Lechler and McNamee

Decentralized Despotism? Indirect Colonial Rule Does Undermine Contemporary Democratic Attitudes Marie Lechler∗ Lachlan McNamee† September 27, 2016 Abstract This paper identifies indirect and direct colonial rule as causal factors in shaping attitudes towards democracy by exploiting a within-country natural experiment in Namibia. Throughout the colonial era, northern Namibia was indirectly ruled through a system of appointed indigenous traditional elites whereas colonial authorities directly ruled southern Namibia. This variation originally stems from where the progressive extension of direct German control was stopped after a rinderpest epidemic in the 1890s, and thus provides plausibly exogenous within-country variation in the form of colonial rule. By analyzing individual-level survey data, we are able to build on the existing literature by disentangling the mechanisms through which different forms of colonialism likely affects contemporary democratic attitudes. Advancing a long-standing debate in the literature, our findings suggest that the ongoing influence of traditional leaders in indirectly ruled areas of sub-Saharan Africa is an important factor in undermining contemporary support for democracy.




1 Introduction “The authority of the chief thus fused in a single person all moments of power: judicial, legislative, executive, and administrative” (Mahmood Mamdani, Citizen and Subject: Contemporary Africa and the Legacy of Late Colonialism, p.23) What factors shape individual and community attitudes towards democracy? There is substantial cross-national and within-country variance in individual support for democratic institutions. This component of the political or ‘civic’ culture of a society has long been shown to play a potentially important role in affecting both the sustainability and success of a democracy (Almond and Verba, 1963; Inglehart, 1990; Putnam et al., 1994; Inglehart and Welzel, 2005). Yet, beyond a number of recent findings that show that support for democracy is endogenous to exposure to national democratic institutions (Persson and Tabellini, 2009; Fuchs-Schündeln and Schündeln, 2015; de Aquino, 2015) we have little quantitative evidence for the factors that cause such variance in political culture. In line with a body of literature that highlights the importance of colonialism for contemporary political and economic outcomes (Acemoglu et al., 2001; Engerman and Sokoloff, 2002; Hariri, 2012), this paper shall demonstrate that the ongoing influence of traditional leaders in indirectly ruled areas is an important factor in shaping contemporary support for democracy. The difficulty in demonstrating the effects of direct and indirect colonialism and its associated legacy of traditional leadership on contemporary democratic attitudes is, of course, that colonial strategies were not assigned randomly. Moreover, even if we believe that indirect colonialism tended to be conducted in pre-colonial states that were more centralized (Hariri, 2012), we cannot rule out that pre-state centralization also affects political culture through channels beyond the form of colonial rule. This paper thus exploits a natural experimental setting in Namibia that provides plausibly exogenous spatial variation in forms of colonial governance. In Namibia, as in sub-Saharan Africa as a whole, colonial authorities instituted systems of direct rule in those areas settled by white Europeans whereas, in those areas where indigenous population was not dispossessed, colonial authorities tended to rule through a indirect system of local ‘tribal’ elites (Miescher, 2012). Unlike elsewhere in sub-Saharan Africa, however, Europeans did not settle and directly rule only the most agriculturally fertile areas of Namibia (Werner and Odendaal, 2010) but rather settled in the more arid southern areas of Namibia which were hardest hit by an 1897 rinderpest epidemic. In order to protect German herds from future epidemics, a veterinary cordon fence was introduced at the spatial extent of direct German control in 1897 that divided northern and southern Namibia. In the face of stringent financial constraints, the German colonists then never completely expanded their settlement territory to the wealthier and more densely populated northern areas of the country (Eckl, 2007) but rather ruled indirectly through a system of appointed traditional authorities. Hence, whilst indirectly ruled areas of Namibia were governed through a system of appointed traditional authorities, traditional authorities were given little or no political role in the directly ruled central and southern areas of Namibia. 2 After Namibian independence in 1990, these regional differences in the influence of traditional leaders still persist; traditional leaders play an extremely important formal role in land allocation and customary law enforcement in northern Namibia whilst playing a largely symbolic role in central and southern Namibia (Keulder, 2000). Given that this colonial-era dividing line, progressively formalized throughout the 20th century, was drawn with little reference to existing indigenous communities, Namibia provides an ideal setting to examine the effect of direct and indirect colonialism on contemporary democratic attitudes. Moreover, by analyzing individuallevel survey data, we are able to disentangle the mechanisms through which direct and indirect colonial rule likely affects contemporary democratic attitudes. Specifically, we will seek to test whether contact to unelected traditional leaders weakens contemporary democratic political attitudes and thus whether there is a necessary trade-off between the consolidation of ‘traditional’ and ‘modern’ institutions. Such a theory was first powerfully advanced by Mamdani (1996) and developed by Englebert (2000), however recent work in the literature has instead argued that there need be no necessary trade-off between the legitimacy of traditional and modern democratic institutions in sub-Saharan Africa (Williams, 2004, 2010; Logan, 2008, 2009). Rather, because electoral and non-electoral accountability mechanisms can both be effective in keeping political leaders responsive (Baldwin, 2015), it may be that traditional and democratic institutional legitimacy is instead a ‘rising tide that lifts all boats’ (Logan, 2013). We adjudicate between these competing theoretical perspectives and find that that the ongoing influence of traditional leadership structures undermines support for central democratic tenets. The paper is structured as follows: we first present related literature and describe the historical background in Namibia. We then discuss the data and the OLS estimation strategy we apply to identify the effect of indirect rule on political attitudes. Finally, we present 2SLS results to isolate the channel of causality. 3 2 Related literature It has long been theorized that the legacies of direct and indirect colonialism have played important roles in shaping political phenomena in the post-colony. Directly ruled or ‘settler’ colonial countries tend to be both more democratic and have higher contemporary levels of education and income today (Acemoglu et al., 2001; Engerman and Sokoloff, 2002). Hariri (2014) suggests, in a classical modernist fashion, that directly ruled countries are more democratic because direct European colonialism better disrupted the bonds of traditional authority and thereby enabled the growth of a participatory, egalitarian political culture. Yet, we have little sense in this account of precisely what components of ‘traditional’ culture need to be disrupted for a democracy to subsequently flourish. Moreover, this account of the political effects of colonialism is analytically suspect because indirect colonial rule did not leave traditional forms of governance intact but rather often radically reshaped traditional systems of governance to better suit the administrative requirements of indirect rule. Specifically, it is a common misconception that traditional leadership in its common form has always been a historical component of governance in sub-Saharan Africa. Contemporary systems of traditional authority are instead actually the legacy of indirect systems of European colonial rule that radically changed existing African indigenous forms of governance (Newbury, 1988; Mamdani, 1996; Boone, 2014). In extending their control over indirectly ruled colonies, colonial authorities re-fashioned the existing political landscape by recognizing and bolstering the coercive power of supportive elites, detaching the authority of traditional leaders from the consent of local clansmen, and by creating hierarchies of control with different salaried ranks of ‘headmen’ and ‘chiefs’ where previously there existed only amorphous and territorially dispersed clan-based loyalties (Newbury, 1988; Mamdani, 1996). It is thus somewhat incorrect to see ‘tribal’ African societies as gradually progressing over time to a modern European ideal, assisted by the disruptive effects of direct colonization, when the bonds of traditional authority so salient in many contemporary African states are themselves the products of institutional legacies of indirect European rule. Indeed, this paper identifies colonially constructed traditional structures – the modal form of indirect rule in sub-Saharan Africa – as a stumbling block to contemporary individual support for democracy. Moreover, although substantial attention has been paid to national political institutions in shaping individual democratic attitudes, there is a relative lack of analytical attention to sub-national colonial legacies in shaping democratic attitudes in the context of governance in sub-Saharan Africa. Traditional leaders1 or ‘tribal chiefs’ were the key administrative stakeholders in indirectly ruled colonies and still today often enjoy unparalleled political, social and economic authority in their localities (Düsing, 2002; De Kadt and Larreguy, 2014; Baldwin, 2014; Acemoglu et al., 2014). Governance in areas under customary law enforced by traditional authorities 1We do not mean to imply an endorsement of claims to traditional notions of legitimacy when using the term traditional leader. Rather, we follow Baldwin (2015) by defining traditional leaders with reference to contemporary customs i.e. as "rulers who have power by virtue of their association with the customary mode of governing a place-based community" (p.21) 4 therefore differs qualitatively from governance by local bureaucratic actors with reference to formal legal or democratic standards because the mantle of traditional leadership is usually in some form hereditary and the interpretation and implementation of customary law is conducted in an informal basis with no reference to external impersonal standards. As highlighted by many African scholars and political leaders (Mboya, 1956; Luthuli, 1962; Ntsebeza, 2005; Meer and Campbell, 2007), the institution of traditional leadership is particularly incongruous with democratic notions of rule of law, the primacy of individual over group rights, and electoral accountability of authority; indeed, Mamdani goes so far as to call traditional leadership a system of ‘decentralized despotism’ (Mamdani, 1996). Is the ongoing political influence of these traditional authorities in the post-colony a significant block to democratic consolidation in sub-Saharan Africa? Such a debate has a rich intellectual history; if we share Diamond (1999)’s influential view that democratic consolidation is the process of achieving "broad and deep legitimation, such that all significant political actors, at both the elite and mass levels, believe that the democratic regime is the most right and appropriate for their society, better than any realistic alternative they can imagine" then individuals in a consolidated democracy must believe that electoral democracy is the ‘only game in town’ and in its its inherent superiority to all other forms of governance (Linz and Stepan, 1996). Yet, because political attitudes are endogenous to exposure to forms of governance (Fuchs-Schündeln and Schündeln, 2015), the ongoing exposure of individuals in rural sub-Saharan Africa to local leadership structures that rely not on electoral bases of legitimacy but on hereditary and customary notions of legitimacy would be expected to lower support for central democratic tenets. Indeed, de Aquino (2015) recently replicated Fuchs-Schündeln and Schündeln (2015)’s findings in the context of sub-Saharan Africa by showing that exposure to national level democratic institutions increases support for democracy in sub-Saharan Africa. Mamdani (1996); Englebert (2000); Ntsebeza (2005); Ribot (2001), among others, have in this vein suggested that African states have been engaged a struggle with local traditional leaders over bases of power and political legitimacy amongst subject populations in the post-colonial context. Yet, the idea that there is a trade-off between traditional forms of governance and modern democratic consolidation has also recently come under sharp criticism. Traditional leaders are often the most widely supported and trusted political actors in surveys in sub-Saharan Africa (Logan, 2008) and thus appear to have an independent source of institutional legitimacy.2 Williams (2004) powerfully argues for approaching the study of traditional leadership from a ‘multiple legitimacies’ framework in which communities need not make an either/or choice between chieftancy and democracy but rather that the two may eventually come to inform each other in a new hybird form of legitimacy. In particular, because local political actors may be kept accountable and good governance achieved through both electoral and non-electoral means (Baldwin, 2015) and good governance is a critical part of the legitimation process (Williams, 2010), there 2Logan (2008) explores a number of reasons for this support including the greater symbolic resonance, responsiveness, proximity to and overall effectiveness of traditional leaders at performing governing functions in their communities compared to elected officials 5 may be no necessary trade-off between support for traditional leadership and elected leadership. Rather, insofar as good governance also requires co-operation between traditional authorities and elected officials, it may be that legitimacy is a a rising tide that lifts all boats (Logan, 2013) and thus simply "‘commonsensical’ that the institution of the chieftaincy and democratic elections can, and should, coexist” (Williams, 2004) in the post-colonial context. Such a theoretical argument has received empirical support from Logan (2008, 2013) who has used cross-national individual survey data to illustrate that greater trust and support for traditional authorities does not negatively correlate with support for core democratic tenets but rather that the legitimacy of traditional leadership and the state positively correlate and may thus reinforce one another. Yet, to the authors’ knowledge, no paper has previously tried to adjudicate between these competing theories by exploiting exogenous variation in the influence of traditional leaders - something that is essential to conduct causal inference given that the institutional influence of traditional leadership across different ethnic groups is far from assigned randomly. This paper instead exploits exogenous variation in indirect colonial rule amongst members of the same ethnic groups to try to disentangle the mechanisms through which indirect colonial rule likely undermines contemporary democratic consolidation and finds support for the hypothesis that greater contact with unelected traditional authorities indeed undermines individual support for central democratic tenets. 6 3 Historical background Namibia, or South-West Africa as it was formerly known, was colonized progressively by Germany over the second half of the nineteenth century in the well-known ‘Scramble for Africa’. Prior to colonization, the dominant ethnic groups in Namibia were Ovambo (Ambo), Herero, Nama (Heikum), Bushmen (Kung) and Damara (Bergdama) (see figure 3 in the Appendix). They had qualitatively similar political structures as measured by traditional form of succession of the local headman (patrilineal heirs) and none of these groups had individual property rights. However the means of subsistence differed. While the Ovambos depended on agricultural farming, Herero and Nama depended on animal husbandry and Bushmen and Damara on gathering and hunting 3 . In the face of stringent financial constraints that discourFigure 1: Map of 1907 aged large-scale military expansion, German colonization initially focused on the less densely populated southern and central regions of Namibia where land could be more easily acquired. German colonial authorities gradually expanded their territorial remit from the coast by progressively playing off warring local factions and remunerating a number of indigenous elites in central Namibia for lost landholdings (German Colonial Office, 1919; Ofcansky, 1981). In 1897, a critical event occurred that was to shape the spatial incidence of direct and indirect rule; a rinderpest epidemic killed 95 percent of the cattle herds in central and southern Namibia. The epidemic particularly devastated cattle-dependent indigenous communities in central and southern Namibia because, unlike crop-dependent communities in fertile northern Namibia, the arid nature of the land prevented agriculture from being used as a feasible food-source substitute (Miescher, 2012; Eckl, 2007). The rinderpest epidemic thereby provided a key opportunity for German colonists to acquire large tracts of land in central and southern Namibia relatively cheaply with lessened collective resistance from weakened indigenous communities. However, the epidemic also presented a dilemma to colonizers - there was little prospect of quickly extending direct German rule to the relatively unaffected northern areas of Namibia, yet continuing to allow free animal movement across South-West Africa would be to potentially expose German herds to future devastating epidemics. Shortly after the epidemic in 1897, therefore, the German colonial government set up a veterinary cordon fence at the boundaries of where at the time its direct control extended in order to protect southern and central Namibian cattle herds from future potentially rinderpestinfected animals from northern Namibia (Miescher, 2012). The area north of this fence was left relatively untouched for the remainder of the German administration as the Germans focused on consolidating political and military control over central and southern Namibia. However, 3 Information on local headmen taken from v72, data on property rights from variables v74 and v75 and information on economic structures from variables v1-v5 in (Murdock, 1967). 7 after the South Africans began to administer South-West Africa after World War 1, the South Africans began to try to establish more regular administrative structures through which to indirectly rule the areas north of the veterinary cordon fence. Yet, the often amorphous and territorially fluid indigenous political structures did not provide the tribal ordering colonial offi- cials had been conditioned to expect, and initial attempts to try and co-opt the paramount chief of areas such as Kaokoland were met with puzzling failure; no clear hierarchical political order could be found (Bennett, 1998). In 1927 the South Africans formally appropriated the power to create and dissolve tribes and set about appointing persons as chief or headman of such tribes. As Friedman (2006) points out, the bases of consequent appointments to traditional leadership in South-West Africa were often contradictory - the government recognized particular persons as traditional leaders ’because they were looked upon as such by the people’, that is, because their authority was derived ’traditionally’. On the other hand, many leaders were often officially warned, for example, that ’unless they carry out instructions issued to them by officials of the Administration and do everything possible to assist these officials in future, the Administration...will be forced to consider whether they should not be deprived of their status’" (Friedman 2006, pp.29-30). Traditional leaders in areas north of the veterinary cordon fence were thus forced to balance the competing prerogatives of the administrative necessities of implementing indirect rule and maintaining an ongoing claim to customary authority. The veterinary cordon fence therefore formed the dividing line between ‘white’ and ‘black’ Namibia – a dividing line between the area directly settled and ruled by German and later South African authorities, and the area indirectly ruled through a system of appointed indigenous elites who had jurisdiction over a number of racial ‘homelands’. The German forces only guaranteed police protection for settlers living within the zone south of the veterinary cordon fence (the ‘Police Zone’). The country was arbitrarily divided into an indirectly and a directly ruled area (see figure 1) and the more prosperous, densely populated northern region had little or no European settlement. Reflecting the historical experience of other colonies, a within-country ‘reversal of fortune’ (Acemoglu et al., 2002) gradually occurred in Namibia whereby different colonial institutions were set up in the relatively densely populated areas of northern Namibia, which are now the poorest in the country (Namibia Statistics Agency, 2011). Tribal leaders in the north were given a great deal of political autonomy, such as the responsibility for administering communal land and settling disputes, and were given no formal political role in the south (Keulder, 2000). This spatial division was later formalized by the South African authorities through the Odendaal Commission of 1964 which created a number of racially demarcated ‘Homelands’ in northern Namibia to be administered by officially recognized ‘tribal chiefs’. While the north was ruled by traditional authorities, the indigenous population in the south was exploited by the German and later South African colonizers through a system of temporary contract labor on white-owned farms and factories (Odendaal, 1964; Moorsom, 1977; Melber, 1996). Under effective apartheid, rule of law and electoral suffrage only extended to the white population and the vast bulk of laborers were returned to their racial ‘homeland’ after one or two years working 8 in the south. After Namibian independence in 1991, formal political structures across the country were homogenized but the informal influence of tribal leaders in the north persists to the present day. These tribal leaders (or traditional authorities) have proven extremely active and successful in mobilizing to protect colonial-era institutional privileges (Düsing, 2002) and so traditional authorities are still highly influential in enforcing customary law. Moreover, individuals in the north are often extremely supportive of their traditional authorities (Keulder, 2000). As the institution of traditional leadership is hereditary, unaccountable and undemocratic, it is reasonable to presume that individuals living in previously indirectly ruled areas have become socialized to accept the legitimacy of non-electoral mechanisms for selecting political leadership. On the other hand, Namibians living in the former Police Zone have only experienced a democratic governance system since independence in 1990 at all levels of government. The Namibia government invested heavily in the northern regions after independence in order to support the convergence of living standards in the two parts of the country (Development Expenditure Report by National Planning Commission Namibia). The Namibian Household Income and Expenditure Survey (NHIES), which was first conducted in 1993, documents this convergence process in its 2010 report by showing that differences in terms of poverty rates between northern and southern regions have declined. 4 Data We use the original map published by the Odendaal Commission in 1964 as digitized by Mendelsohn (2002) to identify regions directly controlled by the colonizers and those that were governed by traditional authorities during colonial times. The political attitude data used in this paper stem from the Afrobarometer survey. Between 1999 and 2008, four survey rounds (1999, 2003, 2005, 2008) were conducted, which covered questions about attitudes towards politics, the economy and civil society. We limit our analysis to the indigenous population in both the formerly directly and indirectly ruled areas and therefore exclude whites from the sample. Afrobarometer uses random sampling methods, which are conducted with probability proportionate to population size (i.e. more densely populated areas have a higher probability of being sampled). Thus, "the sample design is a clustered, stratified, multi-stage, area probability sample.” (Afrobarometer.org). In order to quantify the influence of traditional leaders in their communities we use a Afrobarometer question about the frequency of contact with traditional leaders. The relevant question about "demand for democracy" (Bratton, 2004; de Aquino, 2015), our main oucome variable, asks about support for democracy (see appendix for original questions). As a broader attitudinal outcome, we analyze attitudes towards authority using a question which asks whether authorities should be respected or whether one should be allowed to question them. Finally, to corroborate the importance of differences in attitudes to traditional leadership, we use a question about whether the individuals trust tradi- 9 tional leaders as our third outcome4 . We include as important control variables the individuals’ evaluation of the performance of the local government council, in order to control for differences in individual politicization and local government effectiveness, which Williams (2004) and Logan (2013) see as an important component of democratic and traditional authority legitimacy. We also use measures for education (highest level of education), income (food consumption) and age as controls.5 Finally, we also include geographic controls (carrying capacity and average rainfall) (Mendelsohn, 2002). The geographical location of the surveyed individuals is identified by enumeration area. The Namibian Statistics Agency divided Namibia into 4080 enumeration areas for the 2001 census (see figure 2), each comprises between 80 and 100 households. Therefore, there are more enumeration areas in more densely populated regions. Table 1 indicates that people living in indirectly ruled areas (outside the former Police Zone boundary) do have statistically significant more contact with traditional authorities. Moreover, people in the southern directly ruled part of Namibia tend to have higher support for democracy, trust traditional authorities less and respect authorities less than people living in the northern indirectly ruled areas. To put the results in Table 1 into perspective, Table 9 in the Appendix compares means of the main variables of interests for Namibia with means for 19 other African countries for survey round 4. Contact with traditional rulers is on average even higher in other African countries, which demonstrates the importance of traditional leadership on the continent. Table 1: Descriptive Statistics (1) (2) (3) Direct rule Indirect rule Difference Contact traditional leader 0.23 0.70 -0.47*** [0.60] [0.99] (0.027) Support for democracy 2.47 2.40 0.068*** [0.79] [0.83] (0.025) Trust traditional leaders 1.46 1.87 -0.41*** [0.98] [0.94] (0.036) Respect authority 2.32 2.45 -0.13*** [0.98] [1.01] (0.032) Observations 1,824 3,120 4,944 We then created a 100km buffer zone around the plausibly exogenous boundary between these two zones (see figure 2) and only focus on observations within this buffer to ensure comparability6 . We chose a 100km buffer because individuals living in this zone face a similar geographic, political and cultural environment. We also believe that it is useful to exclude individuals who 4We econometrically treat contact with traditional leaders as the treatment and trust in traditional leaders as an outcome. An individual’s frequency of contact with traditional leaders exogenously varies across the internal colonial border. Trust in turn is an attitude that individuals form based on their experiences and which we show is affected by contact with traditional leaders 5The income and education variables are discrete and to allow for a flexible estimation we include dummies for each income group and education group 6We excluded Etosha National Park from the buffer area. 10 Figure 2: Enumeration areas and buffer live in the desert and those who live close to economic centers (such as as the capital region Windhoek). While the 100km is our preferred buffer size we also included estimations using observations from the entire country and observations from a 50km buffer zone as robustness checks 7 . We only focus on the northern part of the former Police Zone boundary as this part still represented the original boundary drawn by the Germans when the Odendaal Commission of 1964 formalized the border. Other parts of the border were changed over time due to political and economic reasons and is less plausibly exogenous. The number of enumeration areas within the 100km buffer zone is 1247. Out of these 1247 enumeration areas, the Afrobarometer survey covered between 42 and 47 in in each round. This constitutes a random sample of all 7The 100km buffer is our preferred buffer as it constitutes the optimal solution for the trade-off between number of observations and comparability 11 enumeration areas in the buffer zone. There are more enumeration areas in the northern part of the buffer as this part is much more densely populated than the southern part. We observe eight individuals per enumeration area in each survey round. This gives us a maximum number of 1426 observations for the 100km buffer. This number however differs between specifications as not each question is asked in every survey round and we eliminated observations, where the responded answered “don’t know” to the Afrobarometer question. We can thus link information about the colonial ruling style with political attitude data. For detailed summary statistics of the variables of interests see Table 10 in the Appendix. 5 The effect of direct vs indirect colonial rule on political attitudes We identify the effect of indirect colonial rule on democratic attitudes by OLS estimation. The treatment of interest is indirect vs direct colonial rule which is independent of other factors affecting political attitudes for observations close to the colonial border when controlling for ethnicity fixed effects. The northern part of the border between directly and indirectly ruled territories was shaped by the spatial extent of direct German control at the end of the rinderpest epidemic of 1897. The border zone where the progressive extension of direct German rule was frozen in 1897 can thus be considered exogenous to pre-colonial political attitudes. Pre-colonial political structures and attitudes were ethnic-group specific. There is no evidence in the Murdock (1967) data that there were differences across ethnic groups in Namibia. We nevertheless include ethnic fixed effects in all specifications so as only to compare individuals from the same ethnic group and thereby ensure that pre-treatment attitudes did not differ between the direct and indirectly ruled areas. All ethnic groups are represented in both parts of the buffer. Survey round fixed effects are included in order to account for the different timing of the Afrobarometer survey rounds. The border also cuts through seven (out of 14) administrative regions so that we can compare individuals who face the same regional institutions with each other by including region fixed effects. We also control for the performance of local governance councils to ensure that our estimated effects are not driven by differences in institutional quality. Differences in terms of income, education and age between regions very close to the border should be mitigated today. Table 6 and table 7 in the appendix provide evidence that there do not exist significant differences in terms of income and education between the northern and the southern part of the buffer. We nevertheless control for these factors as they are important determinants of political attitudes. We include dummies for education and income and hence only compare individuals with similar income and educational levels. Standard errors are clustered on an enumeration area level. The baseline estimation equation is: Yider = α + βIndirectruled + X 0 iderγ + ηe + µr + ider 12 Y expresses demand for democracy of individual i, living in enumeration area d, belonging to the ethnic group e, being surveyed in round r. Indirectrule is a dummy variable indicating whether the individual lives in an enumeration area which belonged to the indirectly or the directly ruled part of Namibia. X is a set of individual-level control variables, which includes the performance of the local government councils, age and dummies for income and education8 . ηe are ethnicity fixed effects and µr are survey-round fixed effects. In most specifications we also include region fixed effects. OLS estimates are reported in Table 2. The columns present different specifications including different sets of controls. Living in the formerly indirectly ruled part of Namibia decreases the probability that people think that a democratic government is preferred to any other type of government (table 2). The magnitude of the effect is in the range of a fourth of a standard deviation of the dependent variable (i.e. living in the formerly indirectly ruled areas decreases support support for democracy by 0.2 on a scale from 1 to 3). The size of the effect increases to a third of a standard deviation when applying more conservative estimation specifications such as including region fixed effects and individual-level control variables (column 2), adding performance of the government as a control variable (column 3) and adding geographic control variables (column 4). Finally, we also applied an ordered probit model, which confirms the previous findings. The results in Table 11 and Table 12 in the appendix show alternative outcome measures, which indicate that contact with traditional leaders may be a potential channel of causality. People living in the north tend to think that authorities should be respected rather than questioned. Questioning leaders is an important feature of democracy but does not seem to be as prevalent in the political culture of people living in the northern part of the buffer as in the culture of people living in the southern part. This might be connected to the role that traditional leaders play in society. Moreover, people living in formerly indirectly ruled areas tend to trust traditional leaders more than people living in formerly directly ruled areas. This might be due to the historically more important role of traditional leaders in the administration of the community in indirectly ruled areas. This channel of causality will be examined in the following section. All results also hold when not only focusing on observations in the 100km buffer zone but using a sample from the entire country and also when using a 50km buffer zone (see Appendix table 14). As an additional robustness check we clustered the standard errors on a constituency level, which reduces the number of clusters from 165 to 40 (see Appendix table 15). The results still hold. 8Xider = P4 n=0 incomei ider + P8 m=0 educationi ider + ageider 13 Table 2: Effect of indirect rule on support for democracy (1) (2) (3) (4) (5) Ordered VARIABLES OLS OLS OLS OLS probit Indirect colonial rule -0.178** -0.250** -0.311*** -0.331*** -0.465*** (0.0746) (0.101) (0.103) (0.125) (0.158) Performance government -0.00667 -0.00705 -0.00881 (0.0288) (0.0287) (0.0419) Carrying Capacity -0.0241 (0.0362) Av Rainfall -0.00228 (0.0393) Constant 2.493*** 2.808*** 2.892*** 3.074*** (0.0908) (0.343) (0.350) (0.606) Observations 1,347 1,329 1,274 1,274 1,274 R2 0.019 0.043 0.043 0.043 Ethnicity FE yes yes yes yes yes Survey round FE yes yes yes yes yes Region FE no yes yes yes yes Controls no yes yes yes yes # clusters 165 165 165 165 165 Results from OLS regressions including ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 6 Channel of causality: Contact to traditional leaders In order to identify the mechanisms through which indirect colonial rule exerted persistent effects on democratic attitudes we conduct a 2SLS analysis. Specifically, we seek to adjudicate between the competing theoretical perspectives of Mamdani (1996); Englebert (2000) and Williams (2010); Logan (2013) by testing whether contact to traditional authorities is an important mechanism for persistence in the effects of indirect colonial rule in contemporary democratic consolidation in sub-Saharan Africa. The exogenous division of Namibia caused substantially differing degrees of contact with traditional authorities as people in indirectly ruled areas had much closer contact with their tradi- 14 tional authorities during colonial times. Hence, after independence there are more traditional leaders present in formerly indirectly ruled areas than in formerly directly ruled areas, who exercise power. Our first-stage results (table 3) confirm that contact to traditional leaders increases by 0.37 to 0.6 (on a scale of 0-3) if an individual lives in the north rather than in the south. The partial F statistics is larger than 10 for each specification, which indicates that we do not have a weak instrument problem. These results hold even when including a number of controls and fixed effects (columns 2, 3 and 4). Table 3: First-stage results (1) (2) (3) (4) Contact Contact Contact Contact VARIABLES traditional leader traditional leader traditional leader traditional leader Indirect colonial rule 0.555*** 0.391*** 0.367*** 0.393** (0.0700) (0.104) (0.109) (0.173) Performance government 0.104*** 0.101*** (0.0384) (0.0383) Constant 0.240*** 0.887*** 0.126 -0.00770 (0.0533) (0.122) (0.321) (0.474) Observations 1,418 1,418 1,336 1,336 R2 0.045 0.142 0.185 0.190 Ethnicity FE no yes yes yes Round FE no yes yes yes Region FE no no no yes Controls no no yes yes F-Test 62.67 18.36 14.04 12.69 # clusters 165 165 165 165 Results from OLS regressions. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 6.1 2SLS results We apply a two-stage-least-square estimation strategy using the geographical location relative to the settlement boundary as an instrument for contact to traditional leaders. Moreover, we again include ethnicity fixed effects so that we only compare political attitudes of members of the same ethnic group. In addition, survey round fixed effects are included in order to account for 15 the different timing of the Afrobarometer survey rounds and regional fixed effects are included in order to account for political and geographic differences between regions. Performance of the government and individual education, age and food consumption are included as controls. All standard errors are clustered on an enumeration area level. The 2SLS results (table 4) confirm that contact to traditional authorities undermines democratic attitudes. An increase in contact with traditional leaders by one standard deviation (which corresponds to a change of 1 on a scale from 0 to 3) decreases support for democracy by around half a standard deviation (0.4) or even more in the more conservative specifications (columns 2 and 3). The (partial) first-stage F-test is greater than 10 for each specification in the 2SLS estimations (see tables 4 and 13), so indirect colonial rule is considered a strong instrument. Table 4: 2SLS estimates (1) (2) (3) Support Support Support VARIABLES for democracy for democracy for democracy Contact traditional leader -0.429** -0.587** -0.768** (0.216) (0.240) (0.348) Performance government 0.0562 0.0652 (0.0443) (0.0507) Observations 1,347 1,274 1,274 Ethnicity FE yes yes yes Survey round FE yes yes yes Regional FE no no yes Controls no yes yes # clusters 165 165 165 First-stage F statistic 20.90 17.32 10.50 Results from 2SLS regressions including ethnicity and survey round fixed effects using indirect colonial rule as an instrument for contact with traditional leaders. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. First-stage Kleibergen-Paap Wald F statistic reported. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. These results are robust to changes in the clustering method and to using observations for the country as a whole (see Appendix tables 18). Finally, the results in table 13 in the appendix confirm that contact to traditional leaders affect 16 other political attitudes beyond support for democracy. Table 13 shows that people trust traditional leaders more and question authorities less when they have more contact with traditional leaders. Contact with traditional leaders was exogenously determined depending on whether individuals live north or south of the veterinary cordon fence, whereas trust in traditional leaders and more general respect for authority are attitudinal outcome measures endogenous to institutional context. These results corroborate the importance of exposure to a nondemocratic form of governance for the development of political attitudes beyond support for electoral democracy. 6.2 Robustness An important assumption for our 2SLS estimation is that the exclusion restriction holds. That means that living north of the border is uncorrelated with any other determinant of political attitudes. One might however argue that there are also other persistent factors, which differed between the northern and southern part and which also affect political attitudes. We are aware of the fact that, as in all historically oriented work, we cannot completely rule out all of these factors but we are confident that they are either negligible or would bias against us. In the following section we discuss the most important other mechanisms that may determine political attitudes and we come to the conclusion that none of them biases our results in our favor. 6.2.1 No geographical differences In our study we only focus on individuals living within a 100km buffer zone of the boundary to ensure comparability between individuals surveyed on several dimensions (proximity to ports, the capital and the national border, geographic characteristics). We also include regional fixed effects, which does not change our results. 6.2.2 No relevant different political socialization Political socialization over one’s lifetime is of course an important determinant of future political attitudes (Fuchs-Schündeln and Schündeln, 2015; de Aquino, 2015). However, the indigenous population did not experience democracy in either the direct or indirectly ruled areas of Namibia during colonial times. Whereas northern Namibia was ruled by authoritarian traditional authorities and, to a lesser extent by the colonial administration, the indigenous population in southern and central Namibia was exploited by the German and later South African colonizers through a system of contract labor (Moorsom, 1977; Melber, 1996; Odendaal, 1964). The “rule of law” and electoral democracy only applied to the white population. Therefore, the indigenous population had a negative experience with western institutions, European settlers, officials and the apartheid democracy of the south. Since the governance system of the south enabled the exploitation of the local indigenous population, if anything, the bias from different experiences with western institutions during the colonial era should bias against our hypothesis. As the corollary to this, we cannot rule out the theoretical potential that the introduction of democ- 17 racy was seen as a greater ‘liberation’ in the south relative to the north. Thus, we test to see whether the effect of indirect colonial rule differs for individuals who experienced liberation and those who didn’t. Table 17 (in the Appendix) demonstrates that there is no interaction effect between age and living in the formerly indirectly ruled areas. That means that the effect of living in the north on democratic attitudes does not differ between young and old people. If different political socialization or the experience of liberation is an important confounder then the effect of living in the north should be much stronger for older people, who experienced the different political socialization between indirectly and directly ruled areas much longer. These results also hold when using a binary age measure9 (see table 17 in Appendix). 6.2.3 No selective spatial sorting During the German rule, permanent migration between the two parts of the country was prohibited. After taking control of Namibia after 1914, the South Africans established a migrant labour system that brought workers from the north to work in the south in order to satisfy white farmers’ demand for cheap farmhand labor. These laborers were required to return to their racial homeland after a period of 18-24 months and re-apply for the temporary labor scheme, and so there was no permanent sorting. We cannot rule out selective sorting after independence in 1990, however we believe this is unlikely to act as an important confounder. In northern areas of Namibia, land is communally held and ties to one’s family, one’s community and to ancestral land rights are extremely close (Paul, 1933; Eirola, 1992). Moreover, migration statistics from the Namibian Statistics Agency suggest that permanent migration from the north, where it has occurred, has been economic in nature and primarily inter-regional as poor laborers move to the larger cities of the south such as Windhoek or Walvis Bay far south of our study area. For this reason, we control for age and education in our specifications - neither of which changes the results. Hence, though it cannot be completely ruled out, it is unlikely that selective sorting explains our results (Moorsom, 1977; Melber, 1996). 6.2.4 Same contemporary institutions We show that, other than the greater importance of traditional leaders in northern Namibia, contemporary institutions do not differ between the northern and southern areas in our sample. In order to ensure that our effects are not different by differing performance of local government officials as theorized by Williams (2010) and Logan (2013), we have previously included controls for the individuals’ evaluation of the performance of local government councils which do not actually appear to have a significant effect on democratic attitudes. Moreover, Namibia is extremely centralized politically because, after independence, the Namibian government made a great effort to homogenize governance between the two parts of the country and improve institutional infrastructure and efficiency in the previously neglected north 9Dividing the sample into those younger than the median age (31 years) and those older than median age. Those younger than median age experienced a large part of their political socialization after 1990. 18 where state capacity was previously low (Werner and Odendaal, 2010; Melber, 2015; Düsing, 2002; Keulder, 2000). The north and in particular the former Ovamboland is now the bedrock of electoral support for the governing party, which regularly obtains over 90% of votes in this region and is thus extremely attuned to the needs and wishes of its supporters in this region (Keulder, 2000), so if anything more contemporary differences in governance after 1990 should bias against our hypothesis. Finally, we can use Afrobarometer results to show that people living north and south of the border do not systematically evaluate the effectiveness of government institutions differently (see Table 5). Table 5: Balancing Table (1) (2) (3) Direct rule Indirect rule Difference Government officials listen 1.22 1.26 -0.048 [1.06] [1.08] (0.11) Trust in police 1.78 1.91 -0.13* [0.85] [0.88] (0.070) Trust in courts 1.83 1.91 -0.085 [0.92] [0.95] (0.067) Fear of unjust arrest 3.93 3.83 0.097 [0.73] [0.93] (0.091) Observations 253 1,163 1,416 Individuals on both sides of the border think that governmental officials listen sometimes to what the people say. The coefficient on fear of unjust arrest, which is an indicator for despotism of officials, does also not differ significantly between formerly directly ruled and indirectly ruled areas of Namibia. As further measures of the reliability of contemporary institutions we use trust in courts and police. Trust in courts does not differ between the two parts. Trust in police is even significantly higher in the north, which would work against us and does therefore not bias our results in our favor. Moreover, we also include fixed effects for the seven regions that the settlement boundary cuts through. This ensures that we only compare individuals living close to each other on the same part of the boundary, who are governed by the same national and regional institutions nowadays. Including these fixed effects does not greatly change our results. 6.2.5 Potential mechanism: income Income differed substantially between areas within and outside the Police Zone during colonial times. After independence however the government introduced policies to reduce the large income disparities between the north and the south. The effect of indirect rule on income should therefore not be highly persistent. We compare only people living close to each other, so that 19 potential income gaps should have closed after independence. Table 6 demonstrates that indirect colonial rule does not have a statistically significant impact on income and thus suggests that the effect of indirect rule on income is not persistent in the buffer zone. Moreover, dummies for different income groups ensure that we only compare individuals with similar income. Table 6: Indirect colonial rule and income (1) (2) (3) (4) VARIABLES OLS OLS OLS Ordered Probit Indirect colonial rule -0.204 0.0398 -0.0247 0.0191 (0.134) (0.163) (0.146) (0.148) Constant 1.307*** 1.216*** 0.739*** (0.126) (0.184) (0.227) Observations 1,417 1,417 1,400 1,400 R2 0.004 0.055 0.123 Ethnicity FE no yes yes yes Survey round FE no yes yes yes Controls no no yes yes # clusters 165 165 165 165 Results from OLS regressions. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 6.2.6 Potential mechanism: education Education does not differ between the northern and southern part of the buffer among the indigenous Namibian population. This is because missionaries founded schools long before the first colonizers reached Namibia. Even during colonial times, missionaries were as active at providing education for indigenous Namibians in the south as in the north and the Namibian government after 1990 has not favored the north or south disproportionately in terms of education. Table 7 shows statistically that areas formerly under indirect rule do not have significantly lower levels of education. Hence, as education does not differ between the directly and indirectly ruled areas of Namibia, it can be ruled out as a likely channel though which indirect colonial rule affects political attitudes. 20 Table 7: Indirect colonial rule and contemporary education levels (1) (2) (3) (4) VARIABLES OLS OLS OLS Ordered Probit Indirect colonial rule 0.0147 -0.274 -0.140 -0.0783 (0.189) (0.193) (0.149) (0.0976) Constant 3.802*** 4.038*** 5.087*** (0.176) (0.225) (0.357) Observations 1,406 1,406 1,400 1,400 R2 0.000 0.025 0.239 Ethnicity FE no yes yes yes Survey round FE no yes yes yes Controls no no yes yes # clusters 165 165 165 165 Results from OLS regressions. The sample consists of observations from the 100km buffer zone. Control variables are age, education dummies and income dummies. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. We are therefore confident that we have identified contact to traditional authorities as the key institutional factor that differs between the formerly directly and indirectly ruled areas of Namibia and which acts as a cause of substantial variance in contemporary political culture. 7 Effects on voter turnout Why does it matter to study determinants of political attitudes? It matters because it has important implications for civic culture and the viability of democratic institutions. Given that voting is the essential participatory exercise in a democracy, one important behavioral outcome that we focus on is voter turnout. We coded voting as a binary variable, indicating whether an individual reported to have voted in the previous national elections or not. The data stem from Afrobarometer survey rounds three and four. 21 Table 8: Effect of indirect colonial rule on voting (1) (2) (3) (4) VARIABLES OLS OLS Probit 2SLS Indirect colonial rule -0.219*** -0.219*** -0.667*** (0.0469) (0.0476) (0.188) Performance government 0.0137 0.0753 0.256* (0.0222) (0.0846) (0.140) Contact traditional leader -1.220** (0.608) Constant 0.624*** 0.579*** 0.125 (0.181) (0.190) (0.501) Observations 723 687 685 687 R2 0.287 0.285 Ethnicity FE yes yes yes yes Survey round FE yes yes yes yes Region FE yes yes yes yes Controls yes yes yes yes # clusters 91 91 91 91 Results from OLS, Probit and 2SLS regressions including ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. The results in table 8 (columns 1 - 3) show that people in the northern part of the buffer vote significantly less than people living in the southern part. Column 4 suggests that this effect, again, works through contact with traditional authorities. This indicates that weaker democratic attitudes are associated with less reported voting - the essential political act in a democracy - and thus that indirect colonial rule indeed presents a block to democratic consolidation both in an attitudinal and behavioural sense. 8 Conclusion The results presented in this paper show that indirect colonial rule has persistent effects on contemporary political attitudes. We have argued that the key mechanism underlying this wellestablished relationship is the ongoing influence of traditional authorities in formerly indirectly 22 ruled areas of Namibia, which acts as a parallel undemocratic governance structure and undermines individual support for democracy. This paper thereby contributes to a long-running debate in comparative politics (Mamdani, 1996; Englebert, 2000; Williams, 2010; Logan, 2013; Baldwin, 2015) - it does appear that the hereditary and unelected system of traditional leadership institutionalized by indirect colonial rule presents a stumbling block to contemporary democratic political consolidation in sub-Saharan Africa. We identify the effect of indirect rule through OLS estimation by exploiting a unique setting in Namibia. Namibia was exogenously divided in 1897 into a southern region directly settled and ruled by colonial authorities and a northern region that was indirectly ruled through a system of appointed indigenous tribal elites, thereby providing plausibly exogenous variation in the form of colonial rule and the influence of traditional leaders amongst members of the same ethnic group. We then use this division of the country as an IV for contemporary contact with traditional leaders and thereby test their impact on democratic attitudes. Whilst we are confident that the institution of traditional leadership plays an important role in shaping individual attitudes towards democracy, we do not wish to imply a mono-causal explanation for variance in contemporary political culture in sub-Saharan Africa. Rather, we wish to highlight the fact that the ongoing parallel existence of undemocratic local governance structures can undermine support for democracy even in the context of a functional, largely successful national democratic polity. This has broad implications for democratization processes in other sub-Saharan African countries, where systems of traditional leadership still play an important role in local governance and national democracy is not as consolidated as in Namibia. However, the fact that contact to traditional leaders may weaken support for core democratic tenets in sub-Saharan Africa does not invalidate the extremely important and valuable governing roles that traditional authorities currently play in their communities. Indeed, it is likely in part because non-electoral mechanisms such as strong social ties have proven so effective in keeping traditional leaders accountable and responsive to the needs of their communities and thus more effective than elected officials (Baldwin, 2015) that support for electoral democracy as a system of government is weakened in areas with influential traditional leaders. Despite the presence of a trade-off between influential local traditional institutions and democratic consolidation, therefore, the policy mechanisms for improving overall quality of governance in sub-Saharan Africa in the future remain more unclear and is a currently fruitful area of research.10 Ultimately, we hope that our findings documented in this paper encourage further research about the potential trade-off between the legitimacy of different institutional configurations and the historical legacies that continue to shape political culture in both sub-Saharan Africa and the wider world. 10As Baldwin and Mvukiyehe (2015) show, introducing elections for traditional authorities may actually have counter-productive effects on community collective action 23 9 Appendix 9.1 Ethnic groups prior to colonization Figure 3: Ethnic groups prior to colonization (Murdock, 1967) 9.2 Comparison to other African countries Table 9 shows Afrobarometer survey results from 2008 (survey round 4) for 19 other African countries 11 in comparison to the Namibia results. Contact to traditional leaders is lower in Namibia than in other African countries. This shows that traditional leadership is an important institutions in many African countries and that it is important to study its implications for the viability of democratic systems. There is no clear difference in support for democracy between Namibians and other sub-Saharan Africans in the sample. 11Benin, Botswana, Burkina Faso, Cape Verde, Ghana, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mozambique, Nigeria, Senegal, South Africa, Tanzania, Uganda, Zambia, Zimbabwe 24 Table 9: Summary statistics of variables of interest for 20 African countries, covered in Afrobarometer survey round 4 (1) (2) (3) 19 African countries Namibia Difference Contact traditional ruler 0.55 0.38 0.17*** [0.99] [0.78] (0.029) Trust traditional leaders 4.37 4.16 0.21*** [1.44] [1.17] (0.042) Support for democracy 2.86 2.86 0.0062 [0.34] [0.35] (0.012) Respect authority 2.22 2.52 -0.30*** [1.13] [1.02] (0.033) Observations 26,513 1,200 27,713 9.3 Summary statistics Table 10 summarizes the main variables of interests for the buffer zone. The number of observations differs as some variables are not available in all four survey rounds. Table 10: Summary statistics for buffer zone Mean SD Min Max Obs Contact traditional leader 0.7 1.0 0 3 1426 Trust traditional leaders 1.9 0.9 0 3 1029 Support for democracy 2.4 0.8 1 3 1352 Respect authority 2.4 1.0 1 4 1373 Performance government 2.9 0.8 1 4 1360 Age 35.8 14.8 18 92 1421 Education 3.8 1.8 0 8 1414 Without food 1.1 1.2 0 4 1425 25 9.4 More outcomes Table 11: OLS estimation: Effect of indirect rule on respect for authorities (1) (2) (3) Ordered VARIABLES OLS OLS Probit Indirect colonial rule 0.221* 0.242** 0.277** (0.114) (0.119) (0.139) Performance government 0.00543 0.00480 (0.0330) (0.0383) Constant 1.619*** 1.827*** (0.286) (0.362) Observations 1,365 1,290 1,290 R2 0.127 0.150 Ethnicity FE yes yes yes Survey round FE yes yes yes Region FE yes yes yes Controls no yes yes # clusters 165 165 165 Results from OLS regressions including ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 26 Table 12: OLS estimation: Effect of indirect rule on trust in traditional leaders (1) (2) (3) Ordered VARIABLES OLS OLS Probit Indirect colonial rule 0.247* 0.266* 0.333* (0.136) (0.146) (0.174) Performance government 0.167*** 0.210*** (0.0398) (0.0500) Constant 1.348*** 0.650* (0.255) (0.388) Observations 1,029 979 979 R2 0.100 0.113 Ethnicity FE yes yes yes Survey round FE yes yes yes Region FE yes yes yes Controls no yes yes # clusters 123 123 123 Results from OLS regressions including ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 27 Table 13: 2SLS estimation: Effect of contact with traditional leaders on respect for authorities and trust in traditional leaders (1) (2) (3) (4) Respect Respect Trust Trust VARIABLES authorities authorities traditional leaders traditional leaders Contact traditional leader 0.636* 0.612** 0.459* 0.412* (0.379) (0.308) (0.240) (0.248) Performance government -0.0554 0.139*** (0.0535) (0.0426) Constant 1.700*** 1.868*** 1.195*** 0.876*** (0.559) (0.521) (0.322) (0.307) Observations 1,348 1,290 1,022 979 R2 -0.240 -0.207 0.058 0.080 Ethnicity FE yes yes yes yes Survey round FE yes yes yes yes Region FE yes yes yes yes Controls yes yes yes yes # clusters 165 165 123 123 First-stage F statistic 8.562 10.22 14.98 14.68 Results from 2SLS regressions using indirect colonial rule as an instrument for contact with traditional leaders including ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 9.5 Further Robustness Checks 9.5.1 OLS OLS results are robust to using observations for the entire country and for a 50km buffer rather than only focusing on the buffer zone (see table 14) and to clustering the standard errors on the constituency level (table 15). 28 Table 14: OLS robustness to buffer (1) (2) (3) (4) VARIABLES Entire country Entire country 50km Buffer 50km Buffer Indirect colonial rule -0.102** -0.135** -0.161* -0.249** (0.0507) (0.0550) (0.0881) (0.103) Performance government -0.00967 0.00926 (0.0176) (0.0442) Constant 2.657*** 2.523*** 2.189*** 1.889*** (0.133) (0.202) (0.151) (0.451) Observations 4,656 3,734 620 511 R2 0.017 0.042 0.044 0.098 Ethnicity FE yes yes yes yes Survey round FE yes yes yes yes Region FE yes yes yes yes Controls no yes no yes # clusters 571 509 77 71 Results from OLS regressions of support for democracy on indirect colonial rule including ethnicity, region and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists observations for the entire country and for the 50km buffer zone respectively. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 29 Table 15: Robustness check OLS: clustering SE on a constituency level (1) (2) (3) Support Support Support VARIABLES for democracy for democracy for democracy Indirect colonial rule -0.178** -0.250*** -0.311*** (0.0768) (0.0841) (0.0948) Performance government -0.00667 (0.0267) Constant 2.493*** 2.808*** 2.892*** (0.0910) (0.348) (0.349) Observations 1,347 1,329 1,274 R2 0.019 0.043 0.043 Ethnicity FE yes yes yes Survey round FE yes yes yes Region FE no yes yes Controls no yes yes # clusters 44 44 44 Results from OLS regressions including ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists of observations from the 100km buffer zone. Standard errors (clustered by Constituency) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. In addition, we tested whether there is an interaction effect between indirect colonial rule and survey round fixed effects. The results in table 16 indicate that there is no such interaction effect. Hence, the effect of indirect colonial rule on political attitudes does not decrease (or increase) over time, which suggests that political attitudes are indeed persistent. 30 Table 16: Interaction between indirect colonial rule and survey round fixed effects (1) Support VARIABLES for democracy Indirect colonial rule -0.369** (0.165) Indirect rule x round 2 0.193 (0.197) Indirect rule x round 3 0.145 (0.197) Indirect rule x round 4 0.315 (0.209) Round = 2 -0.0900 (0.181) Round = 3 0.0289 (0.180) Round = 4 -0.181 (0.193) Constant 2.817*** (0.269) Observations 1,329 R2 0.038 Ethnicity FE yes Survey round FE yes Controls yes # clusters 165 Results from OLS regressions including interaction terms between colonial rule and survey round fixed effects as well asethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists observations for the buffer zone only. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 31 Finally, we analyzed interactions between indirect rule and age using both the reported age and a binary age measure (table 17). The binary measure divides the sample in people older and younger than 31, which is the median age in the sample. Neither of the estimations yields statistically significant effects of the interaction. This demonstrates that the effect of indirect colonial rule on political attitudes does not depend on age. The effect is thus not stronger for older people who experienced colonial rule longer than younger people. Table 17: No interaction effect between indirect colonial rule and age (1) (2) Support Support VARIABLES for democracy for democracy Indirect colonial rule -0.0750 -0.182** (0.159) (0.0828) Indirect rule x age -0.00347 (0.00450) Age 0.00288 (0.00431) Indirect rule x Old dummy -0.000124 (0.107) Old dummy -0.00334 (0.0957) Constant 2.557*** 2.628*** (0.254) (0.225) Observations 1,329 1,334 R2 0.036 0.039 Ethnicity FE yes yes Survey round FE yes yes Controls yes yes # clusters 165 165 Results from OLS regressions including interaction terms between colonial rule and age as well as ethnicity and survey round fixed effects. Control variables are age, education dummies and income dummies. The sample consists observations for the buffer zone only. Standard errors (clustered by Enumeration Area) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 32 9.5.2 2SLS The 2SLS results are also robust to using observations for the entire country and to clustering the standard errors on the constituency level (table 18). We did not include robustness to the 50km buffer because the sample size would be too small for consistent 2SLS estimation of our fixed effect model. Table 18: 2SLS robustness to buffer and clustering method for the entire country (1) (2) Sample: Clusters: VARIABLES entire country constituency level Contact traditional leader -0.296*** -0.587*** (0.114) (0.227) Performance government 0.0170 0.0562 (0.0214) (0.0458) Constant 2.372*** 2.761*** (0.151) (0.356) Observations 3,720 1,274 Ethnicity FE yes yes Survey round FE yes yes Controls yes yes # clusters 509 44 Results from 2SLS regressions of Support for Democracy on Contact with Traditional Leaders using Indirect Rule as an instrument. Control variables are age, education dummies and income dummies. The sample consists of observations for the entire country (column 1) and for the buffer zone (column 2). Standard errors (clustered by Enumeration Area (column 1) and by Constituency (column 2) in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 33 9.6 Afrobarometer survey questions The Afrobarometer survey questions uesed in this paper read: • Support for democracy: Which of these three statements is closest to your own opinion? Statement 1: Democracy is preferable to any other kind of government. Statement 2: In some circumstances, a non-democratic government can be preferable. Statement 3: For someone like me, it does not matter what kind of government we have. 1= Statement 2: Non- democratic preferable, 2=Statement 3: For someone like me, it does not matter what kind of government we have, 3=Statement 1: Democracy preferable • Respect for authority: Let’s talk for a moment about the kind of society we would like to have in this country. Which of the following statements is closest to your view? Choose Statement 1 or Statement 2. Statement 1: Citizens should be more active in questioning the actions of leaders. Statement 2: In our country, citizens should show more respect for authority. 1=Agree very strongly with Statement 1, 2=Agree with Statement 1, 3=Agree with Statement 2, 4=Agree very strongly with Statement 2 • Trust traditional leader: How much do you trust each of the following, or haven’t you heard enough about them to say: Traditional leaders 0=Not at all, 1=Just a little, 2=Somewhat, 3=A lot • Performance of local government councilor: Do you approve or disapprove of the way the following people have performed their jobs over the past twelve months, or haven’t you heard enough about them to say: Your Elected Local Government Councillor? 1=Strongly Disapprove, 2=Disapprove, 3=Approve, 4=Strongly Approve • Contact traditional leader: During the past year, how often have you contacted any of the following persons about some important problem or to give them your views: A traditional ruler? 0=Never, 1=Only once, 2=A few times, 3=Often • Voting: With regard to the most recent national elections, which statement is true for you? 0= You decided not to vote, You could not find the polling station, You were prevented from voting, You did not have time to vote, Did not vote for some other reason 1= You voted in the elections 34 References Acemoglu, D., Johnson, S., and Robinson, J. A. (2001). The Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review, 91(5):1369–1401. Acemoglu, D., Johnson, S., and Robinson, J. A. (2002). Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution. The Quarterly Journal of Economics, 117(4):1231–1294. Acemoglu, D., Reed, T., and Robinson, J. A. (2014). Chiefs: Economic development and elite control of civil society in Sierra Leone. Journal of Political Economy, 122(2):319–368. Almond, G. A. and Verba, S. (1963). The Civic Culture: Political Attitudes and Democracy in Five Nations. Princeton University Press, Princeton. Baldwin, K. (2014). When politicians cede control of resources: Land, chiefs, and coalitionbuilding in Africa. Comparative Politics, 46(3):253–271. Baldwin, K. (2015). The Paradox of Traditional Chiefs in Democratic Africa. Cambridge University Press. Baldwin, K. and Mvukiyehe, E. (2015). Elections and collective action: Evidence from changes in traditional institutions in liberia. World Politics, 67(04):690–725. Bennett, T. (1998). The constitutional base of traditional rulers in south africa. In d’Engelbronner Kolff, F., Hinz, M., and Sindano, J., editors, Proceedings from the workshop: Traditional Authorities in the Nineties - Democratic Aspects of Traditional Government in Southern Africa. Centre for Applied Social Sciences, University of Namibia. Boone, C. (2014). Property and Political Order in Africa: Land Rights and the Structure of Politics. Cambridge University Press. Bratton, M. (2004). The" Alternation Effect" in Africa. Journal of Democracy, 15(4):147–158. de Aquino, J. A. (2015). The effect of exposure to political institutions and economic events on demand for democracy in africa. Afrobarometer Working Paper No. 160. De Kadt, D. and Larreguy, H. (2014). Agents of the Regime: Traditional Leaders and Electoral Behavior in South Africa. Paper presented at APSA meeting 2014. Diamond, L. (1999). Developing Democracy: Toward Consolidation. Baltimore, MD: Johns Hopkins University Press. Düsing, S. (2002). Traditional Leadership and Democratisation in Southern Africa: A Comparative Study of Botswana, Namibia, and Southern Africa, volume 6. LIT Verlag Münster. 35 Eckl, A. (2007). Reports from ’beyond the line’: The accumulation of knowledge of Kavango and its peoples by the German colonial administration 1891-1911. Journal of Namibian Studies: History Politics Culture, 1:7–37. Eirola, M. (1992). The Ovambogefahr: The Ovamboland Reservation in the Making: Political Responses of the Kingdom of Ondonga to the German Colonial Power, 1884-1910, volume 22. Ponjois-Suomen Historiallinen Yhdistys. Engerman, S. L. and Sokoloff, K. L. (2002). Factor Endowments, Inequality, and Paths of Development among New World Economies. Economia Journal of the Latin American and Caribean Economic Association. Englebert, P. (2000). State Legitimacy and Development in Africa. Boulder, CO: Lynne Rienner Publishers. Friedman, J. T. (2006). Making politics, making history: Chiefship and the post-apartheid state in namibia. Journal of Southern African Studies. Fuchs-Schündeln, N. and Schündeln, M. (2015). On the endogeneity of political preferences: Evidence from individual experience with democracy. Science, 347(6226):1145–1148. German Colonial Office, B. (1919). The Treatment of Native and Other Populations in the Colonial Possessions of Germany and England: An Answer to the English Blue Book of August 1918: ’Report on the Natives of South-West Africa and Their Treatment by Germany’. Engelmann. Hariri, J. G. (2012). The Autocratic Legacy of Early Statehood. American Political Science Review, pages 471–494. Inglehart, R. (1990). Culture shift in advanced industrial society. Princeton University Press. Inglehart, R. and Welzel, C. (2005). Modernization, cultural change, and democracy: The human development sequence. Cambridge University Press. Keulder, C. (2000). Traditional leaders. State, Society and Democracy. A Reader in Namibian Politics, pages 150–170. Linz, J. and Stepan, A. (1996). Problems of Democratic Transition and Consolidation: Southern Europe, South America, and Post-Communist Europe. Baltimore, MD: Johns Hopkins University Press. Logan, C. (2008). Traditional leaders in modern africa: Can democracy and the chief co-exist? Michigan State University, Afrobarometer Working Paper No. 93. Logan, C. (2009). Selected chiefs, elected councilors, and hybrid democrats: Popular perspectives on the co-existence of democracy and traditional authority. Journal of Modern African Studies, 47(1):101–128. 36 Logan, C. (2013). The roots of resilience: Exploring popular support for african traditional authorities. African Affairs, 112(448):353–37. Luthuli, A. J. (1962). Africa and Freedom. Africa Today, 9(1):7–9. Mamdani, M. (1996). Citizen and Subject: Contemporary Africa and the Legacy of Late Colonialism. Princeton University Press. Mboya, T. (1956). The Kenya Question: an African Answer. Number Nr. 302 in Fabian tract. Fabian Colonial Bureau. Meer, T. and Campbell, C. (2007). Traditional leadership in democratic South Africa. Durban and Cape Town: Democracy Development Program. Melber, H. (1996). Urbanisation and internal migration: Regional dimensions in post-colonial {Namibia}. Namibian Economic Policy Research Unit. Melber, H. (2015). Understanding Namibia: The Trials of Independence. Oxford University Press. Mendelsohn, J. (2002). Atlas of Namibia: A Portrait of the Land and its People. New Africa Books Pty Ltd. Miescher, G. (2012). Namibia’s Red Line: The History of a Veterinary and Settlement Border. Palgrave Macmillan. Moorsom, R. (1977). Underdevelopment, contract labour and worker consciousness in Namibia, 1915–72. Journal of Southern African Studies, 4(1):52–87. Murdock, G. P. (1967). Ethnographic atlas. Ethnology, pages 109–236. Namibia Statistics Agency (2011). Namibia 2011: Population and Housing Census Main Report. Namibia Statistics Agency, Windhoek, Namibia. Newbury, C. (1988). The Cohesion of Oppression: clientship and ethnicity in Rwanda, 1860- 1960. Columbia University Press. Ntsebeza, L. (2005). Democracy Compromised: Chiefs and the Politics of the Land in South Africa. Brill, Leiden. Odendaal, C. (1964). Report of the Commission of Enquiry into South West Africa Affairs, 1962-1963. Pretoria, Republic of South Africa. Ofcansky, T. P. (1981). The 1889-1897 Rinderpest Epidemic and the Rise of British and German Colonialism in Eastern and Southern Africa. Journal of African Studies, 8(1):31–38. 37 Paul, J. (1933). Wirtschaft und Besiedlung im südlichen Amboland von Johannes Paul. Veröffentlichungen, Wissenschaftliche, des Museums für Völkerkunde zu Leipzig. Vertriebsstelle: Ferdinand Hirt & Sohn, Leipzig. Persson, T. and Tabellini, G. (2009). Democratic Capital: The Nexus of Political and Economic Change. American Economic Journal: Macroeconomics, 1(2):88–126. Putnam, R. D., Leonardi, R., and Nanetti, R. Y. (1994). Making democracy work: Civic traditions in modern Italy. Princeton University Press. Ribot, J. (2001). Local actors, powers and accountability in african decentralizations: A review of issues. United Nations Research Institute for Social Development (UNRISD), October. Werner, W. and Odendaal, W. (2010). Livelihoods after land reform: Namibia Country Report. Williams, J. M. (2004). Leading from behind: Democratic consolidation and the chieftaincy in south africa. Journal of Modern African Studies, 42(1):113–136. Williams, J. M. (2010). Chieftaincy, the State and Democracy: Political Legitimacy in PostApartheid South Africa. Bloomington: Indiana University Press. 38