Authority Structures and Single-Party Dominance in Indigenous Communities in Taiwan

Across the developing world, the breakdown of authoritarian rule and single-party dominance mark political transitions, and party competition has been widely recognized as a sign of democratization (Diamond 2002; Magaloni 2006). To understand how single-party domination endures during democratic transitions, scholars explore the basis of authoritarian legacies and factors weakening potential opposition forces (Darden and Gryzmala-Busse 2006; Greene 2007). In particular, recent scholarship (Riedl 2014; Schmidt 2020; Slater 2010) highlights the vital roles of local elites and their problem-solving networks in sustaining party identification. Building on these explanations, I establish the need to examine organized social relations to flesh out why single-party support lasts in some cases but not in others.

During political transitions, the former ruling regimes usually enjoy resource advantages from authoritarian legacies. These advantages can sustain single-party support, as voters continue to identify with the party in power since authoritarianism. The authoritarian legacies generate lasting single-party support because of ideological roots and material exchanges. First, the incumbent party’s agenda, such as mass schooling, can shape nationalism and political belief before party competition becomes common (Darden and Gryzmala-Busse 2006:90). And in addition to this ideological foundation, authoritarian regimes capitalize on existing ties between parties and local networks, or “quid pro quo direct exchange of material benefits for political support” (Kitschelt and Wilkinson 2007; see also Shefter 1977). Such ties are the foundation for the “poor voters backing elite parties” paradox. The political machines in urban Egypt and Mexico (Masoud 2013; Yashar 2005) are two examples where former elite parties have retained state resources since the authoritarian era and thus secure their votes through service-based voter mobilization.

But in developing communities, which are less accessible by party organizations, poor voters backing elite parties remains puzzling. Attempting to resolve this puzzle, recent scholarship identifies local elites as important in establishing the foundation of clientelism (Galvan and Sil 2007; MacLean 2010). Specifically, by offering the poor rudimentary welfare services, the elite parties recruit disadvantaged constituencies to their support while retaining the loyalty of the party’s elite core (because the limited redistribution does not threaten the elite). For example, while illustrating poor constituencies’ support for the right-wing Bharatiya Janata Party (Indian People’s Party), Thachil (2014) demonstrates that the party privately distributes public goods through outsourced local organizations to secure single-party support.

Although current scholarship has identified that both historical legacies and informal organizations shape single-party support in underprivileged communities, how these legacies and clientelist ties interact with existing power relations in these communities remains underexplored. In answering the question, scholars point to social and economic cleavages developing from class, ethnicity, and religious ties that set up differences for political engineering (Lazarsfeld, Berelson, and Gaudet 1944; Manza and Brooks 1999; Wedeen 2008). To understand these differences, it is crucial to contextualize power relations within and across local organizations. But “the politicization of [cleavages such as] ethnic groups does not translate directly into a particular type of political behavior” (Riedl 2014:7). Rather, to build informal political connections for party identification, local elites play a determining role. Research has found that the stronger the elite cohesion, the less likely it is that the incumbent party’s support can be undermined (Riedl 2014:218), but how these local elites foster affinities between parties and social groups requires further analysis.

Notably, the extent to which social and economic cleavages get politicized also indicates whether there is room for challengers to develop and engineer political competition. In particular, the stakes for building alternative problem-solving networks are higher for the opposition during political transitions because the former authoritarian regimes hold and occupy existing networks. As political developments in Malaysia (Slater 2010:23), Turkey (Muftuler-Bac and Keyman 2012), and Africa (Riedl 2014) show, even though opposition parties may gain some temporary victory, the parties associated with former authoritarian regimes have deep connections with the population through cohesive internal organizations that can “overcome subsequent disputes over power and position” (Riedl 2014:215) and thus have the power to retain single-party dominance. Meanwhile, a question yet to be answered is why some constituencies offer more access to party competition than others.

In sum, studies exploring authoritarian legacies find that informal political organizations play a salient role in sustaining single-party support. However, to further understand party dominance (and how it fractures), a central puzzle is: What kinds of organizations strengthen local elite cohesion across constituencies during political transitions, so that opposition parties struggle to establish alternative problem-solving networks or recruit local elites by different ideologies? While voter mobilization drawing on partisan narratives and social cleavage explains single-party dominance, my analysis underscores the interactions between the organized social relations and electoral outcomes that enable party support to persist. Building on existing studies, this research identifies structural patterns—namely, the chief and big man systems—that explain the mechanisms that shape the unevenly enduring authoritarian legacies and subsequently determine the access of the political opposition to these constituencies in rural areas.

The authority and leadership structures of a community often indicate the social relationships among the community members, as well as how resources are mobilized or transferred within the channels of social networks. Research on how lasting social bonds shape party loyalty and political support identifies the critical role of local authorities as intermediaries between political parties and the populace (Cornell and Kalt 2000; Jolly and Mosko 1994; LeVine 1989). For example, elected politicians attempt to form coalitions with chiefs; successful coalitions, in turn, often enable the chiefs to retain land ownership, and the politicians to attain legitimacy to rule (Baldwin 2014; Lawson 1996). However, the analyses of these authorities have not explicated (1) the cross-group patterns through which authorities establish or retain their authority among their fellow community members, or (2) whether various kinds of existing social relationships experience similar development. By looking into the contrasting authority types—the chief and big man systems—I explain the different mechanisms for sustaining party identification and the processes by which such community-level power relations fracture.

The chief and big man systems are contrasting authority structures across the Austronesian communities in the Pacific Islands, identified as an ideal contrast for comparisons by Sahlins (1963:285). Across the social sciences, scholars have recognized Sahlins’s framework and generated organizational studies confirming this structural difference (Boehm et al. 1993; Burrell and Morgan 1979; Martin 2009; Powell 1990). The Polynesian chief system is characterized by an ascribed elite stratum of inherited nobility that continues to dominate for generations, while the Melanesian big man system is open to any aspiring member to compete for spontaneous leadership (Hennings 2007; Sahlins 1963). The chief system, with its inherited hierarchy, creates fewer incentives for members to compete for leadership, compared with big man communities (Finney 1973; Kaplan 1998). Chief and headman families are entitled to political authority, while commoners are obliged to follow their leadership with its inherited social prestige. Thus, political entrepreneurs (contenders) struggle to challenge the hereditary command structures. In contrast to the preexisting inequality in chief communities, the big man system is characterized by redistribution, a process during which “those of great prestige have neither power over others nor [have monopolized] any resources” (Martin 2009:216–17).

The big man plays a mediating role across extended families (or phratries) as political affairs are negotiated across clans. Traditionally, household heads gather to select the big man based on his ability and resources. In the absence of an ascribed status, one needs to demonstrate generous reciprocity—or even “autoexploitation”—to achieve a big man title (Lederman 2015; Sahlins 1972:136). Even after being recognized as a leader, a big man needs to constantly display his qualifications, so he strives to garner as many resources as possible for “giving away” to fellow members (Hayden and Gargett 1990; Martin 2009:216). This need to expand the number of followers creates a mutual dependence—and contemporary democratic elections provide an arena to showcase such “giving-away” for potential big men (Godelier and Strathern 1991; Sahlins 1963; Stewart 1990).

The fundamental difference between the two structures is whether the local power relations have been centralized: leaders with inherited social prestige continue to hold decision-making authority in chief communities, while in villages with the big man system potential leaders can attain social recognition after “jump-starts” (Martin 2009:216). In particular, during state incorporation and capitalist development, contemporary indigenous communities faced rapid social transformations, especially the settlement of foreign institutions (Huang 2012). One common foreign influence is the establishment of Christian churches (including Catholic, Protestant, evangelical, and other denominations). Aborigines have shown resistance to the settlement or reformulation of native value systems (Huang 1984; Kaplan 1998). Like their attitudes to alternative mobilization networks under multiparty systems, how indigenous peoples react to these exterior institutions, and the processes of institutionalization, thus offer sites to delineate the mechanisms that facilitate competition for political allegiance within communities (Stewart 1990).

The different authority structures find their authentic archetypes coexisting across the Austronesian-speaking indigenous communities in Taiwan, and studies have identified the island as an source of the Polynesian chief and Melanesian big man groups (Bellwood, Fox, and Tryon 2006:1; Bellwood and Dizon 2008). The authority structures—initially studied by Japanese scholars who investigated the indigenous tribes for colonial governance—and their representativeness in Taiwan offer a suitable ground for controlled comparisons (Mabuchi 1951, 1960; Huang 1984; Scaglion 1996). The indigenous communities settled on the island 6,500 years ago, and some of their ancestors brought their linguistic roots and social structures to other Pacific islands (Bellwood, Fox, and Tryon 2006; Manning 2003:35). Meanwhile, as ethnic minorities residing in rural areas, aborigines in Taiwan have suffered from social exclusion—for instance, 60 percent of the 600,000 indigenous people live below the island’s poverty threshold (Lin 2012).

Taiwanese indigenous communities’ typical structural patterns also show endurance in scholarly accounts (Blundell 2000; Huang 2012) and provide an ideal site to compare the authority structures.⁴ Researchers have demonstrated that these structures existed before the colonial rule of the early twentieth century, and their differences have persisted as the tribes’ geographical remoteness prevented formal incorporation into the state system until the recognition of sixteen indigenous groups in the last decades of the twentieth century (Huang 1984, 2012; Wang 2003). While these tribes have a variety of languages and rituals,⁵ the two types of authority structures are present in all of them, and thus offer a unique basis for comparison. The identified ethnic distribution and their respective authority structures have not only continued to be observed in scholarly accounts but have also been documented in official records (Indigenous Peoples Council 2014). Therefore, the persistence of their structural differences warrants this comparative study of authority structures.

The voting patterns of the indigenous peoples provide a particularly suitable ground for comparing authority structures during political transitions, given the vibrant democratization in Taiwan. The KMT government moved to Taiwan in 1949, having been defeated by the Chinese Communist Party in Mainland China. In establishing a political base at the onset of authoritarian rule, the KMT used its lessons from the loss in the Mainland—most notably the party’s lack of support in rural areas. Thus it built local party branches, assigned party representatives in villages, and fostered pro-KMT ideology in the education system in Taiwan (Dickson 1993; Jacobs 2012). However, the KMT faced a challenge from the DPP toward the end of the twentieth century, and Taiwan has experienced three power turnovers in national elections in the past two decades. The fervent party rivalries reflect competition between the two major parties in voter mobilization strategies. The DPP has also secured resources and supporting networks at national and local levels during its members’ incumbencies.

But the indigenous constituencies remain exceptions to this political transition. The KMT has retained support among most indigenous voters, and party competition has barely been present in aboriginal constituencies (Figures 1 and 2). Though the DPP has promised more recognition for indigenous peoples at the national level—socioeconomically disadvantaged communities having faced obstacles to full civic inclusion ever since Japanese rule—the problem-solving networks the DPP has been able to secure elsewhere have struggled to make their way into most indigenous villages.

While state rule and political parties claim control across the indigenous areas, church settlements also underlie processes of social change in the tribes. As in many countries in Asia and Africa, church settlements bring democratic ideas, particularly in the case of the Presbyterian Church (Woodberry 2012:244–45). In Taiwan, the Presbyterian Church played a major role in initiating democratization (Lin 1999). Even though only 5 percent of the national population identify themselves as Christians (predominantly Protestant or Catholic), Christian churches are a dominant presence in indigenous areas—particularly Presbyterian Churches, which have appeared in more than 95 percent of the villages and shown a steady growth in attendance (Christian Resource Center 2013; Presbyterian Church in Taiwan 2009). This dominance of the Presbyterian Church not only provides a case to examine the internal power relations in the tribes but also makes the lasting single-party support for the KMT in indigenous communities more puzzling.

The first part of my findings is based on fieldwork and interviews. Subnational comparative ethnography is suitable for disentangling the uneven development of political and socioeconomic transformation (Snyder 2001:94). The linguistic, cultural, and developmental variations across aboriginal tribes further justify using the controlled comparative method to avoid inappropriate regional averages and focusing on one unrepresentative case. Ethnography also helps situate me in “the context in which practice takes place” and further enhances process tracing and mechanism building (Simmons and Smith 2015:14).

My focus on comparing authority structures to uderstand political development was informed by my ethnographic research and interviewees. This process exemplifies how comparative ethnography identifies meanings that are not detected and thus not controlled before fieldwork (Simmons and Smith 2015:14). Based on my initial fieldwork, in which I found that leaders with inherited titles (i.e., chief communities) play major roles in the persistence of support for the KMT, I generated a hypothesis that single-party support is more likely to decline in communities without a preexisting hierarchy and those which are already organized to have a place for competition (i.e., big man communities). Notably, the structural types were identified and categorized by the earliest scholarship on the indigenous communities, and I find communities based on their structural patterns according to recent research and official records (Blundell 2000; Huang 2012; Indigenous Peoples Council 2014; Wang 2003).

Once I identified authority structure as a key explanatory variable to understand the persistence and the erosion of single-party support, I selected atypical cases—where party competition has occurred—to maximize my capacity to find counterintuitive results. I compare the three anomalies across indigenous communities—two big man communities and one chief community, where the DPP has won at least 50 percent of the votes in one or more elections since the 1990s—to other neighboring communities whose support for the KMT has persisted. Table 2 (C for community) shows the case selection scheme based on the two comparative strategies.

Between August 2012 and early 2016, I carried out formal and semi-structured interviews in addition to ethnographic work in indigenous communities. By January 2016, I had gathered responses from 116 indigenous people. In chief communities, I usually interviewed chiefs and headmen in their family houses, but commoners preferred to be interviewed outside their tribes; in big man communities, I told respondents which family I was staying with, and they often chose to come to the family’s house to do the interviews in the yard.⁶

In addition to these interviews, I observed and participated in campaign organizing, townsmen reunions, family reunions, weddings, funerals, church services, marriage negotiations, and meetings of various community-based organizations. During fieldwork visits where I did detailed ethnography in some communities and observation in others, I stayed in the indigenous areas for about half a year cumulatively. When I was not at the field sites, I conducted follow-up interviews on the phone or through social media. In areas where nonindigenous people are in the minority, participating in everyday events not only allowed me to notice the details of interactions between individuals and across households but also let community members get to know me as a researcher studying political development across indigenous communities. My primary sampling method is snowball sampling, but after villagers got to know me, I found respondents from different social backgrounds (Table 3).

Based on my fieldwork and interviews, I argue that authority structures explain how party identification with the KMT has endured in indigenous constituencies. Local politicians⁷ and civil servants (soldiers, bureaucrats, and teachers) serve as intermediaries for mobilizing KMT support as they act as communal leaders, but such support could not become a diffused party identification without local structures. The KMT-affiliated local elites, despite having been nurtured through the party’s political machine or having received similar pro-KMT career training, have different degrees of influence over other members’ party identification, depending on their structural position. In chief communities, leaders with inherited status receive education and occupational training to support the KMT. During the past decades of democratization, the legacy of the party’s authoritarian rule and the leaders’ party identification have affected commoners primarily because of preexisting patron–client ties, operating outside a formal system of government (Weingrod 1968:378–79). In contrast, the political representatives in big man communities do not necessarily come from the most prominent family—families recognize the big man based on evaluation of his talent. Thus, the big man’s impact on voter mobilization depends on his temporary personal effort, which may have limited influence on the younger generations or weaken as soon as he loses leadership.

Across chief communities, local elites, especially elected officials, have been leaders with inherited prestige, for these leaders have the wealth and the social capital to take ruling positions in contemporary formal institutions. The inherited wealth and property of aristocratic families also granted their descendants easier access to education and connections with modern institutions in transitional societies, as compared with commoners. One headman explained, “We [leaders with inherited social prestige] had to acquire knowledge from the cities and non-aborigines to rule the tribe well, as our communities were facing rapid social change.” These successors of leadership later return to the tribes with the skills and qualifications to run local governments, as either elected officers or bureaucrats. Because these leaders received pro-KMT education and career training, they developed a strong party affiliation with the KMT. Subsequently, their followers demonstrated political loyalty to the KMT. While the authoritarian practice of enlisting local elites to mobilize communities was not unique to indigenous areas, the influence of this KMT mobilization system on election outcomes has declined in most nonindigenous communities in Taiwan (Wang 2004). However, chief communities demonstrate enduring support for the KMT given the continuous influence of the inherited leaders within and outside political institutions, and even from one generation to the next.

As in chief communities, the few people who manage to earn stable income as public servants in big man societies become suitable candidates to run for office primarily because of their skills in handling administrative and bureaucratic procedures. These contemporary big men also identify with the KMT through their occupational training. Although these recognized intermediaries between the members and institutionalized political actors affect other voters’ party identification when the social networks are relatively closed in indigenous areas and thus single-party support is established, the impact of the big men on political affairs decreases as soon as these leaders are no longer recognized as representatives of the community. Furthermore, young people in big man communities make political decisions under the influence of leaders recognized in their generation. One respondent in his thirties said:

Developing from the structural difference (i.e., whether the community has a preexisting centralized power structure), the impact of authoritarian legacies presents divergent persistence across the indigenous areas. In chief communities, hereditary leaders pass on their gratitude for their KMT education and career training to their descendants. In contrast, when the older generations gradually give way to younger big men who can gather resources, these rising leaders have little attachment to their seniors’ party identification. Similar uneven effects are evident from macrolevel social processes such as urbanization and mass education. Unlike villagers in stratified chief societies—where the social hierarchy continues to have an influence in urban indigenous churches and communities, and people from the same village stay together—the members of big man communities usually live with their households and seldom visit their tribes of origin.

To sum up, although community leaders are the basis for the lasting support for the KMT, the structural positions of these leaders in relation to other members determine how political loyalty has been established. In contrast to the ephemeral and tenuous power system of the big man communities (Finney 1973; Stewart 1990), the chief system grants its leaders inherited titles and recognition. In big man villages, where leaders are recognized based on their personal ability and therefore power turnovers have been frequent and common, authoritarian legacies, although present, may not be as lasting as in chief societies. This key structural difference also suggests that authority structures account not only for the origins of persistent loyalty to the KMT but also for the development of electoral competition.⁸

If authority structures shape support for the KMT, how can we account for the decline of such support in some communities? I further assess the mechanisms that explain the relationship between authority structure and defection from single-party support through two phenomena: first, whether there are opportunities for DPP-affiliated contenders in the community to gain votes, and the extent to which mobilization is deemed legitimate; and second, how competition for political allegiance in community-based institutions develops from and reshapes the preexisting social relations.

Across aboriginal areas, some community members with political ambition have aligned with the DPP. These alignments and their divergent electoral outcomes offer the first hints of the differing opportunities for building alternative political support between the two structures. In chief communities, political entrepreneurs who identify with and stand as candidates of the DPP have a marginalized inherited status. They have monetary resources but little recognition in their home tribe. In big man societies, potential big men might depend on the DPP’s support to amend issues when dealing with modern government institutions. They also mobilize voters for the DPP as the party helps resolve the issues; as a result, voters defect, at least temporarily, from single-party support for the KMT.

Political entrepreneurs who run for office as DPP candidates in chief communities usually have marginalized inherited status, because aristocratic families with higher status have mostly affiliated with the KMT. As a local director of a DPP indigenous branch put it, “I joined the DPP predominately because the KMT did not want to nominate me for the local election.…The KMT has so many loyal headmen as their potential candidates, why would I get a chance to compete, as someone with lower status?” On the one hand, these DPP-affiliated members seek monetary support from the party; on the other hand, some of them share the DPP’s ideology when they work or study in the cities. But their scant experience of participating in community affairs usually prevents these politically minded members from gaining either attention or support. An indigenous movement leader, who was one of the earliest aboriginal supporters of the DPP, recalled his participation in politics:

In contrast, party delegates in big man communities have had some electoral success for the DPP. In C10, for instance, where the DPP won two-thirds of the votes in 2012, many voters shifted away from identification with the KMT because a DPP legislator helped a local leader with a lawsuit. A distant relative of that big man told me, “While we sought help from both parties, only the DPP legislator put time and effort into the case, and his help indirectly benefited many households in the community—though the local elite was accused of embezzling public funds, villagers benefited from the construction of public facilities.” But other respondents pointed out that “after repaying the favor, the same local politician, in fact a KMT supporter, mobilized his fellow community members to ‘return’ other favors from the KMT by voting for the KMT.” This description was supported by the 2016 presidential election, when a majority of that constituency voted for the KMT.

Although DPP-affiliated political entrepreneurs have increased in number since democratization, the indigenous communities have shown different degrees of change in party identification across the two authority structures. Members in a chief community who have lived outside the tribe have incentives and experiences to mobilize support for the DPP, yet they usually fail because of their lack of inherited prestige. The DPP-affiliated movement leader’s description also demonstrates how his advocacy for nationwide indigenous rights does not translate into reducing the lasting single-party support in the village he comes from. Big men, whose main goal is to distribute resources in exchange for authority, might change their party identification depending on the sources of the resources, and this could fracture single-party support. But given the fluidity of the big man structure, such a fracture might only be temporary. This observation advances the literature by showing how organized social relations create opportunities for temporary victories by the opposition during political transitions (Muftuler-Bac and Keyman 2012; Riedl 2014; Slater 2010). Lasting changes in party identification result from internal competition within communities, as the following examples illustrate.

In the indigenous communities, Christian churches serve as local organizations indicative of power contestation. Around the mid-twentieth century, the indigenous communities encountered church settlement in addition to state incorporation. The responses of community members to church settlement have also differed between the two structures and further indicate whether they foster political opportunities for competition. In chief communities, leaders with social prestige regard the churches as contenders for authority. Even where these leaders converted to Christianity, both church leaders and leaders with inherited prestige emphasized the political legitimacy of inherited leadership. In big man communities, leaders initially joined the church to demonstrate their devotion and accepted the institution to consolidate decentralized communities. But gradually, potential big men established new churches to compete against others in power.

When I asked community members about who supports the DPP in chief societies, many informants directed me to staff or presbyters of the Presbyterian Church.⁹ One priest confirmed that she supports the DPP and has been frustrated by the enduring single-party support for the KMT in chief communities. She explained,

Most headmen and chiefs did not go to church; and even if they do at present, they emphasized that the churches have no right to interfere with political life. As one aristocrat noted, “Why should we listen to the presbyters for decision-making of the community? Commoners have no authority there.”

While preexisting hierarchy continues to shape the political landscape in chief communities, in spite of challenges from established foreign institutions, members of big man communities initially saw Christian churches and their economic assistance as an arena for authority competition. A woman in her sixties recalled, “Church leaders had to be elected, so those who got the position were recognized by their monetary and spiritual devotion; these leaders were usually elected local politicians.” In other words, local big men who had the ambition for power came to the church to compete for recognition.

Gradually, the competition for leadership became so intense that some leaders eventually walked out and established new churches. The new churches signify the decentralizing power structures that have been common and frequent in villages with the big man system. In the two big man communities where the DPP has been able to win more than half of the votes, voters explained the fracture of single-party support by associating party competition with competition between churches. One villager observed, “The party competition reflects the divide between the Catholic and Presbyterian churches. The Catholic church was established earlier in the tribe, but DPP followers eventually established the Presbyterian church.” In other words, the churches offer institutionalized alternative networks for resource distribution, and over time partisan affiliation corresponds with church membership.

The power dynamics of the authority structures are reflected in the interaction between indigenous communities and Christian churches, the most common foreign institutions in these communities. Although most of the residents are Christians, and the leaders of the Presbyterian Church attempted to counter the chief societies’ single-party support for the KMT, the leaders with inherited recognition remain authoritative in political decision-making. In big man communities, potential big men compete by forming and leading churches, a process that might increase the number of churches in each community. Such church establishments also exemplify the fluidity of that authority structure.

To test the generalizability of the observed differences between chief and big man communities regarding both lasting single-party support for and defection from the KMT, I create a novel data set using township-level electoral data from the Central Election Commission of Taiwan. I also collect demographic data from government bureaus for all 368 townships in Taiwan.

The KMT vote share in each township is my dependent variable. It captures the degree of party competition because of the general single-party support for the KMT in indigenous areas as compared with the national two-party rivalry (see Figure 1 for the persistently different trends in party identification between indigenous and nonindigenous constituencies). I analyze the electoral data from the 2016 presidential election because I conducted my fieldwork from late 2012 until the 2016 election: the outcome of the 2016 election reflects the development I observed. I choose to analyze national elections because they capture party identification more precisely than local elections; national elections avoid confounding effects from candidates’ personal ties with communities.

In Taiwan, there are 368 townships; of these, 55 are indigenous townships. I identify the tribes in each township based on the List of Authorized Tribes released by the Executive Yuan Indigenous Peoples Council. Following official government records and the typology of the indigenous groups in the literature (Huang 1984; Indigenous Peoples Council 2014; Kirch 2000; Mabuchi 1960), I use a categorical variable based on the respective authority structure of each tribe. Tribes with the chief system are coded as 2, and those with the big man system, 1; the rest of the townships are nonindigenous and are coded as 0.

The first control in this study is the proportion of aborigines in the township. How much of the population of a township is indigenous strongly correlates with the electoral outcome. Next, because each township has a different timing and experience of economic development, industrialization, and urbanization,¹⁰ I consider related factors that might have affected party identification. I control for the proportion of residents with high school degrees, because education is a prominent factor in party identification (Berglund et al. 2005; Inglehart and Klingemann 1976; Petrocik 1974). And I control for population, because studies show that the size of the community affects voting behavior (Panagopoulos 2010). In the context of this study, chief societies tend to be larger than big man communities (Huang 1984:4; Sahlins 1963:287), and nonindigenous townships are more populous than the indigenous ones.

I control for economic conditions through tax reports, using data from the Ministry of Finance. From the annual tax reports, I use the average tax paid by each household in every township. The amount paid reflects household earnings. Like education, the average tax paid by each household may positively correlate with opportunities for party competition, or higher vote share for the opposition party (Converse 1976). Moreover, I control for the degree of development in Taiwan. I follow the most recent measurement from the Executive Yuan, which estimates the proximity to the nearest city, socioeconomic conditions, and access to modern institutions, such as the education system, of all townships in Taiwan on a scale from one to five (Hsieh 2012). Furthermore, the location of a township and its proximity to cities largely determine what kind of information residents receive and with whom they interact. I control for the region where the township is—the north, central, or south region of the island. Controlling for the region is particularly important because the south, in general, has been more pro-DPP and the north, more pro-KMT (Lee and Hsu 2002).

Considering my ethnographic observations and the literature that suggests that competitive religious membership may affect party affiliation (Wittenberg 2006), I also control for the number of churches per ten thousand people in the township. Using the most comprehensive data available on the number of churches in each township (including Presbyterian, Catholic, True Jesus, Seventh-Day Adventist, and other denominations), from the Christian Resource Center, I control for church presence to evaluate the genuine and independent effect of authority structures on party identification. Table 4 shows the definitions of the variables used in the analysis.

I estimate a series of ordinary least squares (OLS) models to test the relationships between authority structure and KMT vote share. To test whether the factors that previous scholarship has considered affected party identification, I begin with baseline models that have all the controls. I then test whether, in the indigenous areas, authority structures would have an independent effect on vote share by adding my predictor—community structure—into the model. Given that the effect of church presence and membership might differ as a result of the preexisting indigenous social structures (Huang 1984:15–16; Inglehart and Welzel 2005:117; Woodberry 2012), I further test the association between party competition and the number of churches per ten thousand residents in each township.

Table 5 reports the descriptive statistics of the variables used in the analysis. First, authority structure and the proportion of aborigines positively correlate with KMT vote share. Both correlations support the indigenous peoples’ relatively strong party identification with the KMT. Table 6 shows the results of the OLS regression of authority structure versus support for the KMT. Model 1, the baseline model, first shows that the proportion of aborigines significantly correlates with KMT vote share (p < 0.001).¹¹ Model 1 also demonstrates the effect of other controls on vote share. Consistent with findings in the literature, region, education, income, and development are associated with electoral results. Townships with a higher proportion of high school graduates tended to cast more votes for the KMT. Degree of development is also positively associated with higher KMT vote share. Voters in the center and the south of the island also displayed significantly lower support for the KMT than voters in the north.

Is there an independent effect of authority structure (i.e., chief versus big man system) on KMT vote share? Estimates from OLS regression models are presented in Model 2.¹² I find that tribes with the chief system have significantly higher KMT support in the 2016 presidential election than big man and nonindigenous constituencies (p < 0.001). In fact, Model 2 demonstrates that in contrast to tribes with the chief system, the big man communities had shifted away from the KMT (11 percent fewer votes) more than nonindigenous communities (7 percent fewer votes). This finding supports the patterns observed in the fieldwork and interviews and suggests that the big man communities are more likely to shift away from single-party support, as compared to other authority structures. The effects of educational attainment, income, and region on KMT vote share are the same in Model 2 as in Model 1.

To evaluate whether the number of churches differs between structural types, Figure 3 shows show a bar graph of the church-population ratio. Since the church is an indicator of formal and competing institutions and thus has implications for power decentralization, the more churches in an indigenous township,¹³ the more likely the community has a stronger tendency toward party competition. Figure 3 supports this hypothesis from the field, as the church-population ratio is higher in big man communities than in townships with the chief system.

Based on qualitative and quantitative analyses, I argue that the influence of the leaders in chief societies rests on their inherited status, which contenders and the formal institutions established in recent decades have found difficult to challenge. In contrast, big men develop their influence on personal grounds, which creates room for contestation and the emergence of internal competition for political allegiance. Table 7 summarizes the main findings of the implications for political development across the two authority structures.

Authority structure not only offers a robust explanation for the variation in democratization but also enables competing hypotheses by comparing the two patterns. Even though leaders in both chief and big man societies support the KMT, based on economic and ideological legacies, patron–client ties have been stronger in the stratified chief communities. Similarly, even if ethnic conflict underlies indigenous peoples’ previous reluctance to support the DPP, it affects electoral outcomes unevenly. Overall, the findings support that preexisting social bonds determine how lasting the authoritarian legacies and memories of ethnic conflict can be.

How collective action develops in response to environmental and economic shocks across indigenous communities further strengthens the robustness of the findings. As the tribes live in rural and mountainous areas, many villages have been hit by disasters and had to be relocated. In 2009, Typhoon Morakot brought disastrous landslides, and many villages had to relocate. While the relocation took most members of the big man communities by surprise, many chief villagers had social movements that had been requesting relocation for decades. During the relocation processes of the 2010s, most big man communities struggled to consolidate their efforts to negotiate for the best conditions to meet the needs of the villagers. In contrast, the chiefs, as local politicians, represented the communities in negotiations with the government. As a result, the villagers in reconstructed big man communities often expressed envy of the more favorable relocation and reconstruction outcomes of the chief societies. Lo and Fan’s (2020) recent comparison of those reconstructions provides strong support for stronger elite cohesion in chief societies than in big man communities: amid post-disaster relocation, the village where grass-roots resistance to unresponsive social connections failed in that study was a big man community, while the other two, which cultivated awareness of the state’s symbolic violence through bonding networks and bridging ties, were villages with the chief system.

As party competition develops and becomes common, authority structure determines how effective local institutions are in mobilizing political support during political transitions. The Christian churches, for instance, have struggled to mobilize voters in chief communities but gained much influence in big man villages. The findings also provide a robust counterexample to modernization predictions (Almond and Verba 1963; Boix 2003; Lipset 1960; Soifer 2009). In the context of Taiwan, chief communities have been wealthier and have more access to institutional resources, compared with big man villages. However, party competition has been more common in big man societies. This example echoes the ongoing scrutiny of modernization theory, which has underexplored the “existence of structures reflecting historical forces, class interests, and power” across the developing world (Portes 2015).

The literature on democratization often regards lasting single-party support as a result of the incumbent advantages of formerly authoritarian regimes and the restricted identity on which oppositional parties were built. These explanations, however, cannot explain subnational variations in the continuity or change of party identification. Even when studies have explored how local social bonds shape electoral outcomes in transitional societies, they have paid little attention to patterned differences between types of relations. Closing these gaps in the literature by highlighting the dynamic power relations in underprivileged ethnic minorities, I demonstrate that the authority structures of communities determine the persistence or fracture of party loyalty during democratization, a process where single-party support originates from community members’ responses to changing political systems.

In contrast to the common nationwide party competition, indigenous constituencies in Taiwan have strongly supported the party that ruled the island before its democratic transition. As the analysis shows, the variation in local authority structure has determined how authoritarian legacies are embedded in indigenous communities in Taiwan. In chief communities, leaders with inherited prestige enjoy priority access to education and contemporary administrative institutions. Having received pro-KMT information during their education or occupation training, these leaders became hubs reinforcing identification with the KMT. In contrast, without centralized power under the control of a few families, the big man system has a volatile mechanism for generating leaders and thus offers more opportunities for potential infiltration of the opposition party—in this context, the DPP.

Authority structure also accounts for the different degrees of defection from single-party support during democratization. In chief communities, there are limited kinship networks that could support an organized opposition. In big man tribes, there are multiple politically relevant kinship networks, so there is a better structural foundation for political competition. In the environments that foster alternative problem-solving networks, such as Christian church settlements, churches serve as vehicles for establishing recognition and mobilizing support in communities without preexisting hierarchy (big man communities). Although a change in party identification is often conceptualized as a result of successful party strategies, a focus on local power structure shows that members with political ambition can more easily challenge established political support in one kind of authority structure (big man) than the other (chief). These structural differences and their theoretical implications are further supported by regression analyses. Controlled comparisons suggest the generalizability of the observed association beyond the scope of existing single-case studies that show the crucial role of local elites in mobilizing community members and determining their relationships with the party and the state.

This comparative study of party identification across disadvantaged minorities helps explicate how a “diverse set of linkage strategies between political elites, local power brokers, and the masses” generate subnational variation in electoral outcomes and democratization processes (Riedl 2014:231). The significant role of local authority structures can help structure our analysis of similar phenomena outside indigenous communities. In this instance, insights from the chief and big man systems uncover the organizational foundations of enduring authoritarian legacies to explain political transitions across much of Africa and Asia, where rural and poor constituencies also demonstrate lasting support for elite parties originating in former authoritarian regimes.

I have been able to conduct this research, over the course of four years, thanks to the generosity and help of the indigenous peoples in Taiwan. My deepest gratitude goes to all the villagers I have met and interviewed. I have also benefited from intellectual support from the academic community. I thank Elisabeth S. Clemens, John Levi Martin, Dong-Sheng Chen, Kai-Shyh Lin, Jenny Trinitapoli, Marco Garrido, Cheol-Sung Lee, Dan Slater, Dingxin Zhou, Dafydd Fell, and Andrew Abbott for invaluable comments on multiple drafts of the manuscript. Marshall Sahlins also provided helpful guidance when I was developing the research design.

I also want to thank participants in the Politics, History, and Society workshop and the East Asia: Politics, Economy, and Society workshop at the University of Chicago for their engagement with this manuscript. The questions and comments I received at invited talks in 2017 and 2014 at the School of Oriental and African Studies, University of London, the 20th North American Taiwan Studies Association Conference at the University of Wisconsin–Madison, and the 9th European Association of Taiwan Studies Conference in Sønderborg, Denmark, also helped me improve the presentation of the theoretical frameworks and empirical evidence. Finally, I am grateful for the constructive comments of the anonymous reviewers.