Southern Politics?
Regional Trajectories of Party
Development in Taiwan*
P
EI-S
HANL
EE A NDY
UNG-M
INGH
SUA common assumption is that regional disparities affect voting be-havior in Taiwan, given the fact that voters in the south show steady sup-port for the Democratic Progressive Party (DPP). The common sense premise is that the south provides a stable support base for the DPP. Al-though a pattern of political clustering appears to be taking shape, em-pirical data are required to test this hypothesis.
The findings of this paper suggest that, first, regional factors are contingent on the influences of election type and the personal attributes of individual candidates. Second, above and beyond what appears
common-©Institute of International Relations, N ational Che ngchi University, Taipei, Taiw an (RO C). Pei-Shan Lee (李佩珊) (Ph.D., Unive rsity of Pennsylvania) is Assistant Professor of Polit-ical Science at National Chung Cheng University, Chiayi, Taiw an. Her researc h interests are in the areas of comparative politics with a regiona l focus on East A sia, comparative democ-ratization, and international political economy. She has published articles on democ ratic governance in Taiwan, as well as issues on constitutional design and ethnic politics. Dr. Lee can be reached at <pollee@ccu.edu.tw>.
Yung-Ming Hsu (徐永明) (Ph.D., University of Michigan at Ann Arbor) is Assistant Re-search Fellow at Academia Sinica, Taipei, Taiw an. His reRe-search analyzes such issues as the Bayesian learning of mass preferences, the role that ideology and ethnic ity play in party realignment, and the dynamics of public opinion. He has published on these topics in vari-ous academic journals and edited books. Dr. Hsu ca n be reached at <polymh@gate .sinica. edu.tw>.
*An earlie r version of this pape r was presente d at the Conference Group on Taiwan Studies at the 2001 A nnual Me eting of the A merican Political Science Association, San Francisco, August 30-September 2, 2001. The authors are grateful to Chris Achen, John Hsieh, Tse-min Lin, and two anonymous re ferees for their comments and suggestions.
sensical, the Kuomintang (KMT) shows greater regional disparities with respect to its support base than does the DPP. This article proposes that party realignment has significantly affected the KMT's support base and has particularly influenced it in the non-southern area.
KEYWORDS: Taiwan party politics; southern politics; north-south divide; electoral geography; electoral stability/volatility.
* * *
The issue of "southern politics" or of the "north-south divide" has been the subject of debate since the 2000 presidential election. Democratic Progressive Party (DPP 民主進步黨) candidate Chen Shui-bian's (陳水扁) slim margin in winning the presidency over James Soong (宋楚瑜) was believed to have been due to his landslide victory in the south. According to the analysis of electorate distribution, southern counties were crucial in Chen's victory. Is there in fact a growing north-south divide with regard to Taiwan's electoral geography? A long-term dynamic analysis is required to assess the impact of regional disparities on electoral stability and vola-tility, however. Data obtained from one or two elections can hardly provide a solid empirical foundation for arguments of political clustering. This paper assumes that regional disparities do, indeed, influence electoral pol-itics in Taiwan, especially for the DPP. The DPP's reliance on loyal and swing supporters in the south has important implications for the rise of distinct southern politics. Meanwhile, the Kuomintang's (KMT, or the Na-tionalist Party, 國民黨) support base and its transformation in the process of democratization are also interesting topics worthy of investigation; therefore, the KMT's electoral support is also analyzed. The combination of the research findings can better portray Taiwan's party support patterns.
Problems on the Surface:
A Preliminary Exploration of Electoral Geography
This paper covers data of six national elections, dated from the pro-vincial gubernatorial election in 1994 to the 2000 presidential election,
which serve as the subjects of our in-depth analysis. Table 1 divides Tai-wan into six regions, of which Yun/Chia/Nan (雲嘉南, i.e., Yunlin County 雲林縣, Chiayi County 嘉義縣, Chiayi City 嘉義市, Tainan County 台南 縣, and Tainan City 台南市) and Kao/Kao/Ping (高高屏, i.e., Kaohsiung City 高雄市, Kaohsiung County 高雄縣, and Pingtung County 屏東縣) are classified as areas of "the south." The six elections that are the subject of this analysis include the 1994 provincial gubernatorial election, 1995 Legislative Yuan election, 1996 presidential election, 1997 county magis-trate/city mayoral election, 1998 Legislative Yuan election, and 2000 presidential election.
The DPP's share of vote in Yun/Chia/Nan and Kao/Kao/Ping in these Table 1
Distribution of Popular Vote for the DPP, 1994-2000 City/County Pei/K ee/Il Tao/Chu/Miao Chung/Cha ng/Tou Yun/Chia/Nan Kao/Kao/Ping Hua/Tung/Peng 2000 1998 1997 1996 1995 1994 Average SD 0.31 0.25 0.31 0.39 0.38 0.17 0.23 0.17 0.16 0.21 0.28 0.13 0.31 0.28 0.28 0.29 0.36 0.16 0.18 0.11 0.13 0.17 0.20 0.08 0.25 0.22 0.22 0.23 0.25 0.16 0.32 0.26 0.28 0.33 0.31 0.17 0.27 0.22 0.23 0.27 0.30 0.15 0.06 0.06 0.07 0.08 0.07 0.04 Total 0.32 0.21 0.30 0.16 0.24 0.30
Pei/Kee/Il (北基宜): Taipe i City (台北市), Taipei County (台北縣), Keelung City (基隆市), Ilan County (宜蘭縣).
Tao/Chu/Miao (桃竹苗): Taoyuan County (桃園縣), Hsinchu City (新竹市), Hsinchu Coun-ty (新竹縣), Miaoli CounCoun-ty (苗栗縣).
Chung/Chang/Tou (中彰投): Taic hung City (台中市), Taichung County (台中縣), Chang-hwa County (彰化縣), Nantou County (南投縣).
Yun/Chia/Nan (雲嘉南): Yunlin County (雲林縣), Chiayi City (嘉義市), Chiayi County (嘉義縣), Tainan City (台南市), Tainan County (台南縣).
Kao/Kao/Ping (高高屏): Kaohsiung City (高雄市), Kaohsiung County (高雄縣), Pingtung County (屏東縣).
Hua/Tung/Peng (花東澎): Hualien County (花蓮縣), Taitung County (台東縣), Penghu County (澎湖縣).
1994: Taiwan provincial gubernatorial election; 1995 and 1998: Legislative Yuan elections; 1997: County magistrate/city mayoral elections; 1996 and 2000: Presidential elections. SD: Standard deviation.
six elections are 27 and 30 percent respectively, which represents the party's best showing among all six regions. Table 1, nevertheless, shows that the DPP's share in all six regions is quite stable, with a standard devi-ation of between 0.04 and 0.08. However, the DPP fares better in the Yun/ Chia/Nan and Kao/Kao/Ping areas than in others.
In comparison, the KMT's share of vote in each region shows greater fluctuation, with a standard deviation between 0.07 and 0.11. The Nation-alist Party's share of vote is generally lower in the north (27 percent), yet its electoral support in the non-northern area is relatively consistent with few disparities. Table 2 demonstrates the KMT's percentage of vote dis-tributed in the six regions.
However, the fact that the DPP has made gains in electoral support in the south does not suffice as solid empirical grounds for the assertion that the rise of southern politics is clearly in the hands of the DPP. It is still pre-mature to make such a claim because tables 1 and 2 do not indicate whether or not the same group of electorates consistently supported the DPP. In other words, the stable share of vote in the south may be attributed to a certain number of swing voters, a factor highly indicative of the hidden dimension of electoral volatility. It is reasonable to assume that due to the different nature of the elections— in magistrate versus presidential and candidate-centered variations, for example, swing voters may shift their Table 2
Distribution of Popular Vote for the KMT, 1994-2000 City/County Pei/Kee /Il Tao/Chu/Miao Chung/Chang/Tou Yun/Chia/Nan Kao/Kao/Ping Hua /Tung/Peng 2000 1998 1997 1996 1995 1994 Average SD 0.18 0.18 0.19 0.20 0.20 0.15 0.27 0.31 0.34 0.35 0.31 0.29 0.32 0.27 0.30 0.26 0.32 0.29 0.34 0.46 0.42 0.45 0.43 0.42 0.20 0.34 0.37 0.35 0.32 0.29 0.32 0.46 0.46 0.39 0.42 0.41 0.27 0.34 0.35 0.33 0.34 0.31 0.07 0.11 0.09 0.09 0.08 0.10 Total 0.19 0.31 0.30 0.41 0.30 0.39 SD: Standa rd de viation.
support to or away from the DPP or otherwise vote strategically. Therefore, a more sophisticated method to unravel the impact of such variables is required.
Hung Yung-tai introduced a method to delineate political clustering by means of aggregate data and clustering analysis. He compiled data ob-tained from individual polling booths, clustered them into geographic units, and finally identified the political inclination of each geographic unit. For instance, one unit may be regarded as a traditional iron-vote area of the KMT whereas another may be considered a domain controlled by specific local factions. The pattern of clustering was obtained by variation analysis. In the next stage, by using cluster analysis, the computer re-judged the clustering results of each unit, thereby creating a map of electoral geog-raphy.1
For example, in the case where one particular party enjoys stable electoral support in one unit— say a majority of 60 percent, this unit can be identified as that party's political domain.
Nevertheless, such an analysis is flawed, in part because this ap-proach does not take into consideration the proportion of swing voters. Gary King's Ecological Inference Model, however, can be applied to cal-culate the proportion of swing voters for specific parties, a feature that can-not be determined by merely comparing the ups and downs in the number of votes one party receives.2
After the measurement of the stability of a party's support, we further run regressions with other variables— namely ethnicity and regional distribution— to explore how electoral support is related to each factor.
The Structure of Analysis
As shown in table 3, we present a framework to accommodate four
1Hung Yung-tai, "Election Prediction: A n Aggregate Data Assisted Model," Xuanju yanjiu
(Journal of Electoral Studies) (Taipei) 1, no. 1 (1994): 93-110.
2Gary King, A Solution to the Ecological Inference Problem (Princeton, N.J.: Princeton
dimensions of electoral support. From a time-series perspective, the term "loyal voters" refers to those who vote for Party A at points t and t-1. The term "swing voters" includes those who flow into or flow out of Party A's camp at either time point. Oppositional voters do not vote for Party A at either t or t-1.
Table 3 contains information as to both electoral stability and the di-rection of support. The measurements of loyal and oppositional voters re-veal the degree of stability of electors for or against specific parties. The difference between loyal and oppositional voters lies in the divergent direc-tions of support they represent. On the other hand, the calculation of swing voters can determine each party's electoral stability and direction of vola-tility. In addition to providing an illustration of cross-level analysis, table 3 also serves as a framework to map out the structure of a party's political support as well as the direction of vote shifts.
Put differently, a cross-tabulation analysis between two elections can provide us with a new analytical perspective. As table 4 shows, by
Table 4
Four Types of Party Development
Votes That "Swing In" High
Low
Low High
Party is expanding its support base Party is consolidating itse lf
Party is replacing its support base Party is decaying
Voters That "Swing Out" Table 3
Four Dimensions of Electoral Support
Election at Time t
Vote for Party A Vote for Party ~A Election at Time t-1 Vote for Party A
Vote for Party ~A
Loyal Voters Sw ing Voters
Swing Voters Oppositional Voters
examining the stability of electoral support as well as the direction of vote shifts, we can discern the patterns and possible trajectories of party development.
If a party enjoys high stability in terms of a stable constituency, and also attracts a high percentage of swing voters, the party can be said to be on the track of expanding its support base. As a party experiences growth in the number of swing voters, albeit at the price of losing its traditional supporters, a pattern of replacement is under way. This means that this party is replacing its support base. The decaying type refers to a situation in which a party's electoral support is not stable and, at the same time, the inflow of swing voters is low. If a party still hangs on to its traditional sup-porters with high stability but attracts few swing voters, it falls into the category of a consolidating party. Using information obtained from elec-toral politics, this framework is conducive to the analysis of Taiwan's party development, especially between two specific elections.
Based on the above framework, the calculation of "loyal" and "swing" voters demonstrates a party's traditional support base together with its efforts to attract new supporters. Barry Burden and David Kimball's approach to the study of ticket splitting is particularly inspiring for our research.3
Here we will proceed with two stages: in the initial phase, we apply Gary King's model and method to calculate the proportion of ticket splitting that each region receives in the Legislative Yuan elections. With the data obtained on the basis of regions as dependent variables, the second stage begins a regression analysis with other independent variables.4
3Barry C. Burden and D avid C. Kimball, "A New A pproach to the Study of Ticket
Splitting," American Political Science Review 92, no. 3 (1998): 533-44, and David C. Kimball and Barry C. Burden, "Seeing the Forest and the Trees: Explaining Split-Ticket Voting within Districts and States" (Paper presented at the a nnual meeting of the American Political Science Association, Boston, Massachusetts, 1998).
4The following w orks have all adopted this method: Burde n and K imball, "A New Approach
to the Study of Ticket Splitting" and "Seeing the Forest and the Trees"; Onathan Cohen, Thad Kousser, and John Sides, "Sincere Voting, Hedging, and Raiding: Testing a Formal Model of Crossover Voting in Blanket Primaries" (Pape r presented at the annual meeting of the American Political Science Association, Atlanta, G eorgia, 1999); Claudine Gay, "Taking Charge: Black Electoral Success and the Redefinition of American Politics" (Ph.D. disser-tation, Harvard University, 1997).
The data in this paper comes from electoral surveys from 359 towns in Taiwan; the time span covers six elections between 1994 and 2000. We determine the proportion of "loyal voters" and "swing voters" in each town by adopting Gary King's model. Regional distribution is then used as an independent variable to gauge its degree of association with the inde-pendent variables. In addition, in order to consider the impact of different types of elections on voting behavior, we classify these six elections into three categories: (1) the executive elections (including the 1994 provincial gubernatorial election and the 1996 and 2000 presidential elections); (2) the Legislative Yuan elections (1995 and 1998); and (3) consecutive elec-tions (1994/1995/1996/1997/1998/2000).
The DPP's Support Base: Loyal and Swing Voters
The DPP's Electoral Support in Executive Elections
As shown in table 5, regression analysis is conducted on the basis of three executive elections. Independent variables are loyal voters (94d96d, 96d20d, and 94d20d) and swing voters (d9496, d9620, and d9420), while Table 5
DPP Electoral Support in Executive Elections
Intercept Hakka Township Minnan Township Mainlander Tow nship Yun/Chia/Nan Region Kao/Kao/Ping Region 94d96d Loyal Voters d9496 Swing Voters 96d20d Loyal Voters d9620 Sw ing Voters 94d20d Loyal Voters d9420 Swing Voters 0.451** –0.103** 0.008 0.073** 0.025 0.110** 0.008** –0.002** 0.000 0.002** 0.001 0.003** 0.933** 0.009** 0.016** 0.008 0.012** 0.011** 0.120** 0.037** 0.098** 0.042** 0.088** 0.059** 0.815** –0.026* 0.070** 0.048** 0.066** 0.068** 0.050** –0.007 0.049** 0.025* 0.062** 0.058**
Note: Numbers shown in the table are e stimated B in the regression.
** = significance at .01; * = significa nce at .05.
dependent variables are single ethnicity-dominated towns and regions.5
The main reason that we incorporate ethnic variables into our model is be-cause the ethnic distribution data collected by Wang Fu-chang (王甫昌) is consistent with our voting booth data based on the same unit of analysis: county/township.6
More importantly, however, we would like to go be-yond the confines of the current literature on regional voting that tends to treat north-south division and ethnic cleavage in a mixed fashion. There-fore, we intended to design a model in which ethnic and regional factors are incorporated; meanwhile, by controlling the ethnic factors, regional variations can be detected.
As 94d96d indicates, since ethnicity townships and regional variables are dummy variables, townships consistent with conditioned variables are recorded as 1, while other towns are recorded as 0. Therefore, B = 0.451 represents non-southern towns7
without significant ethnic concentrations. This means that among DPP supporters in the 1994 provincial guberna-torial election, 45.1 percent of them voted for the DPP again in the 1996 presidential election. Therefore, the degree of loyalty among DPP sup-porters is 0.451 + 0.025 = 0.476. This illustrates that among those who lived in the Yun/Chia/Nan area and voted for the DPP in 1994, 47.6 percent of them voted for the DPP again in 1996.
Figure 1 shows that among the DPP's loyal voters in the 1994 provin-cial gubernatorial election, 45.1 percent of them voted for the DPP again
5We follow Wang Fu-chang's criteria to c lassify the Minnan (閩南人), Mainlander (大陸或
外省人), and Hakka (客家人) area s. See Wang Fuchang, "A Survey of the Social Founda-tions of Ethnic RelaFounda-tions in Taiwan" (Report to the National Scie nce Council by the Institute of Soc iology, A cademia Sinica, in Chinese, 2000), 10. The Hakka-concentrated area totals 42 at the county/tow nship administrative leve l. The Minnan area refers to regions where the Minnan pe ople constitute 90 percent of the e ntire populace or w here the population of the other three ethnic groups is lower than the average. There a re 193 Minnan areas at the coun-ty/township level. The Mainlander area is classified as a region w ith more than 20 percent Mainlander composition. Twenty-one Mainlander areas can be counted at the county/town-ship level across the island.
6Ibid.
7"The south," in our definition, inc ludes Yun/Chia/Nan (Yunlin County, Chiayi County,
Chia-yi City, Tainan County, and Tainan City) and Kao/Kao/Ping (K aohsiung City, K aohsiung County, and Pingtung County). A reas outside of the category of Yun/Chia/N an and Kao/ Ka o/Ping are "non-southern towns/areas."
in 1996. Among the DPP's loyal voters in the 1996 presidential election, 93.3 percent of them voted again for Chen Shui-bian in 2000. Among the DPP's loyal voters in the 1994 provincial gubernatorial election, 81.5 percent supported Chen Shui-bian. In other words, Chen Shui-bian's winning of the presidency in 2000 can be significantly attributed to his ability to attract a large proportion of DPP traditional supporters in the non-southern area. No sharp contrast between the south and non-south was seen among DPP loyal supporters, yet the southern supporters were slightly more in absolute numbers than non-southern supporters.
Figure 2 represents the proportion of swing voters. In non-southern areas, as d9496 shows, less than 5 percent of those who did not vote for Figure 1
Regional Distribution of DPP Supporters in Executive Elections
Figure 2
the DPP in 1994 supported the DPP candidate in the 1996 presidential elec-tion. By contrast, in non-southern areas, 10-25 percent of those who did not vote for the DDP in 1996 voted for Chen Shui-bian in 2000. Also, Chen Shui-bian received 5-15 percent of the swing voters in the same area in 1994. Briefly put, the DPP attracted a significant proportion of the swing voters in the non-southern areas in the 2000 presidential election.
Figures 1 and 2 give us an account of the DPP's development. The support base of the DPP decayed during the 1994-96 period when both the ratio of swing voters and the number of loyal supporters were low. By contrast, in 1996-2000, the DPP expanded. Of course, the pattern shift from decaying to expanding may not necessarily represent a long-term trend, perhaps only having been triggered by the nomination of popular candidates.
Comparison of the 1994 provincial gubernatorial and 1996 presiden-tial elections: Table 6 presents the DPP's loyal and swing voters in both
elections. From the perspective of regional distribution, among all DPP loyal supporters in 1994, the Kao/Kao/Ping area showed a higher pro-portion of loyalty in 1996 than did Yun/Chia/Nan and the non-south. In general, Kao/Kao/Ping provided the DPP with solid electorate support in both 1994 and 1996.
As concerns regional distribution, the overall proportion of swing voters moving to the DPP in 1996 was insignificant, yet in the Kao/Kao/ Ping area the share was slightly higher than in Yun/Chia/Nan and in the non-south. Generally speaking, however, the regional differences among these areas were few.
Table 6
Regional Distribution of DPP Loyal and Swing Voters in 1994 Provincial Gubernatorial and 1996 Presidential Elections
Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9496dpp Loyal Voters 9496dpp Swing Voters 0.451 0.476 0.561 0.008 0.009 0.011
Comparison of the 1996 and 2000 presidential elections: As shown
in table 7, over 90 percent of DPP supporters in 1996 voted for Chen Shui-bian again in 2000 in all three areas. It was evident that the DPP maintained its support base in 2000, with a degree of loyalty of over 90 percent. Yun/ Chia/Nan and Kao/Kao/Ping demonstrated even higher percentages of loyal supporters. Similarly, in addition, swing voters were slightly higher in the south.
Comparison of the 1994 provincial gubernatorial and 2000 presiden-tial elections: Similar to the previous finding, a higher percentage of DPP
loyal supporters in 1994 and 2000 come from the south (see table 8). In both the 1994 provincial gubernatorial and 2000 presidential elections, however, the DPP faced competitive KMT rivals who captured more than 50 percent of the total vote. Conceivably, the DPP's traditional supporters had switched their allegiance to the KMT camp in both 1994 and 1996. It appears that the DPP kept a high percentage of loyal supporters in the south. Of interest, though, is that even though Chen had served as Taipei Table 7
Regional Distribution of DPP Loyal and Swing Voters in 1996 and 2000 Presidential Elections Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9620dpp Loyal Voters 9620dpp Swing Voters 0.933 0.945 0.944 0.12 0.208 0.179 Table 8
Regional Distribution of DPP Loyal and Swing Voters in 1994 Provincial Gubernatorial and 2000 Presidential Elections
Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9420dpp Loyal Voters 9420dpp Swing Voters 0.815 0.881 0.883 0.05 0.112 0.108
city mayor, this political capital did not translate into electoral support for him in 2000. This is worthy of further investigation.
In terms of swing voters, a significant/considerable share of the DPP's traditional supporters shifted toward the KMT in 1994 and 1996, yet shifted their support back to Chen in 2000. The phenomenon is significant in the south, where the highest frequency of this occurred in the Kao/Kao/ Ping area.
DPP Electoral Support in the Legislative Yuan Elections
Consistent with our previous findings, the highest ratios for both DPP loyal and swing supporters were found in the Kao/Kao/Ping area. The Yun/Chia/Nan area followed, but the non-south lagged slightly behind. For those legislative candidates with party labels, their votes represent a party's most loyal supporters. Accordingly, the DPP still had more opportunities in the south to win swing voters in the Legislative Yuan elections. Table 9 demonstrates that the DPP's ratios of loyal and swing voters increased with a decrease in latitude.
Regional Distribution of Loyal DPP and Swing Voters in Consecutive Elections
Figure 3 points out the changes in terms of DPP loyal voters in con-secutive elections from 1994 to 2000. The regional differences were not quite significant, yet the non-south had the least number of DPP loyal voters in consecutive elections. The DPP's support base in the south ap-peared stable and less susceptible to party realignment.
Table 9
Regional Distribution of DPP Loyal and Swing Voters in 1995 and 1998 Legislative Yuan Elections
Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9598dpp Loyal Voters 9598dpp Swing Voters 0.534 0.571 0.629 0.053 0.076 0.106
On account of changes in the number of DPP swing voters, the re-gional differences varied with different elections (see figure 4). In the 1998-2000 and 1997-1998 elections, the ratios of swing voters from the non-south were significantly lower than those from the south.
From figures 3 and 4, clear is that the basic trajectory of the DPP's de-velopment has been expanding, with a higher degree of electoral loyalty and more and more swing voters, except for 1997-98. Why did the DPP electoral base appear to be decaying in the 1997-98 period? This may be Figure 3
Regional Distribution of DPP Loyal Voters in Consecutive Elections, 1994-2000
Figure 4
Regional Distribution of DPP Swing Voters in Consecutive Elections, 1994-2000
attributed to the type of elections because, as a rule, the DPP has generally fared better in executive elections than in those for the Legislative Yuan.
Summary
The above analysis consistently shows that the DPP's support base is concentrated in the south, but the differences across the regions are not overly striking. In addition, the regional distribution of DPP supporters has been influenced by electoral rules and candidate-centered factors. For in-stance, the DPP had more loyal and swing supporters in the executive than in the legislative elections. However, in the 1996 and 2000 presidential elections, candidate-centered factors played a crucial role in both con-solidating loyal supporters and attracting swing voters. Finally, the DPP's electoral support stabilized; party realignment, such as the emergence of the pro-independence Taiwan Independence Party (TAIP, 建國黨), did not affect the party support pattern to any significant degree.
The KMT's Support Base: Loyal and Swing Voters
The analysis of the DPP's support base confirms our presumption that DPP supporters in the south have a higher degree of loyalty. Nevertheless, is the electoral loyalty in the south exclusively directed toward the DPP? Or, do southern voters, generally speaking, have high loyalty to their iden-tified parties no matter who they are? To clarify whether the DPP enjoys exclusive electoral loyalty from the south, the following section conducts a similar analysis of the KMT.
The KMT's Electoral Support in Executive Elections
Figure 5 shows that the KMT had its lowest number of loyal voters in the non-south. About 49 percent of those who supported James Soong had supported Lee Teng-hui (李登輝) in 1996. Among those who supported Lee in 1996, 24.3 percent held on to their loyalty to the KMT. Only 18.7 percent of those who voted for the KMT in 1994 supported Lien Chan (連戰), the KMT presidential candidate in 2000. Clear is that Lien Chan
was unable to attract very much loyalty from the KMT's traditional sup-porters. The KMT's support base in the executive elections varied across regions, but in general the south (particularly the Kao/Kao/Ping area) showed a greater degree of loyalty than in the non-south.
Figure 6 demonstrates that with respect to regional differences, the KMT attracted 49.7 percent of the swing voters in the 1996 presidential election, while Lien Chan only attracted 24.3 percent in 2000. Regional differences were almost nonexistent.
In a similar vein, from figures 5 and 6, we note that in 1994-96 the KMT's development fell into the category of replacement given that the KMT attracted a large proportion of swing voters but the degree of loyalty
Figure 6
Regional Distribution of KMT Swing Voters in Executive Elections Figure 5
Regional Distribution of KMT Loyal Voters in 1994 Provincial Gubernatorial and 1996 and 2000 Presidential Elections
from traditional supporters was not significant. This pattern may not have been reflected in the total share of votes the KMT received but is revealed in our analytical framework. Lee Teng-hui's leadership during this period may have been a triggering factor in the replacement of the KMT's sup-port base. Even so, in 1996-2000, the KMT's development appeared to be decaying with low electoral stability and few swing voters. The consecu-tive development from replacement to decay may not represent a long-term trend since candidate-centered factors may have played an intervening role. However, worth noting is that although some of the votes the KMT lost in 1994-96 shifted back to the KMT in 1996-2000, the number of voters turning away from the KMT was more significant. In short, the KMT has encountered a hollowing out process by losing both of its traditional sup-porters and swing voters.
Comparison of the 1994 provincial gubernatorial and 1996 presiden-tial elections: In 1994 and 1996, the KMT fielded competitive candidates
whose mobilization abilities resulted in their overwhelming victory in both elections. The KMT had nearly 50 percent of all loyal supporters in each region. In other words, Lee Teng-hui consolidated almost one-half of the loyal voters who had supported James Soong in 1994. The former president's ability to rally support was effective across all regions. In addi-tion, Lee Teng-hui's ability to win over swing voters was also significant (see table 10).
Comparison of the 1996 and 2000 presidential elections: In
com-paring Lee Teng-hui and Lien Chan as KMT presidential candidates in 1996 and 2000 respectively, table 11 indicates that only 20-30 percent of
Table 10
Regional Distribution of KMT Loyal and Swing Voters in 1994 Provincial Gubernatorial and 1996 Presidential Elections
Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9496kmt Loyal Voters 9496kmt Swing Voters 0.497 0.519 0.543 0.381 0.41 0.415
voters remained loyal to the KMT in 2000. Interestingly enough, the Kao/ Kao/Ping area was the most loyal, but the differences between Yun/Chia/ Nan and the non-south were slim. Lien Chan only won 10-20 percent of swing voters, and again regional differences were not significant.
Comparison of the 1994 provincial gubernatorial and 2000 presiden-tial elections: Table 12 illustrates that only 20-25 percent of James Soong's
supporters voted for Lien Chan in 2000. The huge disparity can in part be attributed to Soong's own effective mobilization ability.8
In addition, Soong versus Lien in the 2000 presidential campaign constituted a crucial choice for both traditional and loyal supporters of the KMT. The 20-25 percent share of KMT supporters represented a consolidated support base for the KMT. The Kao/Kao/Ping area still had the highest proportion of loyal voters, while Yun/Chia/Nan and the non-south did not show any
sig-8Hsu Yung-Ming, "The Phenomenon of James Soong: Elite Splitting, Electoral Mobilization,
and Party Realignment," Lilun yu zhengce (Theory and Policy) 14, no. 2 (2000): 93-118. Table 11
Regional Distribution of KMT Loyal and Swing Voters in 1996 and 2000 Presidential Elections Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9620kmt Loyal Voters 9620kmt Swing Voters 0.243 0.245 0.299 0.111 0.112 0.111 Table 12
Regional Distribution of KMT Loyal and Swing Voters in 1994 Provincial Gubernatorial and 2000 Presidential Elections
Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9420kmt Loyal Voters 9420kmt Swing Voters 0.187 0.206 0.257 0.15 0.154 0.158
nificant disparities. The fact that the KMT captured less than 30 percent of the KMT's loyal voters reflected the general weakness of Lien Chan in assembling support. The regional distribution of KMT loyal voters showed a clear north-south difference, which was more significant than the DPP's support pattern.
The KMT's Electoral Support in the Legislative Yuan Elections
As table 13 shows, 75-80 percent of voters supported the KMT in both the 1995 and 1998 Legislative Yuan elections. It appears that the KMT's electoral support in the legislative elections was quite stable (see table 14). In terms of regional distribution, the Yun/Chia/Nan area had the highest degree of loyalty. This finding is different from the KMT's Table 13
Regional Distribution of KMT Loyal and Swing Voters in 1995 and 1998 Legislative Yuan Elections
Area Non-Southern Areas Yun/Chia/Nan Kao/Kao/Ping 9598kmt Loyal Voters 9598kmt Swing Voters 0.761 0.807 0.743 0.129 0.135 0.118 Table 14
KMT Electoral Support in 1995 and 1998 Legislative Yuan Elections
Area Intercept Hakka Township Minnan Township Mainlander Township Yun/Chia/Nan Region Kao/Kao/Ping Region 95k98k Loyal Voters k9598 Swing Voters 0.761** –0.043* 0.022 0.054* 0.046** –0.018** 0.129** 0.000 0.037** 0.040* 0.006 –0.011
Note : Numbers shown in the table are estimate d B in regression.
** = significance at .01; * = significanc e at .05.
performance in the executive elections. This discrepancy may have re-sulted from the prevailing influences of local factions in the Yun/Chia/ Nan area in which the electorate voted on the basis of candidate-centered considerations. From the perspective of ethnicity, the KMT had the highest degree of loyal supporters in towns where Mainlanders (大陸或外省人) are the dominant ethnic group. The KMT had fewer loyal voters in Minnan (閩南人) towns, while the Hakka ( 客家人) towns exhibited the lowest degree of loyalty to the KMT.
In terms of swing voters, only 10-20 percent of the vote shifted in favor of the KMT in 1998. The KMT's support pattern in the Legislative Yuan elections appeared stable; thus, the space to attract swing voters may have been limited. The regional distribution of swing voters demonstrates that the Yun/Chia/Nan area had the most swing voters behind the KMT, but the regional differences were not significant.
The KMT's Electoral Support in Consecutive Elections
Figure 7 shows that the KMT enjoyed the highest ratios of loyal voters from the Yun-Chia-Nan area in two Legislative Yuan elections (1995 and 1998), but suffered the lowest degree of loyalty from the same area in the 1997 executive election. One possible explanation lies in the different nature of the elections. The legislative election is a non-zero-sum game that leaves sufficient room for local factions to maneuver. In addition, the Figure 7
KMT had been cooperating with local factions most successfully in agri-cultural counties in the Yun/Chia/Nan area. It follows that this Yun/Chia/ Nan area provided the KMT with a stable support base. In the 1997 execu-tive election, however, the zero-sum nature of the election triggered fac-tional friction and splits in the KMT camp; the degree of electoral loyalty decreased as a result.
Figure 8 shows that the highest ratios of swing voters were in the Yun/Chia/Nan area in the 1995 and 1998 Legislative Yuan elections, but this area also had the lowest proportion of swing voters in the 1997 execu-tive election. This is in part due to the functioning of local factions whose campaigns were able to attract some non-party-oriented voters. In 1997, friction among local factions directly resulted in a split in the KMT camp in the Yun/Chia/Nan area; thus the inflow of swing voters decreased. In the 1996 presidential election, the proportion of swing voters was unusually high, which means that Lee Teng-hui did attract a significant number of non-KMT supporters.
Figures 7 and 8 demonstrate that, except in the 1995-96 period, the KMT has been decaying since 1994. Not only has the ratio of swing voters supporting the KMT been decreasing, but the degree of loyalty from its traditional supporters has also dropped. The notable exception in 1995-96 could be explained by a contingent factor that presidential candidate Lee Teng-hui was able to consolidate traditional supporters while attracting Figure 8
swing voters. In this period, the KMT showed a pattern of replacement. Since 1996, however, the continuous decay has been apparent. This may reflect the less obvious point that although Lee Teng-hui was believed to have facilitated the KMT's replacement process by appealing to the broad support of the native people, his personal charisma did not pass on to suc-cessive KMT leaders. Therefore, the KMT's replacement effort pursued by Lee Teng-hui was not firmly established and could not be sustained. From this perspective, Lee Teng-hui's attempt to transform the KMT into a native party may well have led to a failed coup d'état.
Summary
The analysis of the KMT's support base in the south finds that the Kao-Kao-Ping area provided the most loyal voters for the KMT in execu-tive elections, while the electorate in the Yun/Chia/Nan area displayed the highest degree of loyalty to the KMT. Although electoral loyalty had re-gional disparities, the nature of the elections and candidate-centered factors also played important roles in affecting voting behavior.
Reflections
This paper is by nature preliminary in that the study of regional voting is not as well developed as research on ethnic politics and party system in terms of literature adequacy and data accumulation. As a first-step effort, we have attempted to explore whether or not party support base in Taiwan has anything to do with regional factors. Moreover, since the structure we adopted in presenting the data is a cross-election analysis in a time-series framework, we also sought to construct a four-cell typology as a theoretical framework to make sense of the implications contained in the data. We have found that, with the exception of one study,9
there is a theoretical
9The study by Carmines and Stimson is the only theoretic al piece in both the English- and
Chinese-language lite rature. See Edw ard Carmines and James Stimson, Issue Evolution:
Race and the Transformation of American Politics (Princeton, N.J.: Princeton University
lacuna in the current literature on developing typologies that may help researchers sort out different patterns and delineate possible trajectories. Therefore, we have attempted to present such a framework that could serve to trace party changes and to understand how regional variations are related to party development.
We have analyzed two dimensions of electoral support: electoral sta-bility and electoral volatility. Together with the regional factor, we found that Taiwan's party politics has regional differences and southern voters have shown significant support for the DPP. This tendency has prompted researchers to explore the phenomenon of "southern politics."
As our analysis has indicated, the loyal voters of both the DPP and KMT all come from the south. The non-southern area showed a low degree of partisan loyalty. Southern voters appear to have a more stable partisan identification, thus displaying higher degree of loyalty toward specific parties. Therefore, although the data confirms our presumption that the south presents the highest degree of support for the DPP, the KMT also has gained a higher degree of loyal supporters from the south. The southern voters seem to have stable party patterns that are less vulnerable to the influences of party realignment.
On the other hand, the DPP has gained the most swing voters from the south regardless of the nature of the elections. This means that the DPP has shown a greater capacity to mobilize swing voters in the south. For the KMT, however, there has been no specific regional concentration of swing voters.
To sum up, the notion of the rise of "southern politics" is not sup-ported by sufficient empirical grounds. Although the DPP seems to have more electoral support in the south, equally puzzling is that other candi-dates like James Soong also proved effective in gaining support from this same region. Therefore, further research into this puzzle is required.
Other References
Huang, Chi, and Yi-Chiau Chang. "Splitting Vote: An Analysis of the 1997 Chiayi City Mayor and Legislative Elections." In The Effects of Level-of-Analysis
on Political Research, edited by Yung-Ming Hsu and Chi Huang, 183-218.
Taipei: Weber Publication, 2001.
Lough, Ming-Ching. "The Provincial and Gender Differences of Educational Ac-complishments." Taiwan Economic Review 29, no. 2 (2001): 117-52. Lublin, David I. "Context and Francophone Support for Quebec Sovereignty."
Working Paper, American University, Washington, D.C., 2000.
Herron, Michael C., and Kenneth W. Shotts. "Using Ecological Inference Point Es-timates in Second Stage Linear Regressions." Manuscript, 2000.
Voss, D. Stephen, and David Lublin. "Ecological Inference and the Comparative Method." APSA-CP: Newsletter of the APSA Organized Section in
Compara-tive Politics 9, no. 1 (1998): 25-31.
Voss, D. Stephen, and Penny Miller. "The Phantom Segregationist: Kentucky's 1996 Desegregation Amendment and the Limits of Direct Democracy." Working Paper, University of Kentucky, 2000.
Wolbrecht, Christina, and J. Kevin Corder. "Gender and the Vote, 1920-1932." Pa-per presented at the annual meeting of the American Political Science Asso-ciation, Atlanta, Georgia, 1999.
