empirical support because researchers usually assume
that status inconsistency defines network heterogeneity (Horan, 1971; Mutz, 2002a;
Pool, 1965). Advanced measurement of network heterogeneity is concerned with respondents’ own perception of their difference from other personal discussants in terms of party support (Mutz, 2002b; Ulbig & Funk, 1999). Beck’s (2002) work shares this measurement. The focal behavioral consequence of such interpersonal incongruence is partisan defection or voter defection. According to Beck, even to a lesser extent, if the partisan personal network failed to provide enough support for its candidate, then the partisan is likely to defect.
This social support theory is partially supported by data collected in Taiwan’s 2002 mayoral election: it explains Kaohsiung voters’ partisan defection, but not Taipei voters. It was found that for Taipei partisan voters, frequent political discussion decreases the likelihood of partisan defection, implying that Taipei partisan voters who interact frequently with their communication networks are more likely to strengthen their existing vote preferences (F. C.-S. Liu, 2006). In another similar study using 2006 Taipei and Kaohsiung mayoral election, where the two datasets were combined for the sake of sample size, Beck’s hypothesis is supported (F. C.-S.
Liu & Chiu, 2011).
Regarding the effect of political discussion within heterogeneous networks, recent studies in experimental political science and experimental psychology have
shown that “in the contexts of individual decision making, people tend to look for positive confirmations of hypothesized patterns, while disregarding or failing to look for negative information that does not fit the expected pattern” (Dickson, Hafer, &
Landa, 2008; Taber & Lodge, 2006). Dickson et al (2008) suggests that it is likely that the incongruence between a listener and a speaker in a discussion or an exchange of arguments setting can result in a listener either adjusting towards or away from the speaker’s position. However, in their findings of their experiment Dickson et al (2008) found that voters tend to use their prior beliefs despite discussions to convince otherwise. That is, when there is incongruence in exchange of ideas or arguments, more often than not voters hold on to their a priori positions rather than get swayed to a different position. What we can infer from these studies, in terms of our understanding of vote choice as an individual decision making event, is that in situations when opinions are heterogeneous hold on to their positions or seek out opinions that will reinforce their positions.
Besides partisan strength and network heterogeneity, alternative variables accounting for changes in party choice include political-psychological variables—the perception of social support, perceptions of dominant parties, subjective evaluation of candidates, knowledge of the other voters’ partisan preferences, and retrospective views about the economy’s status—and social context variables—interaction with
political party networks and social class networks (Alvarez, Nagler, & Bowler, 2000;
Bartle, 2003; Paul A. Beck, 2002a; Burbank, 1997; Fournier, Blais, Nadeau, Gidengil,
& Nevitte, 2003; Kenny, 1998; Weisberg, 2002; Wekkin, 1991). In general, it is expected that unfavorable evaluations of the incumbent, favorable evaluations of the challenger, and negative views about the economy’s status would increase the likelihood of changes in party choice in the voting booth.
It is worthwhile to mention the role of perception of the level of competition of an election. Past studies have suggested that candidates at the state and local levels, for instance, are less visible than those in national elections and, therefore, the electorate are not as interested as they are when it comes a national election. In such a circumstance, even loyal voters are likely to defect (Jacobson, 2001; D. E. Stokes &
Miller, 1966). Likewise, an incumbent candidate that has a good performance record or challengers’ possess a dismal image, may result in the incumbent likely gaining support from the challengers’ party supporters (P. E. Converse & Georges, 1966;
Fiorina, 1981; Key & Cummings, 1966).
The literature of party realignment further suggests that “voters unhappy with their party often defect to the opposition without changing their party identifications, while new voters are bringing their party into line with their vote” (Paul A. Beck, 2002a, p. 310). All of these factors help to draw a basic understanding about the
extent to which the theory of social support explains sub-national elections like Taiwan’s 2011 mayoral election: we are likely to find weak evidence in elections that are less competitive and when an incumbent candidate is positively viewed.
Data
The datasets used for this study are Taiwan’s Election and Democratization Study for the 2010 Taipei, Taichung, and Kaohsiung mayoral elections (TEDS2010C, N = 1,131, 1,168, and 1,177, respectively).1 Election Day was November 27, 2010, and the survey was conducted from January 5, 2011, to March 5, 2011. Except that some questions were added to Kaohsiung questionnaire and accordingly question number updating, the wording and the sequence of question are exactly the same across the three questionnaires.
There are three strengths of using these datasets for this study. First, the questions used for coding and recoding the variables in TEDS2010C are consistent with previous TEDS series surveys. Therefore, the results of this paper are comparable to earlier studies using TEDS2002C and TEDS2006C. Second, the three parallel datasets allow a researcher to compare voter behavior across these cities in
terms of the patterns of partisan defection. Third, TEDS2010 provides “qualitative” or
1 Data analyzed in this article were collected by the research projects of TEDS 2010 (NSC 99-2420-H- 031-002-), directed by Shiow-Duan Huang. Public Opinion Survey Center, National Chung-Cheng University, is responsible for the data distribution. The author and colleagues thank the institute and individuals previously mentioned for providing data. The views expressed here are the author's own.
open-ended questions to explore reasons of voting. This questions help understand what the partisan voters who change their voting orientation thought. Appendix 1 gives the details of the coding scheme.
Note that a number of important variables about communication networks and the control variables suggested by the literature are unavailable in TEDS2010C, including the size of networks, the details of discussants’ political background, economical and residential stability, and ways of evaluating candidates and political parties (Paul A. Beck, 2002b; Coleman, 2004; Fournier et al., 2003). The coding scheme of the major variables are put in the Appendix.
Patterns of Partisan Defection Described
[Table 1 is about here]
The overall pattern of defection across the three cities suggests that few partisan individuals change their vote choices, consistent with earlier empirical studies. As Table 1 shows in the bottom two lines, Kaohsiung has more partisan defectors than Taipei and Taichung does. Not a surprise that Kaohsiung stands out as an unique case as it has three major candidates (Yang Chiu-Hsin “defected” from DPP) while Taipei and Taichung are characterized by the competition between the
blue and green camps.
The result of cross-tabulation analysis in Table 1 suggests another pattern that is consistent with our hypothesis: More than 64% of individuals who defected are likely to be those exposed to heterogeneous networks. But the pattern blurs if we look into those who did not change their votes: 1/3 of voters who did not defect are associated with the homogeneity of communication network; more than half of those defected are not explained by perceived network heterogeneity but by other variables.
Therefore, we expect at this stage that, although the two variables are statistically significantly related, our hypothesis may not be fully supported by samples of these cities in further analysis.
[Table 2 is about here]
According to theory, partisanship stands out as an alternative explanatory variable for the above pattern. Following pundits saying that the defection of Yang in Kaohsiung may hurt DPP, we suspected that in Kaohsiung supporters of DPP are more likely to defect, while party ID may not play a critical role in Taipei and Taichung. Two of our expectations hold: Party ID matters in Kaohsiung but not in Taichung. What surprised us are another two results shown in Table 2: that Party ID matters in Taipei and that it is KMT supporters in Taipei and Kaohsiung are volatile
in their vote choices, not DPP supporters. It seems clear at this stage that Yang’s defection in Kaohsiung had to a greater extent influences the loyalty of Kaohsiung KMT supporters. Not that it is incorrect and too quick to conclude that “party ID influences the likelihood of defection” (that is, KMT supporters are more likely to defect). Rather, what we would say about this pattern is that in the 2011 election KMT supporters overall are more volatile and more likely to change their vote choices than their DPP counterparts.
Patterns of Partisan Defection Explained
We provide both reduced and full models for each sample to examine the hypotheses that the level of partisanship and perceived network homogeneity influences the likelihood of partisan defection. Reduced models are those include most important variables provided by the literature. More control and demographic variables are then added into the reduced models to check if the hypotheses still hold.
Note here that all of the variables included into the models are not seriously linearly correlated; this means the variables were included into the model not in expense of other variables’ explanatory power.
For the hypothesis that perceived social supported within a communication network decreases partisan defection, we find that, with all things considered, it is
supported in the sample of Taipei, but not in Taichung and Kaohsiung. As Table 3 shows, for Taipei voters the higher level of heterogeneity or the less level of network homogeneity suggests a higher probability of partisan defection, even that partisan strength is controlled. For the same hypothesis, we find a negative relationship in Taichung at the weak significant level (0.1), but this result suggests a spurious relationship between network heterogeneity and partisan defection in Taichung.
Along with the fact that this relationship is not supported at all in Kaohsiung, further explanations for such “exceptions” are necessary in the coming context.
The second hypothesis regarding partisan strength is robustly supported across the six models, except that the full model of Taipei. This is an interesting pattern differentiating Taipei voters from Taichung and Kaohsiung voters. For the latter whose defection is well explained by partisan strength, Taipei voters are less
constrained by their party ties, the fact that leaves the second hypothesis spurious. Up to this stage, we could summarize that the theory of social support finds supported in the Taipei sample, while the convention wisdom of partisanship holds in Taichung and Kaohsiung.
Shift attention to the middle section of Table 3. How voters support for
individual candidates and their partisanship do play a role of explaining why the two hypotheses are not fully supported at the mayoral level election. In Taipei, voters who
defected are more likely to be those disliking the incumbent (Hau Lung-Bin 郝龍斌) and those liking KMT, a pattern that is found only among KMT supporters, but not among DPP identifiers. This seemingly contradictory findings is in effect consistent with the pattern shown earlier in Table 2. Voters in Taipei who like KMT are not those simply based their vote choices on partisan loyalty. The more they like this political party, the higher expectation on it, the more likely they would defect if their candidate does not fit their high expectation.
A similar but even more apparent pattern is found in Kaohsiung. Defectors are primarily those liking KMT but dissatisfied with their candidate (Huang Chao-Shun
黃昭順). Although it is apparent that individuals liking Yang and/or disliking Chen changed their vote choices, DPP voters in Kaohsiung are significantly more loyal and less likely to defect. Therefore, we could confirm with Table 3 that KMT supporters are less partisanship-bounded than DPP identifiers. Their satisfaction with their party’s candidate did not offset their party loyalty but drove their vote choices.
Taichung stands out a puzzling case that does not fit into the above discussion.
None of candidate and partisanship plays any role in the models, although the conventional wisdom holds: weaker partisans and younger voters defected. While Taichung is not as urban as Taipei, and it is not as south and unique as Kaohsiung where voters stick with their emotional attachment to individual candidates, the
puzzle here further suggests that it is more difficult to describe and predict vote decision of Taichung voters.
Reasons of Defection
TEDS2010C provides a qualitative or open-ended question about the major reason why a respondent voted for a particular candidate. We employed this question and subset the samples to filter out the reasons held by partisan defectors. Although not all of the partisan voters who defected revealed their reasons of voting, we see the answers provided in the following tables are representative enough of those held by other defectors. We present these reasons in the sequence of Taipei (Table 4), Taichung (Table 5), and Kaohsiung (Tables 6 & 7). In general these reasons are candidate-centric and focusing on their ability and past performance.
[Table 4 is about here]
[Table 5 is about here]
Table 4 shows that KMP partisans in Taipei have more reasons to defect and vote for Su Chen-Chang, most of whom saw Su has much ability (2), has been doing well in his office (2), or are influenced by their family members (2). Table 5 shows a
similar list of reasons, but it further shows that KMT voters who defected are those view Su Chia-Chung as a better alternative (9), see Hu has not been doing well in his terms (2), and are influenced by their family members (2).
[Table 6 is about here]
[Table 7 is about here]
Kaohsiung had three candidates and both KMT and DPP found some partisans defected. Table 6 shows that the major reasons KMT voters defected are
acknowledging that the incumbent mayor Chen Chu’s good performance in her previous term (13), dislike the candidate of their party (4), or are influenced by their family members (2). As we find that DPP is characterized by its strong partisan magnet, the only reason of DPP defector we can find is because he/she is associated with Huang’s family.
As to the reasons of voting the ex-DPP member Yang Chiu-Hsin, Table 7 shows a more comprehensive list of why KMT members defected. Besides disliking their own candidate Huang Chao-Shun (24), KMT partisans found Yang has an image that they are looking for that may not be emphasized by their candidate: acknowledging Yang’s performance in the past terms in the office of Kaohsiung County (39),
incorrupt (9), or good fame (5). DPP identifiers who were ready to change their vote choice and vote for Huang turned out to vote for Yang because they dislike Huang as an alternative (8), or they have been acknowledging the good performance of Yang (8).
Drawn from TEDS2010C additional set of questions for Kaohsiung, the
following information shows that Yang is the alternative for those who like KMT but dislike Huang Chao-Shun. Among KMT defectors, 68.2% (30) does not want Chen to win; 22.7% (10) did not want Yang to win (E02F). Among these KMT defectors only 39.3% (22) made their mind one month before the election day, 32.1% (18) decided two weeks before the election day, and 28.6% (16) decided in the last two days (E02G), showing a great level of uncertainty on how to win. This is evidenced by the figure that 84.9% (45) expected Chen would win before they knew the election result and only 1.9% (1) thought Huang would (E2H).
This section of qualitative data supplements to our earlier generalization that in Taipei and Kaohsiung KMT supporters can support for their party on the one hand but are more volatile than their DPP counterparts when a better alternative becomes available.
Conclusion and Discussion
This paper is initiated with our theoretical inquiry about how well will the
theory of social support explains partisan defection in Taiwan’s 2011 mayoral election, particularly the Kaohsiung case where three candidates compete for a seat.
While our findings can be seen contradictory to some earlier works and, as expected, evidence is mixed, our findings shed light on better understanding of Taiwan’s voters resided in metropolitan areas. We almost come to a conclusion that the theory of social support is conditioned on other factors that influence a sub-national election. It explains sometimes, but it is conditioned on campaign situations.
We find that, first, KMT supporters in Taipei and Kaohsiung are more volatile not in terms of their party identity but of their voting preferences. Second, DPP voters in Kaohsiung are more loyal to both their party and their mayor Chen Chu. This explains why they become more united in the face of the defection of Yang Chiu- Hsing but also justifies the election results that Yang draws much more votes from KMT than from DPP. Third, in a campaign that is “cold” enough to discourage political discussion and therefore extinguish the influence of communication networks like Taichung, only a few variables such as partisan strength and age would explain partisan defection.
We find these results leave plenty of room for discussion. First, it can be true that partisanship’s effect is associated with communication network effect. Network heterogeneity can offset the influence of strong partisanship in a presidential election
in Taiwan (F. C.-S. Liu, 2011).That is, one may expect that the significance of one discourages the other, as seen in the Taipei and Kaohsiung models. “In the short term, such a [communication network] reinforcing effect may mask the impact of weakening partisan attachments by making weak party identifiers behave more like strong identifiers. In the long run, though, the weakening of partisan attachments in the electorate will likely undermine the impact of contextual effects because the social context appears to enhance, rather than create, partisan attachment” (Burbank, 1997 p.
127). In our present study, we do see that in Kaohsiung psychological attachment to individual candidates weighs more than social environment regarding explaining vote defection. But in Taipei and Taichung, the effect of the likert scales for individual candidates vanishes, suggesting that the relationships between favoring candidates, identifying with parties, and partisan defection may be more complicated than we initially thought. Let’s assume that partisanship strength can be indicated or measured by the level of liking/disliking the political party. What surprised us is that in Taipei and Kaohsiung, voters liking KMT are able to divert their vote choices to other candidates while remain liking this political party. Future studies are encouraged to look into this interesting pattern that provides more explanation.
Second, setting aside theoretical discussion, the comparison among voters in Taipei, Taichung and Kaohsiung regarding the relationship of heterogeneity network
to their voting defection is also shed light on our understanding of Taiwan politics.
That the relationship found spurious in Kaohsiung challenges well-ingrained stereotype: that “southern” partisans are more susceptible to interpersonal relationship than “northern” partisans. In effect we see that defection due to family influence is not restricted to Kaohsiung (Tables 4 to 7) but Kaohsiung voters are indeed more constrained by their partisan- and candidate-orientation. Partisan voters in Kaohsiung are less likely to defect simply because of living in a heterogeneous environment;
voters in Taipei and Taichung, instead, are more subjective to the heterogeneity of their communication network during an election, supporters of different party, specifically.
Third, we found that frequent political discussion and partisan defection is not statistically significantly related across all of the models (as seen in Table 3).
Although the effect of political discussion is beyond the scope of this paper, we included the frequency of political discussion as an important control variable and expected a negative relationship. If it is true that one tends to strengthen his or her prior beliefs when discussing politics with network members, as found in Taipei voters in the 2002 mayoral election, frequent political discussion should lead a stabilization of one’s voting preferences. In other words, frequent political discussion has a negative impact on the likelihood of partisan defection. Apparently, our findings
concur with Dickson et al’s (2008) experimental studies suggesting that voters in heterogeneous network would seek out positive reinforcement of their a priori choices. More explanations on such discrepancies are welcomed in future research.
Our preliminary explanation is that political discussion was not heated during the 2012 mayoral election in the three cities, the circumstance that prevents the mechanism of political discussion found in experiments from occurring.
Finally, like most other studies on this subject the scope of this study is limited by the variables chosen. Although the chosen datasets provide most important variables for constructing the models, there remain some that we did not bring into control and discussion, such as interaction with political party networks and social class networks, the perception of partisan dominance, exposure to the mass media, favorable evaluations of past governing performance of the opposing party, evaluation of incumbents and challengers’ stances on political issues, incumbent performance evaluations regarding an issue, and evaluation of candidate credibility, visibility and appeal (or favorable evaluations of the incumbent and challenger) (Fournier et al., 2003).
Table 1: Cross-Tabulate Analysis of Partisan Defection by the Level of Heterogeneity
Perceived Network Homogen
eity
Taipei Taichung Kaohsiung
No Defection
Defection No
defection
Defection No
defection
Defection
Heterogene ous
625 (64.2)
25 (89.3)
403 (66.4)
28 (90.3)
264 (55.0)
122 (64.2) Homogene
ous -Supported
348 (35.8)
3 (10.7)
204 (33.6)
3 (9.7)
216 (45.0)
68 (35.8)
Sum 973 28 984 31 480 190
Total 1,001 638 670
Note: (1) in parentheses are column percentages. (2) A voter is coded as defect when he/she voted for a candidate of any other parties or cast a waste ballot. (3) Pair-wise deletion is used in this analysis; that is, the number of observations coded as “not available” or N/A, e.g.., “don’t know,” “forget,” or
“refuse to answer,” etc. are not reported here. (4) The chi-square tests of the Taipei and Taichung samples are significant at the 0.01 level; the chi-square tests of the Kaohsiung samples are significant at the 0.05 level.
Source: TEDS 2010C.
Table 2: Cross-Tabulate Analysis of Partisan Defection by Party Identification
Taipei Taichung Kaohsiung
Party ID No defection Defection No defection Defection No defection
KMT 502 (63.4) 18 (64.3) 375 (58.8) 24 (68.6) 167 (33.5)
DPP 262 (33.1) 6 (21.4) 251 (39.3) 11 (31.4) 326 (65.3)
NP 14 (1.8) 2 (7.1) 1 (0.2) 0 (0.0)
PFP 6 (0.8) 0 (0.0) 6 (0.9) 0 (0.0)
TSU 2 (0.3) 2 (7.1) 1 (0.2) 0 (0.0)
GP 6 (0.8) 0 (0.0) 3 (0.5) 0 (0.0)
No ID 0 (0.0) 0 (0.0) 1 (0.2) 0 (0.0)
Sum 792 28 638 35
Total 820 673 697
Note: (1) in parentheses are column percentages; (2) a voter is coded as defector when he/she voted for a candidate of any other parties or cast a waste ballot; (3) the number of observations coded as “not available” or N/A, e.g.., “don’t know,”
“forget,” or “refuse to answer,” etc. are not reported here; and (4) the chi-square tests of the Taipei and Kaohsiung samples are significant at the 0.001 level, while the chi- square test of the Taichung sample is NOT significant even at the 0.1 level.
Source: TEDS2010C.
Table 3: Logistic Regression Models of Partisan Defection
Taipei Taichung Kaohsiung
Reduced Full Reduced Full Reduced Full
(Intercept) -0.731 (1.0
46)
-0.984 (2.
01 9)
0.427 (1.0
22)
4.418* (2.
16)
-0.833 (0.6
04)
-1.951+ (1.
02 3) Homogeneous
&
Support ed
-1.615* (0.6
43)
-1.934* (0.
78 1)
-1.244+ (0.6
39)
-1.098+ (0.
65 8)
0.021 (0.2
20)
0.002 (0.
25 1) Partisan
Strengt h
-0.937**
(0.3 38)
-0.620 (0.
38 4)
-0.760* (0.3
00)
-0.988**
(0.
36 3)
-0.622 ***
(0.1 48)
-0.783***
(0.
18 0) Discussion
Frequen cy
.0146 (0.2
44)
0.173 (0.
29 5)
-0.304 (0.2
46)
-0.070 (0.
30 7)
-0.135 (0.1
30)
-0.202 (0.
15 5) Like Su Jen-
Chang
-0.012 (0.0
91)
-0.100 (0.
16 3)
- - - -
Like Hau Lung-
Bin
-0.152 (0.0
97)
-0.448* (0.
18 1)
- - - -
Like Su Chia- Chuang
- - -0.076
(0.0 93)
-0.053 (0.
14 5)
- -
Like Hu Chi- Chang
- - -0.134
(0.0 90)
-0.119 (0.
14 5)
- -
Like Huang Chao-
Shun
- - - - -0.069
(0.0 51)
-0.216**
(0.
06 6) Like Yang
Chiu- Hsin
- - - - 0.515***
(0.0 66)
0.496***
(0.
07 7) Like Chen
Chu
- - - - -0.256 ***
(0.0 44)
-0.148* (0.
06 3)
Like KMT - 0.421*
(0.
20 5)
- 0.195
(0.
15 4)
- 0.243***
(0.
07 3)
Like DPP - 0.126
(0.
17 9)
- 0.054
(0.
14 0)
- -0.150* (0.
06 9) Retrospective
Econom y
- -0.227
(0.
44 6)
- -0.604
(0.
02 2)
- 0.333
(0.
21 4)
Age - -0.019 (-
0.0 20)
- -0.060**
(0.
02 2)
- -0.001
(0.
00 9)
Education - -0.02
(0.
10 1)
- -0.173
(0.
11 0)
- 0.053
(0.
04 9)
Sex (Male) - 0.214
(0.
46 6)
- -0.748
(0.
45 9)
- -0.234
(0.
23 5)
Income - 0.033 - -0.023 - 0.022
(0.
09 0)
(0.
08 5)
(0.
04 5)
N 773 683 608 496 637 547
Nagelkerke’s R- square
0.13 0.161 0.094 0.176 0.313 0.398
AIC 224.98 186.26 230.55 194.39 622.84 513.98
Note: + p<.1; * p<.05; ** p<.01; *** p<.001;
Source: TEDS2010C
Table 4: Taipei Voters’ Reasons of Defection
KMT Identifiers (Voting for Su
Chen-Chang)
DPP Identifiers (Voting for Hau Lung-Pin)
Check-n-balance [106] (1) Hau is incorrupt [203] (1)
Su diligent at politics [202] (1) Hau has much ability [207] (1) Su has much ability [207] (2) Identify with Hau’s
policy platform
[208] (1)
Su more experienced [210] (1) Hau was doing well [302] (1) Su a good alternative [212] (1) Family support Hau [402] (1) Su was doing well [302] (2) Did not want to be the
same with family and friends
[409] (1)
Hau was doing poorly [310] (1)
Persuaded by others [401] (1)
Family support Su [402] (2)
Same school with the candidate [408] (1)
Su is better [503] (1)
Note: In the braces are the coding schemes of the variable E02C. In the parentheses are the counts of voters who revealed their reasons.
Source: TEDS2010C
Table 5: Taichung Voters’ Reasons of Defection
KMT Identifiers (Voting for Su
Chia-Chuan)
DPP Identifiers (Voting for Hu Chih-Chang)
Su diligent at politics [202] (1) Support the KMT [101] (1) Identify with Su’s policy
platform
[207] (1) Hu more experienced [209] (1)
Su a good alternative [211] (9) Hu has good reputation
[212] (1)
Good reputation [212] (1) Has better Image [214] (1)
Like Su [215] (1) Hu was doing well [301] (2)
Honest [216] (1) Give Hu one more
chance
[306] (3)
Hu was doing poorly [309] (2) Family support Hu [402] (1)
Knows what we need [312] (1)
Family support Su [402] (2)
Strategic voting [503] (1)
As a farmer like Su [505] (1)
Note: In the braces are the coding schemes of the variable E02C. In the parentheses are the counts of voters who revealed their reasons.
Source: TEDS2010C
Table 6: Kaohsiung Voters’ Reasons of Defection (I)
KMT Identifiers (Voting for Chen
Chu)
DPP Identifiers
(Voing for Huang Chao-Shun)
Dislike Huang Chao-Shun [207] (4) Associated with Huang and her family
[404] (1)
Identify with Chen’s policy platform [208] (1)
Charismatic [211] (1)
Good Image [215] (1)
Give a chance for the second term [301] (2)
Has been doing well [302] (13)
Pave the way for the future [307] (1) Emphasize area balance [310] (1)
Family support Chen [402] (3)
Associated with Chen and her family [404] (1) Affected by the campaign [409] (1)
More likely to win [410] (2)
Depend on the mood [501] (2)
Note: In the braces are the coding schemes of the variable E02C. In the parentheses are the counts of voters who revealed their reasons.
Source: TEDS2010C
Table 7: Kaohsiung Voters’ Reasons of Defection (II)
KMT Identifiers (Voting for Yang
Chiu-Hsin)
DPP Identifiers (Voing for Yang Chiu-Hsin)
Check-n-balance [106] (1) No party affiliation [112] (1)
Dislike DPP [108] (1) Steady and working hard [201] (1)
Hard to choose between the other two candidates
[111] (2) Dislike Huang [207] (8)
No party affiliation [112] (1) Charismatic [211] (1)
Caring about people [202] (2) Time to change [212] (1)
Incorrupt [203] (9) Good fame [213] (1)
Dislike Chen Chu [206] (2) Geographic factors [216] (1)
Dislike Huang [207] (24) Like Yang [217] (1)
Identify Yang’s policy platform [208] (3) Honest [218] (2) Well educated and experienced [210] (3) Was doing well [302] (8)
Time to change [212] (5) Persuaded by others [401] (1)
Good fame [213] (5) Family support Yang [402] (2)
Good image [215] (2) Associated with Yang and
his family
[404] (1)
Geographic factors [216] (3)
Like Yang [217] (4)
Honest [218] (2)
Was doing well [302] (39)
Emphasize area balance [310] (1)
Locally-minded [311] (1)
Persuaded by others [401] (5) Family support Yang [402] (4)
Associated with Yang and his family
[404] (1)
Same school with Yang [408] (1)
Likely to win [410] (2)
Dependent on mood [501] (3)
Note: In the braces are the coding schemes of the variable E02C. In the parentheses are the counts of voters who revealed their reasons.
Source: TEDS2010C
References
Alvarez, R. M., Nagler, J., & Bowler, S. (2000). Issues, Economics, and the Dynamics of Multiparty Elections: The British 1987 General Election.
American Political Science Review, 94(1), 131-149.
Bartle, J. (2003). Partisanship, performance and personality: Competing and
complementary characterizations of the 2001 British General Election. Party Politics, 9(3), 317-345.
Beck, Paul A. (2002a). Encouraging political defection: The role of personal
discussion networks in partisan desertions to the opposition party and Perot votes in 1992. Political Behavior, 24(4), 309-337.
Beck, Paul A. (2002b). Encouraging political defection: The role of personal
discussion networks in partisan desertions to the opposition party and Perot votes in 1992. Political Behavior, 24(4), 309-337.
Beck, Paul Allen, Dalton, R. J., Greene, S., & Huckfeldt, R. (2002a). The social
calculus of voting: Interpersonal, media, and organizational influences on presidential choices. American Political Science Review, 96(1), 57-73.
Beck, Paul Allen, Dalton, R. J., Greene, S., & Huckfeldt, R. (2002b). The social calculus of voting: Interpersonal, media, and organizational influences on presidential choices. American Political Science Review, 96(1), 57-73.
Berelson, B., Lazarsfeld, P. F., & McPhee, W. N. (1954). Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago, IL: University of Chicago Press.
Boyd, R. W. (1969). Presidential elections: An explanation of voting defection. The American Political Science Review, 63(2), 498-514. doi:10.2307/1954703 Burbank, M. J. (1997). Explaining contextual effects on vote choice. Political
Behavior, 19(2), 113-132.
Campbell, A., Converse, P., Miller, W., & Stokes, D. (1960). The American voter.
New York: Wiley.
Coleman, S. (2004). The effect of social conformity on collective voting behavior.
Political Analysis, 12(1), 76-96.
Converse, P. E., & Georges, D. (1966). DeGaulle and Eisenhower: The public image of the victorious general. In A. Campbell, P. E. Converse, W. E. Miller, & D.
E. Stokes (Eds.), Elections and the Political Order (pp. 292-345). New York:
Wiley.
Dickson, E. S., Hafer, C., & Landa, D. (2008). Cognition and strategy: A deliberation experiment. The Journal of Politics, 70, 974.
Eveland, William P., J., & Hively, M. H. (2009). Political discussion frequency, network size, and “heterogeneity” of discussion as predictors of political knowledge and participation. Journal of Communication, 59(2), 205–224.
Fiorina, M. (1981). Retrospective voting in American national elections. New Haven:
Yale University Press.
Fournier, P., Blais, A., Nadeau, R., Gidengil, E., & Nevitte, N. (2003). Issue Importance and Performance Voting. Political Behavior, 25(1), 51-67.
Horan, P. M. (1971). Social positions and political cross-pressures: A re-examination.
American Sociological Review, 36(4), 650-660. doi:10.2307/2093595
Huckfeldt, R., Mendez, J. M., & Osborn, T. (2004). Disagreement, ambivalence, and engagement: The political consequences of heterogeneous networks.
Political Psychology, 25(1), 65-95.
Jacobson, G. (2001). The politics of congressional elections (5th ed.). New York:
Longman. Retrieved from http://polisci.ucsd.edu/faculty/jacobson.htm Kenny, C. (1998). The behavioral consequences of political discussion: Another look
at discussant effects on vote choice. Journal of Politics, 60(1), 231-244.
Key, V. O., & Cummings, M. C. (1966). The responsible electorate: Rationality in presidential voting, 1936-1960. Cambridge,: Belknap Press of Harvard University Press.
Krosnick, J. A. (1991). The stability of political preferences: Comparisons of
symbolic and nonsymbolic attitudes. American Journal of Political Science, 35(3), 547-576.
Lazarsfeld, Paul Felix, Berelson, B., & Gaudet, H. (1944). The People’s Choice: How the Voter Makes Up His Mind in a Presidential Campaign. New York: Duell, Sloan and Pearce.
Liu, F. C.-S. (2006). Communication networks and changes in electoral choices: A study of Taiwan’s 2002 mayoral elections. Journal of East Asian Studies, 6(1), 139-158.
Liu, F. C.-S. (2011). Perceived partisan heterogeneity in communication networks and changes in party choice in a national election: evidence from Taiwan.
International Political Science Review, 32(1), 61-78.
Liu, F. C.-S., & Chiu, A. S.-Y. (2011). Network heterogeneity, partisan defection, and voter turnout: Examine theory with empirical data from Taiwan.
International Journal of Business and Social Science, 2(19), 258-272.
Liu, J. H., Ikeda, K., & Wilson, M. S. (1998). Interpersonal environment effects on
political preferences: The “middle path” for conceptualizing social structure in New Zealand and Japan. Political Behavior, 20(3), 183-212.
McClurg, S. D. (2006). The electoral relevance of political talk: Examining disagreement and expertise effects in social networks on political participation. American Journal of Political Science, 50(3), 737–754.
Mutz, D. (2002a). The consequences of cross-cutting networks for political participation. American Journal of Political Science, 46(4), 838-855.
Mutz, D. (2002b). Cross-cutting social networks: Testing democratic theory in practice. American Political Science Review, 96(1), 111-126.
Pattie, C., & Johnston, R. (2001). Talk as a political context: Conversation and
electoral change in British elections, 1992-1997. Electoral Studies, 20(1), 17- 40.
Pool, I. de S. (1965). Candidates, issues and strategies: A computer simulation of the 1960 and 1964 presidential elections. The M.I.T. paperback series, MIT 33 (Rev. ed.). Cambridge: Massachusetts Institute of Technology Press.
Scheufele, D. A., Nisbet, M. C., & Brossard, D. (2004). Social structure and
citizenship: Examining the impacts of social setting, network heterogeneity, and informational variables on political participation. Political
Communication, 21(3), 315–338.
Schickler, E., & Green, D. P. (1997). The stability of party identification in western democracies - Results from eight panel surveys. Comparative Political Studies, 30(4), 450-483.
Stokes, D. E., & Miller, W. E. (1966). Party government and the saliency of Congress. In A. Campbell, P. E. Converse, W. E. Miller, & D. E. Stokes (Eds.), Elections and the Political Order (pp. 194-211). New York: Wiley.
Taber, C. S., & Lodge, M. (2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political Science, 50, 755-769.
doi:10.1111/j.1540-5907.2006.00214.x
Ulbig, S. G., & Funk, C. L. (1999). Conflict avoidance and political participation.
Political Behavior, 21(3), 265-282.
Weisberg, H. F. (2002). Partisanship and incumbency in presidential elections.
Political Behavior, 24(4), 339-360.
Wekkin, G. D. (1991). Why crossover voters are not mischievous voters: The
segmented partisanship hypothesis. American Politics Quarterly, 19(2), 229- 247.
Appendix 1
Variable Names Variable Coding Scheme
Dependent Variable
Partisan defection 1 = a respondent’s vote is inconsistent with his/her partisan orientation. An identifier of the major political parties
DPP/KMT/PFP/NP/TSU voting for candidates other than ones nominated by their parties or casting a waste ballot); 0 = otherwise. (P01, P01A, P01B)
Independent Variables Homogeneous &
Supported
1 = the majority (more than half) of respondent’s network of political discussion support for a political party that is the same as the respondent’s; 0 = half to all of network members
supporting for parties different from the respondent’s. (B01, B01A, B01B)
Partisan Strength “How strongly do favor this political party?” 2=Very strong;
1=median level; 0=just a little bit. (P01C) Retrospective Family
Economy
“Would you say that your family’s economy getting better than the previous year (2009), getting worse, or hasn’t made much difference either way?” 2 = better; 1 = about the same;
0=worse. (K05) Frequency of Political
Discussion
“How frequently do you discuss politics or campaign issues?
3=very often; 2=sometimes; 1=seldom; 0=never. (B01) Favorability of
Candidates
[Taipei] “From 0 to 10, what score would you give Hau Lung- Bin / Su Cheng-Chang?”(Q02A, Q02B)
[Taichung] “From 0 to 10, what score would you give Su Jia- Chyuan / Jason Hu (Hu Chih-Chang)?”(Q02A, Q02B)
[Kaohsiung] “From 0 to 10, what score would you give Yang Chiu-Hsing / Huang Chao-Shun / Chen Chu?”(Q02A, Q02B, Q02C)
Favorability of KMT/DPP
“From 0 to 10, what score would you give KMT / DPP?”(P02A, P02B)
