Judiciary institutions and violent crime in American Indian nations

In many American Indian nations the security situation is dire. While scholars have studied how institutions shape economic development in American Indian and Alaska Native (AIAN) nations, the role of AIAN institutions for security and violent crime has received much less attention—despite the extensive literature highlighting the important role of effective and legitimate institutions in the long-term decline of violence. We analyze how varying types of American Indian poli-ties and judiciary institutions fare in tackling violent crime using data across 146 American Indian polities. Our findings indicate that more autonomous American Indian criminal justice institutions with specialized court systems are associated with lower violent crime. However, customary justice institutions do not appear to be effective in reducing violent crime, highlighting the problem of cultural mismatch between traditional and formal justice systems. We argue that analyzing AIAN nations provides important insights into how institutional legitimacy can shape violent crime.


| INTRODUCTION
We study how institutional differences across American Indian nations can shape the quality of security provision and explain variation in the incidence of violent crime. The security situation is dire in many federally recognized tribal entities. 1 A National Congress of American Indians (NCAI) brief notes that "Indian reservations nationwide face violent crime rates more than 2.5 times the national rate, and some reservations face more than 20 times the national rate of violence." 2 American Indian women "are 10 times as likely to be murdered than other Americans … [and] raped or sexually assaulted at a rate four times the national average" (Williams, 2012), an issue reflected in the activism surrounding the Missing and Murdered Indigenous Women and Girls movement (Indian Law Resource Center, 2019; Lucchesi & Echo-Hawk, 2018). These high levels of violent crime and insecurity ensue in a context of poverty and inequality. In 2012, 29.1% of the approximately 5.2 million people identifying as American Indian or Alaska Native (AIAN) (inside and outside Indian lands) were estimated to live in poverty (Krogstad, 2014). 3 Political scientists to date have paid limited attention to patterns of security and violent crime across AIAN nations and the plausible impact of institutions. We study how varying types of American Indian polities and judiciary setups help explain patterns of violent crime across American Indian nations. 4 We find lower rates of violent crime where American Indian judiciary institutions feature specialized and autonomous capacity for criminal sentencing.
The comparatively weak capacity of local tribal entities relative to federal units such as the Federal Bureau of Investigation (FBI) make this divergence all the more remarkable, as it is difficult to attribute a positive impact to greater enforcement capacity. The content of institutions also matters: the effect is limited to institutions relevant to violent crime. We find no general spill-over effect of local institutions. Finally, a greater reliance on more customary types of dispute resolution mechanisms seems to go together with higher violent crime rates. We argue that this may be less the result of the incapacity of indigenous institutions, and rather a problem of the incongruence and compatibility of parallel indigenous and formal approaches to combating violent crime, in a context in which the sovereignty of American Indian institutions is both subject to federal authority and often challenged by overlapping jurisdictions from empowered individual states and their actions toward tribal entities (see Reid & Curry, 2021, p. 76).
Our study complements research on patterns in economic development across American Indian reservations and tribal jurisdictional areas, which highlights various success stories attributed to "local control, strategic thinking, effective governing institutions, and leadership" (Cornell, 2001, p. 84). The fact that American Indian institutions can enhance development is all the more impressive, as these institutions are generally resource poor, and suggests that success is more likely to be due to greater legitimacy, responsiveness, or "cultural match" (Cornell & Kalt, 2000), rather than financial institutional capacity. Recent studies on indigenous authorities outside of the US mirror these observations, showing that indigenous leaders can be effective in providing public services (Baldwin, 2015;Díaz-Cayeros et al., 2014). Dimitrova-Grajzl et al. (2014) examine how changes after Public Law 280-transferring jurisdiction from the federal government to states-are associated with increasing crime in communities subject to PL-280. However, they focus on counties with significant AIAN populations rather than tribal areas. Our research speaks to previous works on the link between institutions and governance, as well as the role this has played in the long-run decline of violent crime (Nivette, 2014;Nivette & Eisner, 2013). Ferguson (2016) and other participants in a symposium in Perspectives on Politics argue that social science has overlooked issues concerning AIAN nations, and comparative research on AIAN governance is rare. Comparing AIAN polities represents a neglected opportunity to study how variation in governing institutions across units can shape variation in violent crime, given the severity of the problem of violent crime in many American Indian nations and the widespread scholarly attention to the link between institutions and violent crime rates across societies outside AIAN communities. We review three streams of literature to which our research contributes, highlighting the important potential of the comparative empirical study of AIAN institutions.

| Violent crime and institutional legitimacy: Global trends
Scholars have examined the evolution in violent crime rates over long historical periods for European cities and countries and found a dramatic decline (Gurr, 1981). Eisner (2003, pp. 84-85) argues that homicide rates in England in the High and Late Middle Ages appear to have been as high as 20 per 100,000, and then declined to about one per 100,000 inhabitants in the twentieth century. Data for other European countries suggest even larger declines than the 20:1 seen in England, in some cases up to 50:1. Outside Europe, Roth (2001) estimates a homicide rate among European settlers in colonial New England of over 100 per 100,000 during the era of frontier violence before 1637, subsequently declining to levels comparable to England.
Many explanations have been proposed for this dramatic decline in violent crime. Gurr (1981, p. 295) sees the decline as "a manifestation of cultural change …, especially the growing sensitization to violence and the development of increased internal and external control on aggressive behavior." Although the growth of the state and its coercive capacity is likely to have played a role (Tilly, 1992), there appears to be an important legitimacy component, as crime rates in Northern Europe initially declined with the increasing acceptance or legitimacy of the state, while they remained higher in areas of Southern Europe where popular acceptance of the state was lower (Eisner, 2003). Roth (2001, p. 85) argues that the decline of homicide rates in colonial New England did not correlate with improved economic circumstances, stronger courts, or better policing … [but] with the rise of intense feelings of Protestant and racial solidarity among the colonists, as two wars and a revolution united the formerly divided colonists against New England's native inhabitants, against the French, and against their own Catholic Monarch, James II. Nivette and Eisner (2013) demonstrate a similar relationship between higher legitimacy, better governance and lower violent crime rates in a global sample (see also Neumayer, 2003;Nivette, 2014). LaFree (1998) argues that social and political institutions can shape human actions through rewarding desired behavior and controlling unwanted behaviors. When legitimacy breaks down, or members of a society begin to doubt the fairness of their political institutions, … they become less enthusiastic agents for the social control of others, … do less to defend rules and respond less harshly to rule violations … [and] formal punishment by the legal system is less threatening and carries less of a stigma (LaFree, 1998, p. 80).
If a state engages in perceived illegal or corrupt actions, individuals are likely to increase their use of "self-help" or "private justice" to solve interpersonal conflicts or grievances (Eisner, 2009).
Although the consequences of a breakdown in legitimacy are clear, we know less about what influences popular perceptions of legitimacy in the first place. It seems plausible that legitimacy is enhanced when institutions reflect the identity of a subject, rely on forms of tacit consent, are not externally imposed, and are seen as generating favorable outcomes (Johnson and Ridgeway, 2006;Tyler, 2006). Taken together, these arguments suggest a possible role for AIAN institutions to influence the prevalence of violent crime. While AIAN institutions are indigenous institutions in that they encompass tribes that predate US territorial conquest, many AIAN nations are also modern creations and may be far from the traditional homeland of the commu nity. Modern indigenous institutions can be more legitimate or representative in the sense of reflecting the members of the community and help foster a sense of autonomous leadership and governance by the community relative to institutions of the federal government.

| Governance in AIAN communities
There are many highly specialized comparative studies of AIAN institutions, including on gaming regulations (Boehmke & Witmer, 2004;Witmer & Boehmke, 2007), historical contractual relations between American Indian governments and US federal authorities (Spirling, 2012), contemporary interactions with local non-native state authorities (Evans, 2011(Evans, , 2014, legislation addressing American Indian issues (Witmer et al., 2014), AIAN lobbying efforts (Boehmke & Witmer, 2012), or membership rules in tribal constitutions (Gover, 2010). Yet, few compare how the various forms of governance and polities can shape socio-economic outcomes on the tribal area. Notable exceptions include Cornell and Kalt (1995), Cornell and Kalt (2000), and Akee et al. (2015), all focusing on how different institutional setups of American Indian polities are associated with economic well-being. Cornell and Kalt (1995) compare the trajectory of economic development between the White Mountain Apache and the Pine Ridge Oglala Sioux. While both nations are administered with similarly structured, centralized tribal government, these polities were established with the Indian Reorganization Act (IRA) of 1934. While the previous Apache polity mirrored the IRA setup quite well, the traditional Sioux organization was "almost precisely the opposite" (Cornell & Kalt, 1995, p. 424). They credit the comparatively better developmental performance of the Apache to the less disruptive "match" of formal, externally imposed institutions and internal political culture and organization. Cornell and Kalt (2000) examine the sources of development in American Indian economies both qualitatively and statistically for 63 nations, finding that "tribal constitutional forms appear to be make or break keys to development" (2000, p. 442). They argue that self-governance promotes economic development, and that a "cultural match" between a community's formal institutions and pre-constitutional norms of political legitimacy is crucial for development (2000, p. 466). They assess institutional variables of American Indian polities, such as whether nations elect their leadership directly or indirectly, whether a judiciary is defined in the constitution and whether the latter is independent, and the term length of the tribal chairs. They find that employment and growth are positively affected by having a strong legislature, that is, a council that can hold chairpersons accountable, as compared to a legislative made up of all tribal members in a general council. Cornell (2001) provides further comparative qualitative evidence for how positive development trajectories in American Indian communities emerge through tribal self-governance, effective institutions, following a strategic development plan, and if the leadership is perceived to be legitimate. Akee et al. (2015) consider over 70 American Indian constitutions of nations with democratic forms of government and a population greater than 750 people, focusing on "implications of the external political conditions at the time of adoption of American Indian constitutions and how they affect long-run economic development for those tribes" (2015, p. 845). Among other things, they code the rules for executive leadership selection, independence of the judiciary, staggering of elections, and term length as defined in the constitutions. They show that differences in party ideology of the US president when constitutions came into effect shaped tribal institutional arrangements and long-run economic outcomes: American Indian nations adopting constitutions under Democratic presidents perform better in later economic development. Moreover, " [u] nder Democrat US Presidents American Indian tribes tended to adopt an indirectly elected chief executive (parliamentary-type system) and under Republican US Presidents they tended to adopt a directly elected chief executive (presidential-type system)" (2015, p. 845). Their results suggest that at critical junctures, the party of the US president has shaped constitutional design, and that the indirect form of government leads to higher labor force participation and per capita income.
These comparative analyses provide essential insights on the effect of institutional setups across American Indian communities. Yet, they focus mainly on development. Poverty and marginalization are often seen as root causes of crime, and may account in part for the high violent crime rates in AIAN communities. Yet, existing research on institutions and crime in our view provide compelling reasons to expect that variation across AIAN governance institutions should exert an independent, direct influence on violent crime.

| Violent crime in non-state societies and AIAN communities
The role of institutions and crime has been examined beyond nation states, with many studies considering non-state societies or societies with traditional governance institutions (Nivette, 2011). Levels of violence in stateless societies appear to generally exceed those of modern states. However, the variation does not fit the traditional Hobbes/Leviathan versus Rosseau/"Noble Savage" dichotomy often invoked (Fry, 2006). Nivette (2011, pp. 508-510) provides evidence that non-state societies with stronger social institutions and integration see less crime, reminiscent of the evidence for states. However, many studies admittedly suffer from poorly defined, subjective measures (i.e., classifications of high and low crime), and there are often concerns about limited data and selection biases.
Beyond stateless societies, there is some evidence that colonially imposed or post-colonial inherited institutions have been less effective in addressing crime. This has been attributed to a fundamental lack of legitimacy, or failure to "express the accepted norms, values, and customs of the indigenous people" (Tankebe, 2008, p. 73). This is highly relevant for AIAN units, since the US Federal Government exercises ultimate authority, and many nations originally inherited governing institutions based on a template reflecting US practices.
Most existing analyses of crime in AIAN communities have focused on either individual level risk factors or AIAN as a population rather than local outcomes in tribal units (Lainer & Huff-Corzine, 2006;Sapra et al., 2014;Yuan et al., 2006), or the role of acculturation or alienation at the community level rather than institutions per se (Levy & Kunitz, 1971;Levy et al., 1969). This is an important omission, as AIAN communities can be seen as an enduring case of internal colonialism, given their relations with US federal authorities. Moreover, there is significant variation across AIAN institutions, where some nations rely much more on self-governance than others. We leverage institutional variation across American Indian institutions in criminal justice for a comparative analysis of the types of institutions likely to generate the legitimacy to reduce crime rates.

| THEORY AND PROPOSITIONS
The setup of American Indian reservations and tribal jurisdictional areas as federally recognized lands means that American Indian communities have two possible institutional influences on crime: the internal American Indian polity as well the external institutions of the US Federal Government. Reid and Curry (2021, p. 70) note that AIAN polities exist in an "odd semisovereign status," where federal law and policies hold that tribes are simultaneously sovereign and not sovereign. There is significant variation in how much AIAN polities work autonomously through self-governed institutions in specific areas compared to how much they rely on alternative external institutions, that is, in the provision of security (and other services) by Federal agencies such as the Department of Justice, the FBI, or the Bureau of Indian Affairs (BIA). We argue that the variation in the use of local self-governed institutions offers an opportunity to distinguish (1) material and coercive effectiveness from (2) more intangible features related to the legitimacy of self-governing, autonomous institutions. If all that mattered for crime was the legislation and capacity to exert control, we would expect that reliance on federal institutions should generally be the most effective, given the superior resources and experience of federal law enforcement. However, if institutions that are perceived as effective and legitimate by constituents are helpful for driving down violent crime rates, self-governing AIAN institutions can make up for their lower formal capacity through greater legitimacy, responsiveness, and local presence.
Criminal institutions and jurisdiction in AIAN communities have an added level of complexity beyond what is determined by local laws and legal capacity, since many of the most serious crimes fall under federal rather than tribal jurisdiction. Skeptics may thus question whether legal institutions designed to handle misdemeanors and less serious crime could impact violent crime, ultimately enforced by other criminal agencies than the tribe. We believe that it is reasonable to expect that more legitimate local judicial institutions with competence for lower-level crime have important subsequent consequences in preventing and reducing observed violent crime, even when tribes themselves do not prosecute these crimes. There is an extensive literature in criminology on how improved policing of less-serious crime can reduce the occurrence of more serious crime, since individuals involved are more likely to engage in more serious crime, or a greater prevalence of crime can encourage more criminal behavior and a higher risk of more serious crime. Although the original concept of "broken windows" has often been used to support "zero-tolerance" policing strategies, Kelling and Coles (1998) note that broken windows can also be fixed through community strategies, where policing generates more cooperative relationships with communities and encourage local capacity. In a meta-analysis of empirical studies, Braga et al. (2015) find that policing disorder strategies are associated with a clear reduction in crime, and that the largest effects are found for community and problem-solving interventions. These studies look at policing, but we would expect that their insights can extend to judicial institutions, even if they do not have formal enforcement powers. 5 Beyond legitimacy, institutions are likely to affect crime if they are intentionally designed and dedicated to combating violent crime, either through specialized criminal justice institutions or eliciting compliance with criminal codes. Cornell and Kalt (2000) do not find a significant relationship between economic development and a constitutionally defined and independent judiciary. However, these judicial institutional arrangements are likely to be particularly important for an analysis of levels of violent crime: a functional, self-governed American Indian judiciary without interference from Federal authorities should increase the legitimacy of the judiciary procedure in AIAN communities, promote the rule of law and effectively process varying criminal issues. Thus, we distinguish two features of AIAN judiciary institutions that can shape the effectiveness of self-governance on violent crime: the judiciary specialization and customary judiciary elements of AIAN institutions.

| Judiciary specialization of AIAN institutions
There is much variance in judiciary institutions across AIAN polities. Many AIAN constitutions do not define a judiciary beyond a clause giving the tribal council the authority to appoint judges. Other polities' constitutions have more involved regulations of the judiciary, for example, explicitly defining the latter as independent, or specifying the number of judges and length of their term. Some elect judges into office instead of appointing them, or feature courts with more customary elements. In general, we would expect that the more clearly specialized and specified the judiciary is, the higher the level of security should be on a reservation, village or tribal jurisdictional area, and increased capacity to deal with complex issues related to violent crime. Many AIAN nations also have law and order codes that further define the role of the judiciary. A specialized judiciary-defined as featuring a court system that channels diverse types of crime through specialized channels, judges and staff-may be better able to enforce the rule of law. Thus, we formulate our first hypothesis: H 1 American Indian reservations or tribal jurisdictional areas with a more specialized judiciary will have lower violent crime rates.
In our analysis we focus on specialized judiciary institutions that handle domestic abuse protection orders. We see these as particularly relevant for evolving crime rates, since much violent crime emanates from domestic abuse, given the high recidivism rates for crime. 6 In comparison, we expect to see less impact of specialized institutions less relevant to violent crime (e.g., criminal misdemeanors and traffic violations). American Indian nations that feature such judiciary specialization in self-governing institutions should perform better than units relying on external courts for criminal justice. Again, specialized self-governance may come with more legitimacy and local proximity of the courts, making them more effective judicial institutions in reducing levels of violent crime.

| Customary judiciary elements of AIAN institutions
In AIAN communities, customary conceptions of governance often coincide with externally imposed institutional setups. Institutional legitimacy may therefore not only depend on whether customary forms of governance are de facto applied, but also how they interlink with and contradict more formal constitutional setups. Begay et al. (2007, p. 47) emphasize the importance of cultural match in AIAN communities, that is, "a fit between the formal institutions of governance and the underlying political culture of the society being governed" The authors do not refer to a necessary comeback of traditional institutions, but rather an institutional setup reflecting the contemporary political culture of a community, including but not limited to governance traditions. Thus, even if customary, pre-colonial elements of institutions are maintained or re-established, they may conflict with externally imposed rules, leading to ineffectiveness of these institutions. An effective match, Cornell (2007, p. 73) and Begay et al. (2007, p. 52) argue, is most likely to be achieved through indigenous control, that is, self-governing authority and sovereignty over institutional setups by AIAN authorities.
Examples of the practice of alternative dispute resolution mechanisms in other contexts (Mac Ginty, 2008;Zartman, 2000), sometimes employed jointly or in parallel with more formal mechanisms (as shown in studies on legal pluralism, e.g., Benjamin, 2008;Zips & Weilenmann, 2011), demonstrate the incorporation of indigenous principles of justice. Indigenous practices may provide legitimate pathways to dispute resolution otherwise inaccessible through the application of externally imposed, culturally insensitive procedures. Some AIAN communities have attempted to establish more restorative judiciary institutions with greater cultural match, for example, through peacemaking circles providing alternative forms of recon ciliation. These are restorative forms of justice employed with customary elements. 7 One might assume that such alternative forms of justice should reduce violent crime, given the attempt to generate greater cultural fit and the conciliatory character of these approaches to justice. Moreover, alternative types of dispute resolution are frequently employed on lowest levels of administration, and should thus provide even greater control to AIAN governance authorities. Thus follows our second hypothesis: H 2 If the judiciary on the self-governed tribal level incorporates customary approaches to justice, the violent crime rates on the American Indian reservation or tribal jurisdictional area will be lower.
Yet, the incongruence of co-existing indigenous and formal institutions of justice may also be problematic for the occurrence of violent crime. Some critics see traditional or religious courts as favoring established interests over the rights of victims, in ways that perpetuate rather than combat abuse. For example, parallel Sharia courts tend to discourage women reporting domestic violence and make it difficult for women to obtain a divorce (Zee, 2015). Johnston (2017) argues that local conflict resolution institutions in East Timor have helped to condone abuse of women by identifying the appropriate "fines" instead of discouraging abuse. Traditional conflict resolution institutions alone may not be a panacea in the absence of modern judicial institutions relevant to violent crime.

| EMPIRICAL ANALYSIS
We evaluate the two hypotheses on judicial institutions and crime rates in a comparative analysis of 146 American Indian nations.

| Data and operationalization
Our main outcome variable is a measure of the per capita violent crime rate of each American Indian nation. Our data on total violent crime come from the annual reports by Steven W. Perry for "Tribal Crime in the United States," published by the US Department of Justice, Bureau of Justice Statistics. 8 We use the annual average across the available data over the period 2010-2014. 9 Combining data over multiple years helps to reduce the year-on-year variance and to smooth out potential erratic spikes in relatively rare events for small units over a short interval. In Table S3 we provide the full data and descriptive statistics for all of the 146 American Indian nations in our data. It is difficult to find complete and disaggregated data specific to individual tribal units, but our data cover most of the major tribal units. Most of the missing data appear to pertain to very small reservations and tribal jurisdictional areas with small resident populations. There is an analytical trade-off here in detail versus finding relatively complete data covering a large number of AIAN communities. We acknowledge that the crime data employed below have limitations, since not all tribes may report complete data and many law enforcement agencies have limited jurisdiction over crimes occurring on tribal lands. In addition, tribal authorities rightly remain at times hesitant to share data outside of their communities.
Thus, systematic, nuanced data across AIAN communities is scarce, and we have to make analytical compromises using what is available to make inference regarding our comparative interest. For instance, the time period of our dependent variable (2010-2014) is arguably limited. However, we believe these crime data provide the best possible measure of violent crime specific to the local tribal government, and are sufficiently inclusive to provide useful information on a specific time period and on the potential impact of AIAN institutions. It is instructive to consider how potential undercounts could affect our inferences. In order to bias in favor of our propositions, it would have to be the case that we have a higher undercount for more specialized AIAN institutions with greater capacity-that is, these would need to be less likely to report violent crime. Any systematic bias here would seem more likely in the other direction, namely, that AIAN with less capacity and resources also report less crime, thus making it harder to find support for our propositions.
Since American Indian nations differ dramatically in their total population it is better to look at per capita rates rather examining the absolute number of violent crime events. For population, we use data from the US Census Bureau ( , 2005). 11 This survey covers a range of questions on law enforcement, courts and administration, corrections and intermediate sentences as well as criminal history records and justice statistics. This survey secured high response rates (at over 92%). All but 10 of the American Indian nations for which we have crime data participated in the survey, although not all have answered all questions. We believe that data on institutional variation in 2002 provides a reasonable basis for assessing impact on crime outcomes in 2010-2014. A caveat applies, as the data are now relatively old. However, the 2002 CTJA data allow for a relatively complete and sophisticated measure of institutional variation, and comparable recent data would require further original data collection-which would be cost prohibitive and potentially a further burden on authorities to F I G U R E 2 Violent crime rates per capita for Arizona and Four Corners area, 2010-2014. provide responses to, for example, a survey, if authorities would agree to share this information. Institutions are sticky and rarely change drastically over time-and even if they do, outcomes are likely to reflect lasting effects of previous institutional setups. Thus, it is reasonable to assume that an effect can be manifest between the institutional variation captured in 2002 and crime outcomes in 2010-2014.
First, we measure the judiciary specialization of self-governing American Indian judiciary institutions that are relevant to violent crime by recording if the tribal justice system handles domestic violence protective orders. This is a binary variable based on CTJA item b3_6. 12 The majority of American Indian nations in our sample do handle domestic violence protective orders (86%), but there are a number of nations that do not (14%) such as the Lower Sioux Indian Community in the State of Minnesota. If local institutions have an edge on legitimacy then we should see generally lower violent crime rates when nations have specialized judiciary institutions handling domestic violence protective orders over cases where these are handled by non-tribal institutions. To examine whether more relevant specialized judiciary institutions (e.g., handling domestic violence) have a greater effect on violent crime than less relevant specialized judiciary institutions (e.g., criminal misdemeanors and traffic violations), we create two alternative binary measures of judiciary specialization. Namely, whether or not the tribal justice system handles criminal misdemeanor cases and traffic violations.
Second, we compare the effects of formal, domain specific criminal justice institutions to customary justice institutions to capture the idea of furthering cultural match through alternative forms of justice. To do so we consider two CTJA items. The first asks respondents whether their tribal justice system have a separate peacemaking/circle sentencing court (b4_6). The second considers whether traditional methods and/or forums operate in the tribal jurisdiction (b2_4). We create a dichotomous indicator flagging whether the answers to any of these questions is yes. While the majority of American Indian nations in our sample do not have these customary functions operating in their judicial system (81%), a number of nations do (19%), such as the Navajo Nation in Arizona, New Mexico and Utah.
We also consider potentially important control variables that may be associated with both features of institutions and crime rates. First we consider the overall size of tribal units. The geographic size may be related to problems of governability. We extract a measure of total area in square kilometers from the US Department of Commerce (2015) US Census American Indian/ Alaska Native/Native Hawaiian Areas (AIANNH) National 2014 shapefile data. We log the values, since the marginal impact of size is likely to decrease as the base is larger.
Although our crime rate measures are population adjusted, total population size may be related to volatility in per capita crime. Not all American Indian units may have the capacity to set up autonomous units with complex self-governance institutions, and total size is a plausible proxy for potential resources as very small units are less likely to have the required capacity. There is also some evidence from studies of traditional societies that the size of units tend to be associated with lower crime rates (Nivette, 2011). We control for total population of an American Indian nation using population data from the US Census Bureau (2010), 2006-2010 American Community Survey with the supplements discussed above or missing data. We use logged values to decrease differences for large populations.
Finally, we control for AIAN nations affected by Public Law 280 jurisdiction. The act, established in 1953, transferred legal authority from the federal government to state governments concerning civil and criminal matters involving American Indians and Alaska Natives on reservations and in villages. The law required "mandatory states" including Alaska, California, Minnesota, Nebraska, Oregon and Wisconsin to assume jurisdiction and "optional states" includ-ing Arizona, Florida, Idaho, Iowa, Montana, Nevada, North Dakota, South Dakota, Utah and Washington to assume full or partial state jurisdiction. The law has been widely criticized for creating confusion regarding the role of states in criminal justice systems in AIAN nations, a lack of funding from the federal government and a failure to recognize tribal sovereignty and tribal self-determination (Goldberg-Ambrose, 1997). As Reid and Curry (2021) and Dimitrova-Grajzl et al. (2014) describe, crime rates have increased in communities where PL-280 was enforced. Therefore, we might expect violent crime rates to be higher when the state has authority over criminal matters on American Indian reservations because of decreased institutional capacity and perceptions of illegitimacy among American Indian constituents. We control for Public Law 280 using data from the National Indian Justice Centre (2014) report on Federal Recognized Indian Nations under State Jurisdiction (PL-280). We create a binary variable (similar to Reid & Curry, 2021), assigning a value of 1 to those American Indian nations where criminal jurisdiction lies with the state, and a 0 otherwise.
Crime is related to poverty, and much work argues that greater wealth tends to go together with more responsive institutions and improved governance. We use two measures of variation in poverty and income. Arguably, the most appropriate measure would be that of poverty rates on tribal reservations and tribal jurisdictional areas. However, we have been unable to find systematic comparative data on this. As a fallback, we control for the poverty rates in the county where the tribal seat of administration is located, using data from the 2010 US census. 13 Although this is not tribe specific and may not cover the entire reservation, it should provide a reasonable indication of the number of people affected by poverty. In addition, we consider data on median income data for tribal areas reported in from the US Census Bureau (2010), 2006-2010 American Community Survey, where lower median income would correspond to a more impoverished population. We log the median income values to reduce skew. Table 1 provides estimates for a series of different specifications of a linear regression model of crime rate per capita by tribal area against a series of covariates reflecting our main features of interests and the core control variables. The baseline models 1-2 include just the control variables, without any institutional characteristics. As one would expect, we find a positive association with census county poverty rates and crime in model 1. The results for model 2 shows that we see a drop in the number of observations when we rely on the nation specific median income measure, and we find no evidence that this has a clear covariation with violent crime rates. In light of this, we focus only on the poverty rates measures in the subsequent estimates.

| Empirical results
For total population we find a negative association with crime rates, suggesting that larger tribes may have greater capacity reflected in lower crime rates. For total area the coefficient is positive, although it is not statistically significant.
The dramatic range of variation in crime rates make our results potentially sensitive to outliers. As mentioned above we have two extreme outliers that have annual violent crime rates more than 4 standard deviations above the mean. In model 3 we exclude the two nations with annual violent crime rates above 0.10. As can be seen, the fit as measured by the R 2 of the model increases notably, but we see only minor changes for the coefficients.
In model 4 we introduce our measure of judiciary specialization, that is, whether American Indian jurisdictions handle domestic abuse protection orders. As can be seen, we find a negative coefficient, consistent with our claim (H 1 ) that self-governance through specialized judiciary institutions can reduce violent crime. To put this in perspective, the estimated magnitude of the difference in crime for self-governing institutions is comparable to an over 10% point difference in median income. For example, Navajo Nation has a far above the median poverty rate (38.7 vs. the median 18.20). The predicted value from model 4 conditional on handling domestic violent protective orders is 0.0066. However, the predicted value would increase by almost 75% if the tribe did not handle domestic violence protective orders. Hence, we can see that the coefficient for domestic abuse protection orders entails difference of large substantive magnitude.
We can compare the effects of institutions that are plausibly relevant to crime against generally increased autonomy and specialization in judiciary institutions that one would not expect to be relevant to violent crime. In model 5 (Table 2), we first consider tribal units that have a justice system that handles criminal misdemeanor cases, as opposed to outsourcing this to federal units and in model 6 we consider whether the tribal justice system handles traffic violations. If there was some kind of general spillover effect of self-governing authority we should expect to see similar results here. However, it is clear from the results in column 6 that we find little support for general spillover effects of specialized criminal justice institutions, as the coefficient estimates are negative but not significant.
In model 7 we consider our measure of the use of customary justice institutions. Although such institutions perhaps may have some degree of cultural match and legitimacy among constituents, they do not appear to be effective in reducing violent crime, thus not supporting our hypothesis H 2 . As can be seen, we actually find a positive coefficient estimate for such institutions on the violent crime rate. The results are consistent with a possible tendency to neglect victims and brush domestic abuse and violence under the carpet in the absence of designated institutions to deal with relevant issues. Finally, in model 8 we test the effects of Public Law 280. We find that violent crime rates are higher in American Indian nations where the state has assumed jurisdiction over crimes involving American Indians on the reservation. This finding further supports our argument that greater autonomy for tribal units go together with lower levels of violent crime, excluding justice systems that have a traditional or customary element. 14 We have also considered a number of additional tests for alternative measures of institutions and possible confounding characteristics, but found no evidence that these change our main substantive conclusions. Following Cornell and Kalt (2000), we considered whether a tribal leader is directly (presidential) or indirectly elected (parliamentary), using their coding scheme and a nation's most recent constitution (up to and including versions ratified in 2013). However, we find no evidence that this is associated with clear differences in per capita violent crime rates. Given the role of gambling on many American Indian reservations and its plausible connection with crime, we have also considered data on whether a tribal area has gaming operations (and the number), using data from the National Indian Gaming Commission (NIGC). 15 We find no evidence that this is associated with any consistent differences in the violent crime rate among American Indian units. Finally, we find similar results if we flag all Oklahoma Tribal Jurisdictional Area, which have a distinct background and often include many non-tribal residents not subject to the tribal government (Table 2).

| DISCUSSION AND CONCLUSION
The results above suggest a conundrum for how varying institutional setups can shape the prevalence of violent crime. Self-governance by American Indian authorities with legitimacy  among constituents seems to be crucial: specialized judiciary institutions in American Indian communities-specialized courts dealing with domestic violence protection orders-appear to go together with lower violent crime in American Indian communities, even though the latter rarely have superior resources or enforcement capacity compared to external and US Federal institutions. This is consistent with the idea that local self-governance can increase legitimacy through institutions perceived as more representative and responsive to local concerns (Alesina & Spolaore, 2005;Buchanan & Tullock, 1962). This does not mean that it is straightforward to set up such institutions in the first place, and not all nations may have the capacity to do so, although we stress the control for size as a rough proxy for capacity. Our current data do not allow us to study the evolution of judicial institutions or its relationship with governance capacity, but we hope to be able to extend this in subsequent research. Then again, not every type of local specialization and self-governance of tribal judiciary authority is necessarily beneficial in and of itself unless it is relevant to the subject matter, as seen in the results for judicial institutions handling other issues not clearly related to violent crime such as traffic violations. Our findings even suggest that applying customary forms of justice actually is associated with higher violent crime in American Indian nations. The latter finding is somewhat surprising on the surface, given our propositions about the positive effects of self-governance powers to legitimate authorities and the importance that many have attributed to the "cultural match" of formal institutions and community norms and previous studies on service provision by indigenous authorities. Indigenously controlled and designed judicial institutions should have the greatest cultural match and legitimacy among constituents, thereby providing the most effective provision of local security. However, our findings do not support this proposition, and in American Indian nations these approaches seem to go together with a higher prevalence of violent crime.
On the one hand, this may indicate the problematic nature of indigenous approaches to justice and how these cannot adequately manage the intricate "modern" issues of violent crime, and reconcile customary norms with evolving values about for example, domestic violence. This may be especially problematic given that the economic problems of AIAN communities and the poverty of the public sector of AIAN governance institutions further exacerbate the context for violent crime. AIAN governments might consider indigenous dispute resolution mechanisms the financially cheaper option compared to more specialized justice systems, thereby aggravating the occurrence of violent crime. Moreover, traditional justice mechanisms may be more skewed toward reconciliation in the community rather than protection of victims of violence, and thus insufficient in reducing violent crime sustainably. This finding also highlights the role of institutional effectiveness in generating legitimacy, emphasizing the warning of Begay et al. (2007) about a too simplistic conception of cultural match.
On the other hand, the findings may point to a more complex problem. The parallelism of indigenous and formal justice systems may be a greater obstacle to combating violent crime than having a self-governed, but more singular formal approach to justice. This would not mean that indigenous justice institutions per se lack legitimacy, but that legitimacy of all types of judiciary institutions is undermined by the co-existence of possibly incongruent systems-in the case of AIAN communities, an externally imposed versus an internally fitting one. This type of incongruence complements the idea of a "cultural mismatch" emphasized by Begay et al. (2007, p. 49), according to which externally imposed, formal institutions lack the cultural appropriateness of a community. Yet, our results also suggest that it is less the cultural appropriateness of formal institutions, but rather the mismatch of applied indigenous versus formal institutions that may generate negative outcomes for violent crime.
These findings have repercussion beyond the analysis of AIAN communities: first, in order to reduce violent crime, it is not only the extent to which the modern, formal state expands its reach-even within AIAN nations-but the way these formal institutions can match with indigenous practices of security and justice. Second, where, for example, legal pluralism is practiced, the ambiguity of varying judicial approaches to violent crime, rather than having one formal system, may have detrimental effects on policy outcomes. Third, our analysis highlights the role of legitimacy in reducing violent crime while adding a new comparative perspective on judiciary institutions: self-governing powers of legitimate tribal justice authorities may lessen the level of insecurity in societies. Fourth, our study corroborates previous findings on traditional and indigenous leaders' effectiveness in providing certain public goods. The main difference to arguments presented by Acemoglu et al. (2014) and Baldwin (2015) is that AIAN communities for the most part are elected democratically by the nation's constituents, and thus may draw on both traditional and more formal sources of legitimacy. Moreover, the most effective indigenous authorities in our study seem to be those who are able to match their communities' needs in combating violent crime with a formal type of justice institution.
Regarding AIAN communities suffering from endemic violent crime, a conclusion from our analysis for both Federal and AIAN authorities is to, on the one hand, lobby for and establish even greater judiciary self-governance to create a specialized justice system capable of dealing with a variety of violent crime. On the other hand, AIAN governments should consider the appropriateness of indigenous approaches to justice, and eliminate problematic lack of congruence between such procedures and more formal institutional setups.

ACKNOWLEDGMENTS
We thank the Gerda Henkel Foundation for generously funding this research. We are grateful for helpful comments and discussions with Manley Begay, Nils Petter Gleditsch, Katharina Holzinger, Robert A. Johns, Joseph Kalt, Romain Malejacq, Traci Morris, Mauricio Rivera, and Barbara Zarate. We are also grateful to the members of the American Indian and Alaska Native community who provided critical input regarding this project during our visit in February 2017. A previous version of this manuscript was presented at the International Studies Association (ISA) Annual Meeting (February 2017), the conference "Development in the Face of Inequalities" (Barcelona, May 2017), and at the workshop on "Beyond Political Violence: Causes and Consequences of Crime and Social Violence" at the University of Essex (February 2018).

DATA AVAILABILITY STATEMENT
Data and code to replicate the analyses reported in this article will be made available at the Harvard Dataverse.