‘We were forgotten’: explaining ethnic voting in Bolivia’s highlands and lowlands

ABSTRACT With the election of Evo Morales and his party Movement to Socialism (MAS) in 2005, Bolivia has become the only country in Latin America to have an indigenous party in power. However, it is misleading to take the MAS government as representing all of Bolivia’s ethnic diversity. Its indigenous population can be classified into peoples from the Andean highlands and from the Amazonian lowlands. Research has treated the indigenous population as uniform or focused on the Andean peoples and on parties rather than voters. This paper aims to differentiate this picture by showing that variances between highland and lowland indigenous peoples started with the appearance of Homeland’s Consciousness (CONDEPA) and have increased since MAS came to power. While the highland indigenous peoples have preferred left political parties, parties with pro-indigenous agendas or which used indigenous symbolism, the lowland peoples have tended to support centre-right parties. The paper attempts to explain these differences in the voting behaviour of Bolivia’s indigenous peoples. Ethnic voting is analysed in the time span from 1985 until 2014 within a mixed-methods design.


Introduction
Since 2005, Bolivia has been the only Latin American country to have an indigenous party in power. Bolivia's socio-ethnic panorama is highly diverse. The 2009 constitution has redesigned the country into a 'pluri-national state' and established 36 indigenous languages. The official number of indigenous peoples equates or exceeds these languages. Only few of them live mainly in the Andean highlands and valleysthe largest being Quechua and Aymara -, the majority live in the lowlands. There are few highland peoples with large numbers of members and many lowland peoples with few members. As the Andean indigenous population constitutes the majority, they possess a high electoral weight and visibility at the political level. When speaking of an 'indigenous government', the question arises whether this relates to the overall indigenous population or exclusively to the highland indigenous peoples.
We will demonstrate in this paper that it is incorrect to consider Bolivia's indigenous population as homogenous, as their voting behaviour varies considerably. This gap is most notable between the highlands and the lowlands. Highland indigenous peoples have mostly voted for left and indigenous parties, parties that implemented or were promising policies that were popular amongst indigenous voters as well as parties targeting specific highland peoples in their campaigns. In contrast, lowland peoples have voted for the large mainstream parties with a preference for the centre-right. Significant differences in voting between highland and lowland indigenous voters started in 1989 with the emergence of CONDEPA, which targeted the Aymara municipalities La Paz and El Alto. Surprisingly, when Movement to Socialism (MAS) 1 , a political party with a pro-indigenous agenda, an indigenous president and alliances with indigenous organisations from highlands and lowlands, entered the political stage, the discrepancy between highland and lowland indigenous voters increased even further.
Research on ethnic politics has barely asked whom the 'indigenous parties' represent. It has mainly taken political parties rather than voters into account (Birnir 2004;Madrid 2012;Martí i Puig 2008;Van Cott 2003). Scholars who did analyse indigenous voters' preferences have either considered them as a homogenous unit (Rice 2011) or described their electoral differences while not attempting to explain them (Guzmán Prudencio 2014). In contrast, this article focuses on voters by analysing variation in electoral choice across ethnic groups and takes into account the heterogeneity of Bolivia's indigenous peoples. In order to analyse those differences, theory on party-voter-linkages provides valuable approaches to their analysis. Research on party systems offers convincing explanations for the rise of ethnic parties and the shift of large voter groups from one party to another.
This article seeks to offer a different perspective on ethnic politics. We set out to show and explain the prevailing electoral differences between highland and lowland indigenous peoples in Bolivia. Municipality-level census and election data are analysed based on ecological inference methods, which allows estimating the electoral behaviour of specific groups. The descriptive results of this quantitative analysis are interpreted in their historical context by evaluating the results of guided interviews with experts, indigenous politicians and activists in Bolivia.

Party-voter linkages: theoretical approaches to explain ethnic voting
The question this article attempts to answer is: How can differences in the voting behaviour between highland and lowland indigenous peoples be explained? We base our answer on an analysis of ethnic voting, defined as variation in electoral choice across ethnic groups. The concept of 'indigenous' employed in this article is based on the self-identification of the individual as indigenous, meaning a self-understanding as descendants of the native Americans who lived on the continent before the arrival of the European colonisers. This concept disregards language or colour. Consequently, the article employs data on self-identification.
The party-voter-linkages approach Electoral research turns around the questions of how parties and politicians manage to gather electoral support and what motivates voters to support them. A central issue is the nature of linkages between parties and voters (Kitschelt 1995). For the case of Bolivia, we identify four linkages through which political parties and electorates enter into longterm relationships with each other and which might potentially explain ethnic voting.
First, political agendas and policies have a major impact on the electoral behaviour (Bornschier 2010;Downs 1957). Voters support political parties expecting certain policies that will favour their own ethnic group. As the gap between indigenous and non-indigenous citizens is still enormous, poverty, poor education and discrimination affect indigenous lives in multiple ways (Gigler 2009). Indigenous peoples are the social group that would primarily profit from policies, such as welfare or multicultural policies, and have strong motives to support parties that implement those policies. Parties have focused on particular indigenous peoples throughout Bolivian history, which has sustainably shaped the party preferences of those groups.
Second, collective identity is another important linkage (Campbell et al. 1980). Voters support party candidates according to their ethnic identity, striving for reaffirmation of this specific identity. Candidates, in turn, emphasise this identity and appeal to their electorate on its basis. Identity must previously have become politicised. Indigenous peoples in Bolivia and Latin America have been historically excluded from political representation and socially and economically marginalised. Representation of their ethnic identities at the political level and their recognition at the social level are great motives for ethnic voting, according to the identity theory (Rice 2011;Van Cott 2000).
Another linkage between political parties and voters is through clientelistic networks. Voters would support political parties expecting resources in reward for votes. Members of ethnic groups would not vote for parties according to political contents or ideological positions, but choose a specific party calculating the highest likeliness of gaining positions, favours or material compensation for their group (Chandra 2004;Kitschelt 2000). Clientelism is hard to prove empirically. However, evidence suggests that it is likely to have an impact on ethnic voting in Bolivia.
Social movements have been identified as connectors between groups of voters and political parties. They are considered central actors for the appearance and maintenance of ethnic parties (Rice and Van Cott 2006;Van Cott 2005;Yashar 2005). Madrid (2012) viewed social movements as part of the secret of success of MAS, as the party emerged from a conglomerate of labour unions and indigenous interest organisations in the highlands and lowlands. This way, it managed to reach a broader spectrum of voters across ethnic borders. The breakup of the indigenous movement during MAS's legislative period was accompanied by an increase in ethnic voting.

Alternative explanations
Much of the literature on ethnic politics has focused on the recent indigenous parties, principally Bolivia's MAS, and their electoral strategies (Madrid 2012;Martí i Puig 2008;Rice 2011). Madrid explained the success of MAS by its inclusionary discourse. However, Flesken (2015) demonstrated a decrease in indigenous rhetoric during MAS' government between 2005 and 2010. As we will show in the empirical part of this paper, despite the drop in indigenous rhetoric, the share of highland indigenous voters who supported MAS increased during that period. Hence, it seems that discourse has a minor effect on ethnic voting in Bolivia.
Institutional hurdles, such as restrictive rules for the foundation of parties and penalties for parties with low electoral success, were particularly difficult to overcome for small indigenous parties with few voters and limited resources (Birnir 2004;Van Cott 2003. In the Bolivian case, those hurdles were abolished with the constitutional reform in 2004, when indigenous organisations and civil associations were allowed to run for elections. However, as we will show in the empirical part of this article, ethnic voting increased particularly after this date. Hence, the institutional approach cannot explain ethnic voting in Bolivia.

Data and methods
Bolivia has the largest indigenous population, proportional to the total population, of any Latin American country. According to the 2001 census, nearly two-thirds of the 8,274,325 Bolivians self-identified as belonging to an indigenous people. The largest people are the Quechua, with 30.7% of the population, the second largest the Aymara with 25.2%. Both live mainly in the Andean highlands and valleys. The lowland indigenous peoples constitute approximately 6% of the population (INE 2001) and represent ethnic minorities. The largest lowland peoples are the Chiquitano, Mojeño and Guaraní.
In this article, we differentiate between the largest highland and the lowland peoples. In so doing, we ignore the fact that the number of indigenous peoples in Bolivia is much larger and the ethnic composition of the population more diverse. We explain the reasons for this decision in part A.1 of the appendix. In our analysis we count the number of ethnic group members on the basis of self-identification since it corresponds to a subjectivist understanding of ethnicity that nowadays is standard in social sciences (for recent reviews see Chandra 2006;Wimmer 2008). We expose our reasons for this decision in part A.2 of the appendix. We use the 2001 rather than the 1992 or 2012 census, since it was conducted approximately in the middle of our election time series data and includes more pre-defined categories than the 2012 census. For further explanation, please consult part A.3 of our appendix.

Ecological inferences from aggregate data
In order to estimate the electoral behaviour by each ethnic group, we apply the ecological inference method (King 1997). This method bases its estimates on two main sources of information: Firstly, on the co-variance of the ethnic group size and the aggregate electoral results per ecological unit. Secondly, on information on the boundaries of the possible electoral behaviour of each group, which can be deduced from the combination of the homogeneity of the groups and the uniformity of electoral behaviour. In part A.4 of the appendix we show that most of the municipalities are ethnically homogenous.
We use election results at the lowest possible ecological level between 1985 and 2014. For the period 1985 to 2005, we use data collected by Raúl Madrid (2011); for 2009 and 2014 we collected additional data. For elections prior to and including 1997, the data had to be aggregated to the provincial level since municipalities changed often and did not allow censuses to be matched with election data at this ecological level. We matched the election data with the census data from 2001. The census data covers the size of Quechua, Aymara, Guaraní, Chiquitano, and Mojeño, of whom we aggregated the latter three to one category of lowland indigenous voters, plus a category for 'other' indigenous identities and one for non-indigenous identity. A detailed description and explanation of our analytical strategy based on ecological inference analysis is provided in part A.5 of the appendix. There we also cross-validate our ecological inference estimates with survey data. Finally, a step-by-step description for replication of the ecological inference analysis can be found in part A.6 of the appendix.

Semi-structured guided interviews
For the interpretative analysis, we realised field research in Bolivia in 2016 and conducted semi-structured interviews with experts, indigenous activists and politicians, whom we confronted with the results from the ecological inference analysis on group-level voting behaviour. A balanced selection of the interviewees, including indigenous politicians and activists of the main indigenous organisations as well as scientific analysts, covered the full range of perspectives. The thematic focuses were the linkages between indigenous peoples with political parties and the state, reasons for alliances, and motives for political party preferences. The interviews were systematised and analysed with atlas.ti. Their results were triangulated with the quantitative results on electoral behaviour over time and academic secondary sources. The interviews permitted a contextualisation of the results of the quantitative data analysis.

Why do lowland indigenous peoples vote differently from highland indigenous peoples?
Linkages between highland and lowland indigenous peoples with political parties and, more broadly speaking, the state, evolved in diverging ways. During the colony and after the transition to the republican state, the highland indigenous population served as cheap labour force in the minesthe main source of economic outputand maintained the state by paying excessive taxes from which the European oligarchy was exempted (Albó 2009;Farthing and Kohl 2014;Gigler 2009). Due to their integration into the colonial and republican state and their large history of suppression, the indigenous highland population gathered experience in political protest and political organisation since the colony (Antezana Ergueta 1994;Nasini 2002;Rojas Ramírez 1989).
By the mid-twentieth century, the lowland indigenous peoples had only very loose or no linkages with the state (Yashar 2005, 192). The lowlands were considered empty territories and grounds distributed as pasturelands for cattle farming to large-scale owners, who dominated the region's economy and politics for decades. State policies such as the 1966 Colonisation Law legitimised the appropriation of lowland indigenous lands by highland farmers and landholders (Yashar 2005, 194). Furthermore, the Forest Law, the Fauna Law, the Wildlife Law and the Minerals Law in the 1970s legalised the invasion of their lands for the exploitation of natural resources (Yashar 2005, 198). Political parties ignored the lowland indigenous population due to their small size and low weight in national elections (Gigler 2009). Besides, as parts of them were not officially registered, they did not form part of the body of voters (Nasini 2002;Van Cott 2005).
In the 1950s, the vast majority of Bolivia's population lived in the rural Altiplano (Gigler 2009;Roca 2001). Consequently, they were the main addressees of party appeals and policies. In 1952, the National Revolutionary Movement (MNR) party initiated reforms that would benefit the indigenous population, such as the universal suffrage. Also, large-estate owners were expropriated, indigenous farmers set free from dependent labour on large-scale farms and land was redistributed to them (Lopez Pila 2014). The land reform was applied to a larger extent in the highlands and in some parts of the lowlands, while it didn't reach other parts of it (Farthing and Kohl 2014;Weber 2013). Indigenous farmers were organised in peasant unions and a socialist ideology was imposed on them. These unions had their origins in the Chapare valley near Cochabamba and involved mostly the large highland peoples (Albro 2005).
The military dictatorships of the 1960s and 1970s were the first to politically integrate the lowland indigenous population (Van Cott 2005;Yashar 2005). In 1964, General René Barrientos Ortuño (1964-69) sealed the Peasant-Military Pact with indigenous leaders and established a network of clientelism and political favourism between them and his government, which guaranteed the support of the rural population (Albó 2009). General Hugo Banzer Suárez (1971-1978 and his ADN partyfounded in 1979had its major support base in the lowlands. Indigenous identity in the highlands was politicised in the 1960s. At that time, a rising class of Aymara intellectuals started criticising the ongoing imposition of state interests on the indigenous population and the disregard of their needs. The Katarista movement, named after the Aymara rebel Tupak Katari who resisted Spanish colonial rule, emphasised their cultural otherness to the ruling classes, who were seen as the successors of the European colonisers. At least since the 1974 massacre of peasants in Tolata near Cochabamba, the military government and the rural indigenous population were divided (Toranzo Roca 2006).
The peasant union organisational structures paired with Katarista ideology were the soil from which highland indigenous organisations emerged, among them the Unitary Unionised Confederation of Peasant Workers of Bolivia (CSUTCB) (Medeiros 2001). Socialist ideology and indigenous consciousness are today strongly intertwined in the highlands (Van Cott 2005). Between 1985 and 2014, 20 indigenous parties with Katarista and socialist ideological roots were founded, of which only Movimiento Indígena Pachakuti (MIP) and MAS achieved more than 5% of the votes. The peasant unions remain a powerful tool of political participation especially for Quechua and Aymara until today. As they are closely intertwined with MAS, Quechua and Aymara have higher influence on the government than the lowland peoples do.
Lowland indigenous peoples started mobilising in the late 1970s and founded the Indigenous Confederation of the East and Amazon of Bolivia (CIDOB) in 1982. During the 1980ies, CIDOB, supported by the NGO Support for the Indigenous Peasant of the Bolivian East, build its networks with lowland indigenous organisations (Albó 2009;Lopez Pila 2014;Medeiros 2001;Postero 2007;Van Cott 2005;Weber 2013). Lowland indigenous peoples had until then been widely fragmented and were largely unaware of the fact that they were sharing the region with other peoples with similar needs. Their indigenous identities started politicising only at that time (Albó 2009;Weber 2013). In 1990, CIDOB organised the so-called 'March for Territory and Dignity' over nearly 600 km from Trinidad to La Paz in protest against the unequal distribution of land in the lowlands and the political ignorance of indigenous interests. They were supported by highland activists. It was the origin of Bolivia's latest indigenous movement which existed until the mid-2000s. Until the 2000s, CIDOB remained an important actor for the defence of lowland indigenous interests (Van Cott 2005). Lowland indigenous organisations did not found their own distinct party. Due to their numerical inferiority, they were unlikely to pass the 3% threshold that existed until 2004. Parties that attracted less than 3% of the votes were obliged to pay a penalty. 2 As our analysis demonstrates, in the 1985 elections, shortly after the transition to democracy, voting behaviour was similar across indigenous peoples (see Figures 1-3). They showed a general attachment to the main political forces MNR, ADN and MIR, ADN being the electoral winner. Aymara and Quechua spread their votes more or less equally to those parties, whereas the lowland indigenous peoples preferred right-wing ADN (37%). Because of its pro-indigenous reforms in the 1950iesamong them the land reform -MNR was still popular in the highland countryside and some rural parts of the lowlands (Lopez Pila 2014, 5). But its 1982 government had initiated a disastrous inflation and ADN appeared as the political salvation, which is probably why even Aymara slightly preferred ADN (24%) over MNR (22%). As no party won the absolute majority and ex-dictator Banzer was not considered apt for government, the Parliament nominated MNR as the governing force. MNR initiated a neoliberal adjustment program, which successfully stopped the hyperinflation but also led to mass unemployment in the highlands. This can be considered MNR's first moment of dissociation from its main electoral constituency.
The appearance of CONDEPA in the 1989 elections marked the beginning of ethnic voting in Bolivia. CONDEPA was a neo-populist party; a tendency that emerged in the late 1980s around party leaders that criticised the traditional parties. These parties represented poor and marginalised voter segments, whose political representation was innovative, and allocated themselves on the left of the party system (Mayorga 2002). CONDEPA targeted the urban Aymara communities and won more than 10% of the votes. Aymara voters' first preference was CONDEPA, followed by left and (pro-)indigenous parties, whereas Quechua voters showed a general preference for centre-leftist parties. As Figure  4 shows, CONDEPA attracted mainly Aymara votes (29%) and achieved between 1% and 2% amongst all other ethnic groups, except for the 1997 elections, when CONDEPA became more popular amongst Quechua voters. It disappeared after 1997.
In 1993, CONDEPA gained votes among all ethnic groups but the Aymara, probably because of the campaign of rival Tupac Katari Revolutionary Movement of Liberation  (MRTKL), which was headed by the Aymara Víctor Hugo Cárdenasthe vice-president in the subsequent legislative periodin coalition with MNR. MNRthe electoral winner was still popular amongst the highland and lowland indigenous population. Another neo-populist party, the centre-right Civic Union of Solidarity (UCS), had its best electoral results in the highland cities La Paz and Cochabamba as well as the lowland city Santa Cruz. CONDEPA and UCS gained about a third of the total votes and were the third and fourth largest fractions in parliament.
The MNR-MRTKL government implemented a series of reforms as concessions to the indigenous movement, such as the 1994 Law of Popular Participation (LPP), which was meant to increase the degree of participation of the rural indigenous population. The LPP introduced elections at the municipal level (Van Cott 2005, 62 ff.) and allowed civil participation in decisions on local development and public spending (Postero 2007, 129). Indigenous organisations cooperated with the municipal plan in the implementation of rural development projects and indigenous representatives formed part of the 'provincial development councils', where they negotiated the prioritising of local needs with provincial and municipal authorities (Lopez Pila 2014;Medeiros 2001). This was particularly important for lowland indigenous organisations which until then had had very little say in formal politics. Besides, the law made it easier for small local parties to present themselves for elections, as the requirements for contesting were lower in municipal elections than in national ones (Van Cott 2005). The law changed the perspectives for indigenous actors to enter formal politics. Lowland indigenous candidates suddenly had a real chance to participate in and ideally win elections at local level.
Lowland indigenous constituents did not identify with indigenous parties, such as MRTKL, as they were considered to be dominated by highland indigenous peoples. Instead, they showed an orientation towards the main regional political forces ADN, MNR and MIR in their electoral behaviour. The indigenous parties have generally been parties which ( … ) were born of a Quechua or Aymara perspective. Hence, they are not attractive for an adaptation by the indigenous peoples of Bolivia's Eastern lowlands ( … ).In the country's East, the political parties with the largest influence are centre-rightist ( … ). The indigenous peoples from Bolivia's East support the parties of Bolivia's East, which are more traditional, centre-rightist and culturally legitimated in the political life of the Bolivian East. They are parties of businessmen, some cattle farmers, landowners, who have managed to gain legitimacy, consensus, as those are economically dominant groups. (18) Generally, as lowland indigenous peoples have barely been integrated into the statethey possess loose linkages with political parties and large parts of the lowland indigenous population were not even registered as citizens until the 2000s (Van Cott 2005) -, regional interests and power structures are more imminent than national matters.
Before, ( … ) we, the indigenous peoples, ( … ) did not get involved in matters that affected or did not affect us, any politics. These alliances seem to have translated into vote choices at the national level. The choices for allies at the municipal level were apparently taken according to benefit calculations, that is, the quality of promises party candidates gave to indigenous community leaders.
So instead of looking for an indigenous candidate to support, I believe in you. Why? Because I know that you have the money, you will give me some for my campaign and I and my community will have our place once you get elected. (15) Lowland indigenous voters seem to have followed the political culture of their interest organisation leaders in their electoral behaviour, namely a preference of short-term benefits over long-term political visions. This assumption is supported by the fact that lowland indigenous peoples' matters have never been part of the electoral campaign of any political party, meaning there was no political long-term offer suiting their specific needs.
We, the lowlands, have never appeared in any party programme, even though we have benefitted from some programmes that were applied on us, like health or food provision. (18) Even when lowland indigenous candidates presented themselves for parties with prolowland indigenous agendas, as was the case in 1997 with MBL or in 2014 with PVB, they preferred conventional parties with a tendency towards regionally entrenched centre-rightist parties. Indigenous lowland identity tends to gain relevance when it serves as a tool for the acquisition of resources from the state or NGOs (Weber 2013) or in the defence of territory vice versa highland indigenous migrants (Lopez Pila 2014). Lowland indigenous peoples link it to the regional camba identity, which is essentially a mestizo identity (Lopez Pila 2014).
Cooperation with highland indigenous organisations for a common political party was not considered because of interest differences concerning the use of land. Indigenous lowland peasants see lowland territories as being threatened by coca growers from the Andean valleys.
The territory is limited and both have access to the same territory. The colonizers and the groups of CSUTCB and 'interculturales' 3 need to use (soils in) the lowlands that have owners and they claim them for themselves. So there is rather a conflict for the ownership of the territory. (1) Besides, highland and lowland indigenous peoples differ in their political cultures: while the leaders of the latter have cultivated a cooperative style of negotiation, highland indigenous activists are more confrontational in their political demands (Van Cott 2005). A sense of superiority from highland indigenous leaders towards their lowland counterparts could be felt in the interviews: 'They are like our little brother' (7).
Despite the reforms of the 1990s, indigenous communities continued to experience an invasion of their lands by state and private actors and the socioeconomic gap between indigenous and non-indigenous population prevailed (Gigler 2009;Schilling-Vacaflor 2010). The indigenous voters in both highlands and lowlands turned away from the MNR-MRTKL coalition and, in 1997, Banzer's right-wing ADN won the elections. The party formed a government with support of MIR and CONDEPA, which had increased its electoral share amongst the highland indigenous voters and received most of its votes from Aymara (29%) and Quechua (10%), whereas its results amongst lowland indigenous voters remained poor. ADN had received high shares of votes from the lowland population, including the indigenous voters.
It appears paradoxical that the indigenous lowland population would support a party that was run by the regional oligarchy. But regional seems to overlap ethnic identity and the relations with local power holders appear stronger than those with the nation state that has hardly made an appearance in indigenous lowland communities, who barely benefited from state policies.
We were forgotten. ( … ) We have been persecuted and pushed aside ( … ) under former governments and even now. ( … ) It's a lie ( … ) that we don't want them to build roads for us, that we want to die just like that, that we are some non-contacted people, that we are isolated because we want to. No: We need the presence of the state. (11) In that light, it makes sense that ADN leader Banzer, who descended from a German family of large-scale land owners, was considered as 'one of them' by the lowland indigenous population, as indigenous leaders confirm.
First of all, he was a general, as we say: 'camba'; originated from the region. Hence, it is understandable that there was necessarily sympathy, ( … ) an electoral backup for him ( … ). Secondly, on behalf of this general ( … ) there was a real commitment to the region. (15) Banzer was one of Bolivia's dictators ( … ). But in spite of killing a lot of citizens during his dictatorship, ( … ) the indigenous (lowland) peoples ( … ) felt sympathy for him, as he was originally from the East, not from the high plateau. (16) Secondly, not only Banzer's regional identity had an impact on lowland indigenous voters, but presumably also the entanglement between lowland indigenous and ADN party elites, established in the 1970s. The military junta, which Banzer had headed for seven years, had been the first government to give lowland indigenous authorities a role in politics.
Political exclusion provoked growing dissatisfaction with the prevailing political system and conventional parties amongst the indigenous population in the 1990s. Despite obligatory suffrage having been in place since 1961, large parts of the overall indigenous population abstained from elections in the 1980s and 1990s. In 1989, 62% of the lowland indigenous, 52% of Aymara and 58% of Quechua voters absented themselves from elections. In 1993, 64% of the lowland indigenous population, 42% of Aymara and 48% of Quechua voters absented. In 1997, 22% lowland indigenous, 25% of Aymara and 29% of Quechua voters absented themselves. The large numbers of absentees might be related to the difficulties of reaching remote areas in the countryside and the lack of registration of many indigenous civilians at this time, but possibly also (looking at the variation of abstention in different elections and its relative congruence between highlands and lowlands) to the degree of political disenchantment amongst the indigenous population. The abstention rates were highest amongst lowland indigenous voters and only fell drastically from 2009 on, as Figure 5 shows. Driscoll and Nelson (2014) demonstrated in their analysis of the Bolivian 2011 elections that blank and spoiled votes are employed as an expression of political discontent.
In the late 1990ies, a social movement that comprised indigenous and other civil organisations reached national level. Its existence was facilitated by the pro-indigenous reforms in the same decade (Postero 2007). This movement entered formal politics in the new millennium to replace the traditional elites. One of its products was MAS, which was born out of an alliance between small political parties and civil society organisations and had its roots mainly in the highlands (Farthing and Kohl 2014;Zuazu 2009). The peasant unions of the coca farmers are strong supporters of MAS and the coca leaf formed part of its early discourse on indigenous identity (Grisaffi 2010). It had its first (overwhelming) The years around the turn of the millennium were marked by social riots. Until the 2002 elections, ADN had lost most of its support base, achieved only each 2% of Quechua and Aymara votes and 5% lowland indigenous votes. It had its best results amongst non-indigenous voters (10%). CONDEPA lost political importance after its coalition with ADN, which caused severe damage to its ideological credibility. Besides, the appearance of rival UCS did not permit the growth of either of them (Mayorga 2002). MNR became the strongest force in parliament again with 22.5% of the votes, followed by the newcomer MAS with 20.9%. When looking at the vote distribution of the particular ethnic groups, we can observe a gap between the two parties. MNR had accumulated the highest proportion of the lowland indigenous peoples' votes (41%). MAS had gained a solid portion of the votes of the highland peoples Aymara (20%) and Quechua (30%) and, to a lesser extent, lowland indigenous votes (10%). ADN was replaced by rightist party 'New Republican Strength' (NFR). NFR also achieved high vote shares from Aymara (19%) and Quechua (16%) and far less from the lowland indigenous peoples (5%). A new indigenous party, MIP, which principally aimed at the Aymara electorate, gained 18% of their votes, but only up to 2% from all other groups.
At that time, the demand for a profound change of the political system was strong, as dissatisfaction with a political system based on corruption and favours to the advantage of the traditional elites prevailed. Indigenous parties appeared as political alternatives (Guzmán Prudencio 2014). As the MNR government implemented no reforms, the upheavals continued after the 2002 elections, accompanied by a series of changes of government.
In 2005, MAS achieved an overwhelming victory, reaching the absolute majority with 53.72% of the votes. After the implosion of the former political power structure, MAS had polled large parts of the votes of the highland indigenous peoples (Aymara 63%, Quechua 52%), likewise of the main lowland peoples (22%). A new right-wing civil association, Democratic Social Power (PODEMOS), an ally of ADN, arose simultaneously, attracted the largest part of the lowland indigenous votes (45%) and substantial shares of Aymara (17%) and Quechua votes (30%). Another right-wing oppositional party (United National Front, or UN) appeared, landed in third place and polled almost equal shares of indigenous highland and lowland votes. MNR occupied the fourth position and attracted lowland indigenous peoples' (17%) and other votes, but had lost the highland indigenous constituency to MAS, which provoked its disappearance in the 2009 elections. Figure 6 shows MAS's steady growth since its first participation in national elections in 2002 amongst all ethnic groups until 2009. Thereafter it stagnated amongst lowland indigenous and other voters, compared to a decrease in Quechua, but a sharp drop of Aymara votes.
MAS offered a political alternative, presenting itself as an inclusive indigenous party with a socialist background, promising to abolish the old elites -'decolonize the system' and 'indianize the state'and redistribute the country's resources (Schilling-Vacaflor 2010). In contrast to former parties appealing to the indigenous population, which had recruited its followers mainly amongst the Aymara, MAS was rooted in a variety of civil society organisations and at the beginning of its government had indigenous supporters from highlands and lowlands (Van Cott 2005). Nevertheless, its strongest supporters were the highland labour unions. MAS' politics were backed up by the Pact of Unity, an alliance of the country's largest civil society organisations, which has existed since 2005. 4 The party's attempts to control and incorporate the member organisations provoked the withdrawal of CIDOB and CONAMAQ in 2011 and the breaking apart of both organisations into two wings, of which each one was co-opted by MAS, while the leaders of the non-official wings, just like entire indigenous organisations of the lowlands, have gone underground in recent years. While lowland indigenous organisations could act freely under former governments, they have experienced a process of dissociation with MAS, as their leaders have experienced persecution.
We have recently suffered an infringement of our rights ( … ), as they broke in here and snatched our president. ( … ) We have an international demand pending that we are developing. (11) In contrast, indigenous leaders from CSUTCB, which recruits most of its members amongst Aymara and Quechua, continued to identify with MAS as 'their' government in 2016: As a result of the peasant struggles, we, the CSUTCB ( … ) could found our own political instrument, which is [MAS-) IPSP -Political Instrument for the Sovereignty of the Peoples. ( … ) We have achieved our political banner, ( … ) we have participated in the elections and this is how our president Evo Morales won. We, the parent organisation, have a political party ( … ) and are doing fine: ten years in government! (7) The differences in the voting behaviour of highland and lowland indigenous peoples have increased since the beginning of the MAS government. While its support by the large Figure 6. Estimated turnout as difference between people in voting age and voters by ethnic group.
highland indigenous peoples continues to be high, the majority of lowland indigenous voters support oppositional right-wing parties. This development has gone along with the breaking apart of the social movement that brought MAS to victory. It no longer exists and broke apart into its constituent organisations.
It is difficult to determine whether MAS' source of support among Quechua and Aymara is its ethnic content or its ideological positioning on the left. Vice-president and MAS' chief ideologist García Linera has repeatedly pointed out and published numerous books on the party's merged ideology of socialism and indianism (García Linera and Stefanoni 2008). Its ideological base consists in a mixture of ethnicity and class (García Yapur, García Orellana, and Soliz Romero 2014). Anthropologist Albó observes that MAS has occupied the space of the traditional left and its constituency, which had not been addressed since the 1980s. Following its inclusive strategy, it has not only involved the rural indigenous population, but also urban middle class sectors (Albó 2009). Many indigenous interview partners stated there would never be another Evo. His biography seems to be a point of identification even across ethnic bonds, amongst members of diverse indigenous peoples: His humble descent, his mixed Aymara-Quechua origin and his political activism confer him great credibility, even with members of other indigenous minority peoples. As stated previously, since the beginning of its government, MAS' ethnic appeals have diminished and so did the number of its indigenous representatives (Flesken 2015, 3-4). Its ties with the highland unionswhich represent mainly Quechua and Aymaracontinue to be strong. All organisations based on indigenous ideologies -CONAMAQ (strongly linked to the Aymara), CIDOB and othersare nowadays divided and criticising the government. However, this seems a rather theoretical discussion, as both categories widely overlap: Its core support base, the highland unions, are composed by indigenous peasants, whether they identify primarilyaccording to the left unionist logicas peasants oraccording to the indianist logicas indigenous.
The cleavage between highland and lowland intensified during MAS' legislative period. Conflicts between highland and lowland states (departamentos) evolved on the basis of governmental initiatives. They threatened the lowland elites' privileges, namely the demand to share revenues of the exploitation of natural resources with the national state and the plan to expropriate large land properties. The lowland oligarchy, who denounced their exploitation by the central state, managed to mobilise the regionalincluding the indigenouspopulation for its political projects, emphasising of a regional 'camba' identity, and even advocated a regional segregation (Biggemann, Klimovich, and Thomas 2014;Marca Marca, Baigoria Guzmán, and Velarde Cuellar 2014). The conflict intensified during the years of the constituent assembly and calmed down when state autonomy was included in the 2009 constitution.
The 2009 elections were greatly politicised and highly polarised between highlands and lowlands. MAS surpassed its former result, achieving 64.2% of the vote, gaining large parts of the votes of the indigenous highland population (Aymara 80%, Quechua 66%) and even 47% of the non-indigenous votes, but a comparatively low share of 27% of the lowland indigenous peoples' votes. The majority of the latter gave their votes to an alliance of right-wing oppositional parties named Progress Plan for Bolivia-National Convergence (PPB-CN). PODEMOS had been dissolved in 2008 due to the non-fulfilment of legal requirements. PPB-CN came second with 26.5% of the votes, achieving 59% of the lowland indigenous votes, but only 9% Aymara votes and 18% of Quechua votes. The ballots were basically distributed between these two parties. All other parties achieved single-digit results, as did third-placed UN with 5.7%. The abstention rate had dropped to 12%.
MAS' relations with lowland indigenous peoples have degraded, not only as a consequence of the frictions in the Pact of Unity and the legal persecution of indigenous leaders. In the beginning of its legislature, MAS officials often claimed the indigenous peoples and the social movement had overtaken the government. However, central issues have got stuck in the middle, such as the regulation of highland migration to the lowlands and the annexation of lowland territories by highland farmers (Farthing and Kohl 2014) or the indigenous autonomies established in the 2009 constitution. The latter were meant to improve the situation especially of indigenous minority peoples and give them a greater say, e.g. on resource exploitation on their soils. Besides, lowland peoples criticise an overweight of highland indigenous culture in the government (Postero 2010, 27), a lack of attention by the latter and policies benefitting highland indigenous peoples (Lopez Pila 2014). The government clearly distinguishes between numerically more and less powerful peoples and rarely reacts to requests for dialogue with small indigenous peoples.
They ignore them (the Ayoreo 5 ), because they are few. In contrast, when the Guaraní call (the central government), they don't come the first time. The second time, they do react, because there is pressure and that's it. And they don't only make the minister come: They make the president come. (9) The so-called TIPNIS conflict around the construction of a highway through indigenous territory and the Isiboro-Sécure national park in 2011 illustrates MAS's position towards lowland indigenous interests. The government had failed to respect the prior and informed consent of indigenous residents, as mandated by the 2009 constitution, before signing the construction contract with a Brazilian conglomerate. Local leaders of the TIPNIS region feared the invasion of their lands by coca growers -'colonisers'from the Chapare valley, and organised resistance under the leadership of Fernando Vargas. It harshly suppressed the protest (seven deaths during the 2011 and 2012 marches) (Canessa 2014). Morales was criticised internally and internationally for his attitude towards indigenous rights.
MAS again won the 2014 elections, with 61.4% of the votes. The strongest oppositional force was right-wing Democratic Union (UD), an alliance of UN and right-wing and liberal parties from lowland metropolis Santa Cruz, with 24.2%. The conservative Christian Democratic Party (PDC) finished third with 9% of the votes. MAS had by then established itself as the principal force in parliament. Its major support base was again the highland indigenous constituency (Aymara 70%, Quechua 65%). The lowland indigenous support for MAS was not low either (48%), but their support for UD was far higher (29%) than Aymara and Quechua (12% and 17%, respectively). Despite MAS' continuously high support by the highland indigenous peoples, the Aymara vote share for MAS had fallen by 10% since 2009. This drop is likely to be related to the splitting of CONAMAQ and the threatening of its leaders.

Conclusions
The party-voter linkages approach allowed an analysis of the variance in the electoral behaviour between lowland and highland indigenous peoples. Political agendas and policies have had an impact on the electoral behaviour of indigenous peoples in Bolivia. Even though policies in the observed period have not led to a noticeable increase in ethnic voting in the observed period, previous policies have had repercussions on distinct party preferences of highland and lowland indigenous peoples. Highland indigenous peoples show a preference for leftist parties, which can be attributed to MNR's pro-highland indigenous policies in the 1950s. They equally prefer indigenous parties, parties with pro-indigenous agendas or which employed indigenous symbology, which is related to the Kataristas' influence on highland indigenous peoples' political identity formation. In contrast, lowland peoples have a tendency to vote for rightist parties, which were the first to politically involve them in the Peasant-Military Pact.
Indigenous identity seems to have played a minor role in explaining the differences between indigenous peoples. It appears to have had the largest impact on the electoral choices of the highland Aymara people, who have tended to prefer parties that promote Aymara identity. Regional identity appears to serve as a better explanation for indigenous voting preferences, especially in the lowlands. Lowland indigenous voters seem to be strongly influenced by regional politicians' appeals and their sense of regional belonging seems to have a major impact on their electoral choices. Ethnic identity seems to be a stronger trigger for vote choice among Quechua and Aymara than for members of lowland indigenous peoples, as for them regional entrenchment has stronger explanatory power than indigenous identity.
Clientelistic practices and networks are hard to prove. However, material incentivesnamely, offers by regional politiciansand collective electoral decision making are likely to account for indigenous peoples' vote choices and differences in the voting patterns between highland and lowland indigenous peoples. Highland indigenous peoples' organisations with their close ties to MAS have an impact on highland indigenous peoples' electoral decisions. The latter show a greater orientation towards parties that are supported by their organisations. In turn, lowland indigenous voters have shown a more pronounced orientation towards parties with the best offer for the regional indigenous communities.
Regarding theories on social movements, it is revealing to observe the parallel development of the breaking apart of the indigenous movement during MAS's government and the increase in ethnic voting at the same time. The segregation of the most recent indigenous movementthe linkage of one wing to MAS and the linkage of the second wing to the oppositioncertainly accounts for parts of the differences in the voting behaviour of indigenous peoples.
This article has shown that aggregation of ethnic groups must be handled with care, as one might oversee variation. The present paper contributes to the mentioned discussion, taking into consideration the complexity of indigenous peoples in Bolivia. However, future research might even want to differentiate in more detail between the diverse lowland indigenous peoples.
This paper also contributes to the body of knowledge on the linkages of marginalised groups with political parties and of their participation in political systems. Because this paper has observed an extended period of 29 years, it is able to look at several explanatory variables for the variation of ethnic voting over time. This approach must naturally remain at the surface of specific causal factors, which provide information on the motivations of voters. Diving deeper into voters' motivations must remain a task for future research. Among others, it might be revealing to investigate more in detail the impact of clientelistic networks and practices on differences in the electoral behaviour of indigenous peoples. Notes 1. Originally Movimiento al Socialismo -Instrumento Político por la Soberanía de los Pueblos (MAS-IPSP), but abbreviated here as MAS for simplicity. 2. According to the 1979 electoral law, parties that gained fewer than 50.000 votes had to cover the costs of printing ballots, which was very difficult for indigenous parties with minor budgets (Birnir 2004

Disclosure statement
No potential conflict of interest was reported by the authors.