Types of Communalizing Colonial Policy and Ethnic Violence

While communal recognition can be a goal in itself, colonial officials commonly used CCPs as a means to other ends. Colonial powers sometimes recognized communities in order to empower them. In addition, CCPs commonly recognized communal difference in order to favor some communities over others in a discriminatory fashion. Different combinations of recognition, empowerment, and discrimination, in turn, likely affected inter-communal relations in contrasting ways, suggesting that CCPs might have had diverse effects on ethnic violence.

Within the social scientific literature, there are competing views on the impact of communal recognition on ethnic violence. According to Tajfel [Reference Tajfel1974], the simple act of group differentiation promotes discrimination, prejudice, and competition. He therefore suggests that recognizing difference promotes inter-communal violence. Several scholars of inclusive nationalism take an opposing view that focuses on institutional processes instead of individual interactions. They claim that the recognition and institutionalization of communal difference contains inter-communal competition and antagonisms by preventing one-sided assimilationist policies and symbolically recognizing all communities as members of the nation [Kymlicka Reference Kymlicka1995; McEvoy and O’Leary Reference McEvoy and O’Leary2013; Stepan, Linz, and Yadav Reference Stepan, Linz and Yadav2011]. Both sides of this debate make convincing arguments that seem possible under different circumstances, and we believe that CCPs that simply recognize communal difference have the potential to either promote or deter violence.

In contrast, we hypothesize that highly discriminatory CCPs have strong and positive effects on ethnic violence. Communities that suffer from discrimination resent those who benefit from it. This situation is usually compounded in a colonial setting, as those who face discrimination view the beneficiaries as colonial stooges who betray the true nation. At the same time, communities benefiting from discriminatory CCPs expect to retain their privileged status and mobilize to protect their interests after independence. And the co-presence of one community that is angry and resentful over colonial favoritism and another that is scared and mobilizes to maintain the advantages it gained during colonialism creates a high risk of violence.

Similar to policies that simply recognize difference, we expect that CCPs that empower communities have little or no general effect on ethnic violence. Some claim that colonial empowerment increases inter-communal competition by heightening communal divisions and providing communities with the mobilizational resources [Blanton, Mason and Athow Reference Blanton, Mason and Athow2001], but others recognize that empowering colonial policies can deter ethnic violence by promoting inclusive politics [Wucherpfennig, Hunziker and Cederman Reference Wucherpfennig, Hunziker and Cederman2016]. We believe that empowering CCPs can promote either outcome, and suggest that the direction of their effects depends largely on whether CCPs empower in a discriminatory fashion. CCPs that empower communities to different extents create very powerful grievances, as communal power and autonomy have great practical and symbolic value and provide resources that can mobilize movements to protect the interests of communities and their leaders. Yet when all communities are empowered to similar extents, empowerment is less likely to create grievances, and broad-based communal empowerment can promote inclusive power-sharing arrangements.

Our claims about discriminatory empowerment have implications on other discriminatory CCPs. All types of discrimination can provide strong motives for ethnic violence, but some types motivate violence more than others. We expect that the impact depends on the extent to which discrimination affects the relative power and status of communities. Discriminatory empowerment, for example, commonly promotes conflict over the control of powerful positions, institutions, and the state. Even more, it contributes to contention over the form of the postcolonial nation-state, as communities that were empowered by colonial policies demand a nation-state model that gives them special power and autonomy whereas other communities resent these demands and pursue nation-state models that either ignore or exclude communities previously empowered by colonialism.

Overall, we propose that CCPs were most likely to shape postcolonial patterns of ethnic violence when they were highly discriminatory and empowered communities to different extents. To test these claims, we analyze four CCPs that were commonly used in the British Empire: employing colonial censuses to document communal difference, granting special legislative representation to communities, stacking the colonial police with certain communities, and using indirect rule to provide communities with autonomy. Because these CCPs generally discriminated to different extents and affected relative communal power in different ways, we hypothesize that each had contrasting effects on ethnic violence.

Recognition in a census can empower communities in different ways, and the failure of censuses to recognize communities can be discriminatory. Relatively speaking, however, the use of communal census categories is rarely explicitly discriminatory and therefore has limited direct effects on communal empowerment and discrimination. As a result, we expect that the use of communal census categories has little or no general effects on ethnic violence.

Unlike colonial censuses, communal legislative representation is a highly discriminatory CCP that gives select communities special political representation. And communal legislative representation is an important form of empowerment, as legislative representation affects a community’s ability to pursue communal interests and is a symbol of special communal status. This form of discrimination therefore has the potential to contribute to powerful grievances and intense political competition over the form of the nation-state, and we hypothesize that it increases the risk of postcolonial ethnic violence.

Stacking the colonial police with certain communities is also a highly discriminatory CCP affecting communal empowerment, and we therefore expect that discriminatory police recruitment increases the risk of postcolonial ethnic warfare. Indeed, this policy allows some communities to have greater control over the state’s security forces, which is a source of communal power and prestige.

Finally, indirect rule is a CCP that empowers communities by providing them with some degree of self-rule. Commonly, colonial powers employ indirect rule in a non-discriminatory fashion, thereby providing all communities with power and authority. When this occurs, we expect that postcolonial states must share power with all communities, and indirect rule therefore deters ethnic violence. Yet indirect rule can be used in a discriminatory fashion that promotes stark inequalities in terms of power, status, and self-rule, and we expect that this increases the risk of ethnic violence. Overall, the effect of indirect rule should therefore be mixed.

Multimethod Research Design

Past analyses of colonialism and ethnic violence employ three general research designs. Several assume that the British employed CCPs more than the French and use statistics to explore whether postcolonial ethnic violence varied by empire [Blanton, Mason and Athow Reference Blanton, Mason and Athow2001; Lange and Dawson Reference Lange and Dawson2009; Wucherpfennig, Hunziker and Cederman Reference Wucherpfennig, Hunziker and Cederman2016]. Others provide historical case studies linking CCPs to ethnic violence [de Silva Reference de Silva1986; Idris Reference Idris2005; Mamdani Reference Mamdani2001, Reference Mamdani2009; Pollis Reference Pollis1973]. And still others use statistics to analyze the relationship between CCPs and ethnic violence within the British Empire [Lange and Balian Reference Lange and Balian2008; Ray Reference Ray2018; Verghese Reference Verghese2016]. In this article, we combine the second and third strategies to test our hypotheses, comparing former British colonies to highlight general relationships and using a case study to explore processes and mechanisms.

Our multimethod research design differs from past analyses of colonialism and ethnic violence, which generally use either qualitative or quantitative methods. In addition, our within-case analysis differs from previous case studies because it focuses on the effects of several particular CCPs instead of analyzing colonial “divide and rule” more generally. Finally, our intra-imperial statistical analysis of the British Empire differs from previous analyses in two important ways. First, we analyze four different types of CCPs, whereas previous intra-imperial analyses only considered one. Due to this difference, we are able to explore whether different types of CCPs have different effects. Second, all previous intra-imperial analyses use ethnic violence between civilians as the dependent variable, but we focus on ethnic civil warfare between a state and an organized ethnic opposition. This difference is potentially important: we propose that discriminatory CCPs affect conflict over the postcolonial state and nation-state and that non-discriminatory empowerment affects communal power-sharing, and the literature suggests that all of these shape the risk of ethnic civil warfare in influential ways [Wimmer, Cederman and Min Reference Wimmer, Cederman and Min2009; Wimmer Reference Wimmer2013].

Whereas some claim that multimethod analyses should start by exploring general statistical relationships [Lieberman Reference Lieberman2005], we completed the quantitative and qualitative sections simultaneously. The main reason for this is practical: a simultaneous analysis permits the quantitative and qualitative analyses to inform one another in a back-and-forth manner, thereby increasing the chances of methodological complementarity and decreasing the chances that the analyses produce incompatible results. Our decision to present the quantitative analysis first is therefore based merely on stylistic considerations.

Statistical Analysis of Communalizing Colonial Policies in the British Empire

Our statistical analysis explores whether the use of CCPs is related to the odds of postcolonial ethnic civil war onset among the set of 34 former British colonies that had more than 1 million inhabitants in 2010 and that did not merge with other territories with larger populations at independence.Footnote ¹ Using cross-national panel data (country-year data), we conduct a recurrent-event discrete-time survival analysis. The event of interest is ethnic civil war onset, and time is modeled as the number of years since independence. As such, each country-year is treated as a Bernoulli trial in which an ethnic civil war may or may not occur, and is regressed on different linear combinations of predictor variables through a logit link function. As observations within each country are dependent, we used adjusted standard errors computed with the Eicker-Huber-White sandwich estimator. All statistical analyses were conducted in R3.6.2 using the glm command.

Our dependent variable is a dichotomous measure of whether an ethnic civil war occurred in a country during a given year, and we measure it for each year between colonial independence and 2010. We calculate this variable using the Ethnic Armed Conflict (EAC) dataset, which notes the years and locations of all ethnic-based conflicts resulting in civil wars between 1946 and 2010 [Wimmer, Cederman and Min Reference Wimmer, Cederman and Min2009]. Because our data on ethnic civil warfare begin in 1946 and because some British colonies—Australia, Canada, Egypt, Iraq, New Zealand, South Africa, and the United States—received their independence prior to that date, some cases do not have data for all post-independence years. These missing data potentially bias the statistical findings. In addition, the seven former British colonies that received early independence were unique in other ways that might affect the long-term risk of ethnic warfare independent of CCPs. Recognizing these potential problems, we run the statistical analyses using two different sets: the entire set of former British colonies and a more limited set that excludes countries that gained their independence prior to 1946.

To test our hypotheses, we operationalize four different CCPs: indirect rule, the extent to which colonial powers used communal categories in colonial censuses, communal representation in the colonial police, and communal legislative representation. Our measure of indirect rule is drawn from Lange [Reference Lange2009] and operationalizes the extent of indirect rule by calculating the proportion of total colonial court cases heard in customary tribunals. The variable is available for 31 of 34 British colonies, with scores missing for the Middle Eastern Mandate Territories. For ease of interpretation, we standardize the results so that the regression coefficients of indirect rule represent the effect of a one-standard deviation increase in the level of indirect rule.

Our second focal independent variable measures the use of colonial censuses to document communal difference, a CCP that was relatively non-empowering and non-discriminatory. We use data from Lieberman and Singh [Reference Lieberman and Singh2017] to measure whether colonial censuses gathered information on language, caste, race, religion, and tribe and create a variable that counts the highest number of types of communal categories ever collected on a single colonial census. The scores range from 0—no communal categories collected on colonial censuses—to 5—all types of communal categories collected on a colonial census. Because Jordan and the United States never had a colonial census, no communal categories were ever collected on a colonial census, and we score the variable as zero for both cases. The variable is available for all countries in our set.

Our third focal independent variable is dichotomous and measures a highly discriminatory CCP: communal representation in colonial legislatures. For this variable, we gathered information from primary and secondary sources on whether colonial legislatures at the colony-level either reserved seats or had separate electoral colleges or assemblies for particular indigenous communities at any time during the colonial period. Because of our focus on indigenous communal divisions, we score cases as 0 if legislatures only privilege non-indigenous peoples, such as the reservation of seats for white settlers. This variable is available for all cases in our set.

Our final focal independent variable measures communal police representation, which is also a highly discriminatory CCP. For the variable, we transform data from Ray [Reference Ray2012] to create a continuous measure of the degree to which a country’s largest indigenous community was underrepresented in the colonial police force, with 0 signifying no underrepresentation and 1 representing the total exclusion of the largest ethnic community from the police force. Ray provides data on communal police representation for 24 of the 34 British colonies in our set, and we gathered data for seven additional cases but were unable to find data for Jamaica, Lesotho, and Swaziland. Similar to indirect rule, we standardized this variable for ease of interpretation.

Using these focal independent variables, we employ two standard control variables and two different sets of additional controls. The standard controls are particular to this study and measure two aspects of a country’s history of colonialism. First, we control for the time since colonial independence in all models. Following Beck, Katz, and Tucker [Reference Beck, Katz and Tucker1998], we apply a natural cubic spline to years since independence. Our second standard control measures the non-indigenous African and European populations as a percentage of total population during the late colonial period, and we employ the natural log of the score. We include this variable because these non-indigenous populations are linked to settlement and plantations, shaped social relations in a variety of ways, and might have affected the long-term risk of ethnic warfare independent of CCPs.

Our first set of additional controls includes four variables that are commonly included in past statistical analyses of ethnic warfare: ethnic fractionalization [Fearon Reference Fearon2003], the percentage of the population excluded from formal politics because of their ethnicity [Wimmer, Cederman and Min Reference Wimmer, Cederman and Min2009], the natural log of per capita GDP [Feenstra, Inklaar and Timmer Reference Feenstra Robert, Inklaar and Timmer2015], and the natural log of total population [Ibid.]. We refer to these as the “common” controls. We refer to our second set of additional controls as “historical” controls, as it includes three factors—latitude, mountainous terrain, and precolonial statehood—that preceded colonialism. These precolonial conditions might have shaped both the use of CCPs and postcolonial ethnic warfare, thereby confounding any relationship between colonial policies and postcolonial ethnic warfare. As noted by Alesina et al. [Reference Alesina, Devleeschauwer, Easterly, Kurlat and Wacziarg2003], a country’s latitude measures ecological conditions that shaped the extent of communal diversity, and latitude therefore provides a proxy for historical ethnic fractionalization. Among our set, for example, the correlation between latitude and contemporary ethnic fractionalization is -0.48. Precolonial ethnic diversity, in turn, might have affected both the use of CCPs and the long-term risk of ethnic civil warfare. The extent of a country’s territory that is mountainous shapes the ability of communities to fight civil wars against the state [Fearon and Laitin Reference Fearon and Laitin2003]. At the same time, this environmental condition might have promoted anticolonial resistance, and Ray [Reference Ray2012] finds that such resistance promoted certain CCPs. Finally, precolonial statehood might confound our statistical findings, as the presence of precolonial states is also related to the use of particular CCPs [Ray Reference Ray2012]; Paine [Reference Paine2019], in turn, argues that precolonial states have long-term effects on ethnic warfare by promoting conflict between communities with and without precolonial states. To measure precolonial statehood, we use data from Borcan, Olsson, and Putterman [Reference Borcan, Olsson and Putterman2018] on the extent to which there was a long-standing and autonomous precolonial state that controlled the same territory as a contemporary state. The data range from 0 to 50 and are measured for 50-year periods between 1 and 1950 AD, and we average the scores for all periods between 1001 and 1700 AD.

Table 1 presents the results of our models using discriminatory police recruitment as the focal independent variable. In this table and in all subsequent tables, we include three different sets of control variables and two different sets of cases, for a total of six models. Our first three models use the entire set of former British colonies, with the first model only controlling for the standard controls, the second model adding the common control variables, and the third model replacing the common controls with the historical controls. The final three models replicate the first three models but use the more limited set of former British colonies.

Table 1 Logit analysis of police underrepresentation and ethnic civil war onset

Confidence intervals in parentheses and based on cluster robust standard errors.

Cubic splines are included in all models, but their odds ratios and the constants are excluded from the table.

Sig: * p < 0.1; ^** p < 0.05; ^*** p < 0.01

The results of Table 1 strongly and consistently support our hypothesis that highly discriminatory CCPs promoted postcolonial ethnic civil war onset, as the odds ratios of police underrepresentation are larger than one and significant in all models. The results estimate that an increase in police underrepresentation by one standard deviation increases the odds of ethnic civil war onset by between 1.5 and 2.3. Put differently, the odds of ethnic civil war onset are estimated at three to nine times greater in places with the total exclusion of the largest community from the colonial police relative to places that have no underrepresentation. The set of cases has little effect on the odds ratios of police underrepresentation, but the set of control variables affects the magnitude of the odds ratios: the odds ratios are largest when including the historical controls and smallest when including the common controls.

The relationships of the control variables in Table 1 are consistent with the findings of past analyses. We find that the political exclusion of communities, total population, precolonial statehood, and mountainous terrain are all associated with greater odds of ethnic civil war onset. In all subsequent tables, the odds ratios of the control variables remain substantively identical to those in Table 1.

The models in Table 2 replicate those in Table 1 but substitute communal legislative representation for police underrepresentation. In contrast to Table 1, however, we exclude the control variable measuring non-indigenous population because of multicollinearity.Footnote ² Supporting our hypotheses, the odds ratios of communal legislative representation are very large and significant in all models and estimate that the odds of ethnic civil war onset are between five and sixteen-times greater when a former British colony has a history of communal legislative representation. This suggests that communal legislative representation is an extremely influential cause of ethnic warfare.

Table 2 Logit analysis of communal legislative representation and ethnic civil war onset

Confidence intervals in parentheses and based on cluster robust standard errors.

Cubic splines are included in all models, but their odds ratios and the constants are excluded from the table.

Sig: * p < 0.1; ^** p < 0.05; ^*** p < 0.01

Tables 3 and 4 employ two CCPs as the focal independent variables that were not highly discriminatory or that varied considerably to the extent that they were discriminatory. In Table 3, colonial census categories are significantly related to ethnic civil war onset when only controlling for the standard controls. When adding additional controls, the odds ratios decrease considerably and usually lose significance. However, they remain above one and are moderately sized in some models, suggesting that the use of colonial census categories might increase the odds of ethnic civil war onset.

Table 3 Logit analysis of census category and ethnic civil war onset

Confidence intervals in parentheses and based on cluster robust standard errors.

Cubic splines are included in all models, but their odds ratios and the constants are excluded from the table.

Sig: * p < 0.1; ^** p < 0.05; ^*** p < 0.01

Table 4 Logit analysis of extent of indirect rule and ethnic civil war onset

Confidence intervals in parentheses and based on cluster robust standard errors.

Cubic splines are included in all models, but their odds ratios and the constants are excluded from the table.

Sig: * p < 0.1; ^** p < 0.05; ^*** p < 0.01

The findings in Table 4 focus on indirect rule, and are mixed and sensitive to model specifications. When only including the standard controls, the odds ratios of the extent of indirect rule are near one and insignificant. With the addition of the common control variables, the odds ratios are well below one and significant. Yet when using the historical control variables, the odds ratios of indirect rule are above one and significant. It is therefore very unclear whether the extent of indirect rule has any general effect on ethnic warfare.

Finally, Table 5 includes all four focal independent variables in the same models. The models include the same controls as in previous tables, although we exclude total population in Models 2 and 5 because of multicollinearity.Footnote ³ In all models, legislative representation has very large and significant odds ratios, and the odds ratios of police underrepresentation are also significant and relatively large. The odds ratios of census categories, on the other hand, are now near one and insignificant in four models but significantly below one in two, thereby estimating that colonial census categories either have no effect or negative effects when controlling for other CCPs. Similarly, the odds ratios of indirect rule are now consistently below one and are significant in two models, thereby offering evidence that higher levels of indirect rule either deter postcolonial ethnic civil warfare or have no general effect when controlling for other CCPs. This finding offers some support to claim that indirect rule can limit postcolonial ethnic warfare by promoting communally inclusive power-sharing but suggests that the presence of other discriminatory CCPs counteracts these suppressive effects.

Table 5 Logit analysis of all divisive colonial policies and ethnic civil war onset

Confidence intervals in parentheses and based on cluster robust standard errors.

Cubic splines are included in all models, but their odds ratios and the constants are excluded from the table.

Sig: * p < 0.1; ^** p < 0.05; ^*** p < 0.01

For all models in all tables, we checked for influential cases by dropping cases with studentized residuals greater than 3. Dropping influential observations either did not affect the odds ratios of the focal independent variables or transformed them in ways that supported our hypotheses.

All in all, the statistical findings strongly support our hypotheses that the effects of CCPs depend on the extent to which such policies empower discriminatorily. We find that communal legislative and police representation—which are both highly discriminatory policies affecting communal power and status—are strongly and consistently related to the odds of postcolonial ethnic civil war onset. Alternatively, colonial census categories and indirect rule, both of which were only occasionally employed in a discriminatory fashion, have mixed results and are sometimes related to lower odds of warfare.

Case Study of Myanmar

While supporting our hypotheses, the statistical findings are hardly conclusive given a variety of problems that commonly affect quantitative analyses [Babones Reference Babones2014; Schrank Reference Schrank2013]. This is especially the case for analyses like our own that have a small set of cases and that use variables to operationalize complex concepts. In addition, the statistical analysis does not explore actual processes linking colonial policies to postcolonial outcomes, so it is uncertain whether the relationships are causal or spurious. In this section, we provide a case study of Myanmar for a separate qualitative check of our hypotheses, using the case to explore processes and mechanisms that potentially link CCPs to postcolonial ethnic warfare. Notably, Myanmar experienced many different CCPs and several ethnic civil wars, making Myanmar an extreme positive case. And Goertz [Reference Goertz2016] recommends that multimethod analyses consider extreme positive cases because they are ideal for highlighting mechanisms. Yet, as an extreme case, one must recognize that the effects of CCPs are likely unusually strong in Myanmar.

Colonialism and Ethnic Warfare in Myanmar

Known as Burma until 1989, Myanmar has a long history of statehood extending back nearly one thousand years before British colonial rule. Ethnic Bamar kings ruled these states and sponsored Buddhism as the official religion, and the Bamar language and culture were dominant. At the advent of British colonialism, Bamars made up approximately two-thirds of the population. The remainder of the population consisted of small communities who usually inhabited more peripheral regions of the kingdom, paid tribute to the Bamar king while retaining considerable autonomy, spoke different languages, and practiced a variety of religions.

After conquering Myanmar, the British formally incorporated the region into colonial India, although they ruled it through a relatively autonomous regional administration and made it a separate colony in 1937. Throughout most of the colonial period, many Bamars organized strong resistance to British rule, thereby prompting the British to establish a very mixed and communalized form of rule that attempted to strengthen supporters and weaken opponents. In the former core region, British officials quickly destroyed the monarchy and established a system of direct rule that relied heavily on the military to contain anticolonialism [Callahan Reference Callahan2003]. This military, in turn, was stacked with Indians, Karens, and other “loyal” minorities who were trusted to help combat Bamar resistance [Callahan Reference Callahan2003]. In the peripheral regions of Myanmar, British rule took a very different form. Most inhabitants were minorities and—despite initial resistance—were much less threatening than Bamars and ultimately proved more accepting of British control. The British ruled these regions indirectly and had limited influence on local politics, thereby making possible self-rule for compliant minorities. The British therefore used indirect rule in a discriminatory fashion, giving self-rule to many minorities but not to the Bamars majority.

Shortly after independence in 1948, several ethnic civil wars erupted between the Bamar-led state and different minority communities. The first conflict, which began in 1948 and continues to this day, is a civil war between Karen nationalists and the state. Karen leaders demanded greater political autonomy and organized in defense of Karens, many of whom were suffering violence at the hands of Bamars. Yet the government opposed Karen demands and made little effort to stop ethnic violence against Karens, resulting in escalating conflict leading to civil war. In the indirectly ruled regions, other minorities also fought secessionist wars in an attempt to retain and strengthen communal self-rule. As described below, different CCPs contributed to these wars in a variety of ways.

Communal Legislative Representation and Ethnic Warfare in Myanmar

Similar to several other British colonies, the Burmese Legislative Council reserved seats for select minority communities. Colonial officials justified the implementation of this policy as a necessary step for minorities to participate in democratic governance but also recognized that the policy would have the undesirable effect of strengthening communal divisions in ways that hindered nation-building [Burma Reforms Committee 1921b: 11-12]. In this section, we provide evidence of another related drawback: communal legislative representation promoted the Karen civil war by communalizing politics in ways that fostered antagonisms and increased conflict over the form of the postcolonial nation-state.

Representing 10% of the population, Karens were the largest minority community and the only indigenous community to receive special representation in the colonial legislature. Communal representation of Karens began informally with the appointment of one representative in 1916. It was then formalized and expanded to include five seats in 1923 and twelve seats in 1935. Besides Karens, three non-indigenous communities—Indians, Eurasians, and Europeans—also receiving reserved seats. Notably, minorities from Myanmar’s indirectly ruled regions were ineligible for reserved seats because the legislature only represented people living in the directly ruled regions.

Even before its formalization, communal legislative representation caused conflict between Bamars and Karens. During public meetings in 1921 that considered whether a formal system of communal representation should be created, Karen leaders supported reserving seats, but Bamar nationalists decried this policy as a divisive ploy to weaken the anticolonial movement [Burma Reforms Committee 1921a: 21-24]. In the heated debate that followed, Bamars organized a public protest against the policy and physically prevented some Karens from testifying at the meetings [Keenan Reference Keenan2017].

Once formalized, communal legislative representation became an even more divisive issue by communalizing politics. The policy strengthened communal boundaries by recognizing Karens as a distinct national community and empowering them to look out for the well-being of their community in the Legislative Council. A review of the minutes of the Legislative Council proceedings highlights this communalization of politics by showing that Karen representatives refrained from general policy discussions and only spoke up to pursue Karen interests. Karen representatives, for example, inquired about the representation of Karens in the Military Police [Burma Legislative Council 1932: 583], the Judicial Services [Burma Legislative Council 1936: 146, 262], the Legislative Council [Burma Reforms Committee 1921a: 22, 78], the Imports Advisory Committee [Ibid. 1946: 277], and education and teacher training [Ibid. 1938: 3]. They also demanded greater funding of Karen schools [Ibid. 1938: 5-7], the use of Karen languages in schools and the government [Ibid. 1929: 197], and the suppression of crimes against Karens [Ibid. 1929]. The political demands of Karen representatives, in turn, pushed Bamar politicians to express concern about the privileged position of Karens and to try to protect Bamar interests [Ibid. 1928: 95; 1929: 43, 60, 413; 1930: 130, 368; 1932: 298; 1935: 169; 1936: 269-270].

The communalization of the Legislative Council, in turn, greatly intensified inter-communal antagonisms. In addition to viewing minorities with reserved legislative seats as illegitimate competitors who were only concerned about their own communities, Bamars resented them for opposing Bamar interests. In a debate over the separation of Myanmar from India, for example, a Bamar politician made an open threat to the Indian representatives who opposed Bamar demands for separation by asking, “If we separate [from India] can we drive out the Indians?” [Burma Legislative Council 1933b: 298]. More generally, minority representatives consistently voted in league with colonial officials to oppose Bamar nationalism and anticolonialism, and this voting behavior strengthened popular Bamar anger by solidifying stereotypes of minorities as colonial stooges [Cady Reference Cady John1958: 256-257; Maung Reference Maung1990: 8]. For example, many Bamars were furious when a Karen representative argued in a debate that Myanmar was not prepared for independence and required many more years of British tutelage [Smith Reference Smith Martin1991: 50-51].

In parallel to the Bamars’ anger and resentment, Karens feared what they saw as growing Bamar chauvinism. Karen representatives therefore assertively demanded not only the maintenance of Karen privileges but the expansion of the Karen community’s power and autonomy. Most importantly, Karen politicians began lobbying for the creation of a Karen district that would protect them from Bamar nationalist policies and make possible Karen self-rule. This demand, in turn, only intensified Bamar anger and resentment, especially since Bamars formed the majority of the population in the region that the Karens designated as their homeland.

As independence approached, the non-indigenous minorities with special representation saw the writing on the wall and simply disengaged from Myanmar politics, with the members of these communities usually returning to either the United Kingdom, India, or Bangladesh. Both Karens and Bamars, however, dug in, and the communal and conflictual character of the legislature intensified. Bamar representatives demanded independence, Bamar control of all major institutions, and the founding of a Bamar-based nation-state. Karen politicians, on the other hand, insisted on the maintenance of British rule, the creation of a pluralist nation-state that recognized and protected minorities, the founding of a sub-regional state for the Karens, and a variety of regulations that would reserve spots for Karens in the administration, military, education, and elsewhere. In the end, the British imposed a compromise that proved untenable, and intense communal competition soon transformed into ethnic civil warfare between Karen nationalists and the Bamar state. In this way, communal legislative representation was a very influential determinant of the war, although it was not the only CCP that contributed to it. Another was discriminatory recruitment in the colonial security forces.

Communalized Colonial Security Forces and Ethnic Warfare in Myanmar

Although our statistical findings offer evidence that communal police representation contributes to ethnic warfare, the case of Myanmar does not provide strong within-case evidence that the relationship was causal. That being said, communal police representation is only one component of the colonial security forces, and Myanmar highlights how discrimination in the military had strong effects on ethnic warfare. The case therefore offers evidence that the statistical findings highlight a broader relationship between ethnic warfare and unequal communal representation in the colonial security forces, not simply in the colonial police force.

Bamars were under-represented in the colonial police force by 70%, and several minority communities were over-represented [Callahan Reference Callahan2003; Ray Reference Ray2012]. This situation undoubtedly promoted communal comparisons and affected inter-communal antagonism to some extent, both of which likely influenced the processes leading to ethnic warfare between Bamars and minorities. Yet we failed to find any strong evidence that unequal police recruitment promoted inter-communal conflict and mobilization. Yes, Bamars resented it, and minorities sought to protect their privileged status; but Bamar resentment and minority assertiveness simply intensified preexisting competition and antipathy, and police recruitment was never the focus of inter-communal contention. Our qualitative and quantitative findings are therefore at odds, and this incongruence can be interpreted in two principal ways—either Myanmar is unique, or some aspect of our multimethod analysis is flawed. We are unable to explore other cases to test Myanmar’s uniqueness due to space constraints. Alternatively, there is reason to believe that the statistical relationship between police representation and ethnic civil war onset is at least partially spurious: Ray [Reference Ray2016: 569] finds that communal police and military representation are highly correlated among a sample of British colonies (r=0.84) and the case of Myanmar shows how the communalization of the military had powerful effects on ethnic warfare.

Because the Bamar-led precolonial state strongly resisted British colonialism and because Bamars continued to lead violent anticolonial movements after colonial conquest, the British stacked the colonial military with non-Bamars. British officials judged Chins, Kachins, and Karens as loyal and martial, and these communities made up 81% of the colonial military in 1931 despite making up only 15% of the total population [Selth Reference Selth1986: 489; Callahan Reference Callahan2003: 35]. Alternatively, the British did not trust Bamars and excluded them from the military throughout most of the colonial period. Evidence suggests that this discriminatory military recruitment contributed to ethnic warfare in three influential ways.

First, inequalities in the military promoted strong grievances and competition over its control, and this strengthened the supporters of a highly chauvinistic Bamar nationalism. Relative to the police, the colonial military was more powerful and prestigious, and this made inequalities in military recruitment more glaring. Military representation was also a more contentious issue because Bamars were excluded from the military for most of the colonial period but were merely underrepresented in the police force. Motivated by these grievances, many Bamars organized Bamar militias, and the militias were a major force behind the rise of a more assertive, exclusive, and combative Bamar nationalism [Callahan Reference Callahan2003].

A second way in which military representation affected the Karen civil war was by pitting Bamars and Karens against one another in open conflict, thus promoting intense inter-communal antagonisms. The British used minority military units to combat Bamar-led anticolonial movements, with Karens assisting the British militarily during the wars of colonial conquest, helping to contain the Saya San Rebellion of 1932, and aiding the British fight against Bamars and their Japanese supporters during the Second World War. The resulting violence enflamed relations between Bamars and Karens, and military conflict commonly led to deadly attacks on civilians from rival communities. Notably, tension over minority military support during the wars of colonial conquest had weakened by the turn of the 20th century, but communal legislative representation rekindled these antagonisms [Lewis Reference Lewis James1924]. Alternatively, minority military assistance against Bamar anticolonial movements in the 1930s and 1940s occurred after the colonial legislature had institutionalized inter-communal contention, and this assistance further intensified conflictual relations. The conflict during the Second World War was particularly polarizing: Karens and other minority communities supported the British whereas members of the Bamar militias and other Bamar nationalists initially supported the Japanese, thereby pitting Bamars and minority communities against one another in bloody combat. This military conflict, in turn, transformed into deadly ethnic violence, as Bamar soldiers enacted collective punishment on minority villages.

The third and most direct way through which the overrepresentation of Karens in the colonial military contributed to ethnic warfare was by providing Karens with valuable resources needed to wage and maintain a war against the postcolonial state. During the Second World War, the British gave many Karens weapons to help fight Japanese and Bamar forces, and most Karens retained their arms after the war. Moreover, Karens were over-represented at all levels of the colonial military, and many deserted to form the Karen national military after the government proved unable and unwilling to curb violence against Karens. Weapons, soldiers, and skilled military leaders, in turn, allowed Karen nationalists—who were only a fraction of the total Karen population—to mount and sustain a bloody civil war.

Indirect Rule and Postcolonial Ethnic Warfare

British rule in Myanmar was primarily direct: only 16% of the population was ruled indirectly [Lange Reference Lange2009: 48]. As noted previously, this limited use of indirect rule was discriminatory. We therefore use the case to test our claim that the discriminatory use of indirect rule promotes ethnic civil warfare and find that it did so by promoting antipathy and conflict over the communal character of the nation-state.

Indirect rule was a vital factor shaping the processes leading to ethnic civil warfare between indirectly ruled minorities and the postcolonial state. During the independence transition, the Shan and other indirectly ruled minorities lobbied the colonial government to maintain their autonomy, something Bamar politicians and their nationalist supporters strongly opposed. Prior to independence, an agreement was reached allowing these communities to retain their communal autonomy and to preserve the right to secede from Myanmar at a later date. Yet Bamar nationalists assassinated the Bamar leader—Aung San—who supported and agreed to implement the policy, and Bamar politicians reneged on the agreement after independence, eventually seeking the removal of all local autonomy and implementing assimilationist policies that sought to Burmanize minorities [Brown Reference Brown1996; Geertz Reference Geertz and Clifford1963; Silverstein Reference Silverstein1980]. These state policies sparked minority resistance, demands for independence, and multiple civil wars, showing how indirect rule can promote ethnic warfare when it pits directly ruled communities who control the postcolonial state and resent minority self-rule against indirectly ruled communities who want to maintain their cultural and political autonomy.

In combination with this discriminatory form of indirect rule, a particular form of communal legislative representation also contributed to these wars. As noted previously, indirectly ruled minorities were allowed to rule themselves, but they were excluded from participating in the Burmese Legislative Assembly. Both indirect rule and their exclusion from the legislature, in turn, separated minorities from other communities and made them, for all extents and purposes, autonomous political communities that were not considered—either by themselves or by others—as part of the larger national political community. As a consequence, the indirectly ruled minorities were very concerned about maintaining self-rule and were extremely angry and resentful over Bamar efforts to remove their political and cultural autonomy.

Although Karens were ruled directly, indirect rule also influenced the Karen civil war. With the growing communalization of politics in the 1920s, Karen leaders envied the autonomy of indirectly ruled minorities and lobbied the colonial government for Karen self-rule. One influential determinant of the Karen civil war was the combination of constant and escalating Karen demands for self-rule and growing ethnic-nationalism among Bamars. In this way, indirect rule contributed to the civil war by providing Karens with an example that they assertively sought to replicate, and their privileged positions in the legislature and military made these demands all the more galling to Bamar nationalists.

These findings have implications for our statistical findings: ethnic warfare might be especially likely in countries that experienced a mix of direct and indirect rule, and a non-linear relationship might therefore explain our inconsistent statistical findings on the extent of indirect rule. To test this possibility, we create three dummy variables measuring different levels of direct and indirect rule. One variable measures those colonies that were completely subject to direct rule (extent of indirect rule=0). The second measures those British colonies that experienced mixed rule (extent of indirect rule is between 0.01 and 0.5). Finally, the third measures those British colonies that experienced relatively high levels of indirect rule (extent of indirect rule is between 0.51 and 1.0). Using the more limited set of former British colonies, we rerun models from Tables 4 and 5 while replacing the extent of indirect rule with these variables, with mixed rule being the reference category.Footnote ⁴ As presented in Table 6, the odds ratios of direct rule and high levels of indirect rule are well below one, at least one of the variables is significant in all models, and both variables are significant in two models. The results therefore suggest that direct rule and high levels of indirect rule lower the odds of ethnic civil war onset relative to mixed direct and indirect rule. Paralleling our previous findings on indirect rule, these patterns are strongest when controlling for communal legislative and police representation. The results therefore suggest that British colonies with mixed forms of rule were at greatest risk of ethnic warfare and—in combination with the Myanmar case study—support our hypothesis that a discriminatory use of indirect rule promotes ethnic warfare.

Table 6 Logit Analysis of Mixed Rule and Ethnic Civil War Onset

Confidence intervals in parentheses and based on cluster robust standard errors.

Cubic splines are included in all models, but their odds ratios and the constants are excluded from the table.

Sig: * p < 0.1; ^** p < 0.05; ^*** p < 0.01

Colonial Censuses and Ethnic Warfare in Myanmar

In Myanmar, colonial censuses collected information on a great variety of types of community, including ethnicity, race, religion, tribe, and caste. Communal categories, in turn, commonly became a bone of contention and appear to have contributed to ethnic warfare to some degree. Yet the Myanmar case provides evidence that their effects depended on pre-existing competition and antipathy, something that highly discriminatory CCPs promoted, and the case therefore offers a more fine-grained explanation of our mixed quantitative findings on census categories. Conflict over Karen and Muslim communal categories highlights this complex relationship.

In the late 1920s, Karen and Bamar politicians debated who should be categorized as Karen in the colonial census [Indian Statutory Committee 1929: 37-38]. The boundaries of this communal category were quite blurry, as Karens speak multiple languages, practice diverse religions, have different traditional styles of dress, and vary considerably in their extent of Burmanization. The censuses, in turn, counted the number of Karens in Myanmar, and Bamar politicians recognized that Karens were overrepresented in the legislature relative to their census numbers despite the fact that many non-Karens were “miscategorized” as Karen in the census.

While showing that censuses increased inter-communal contention by facilitating and concretizing inter-communal comparisons, it is clear that communal census categories did not spark communal competition but—instead—became contentious because of preexisting competition and antipathy. Indeed, it is telling that Bamar politicians raised census enumeration as part of their argument against reserving legislative seats for Karens. That being said, censuses provided fuel for this conflict by facilitating communal comparisons and being a source of power and disagreement in a communalized political environment.

Debates about the categorization of Muslims in colonial censuses also occurred at this time. Although this debate was also greatly influenced by other CCPs that ratcheted up inter-communal competition and antipathy, the conflict over the Muslim census category had greater independent effects on ethnic violence because the Muslim category affected citizenship. Prior to British rule, many Muslims had lived in the Arakan region of Myanmar for centuries. After the onset of colonialism, a large number of Indians migrated to Arakan and other regions of Myanmar, including Muslims. Despite their historical presence in Arakan, colonial censuses categorized all Muslims as Indians, and the Indian category was defined as non-indigenous. Many Muslims living in Myanmar disliked being categorized as foreigners and lumped together with Hindus, especially given that Indian immigrants were a target of Bamar aggression. Thus, at the same time that Muslim leaders met with and lobbied the colonial government for Muslim communal legislative representation, they also demanded a new census category—Burman Muslim—that would recognize their indigeneity [Upper Burma Muslims Reference Upper Burma1929]. Bamar leaders learned of these demands and successfully drove the government to reject them [Burma Legislative Council 1933a: 29].

During the Second World War and after independence, most Indians left Myanmar to avoid violence and Bamar rule, although many Muslims in the Arakan region—who are known today as Rohingyas—stayed. Similar to Karens, Rohingyas fought with the British against Bamar nationalists during the Second World War, and this caused severe ethnic violence between Rohingyas and Bamar nationalists during the war. After the war ended, Rohingya leaders lacked confidence in Bamar politicians and asked for the transfer of Muslim regions of Arakan to colonial India, but the British denied this demand. Because Rohingyas were categorized as non-indigenous and because Bamar nationalists viewed Rohingyas as disloyal and dangerous, postcolonial governments refused to revise the colonial census category and grant them citizenship. At one point, the government offered to recognize Rohingyas as non-indigenous citizens, but the Rohingya leaders refused: many had ancestors who lived in Arakan before it was conquered by Bamars, and Rohingyas feared they would face formal discrimination if categorized as non-indigenous. As a result of this refusal, the government denied Rohingyas citizenship and declared them personae non gratae, eventually resulting in the expulsion of these “foreigners” from Myanmar soil [Ibrahim Reference Ibrahim2016].

Thus, colonial censuses categorized Rohingyas as non-indigenous, and conflict over this issue contributed to discrimination, a Rohingya nationalist movement, and extreme violence. Yet the example also highlights how the impact of colonial census categories depended on other CCPs. This example parallels the Karen case in showing that conflict over census categories occurred in an environment with high levels of communal political competition, with people interested in census categories as a means of getting their own communal legislative representation. In addition, the perception of Rohingyas as dangerous and disloyal underlies the unwillingness of the Myanmar government to grant Rohingyas citizenship, and discriminatory colonial military recruitment and violence between Rohingyas and Bamar nationalists during the Second World War are at the heart of these perceptions.

Counterfactual Analysis: The Karen Civil War Without Colonialism?

As noted previously, one potential problem with our statistical analysis is endogeneity: precolonial characteristics potentially shaped both CCPs and postcolonial patterns of ethnic violence, and precolonial factors might therefore underlie a spurious relationship between CCPs and ethnic violence. Our statistical findings do not support this claim, as communal legislative representation, police representation, and mixed indirect rule are all related to postcolonial ethnic civil war onset when controlling for key precolonial conditions. Moreover, the previous sections provide qualitative evidence against claims of endogeneity by highlighting mechanisms through which CCPs affected postcolonial ethnic warfare in Myanmar. While insightful, this evidence is not conclusive, and we therefore complete an additional test of the precolonial hypothesis. For this, we recognize that counterfactuals are an important way of checking rival hypotheses in case studiesFootnote ⁵ and complete a counterfactual analysis of the Karen civil war that considers whether the violence would have occurred without a history of British rule.

The Karen civil war is a classic example of a conflict over the form of the postcolonial nation-state, with Karen nationalists demanding greater political and cultural autonomy and Bamar nationalists pursuing a centralized state controlled by Bamars in the name of Bamars. If precolonial conditions promoted the Karen civil war, one would expect that some combination of assertive Karen nationalism, Bamar resentment and fear of Karens, and inter-communal antipathy and competition would have occurred without the British colonial interlude. Available evidence does not support this view. Prior to British rule, Karens held a marginalized position within the Bamar-dominated socio-political system. While their position could have motivated collective mobilization to counter Bamar dominance, there is no evidence of this ever occurring prior to British rule, largely because Karens were not in a position to counter or threaten Bamar dominance in any way. The main reason for this is that Karens were not organized in a coherent political community prior to British rule and did not share an over-riding Karen communal identity that could frame and inspire such a movement. Indeed, “Karen” only became an accepted communal category during the colonial period and was coined to refer to peoples speaking several unwritten languages, practicing different religions, and wearing different styles of clothing [Jørgensen Reference Jørgensen Anders and Marshall1997: vi].

Although a Karen political community might have emerged and become popular over the past 150 years without a history of British colonialism, this scenario seems very unlikely given the influence of CCPs on the formation of a Karen political community. While colonial censuses made “Karen” a legitimate category of community, legislative representation helped to organize Karens to look out for the well-being of all Karens. Tellingly, Dr. San C. Po, the first Karen representative in the Legislative Council, is recognized as the father of the Karen nation and wrote the “bible” on Karen nationalism after over a decade of fighting against Bamar politicians in pursuit of Karen interests in the Legislative Council [Fink Reference Fink and Po San2001]. Indeed, Karen leaders point to the reservation of legislative seats as awakening Karens as a nation [Burma Reforms Committee 1921a: 232, 238]. And in his early and insightful analysis of Karen-Bamar relations, Lewis [Reference Lewis James1924: 96] notes that Karens had increasingly assimilated into Bamar culture since the late 19th century but describes how the reservation of legislative seats helped to halt this process.

Missionaries also played an important role in molding Karens into a political community, and their influence depended on colonialism. As several scholars note, missionaries sought to create a Karen community as part of their effort to convert Karens [Gravers Reference Gravers and Mikael2002; Lange Reference Lange2017: 116-120; Rajah Reference Rajah2002], with one missionary tellingly declaring, “From a loose aggregation of clans we shall weld them into a nation yet” [Smeaton Reference Smeaton Donald1920: 19]. Missionary activities, in turn, helped to build a Karen nation in a number of ways: they created written languages for Karens, ran schools teaching in Karen languages, helped Karens to organize communal associations, organized activities that brought Karens from diverse regions to meet and consider issues facing all Karens, and emphasized Bamars as dangerous others who had historically exploited Karens. Highlighting the influence of missionaries, nearly all leaders of the Karen nationalist movement have been Christian and Sgaw, the latter being the Karen language group that was most influenced by missionaries [Harriden Reference Harriden2002; Lange Reference Lange2017: 120; Stern Reference Stern1968]. Alternatively, non-Christian and non-Sgaw Karens are most likely to oppose Karen nationalism. Importantly, although missionaries were not formal colonial agents, their presence and influence depended on British rule. Whereas the British allowed missionaries to work in Myanmar and actually offered them considerable assistance, leaders in precolonial Myanmar forbid missionaries from proselytizing and threatened to execute converts.

A comparison of Karens living in Thailand, a country that was never colonized by Europeans, provides additional insight into the counterfactual. Karens are the largest recognized minority in Northern Thailand and were similar to their Myanmar counterparts in that historically they were not organized politically, did not have strong communal identities, were marginalized, and faced considerable discrimination [Buadaeng Reference Buadaeng, James, Thornton and Inman2007]. In the absence of colonial rule, however, Thai officials did not attempt to implement any sort of communalizing policy that recognized communal difference between Thais and Karens, let alone any policy that privileged Karens over Thais. Instead, the Thai state attempted to integrate Karens into the Thai nation while increasing its control over them. For example, the Thai government created government schools for minorities that taught in Thai, used a Thai nationalist curriculum, and thereby facilitated the inclusion of minorities into the Thai nation. Because Buddhism was the official religion and was a source of state legitimacy, Thai officials also limited the influence of Christian missionaries among minorities and countered missionary efforts by sponsoring their own Buddhist missions. And those Christian missionaries that worked with Thai Karens simply helped Karens to attend government schools instead of running their own schools [Buadaeng Reference Buadaeng, James, Thornton and Inman2007]. Due in large part to these differences, Thai Karens never became an assertive political community, and most Karens developed hybrid political identities that were equal parts Thai and Karen [Buadaeng Reference Buadaeng, James, Thornton and Inman2007]. Although one cannot assume that the outcome would have been exactly the same in Myanmar without a history of British rule, a similar outcome seems likely because it is difficult to think of situations in which the Myanmar state would have privileged Karens, recognized them as a distinct community, and opened the door to missionary conversion among Karens. All this suggests that, in the absence of the British colonial interlude, Myanmar Karens would have been more integrated into Bamar institutions and would not have fought a secessionist civil war.