Extraction, Assimilation, and Accommodation: The Historical Foundations of Indigenous–State Relations in Latin America

Abstract

Why do some Indigenous communities experience assimilation while others obtain government protection for their long-standing institutions and cultures? I argue that historical experiences with state-led labor conscription play a key role. In the early twentieth century, Latin American governments conscripted unpaid Indigenous labor to build infrastructure. Community leaders threatened by this conscription were more likely to mobilize their communities to resist it. The mobilization of this collective action later empowered community leaders to achieve state protections for Indigenous institutions and cultures, or “accommodation.” I test this argument using a natural experiment where communities’ eligibility for labor conscription to build a 1920s Peruvian highway was as-if randomly assigned. I develop a measure of accommodation that considers both the existence and enforcement of laws protecting Indigenous institutions and cultures. I evaluate the mechanisms using data on Indigenous mobilization. The findings demonstrate how historical extraction shaped contemporary Indigenous–state relations.

Since obtaining their independence, Latin American governments have often endeavored to eliminate Indigenous groups’ political, economic, and cultural institutions in an effort to promote native communities’ “assimilation” into the dominant society. Even in settings where central states have adopted legal frameworks that “accommodate” Indigenous institutions—recognizing and protecting them as legitimate entities within the nation-state—governments have erected financial and bureaucratic barriers that make the implementation of this accommodation prohibitively costly for many Indigenous communities. This paper explores why some native communities have achieved accommodation—considered the “central demand” of Indigenous groups¹ —while others have experienced assimilation.

To answer this question, I analyze the legacies of extractive labor institutions. Throughout the colonial and immediate postindependence periods, governments in Latin America mobilized unpaid Indigenous labor to work in mines, build infrastructure, and serve in the military. The Spanish Crown relied on partnerships with Indigenous elites to facilitate this labor extraction. Scholars have observed important and enduring negative effects of such institutions on the economic welfare of Indigenous groups (Acemoglu, Johnson, and Robinson 2001; Dell 2010; Mahoney 2010). Yet, following independence, as nation-states grew stronger and sought to deploy their authority in traditionally peripheral areas, they began to engage in more direct forms of labor extraction that effectively circumvented Indigenous elites. This labor conscription, which was particularly common at the turn of the twentieth century, often involved government mobilization of unpaid Indigenous labor to build roads, bridges, and other large infrastructure projects.

I argue that these instances of labor conscription led Indigenous elites, who were not incorporated into the extractive process, to mobilize their communities to resist conscription. This collective mobilization, which occurred during the period of conscription, increased communities’ ability to achieve accommodation over the long term. It resulted in Indigenous leaders assuming new and more powerful positions within their communities, which they later used to convince their communities to bear the cost of accommodation; it reinforced a collective memory of government exploitation, which leaders used to mobilize their communities to pressure governments to implement accommodation; and it created closer linkages with Indigenous organizations that provided community leaders with resources they could use to overcome the bureaucratic and financial costs associated with accommodation.

I examine these enduring effects of labor conscription through an analysis of Peru. During the 1920s, the Peruvian government conscripted unpaid Indigenous labor to build Peru’s first highway. I leverage the fact that, by law and in practice, labor on the road was taken only from municipalities located in provinces that contained a portion of the recently rediscovered, ancient Inca Road, atop which the new road would be built. I identify causal effects using a geographic regression-discontinuity design (RDD), where distance to a border dividing a conscription from a non-conscription province serves as the running variable.² I first demonstrate that Indigenous communities resisted conscription, using data on Indigenous mobilization during the period of conscription. To examine the main outcome, I construct an omnibus measure of accommodation, which includes as components both legal measures of whether communities have obtained communal land titles and have achieved recognition, as well as measures of the persistence of Indigenous languages and institutions. I draw on qualitative observations from my fieldwork, secondary sources, and archival data to analyze the mechanisms of persistence.

The findings provide strong support for my theory. Communities eligible for conscription on Leguía’s road were more likely to rebel against the government during the 1920s. Using a multi-method strategy, I show that leaders of these communities assumed more powerful positions within their communities; they were also more likely to join Indigenous organizations and maintain a collective memory of conscription. Consistent with my argument, these communities have been more likely to obtain accommodation according to the omnibus measure and most individual components of the measure: they have been more likely to preserve their long-standing institutions, achieve recognition, and have a communal land title.

These results speak to long-standing debates around the effects of state coercion on group cooperation. Some scholars argue that coercive behavior by the state, especially through repression, erodes affected groups’ collective mobilization capacity (Francisco 1995; Tilly 1978). Others find that such coercion can increase cooperation among affected groups—for example, ethno-nationalist movements and urban slum dwellers (Aytaç, Schiumerini, and Stokes 2018; Thachil 2020). Most of this existing work focuses on relatively near-term outcomes of state coercion, such as participation in protests or social movements. This paper, however, demonstrates that a specific act of state coercion—labor conscription—can trigger collective resistance that endures long after the coercive behavior stops. Efforts by the state to repress and exploit marginalized communities can thus backfire.

This paper also highlights generally understudied community-level variation in Indigenous–state relations.³ Much of the existing scholarship focuses on important national-level differences in state policy toward native groups (Van Cott 2001; Yashar 2005).⁴ However, across a number of cases, state policy has recognized Indigenous rights, but uptake across Indigenous communities remains uneven and, in some cases, very low.⁵ I develop a theoretical framework around two outcomes—accommodation and assimilation—and devise a measurement strategy to identify these outcomes at a local level within countries. Moving to a lower level of analysis acknowledges the diverse experiences of native groups within a given country and moves beyond policy adoption to examine the crucial but generally underexplored issue of policy implementation. This approach also reflects the level at which many Indigenous people exercise authority and negotiate their relationship to the state (e.g., reservations in the United States and reserves in Canada or Colombia).

My theory and evidence further demonstrate a key path through which governments can project power in areas of historically low state capacity, which O’Donnell (1993) labels “brown areas.”⁶ Problems of state building in peripheral regions have been particularly endemic in Peru (Kurtz 2013; Soifer 2015). I demonstrate, however, that even in this unlikely case and even as early as the 1920s, the Peruvian state successfully projected power into remote areas—through labor conscription as well as accommodation and assimilation.

I begin the paper by situating the concepts of accommodation and assimilation within a broader framework of Indigenous–state relations. The remainder of the paper develops and tests my argument that labor conscription can explain community-level variation in accommodation. The concluding sections explore how my theory may apply to other instances of labor coercion in Latin America.

ACCOMMODATION AND ASSIMILATION IN INDIGENOUS–STATE RELATIONS

At the onset of colonization, Indigenous communities were characterized by a range of economic, political, and social institutions; communal land, traditional political authorities, and Indigenous languages were three of the most important. Subsequently, these institutions, which frequently hold expressive and instrumental value for Indigenous communities, have faced threats from states and private actors. However, in some cases, Indigenous groups have achieved protections for their institutions and cultures.

Different outcomes of Indigenous–state relations in Latin America can be characterized using a two-dimensional typology: the level of policy support or financial assistance the state provides to Indigenous communities (i.e., “support”) and the extent to which states can and do intervene to regulate or otherwise curb the autonomy of Indigenous institutions (i.e., “control”).⁷ For the purposes of this paper, I focus only on outcomes where governments offer a moderate or high amount of support; these are the outcomes for which there is an identifiable policy outcome.⁸ This yields a three-by-two typology of Indigenous–state relations. Of these outcomes, Indigenous leaders generally prefer those that provide less control and more support. Conversely, central state incumbents more often prefer outcomes that provide high control and only moderate support, as a way of establishing authority over peripheral areas at the lowest possible cost.⁹

The outcomes in Table 1 correspond to either assimilation or accommodation. The trio of outcomes that comprise the bottom-right corner of the figure—corporatism, integration, and paternalism—provide different routes through which assimilation has been achieved. Paternalism and integration have been the most common in the historical and contemporary periods. Governments have imposed (paternalism) or offered (integration) Spanish-language schools and private titles to collectively held land.¹⁰ In the mid-twentieth century, governments often used corporatist strategies to encourage Indigenous groups to adopt class as opposed to ethnic identities in exchange for policy concessions from the state. Yet, corporatism allowed governments to exercise greater control over Indigenous-peasant communities and the unions that represented them (Collier and Collier 1979; Yashar 2005, 57).

Table 1. Theorized Outcomes of Contemporary Indigenous–State Relations: Accommodation and Assimilation

Note: Outcomes in bold lettering refer to accommodation; outcomes in italics refer to assimilation.

The three outcomes in the upper left-hand corner of Table 1 correspond to accommodation. Subsidized autonomy grants Indigenous communities the right to make distributive decisions over resources fully provided by the national government with limited state interference; Mexico’s Indigenous municipalities (especially in Oaxaca) have come closest to achieving this outcome. Decentralized autonomy involves replacing subnational governments with Indigenous institutions; this outcome, which is observed in Panama’s Indigenous comarcas, has often required Indigenous groups to fund their own affairs with lower levels of state resources than equivalent non-Indigenous administrative units.¹¹ Bounded autonomy can be observed in Colombia and Brazil under the “reserve” system, where governments often own the land where Indigenous communities reside and thus can ignore Indigenous rights when these entitlements preempt infrastructure production or natural resource extraction. Because of the greater control exercised by governments, this is generally Indigenous communities’ least preferred form of accommodation.

Table 1 and the aforementioned preferences of Indigenous communities and central state incumbents suggest that the former will often seek accommodation, whereas the latter will prefer assimilation.¹² Despite historical power imbalances that favor the state, Indigenous communities have increasingly achieved their preferred outcome of accommodation.

Few attempts have been made in the literature to develop quantitative measures of accommodation, which entails both de facto and de jure protection of Indigenous institutions and cultures. Instead, much of the literature focuses on the adoption of national-level, legal entitlements. Yet, within countries, Indigenous communities have often struggled to achieve meaningful implementation of those guarantees; financial, bureaucratic, and political barriers can prevent communities from achieving accommodation.

To illustrate these inequalities, I develop a novel measure of accommodation using the empirical case of Peru. I construct an omnibus index that includes four component measures.¹³ Two of the measures within this index examine de jure, or “legal,” accommodation, specifically whether a community is registered with the Peruvian government and whether it has a completed title to its communal land. Communities with these protections are more likely to have accommodation, whereas those without them are more likely to experience assimilation.

Two further components of the omnibus measure examine whether these legal protections are enforced: a dummy linguistic measure of whether a community is bilingual and a seven-item index of Indigenous institutions.¹⁴ Bilingualism provides a measure of accommodation; Spanish is necessary to communicate with the state and collaborate with it in defense of Indigenous institutions and customs. Because the state protects Indigenous cultures and institutions, native languages might likewise persist under accommodation.¹⁵ I also consider an index of seven Indigenous institutions, which should be more likely to persist if state protections are enforced. These include traditional authorities, communal land, and social institutions, namely, ayni, minka, and mita—which involve different forms of voluntary service to the community.¹⁶ Using dummy indicators for each Indigenous institution, I construct a seven-item index for each community. Higher index values indicate accommodation, whereas lower values generally correspond to assimilation.

Using these four component measures—the seven-item institutions index as well as dummy indicators for recognition, communal land title, and bilingualism—I construct an omnibus measure of accommodation for each community.¹⁷ Data on each of the four components of the omnibus measure were taken from a 2012 census of Peruvian native and peasant communities (INEI 2014).¹⁸

Figure 1 uses this omnibus measure to explore variation in accommodation—and assimilation—across Peru’s Indigenous communities.¹⁹ I aggregate the results up to the municipal level for purposes of illustration. Municipalities are the lowest administrative tier of government in Peru; communities are nested within municipalities, such that there are on average five communities per municipality. Certain regional trends emerge: assimilation is the most common outcome in northern Peru, whereas accommodation is more prevalent in southern Peru. Central Peru contains a mix of assimilation and accommodation.²⁰ The next section develops a theory to explain why some communities are more likely to achieve accommodation, focusing on the key role of labor conscription.

Figure 1. Indigenous–State Relations in Peru using Omnibus Measure

Note: Community ( $ n=4,993 $ ) scores on an omnibus measure of accommodation, aggregated to the municipal level ( $ n=1,078 $ ). Municipalities are coded based on whether communities in that municipality are—on average—below/above average omnibus score for all communities (2.86). The omnibus measure of accommodation considers whether a community (1) maintains Indigenous institutions, (2) is bilingual, (3) has recognition, and (4) has a completed communal land title. See Supplementary Figures S1 and S2 and Dataverse Appendix Figures D17 and D18 for maps of individual components. Source: Instituto Nacional de Estadística e Informática (2014).

LABOR CONSCRIPTION IN 1920S PERU

Between 1870 and 1930, governments throughout Latin America sought to invest in transportation infrastructure, notably roads and railways, as a way of expanding internal markets and the export sector (Summerhill 2006). Lacking revenue for these large-scale projects, governments in Colombia, Guatemala, Ecuador, Mexico, and Bolivia implemented new laws that forced Indigenous, working-age males to work without pay to build roads and railways.²¹ Importantly, labor conscription was generally short-term, meaning that Indigenous workers returned to their communities once their service was done—barring death. As such, it can be differentiated from other, more sustained forms of coerced labor that are more likely to break up communities, such as indentured servitude or slavery.

Perhaps Latin America’s most notable example of postindependence labor conscription occurred in Peru under President Augusto Leguía (1919–30), who built the country’s first major highway (Chaplin 2015, 65). In determining where to build his road, Leguía followed the route of the Qhapaq Ñan, or Inca Road.²² Built before the arrival of the Spanish in the sixteenth century, the main part of the Qhapaq Ñan traversed the spine of the Andes from present-day Chile to Colombia. The road linked the otherwise-isolated parts of the Inca empire, allowing the emperor in Cusco to exert effective control. The Spanish had little use for the Inca Road, which primarily ran north to south; they preferred east-west routes, which allowed them to transport mineral resources from the Andes to Pacific ports. By the seventeenth century, the Inca Road had fallen into disuse with its location largely unknown. Using historical documents and a team of archaeologists, Leguía reconstructed the Qhapaq Ñan route.

To build his highway, Leguía proposed the Ley de Conscripción Vial, or Road Conscription Law, which required Indigenous men to provide unpaid labor to build roads (Basadre 2014, 207; Davies 1974, 84). The government defined the universe of conscripts using military conscription rolls; every man of working age, that is between 18 and 60, could be drafted to work on the road. However, archival documents from the departments of Ayacucho, Lima, and Ancash suggest that men as young as 16 and as old as 70 were taken to work on the roads (Dirección de Vías de Comunicación 1928). Nearly every community household was touched by labor conscription and observers argue that this period of labor conscription harmed Indigenous communities even more than the notorious extractive practices of the colonial era.²³

Unlike previous instances of state-led extraction, such as the colonial-era mining mita or “Indian tribute,”²⁴ Leguía’s law did not enlist the cooperation of Indigenous leaders; instead, juntas viales, or provincial-level commissions appointed by the national government, were responsible for mobilizing Indigenous laborers.²⁵ For 10 years, juntas viales traveled to Indigenous communities within their provinces to mobilize labor to build roads. Provincial police were enlisted to ensure that community members complied with their obligations.²⁶

Abuses abounded under the road conscription program. Juntas viales forced Indigenous laborers to work far more than their obligation—sometimes for months at a time—without providing food, shelter, clothing, or even documentary proof of conscripts’ service (Basadre 2014, 197; Mallon 2014, 233).²⁷ Stein (1980, 61) notes, “Indians were forced to leave their homes and travel many miles over difficult terrain to the construction sites $ \dots $ Generally the workers received little if any food and no medical attention; deaths among the ‘conscriptos’ were not uncommon.”

Conscription constituted a protracted process that began with a community becoming aware of its eligibility, continued with community members being conscripted—often multiple times—and only ended with the termination of the road conscription program in 1930. Not all communities were eligible for conscription—for reasons I detail in the empirics. Communities learned of their eligibility for labor conscription in several ways. In the early 1920s, working-age men in many communities were required to register in the provincial padrón vial, or road-building registry. Provincial governments circulated convocatorias (announcements) detailing which communities would work on a project—for example, building a bridge and repairing a road (Meza Bazán 1999, 148–9). Ultimately, juntas viales arrived in communities to march workers to road-building sites.

There was often a high correspondence between eligibility for conscription and experiencing it. Available evidence from the padrón vial from the province of Pomabamba, for example, reports that by 1929 over 16,480 workers had built almost 100 km of roads in the province (Dirección de Vías de Comunicación 1928, 249); the 1940 census suggests that there were just over 19,000 Indigenous males in the entire province.²⁸ Because labor conscription repeated on an annual or biannual basis between 1920 and 1930, communities often worked on multiple projects. Araujo and Paulino (1991) estimates that, on average, eligible conscripts worked between 30 and 40 days a year over the span of the 10-year program (50). A community in the province of Canta was forced to work on four different bridge-building projects between February 1929 and April 1930 (Meza Bazán 1999, 149). As I argue in the next section, these prolonged experiences of exploitation were met with resistance by Indigenous communities and their leaders; this collective action provided the basis for future community mobilization to demand accommodation from the state.

Argument

Organization of collective action has been a key historical responsibility of community leaders,²⁹ but not all leaders have been able to mobilize their communities to make demands on the state. For example, a 1969 survey of community leaders in three Peruvian departments—Cusco, Junín, and Pasco—found that 36% felt they could not influence policy, and only 12% reported they could influence policy without outside help (Handelman 2014, 272). Similarly, nearly half of community leaders said that they and their communities could “never” exert pressure on the government (Handelman 2014, 272). In a 2017 survey I conducted in Cusco, over two-thirds of Indigenous community presidents said that they were “not at all” sure their community could obtain government assistance when needed. Explaining why some community leaders can mobilize their members to make demands on the state while others cannot is essential to explaining variation in accommodation.

I argue that labor conscription—and especially communities’ response to it—increased the likelihood of future community mobilization to demand and achieve accommodation (Figure 2). In a first stage, labor conscription generated collective resistance by Indigenous communities. Community leaders had incentives to mobilize their members to resist. Whereas previous instances of extraction relied on cooperation from Indigenous leaders,³⁰ conscription bypassed these leaders and thereby challenged a central domain of their authority: their ability to mobilize community members’ unpaid labor. Conscription likewise increased community members’ incentives to participate in this resistance.³¹ Tilly (1978) argues, “[A]ny mobilization at all is more costly to the poor and powerless; only a threat to the little they have is likely to move them to mobilize” (75). For Indigenous community members, who tended to be very poor, household income vanished during periods of conscription and sometimes after—when a worker was maimed or killed. Because the cost of inaction was high, community members were more likely to engage in defensive mobilization against conscription.³²

Figure 2. Summary of Argument

Collective resistance to conscription changed communities in three enduring ways. First, it redefined power relations within Indigenous communities. Historically, communities had been organized in a relatively horizontal way with community leaders rotating from among the membership base. However, labor conscription encouraged leaders to take on new and more powerful roles within their communities as organizers of collective resistance. Second, Indigenous leaders with a demonstrated capacity to organize collective action—in this case, through resisting conscription—were particularly attractive partners for labor and Indigenous organizations. Ties to these organizations played an important role in facilitating future mobilization.³³ Finally, conscription—and, importantly, resistance to it—generated an enduring collective memory and sustained grievance around which communities could mobilize. Rappaport (1990) observes that “[k]nowledge of the past is $ \dots $ central to efforts at strengthening a communal identity, indispensable in the maintenance of autonomy in the face of European domination” (11–2). Importantly, then, collective memory can facilitate Indigenous mobilization and protest, actions which themselves reinforce collective memory and more broadly, an Indigenous consciousness.³⁴ Furthermore, past resistance to the government—and the memory of it—can provide an important model for future mobilization against the state (Tarrow 2011, 29–30).

Together, these changes provided Indigenous communities the tools to achieve accommodation. Incumbents at the subnational and national levels have often resisted accommodation. State officials have long sought to establish relatively direct control over peripheral areas, where state-led governance and security have been historically weak (Mazzuca 2021; O’Donnell 1993; Soifer 2015). While offering accommodation to locally legitimate Indigenous leaders may promote order and good governance,³⁵ it is generally viewed as a second-best option for incumbents—who have traditionally preferred more direct forms of rule that might be more easily achieved through assimilation. As such, even when national governments have adopted policies that offer accommodation, they have often simultaneously or subsequently erected financial and bureaucratic barriers that result in an uneven and generally low level of implementation by individual communities. Decade-long wait lists, prohibitively expensive registration fees, and impenetrable bureaucratic paperwork have all hindered communities’ achievement of accommodation.

The three mechanisms of persistence that arise from resistance to conscription can help Indigenous community leaders overcome these barriers. Indigenous leaders who have more power within their communities can compel members to contribute time and money to lobby for and fund accommodation. In cases where governments refuse to recognize accommodation despite legal commitments to do so, leaders can draw on the collective memory of conscription to organize protests and marches for accommodation—demanding the government atone for past wrongs.³⁶ Memory of these grievances can also encourage communities to invest in preserving historical documents (e.g., colonial-era land titles and nineteenth-century legal judgments) that are essential to community survival and provide a documentary basis for Indigenous rights claims (Medrano 2011, 289). Finally, Indigenous leaders can draw on their ties to Indigenous organizations to obtain financial and legal resources necessary to achieve and maintain accommodation.³⁷

Once achieved, accommodation presents a mutually reinforcing outcome.³⁸ The recognition of traditional Indigenous leaders increases the legitimacy and power of their position, reinforcing their ability to mobilize collective action and maintain accommodation. The titling of Indigenous communal land has similar effects on leaders’ authority. Indigenous elites mediate access to collectively held land and thus can compel community members’ compliance with their dictates: members who do not participate in collective action, for example, risk losing their access to communal land.³⁹ As Yashar (2005) argues, “The legal registration of communities and granting of community-based property created a legally defined, state-sanctioned, geographic area that allowed for the growth and/or maintenance of politically autonomous local enclaves, Indigenous culture, and political practices” (63).

The theory above highlights a key long-term implication of labor conscription—that conscripted communities have been more likely to achieve accommodation. Other experiences with extraction, such as the expansion of large estates, might have triggered similar reactions, but no extractive event in the twentieth century was as systematic, enduring, and widespread among Indigenous communities as labor conscription. Furthermore, unlike many other forms of extraction, Indigenous elites were not coopted into conscription and therefore played an important role in organizing responses to it; this collective mobilization—and the memory of it—increased the likelihood that communities achieved accommodation. Indigenous communities that did not face conscription were less able to pressure state officials to implement accommodation, increasing the likelihood that they experienced assimilation. The theory thus not only explains variation in community mobilization to demand rights, but also highlights how Indigenous communities can shape the deployment of state power in peripheral areas through accommodation and assimilation.

IDENTIFICATION AND ESTIMATION STRATEGY: A NATURAL EXPERIMENT

Testing the above theory presents a key empirical challenge: national governments may target labor conscription toward communities with higher levels of preexisting collective action capacity and stronger institutions—either to weaken these communities or to use their institutions to facilitate conscription.⁴⁰ Furthermore, if communities are drafted based on their proximity to a road or railway, the treatment may be bundled in a way that makes it difficult to isolate the effect of labor conscription. As such, subnational comparisons across communities may be confounded.

To address this problem and causally identify the long-term effects of labor conscription, I analyze a novel natural experiment in which Peruvian Indigenous communities’ exposure to labor conscription varies as-if randomly. According to the Road Conscription Law, laborers were required to work on only the sections of road that were built in the province where they lived. As the law (Ley 4113) states, “Conscripts will be taken, except in exceptional cases, from the same [municipality], and cannot be taken from one province to work in another.” This requirement served a logistical purpose: transporting Indigenous workers to distant work sites was costly and, thus, government officials preferred to keep workers in their own municipality—and certainly in their own province. Furthermore, juntas viales were in charge of labor mobilization and their jurisdiction stopped at provincial borders. There was also a political logic to the requirement that laborers not be taken to other provinces. Provincial officials and the landlords they represented were protective of Indigenous communities under their jurisdiction because “if [they] were out building highways for the president, they were not toiling in their landlords’ fields” (Heilman 2010a, 66). Documents from the period suggest that community members were not recruited to work in provinces other than their own (Dunn 1925, 79; International Labour Office 1929, 137–8).

My empirical strategy leverages this selection procedure for labor and specifically, a geographic RDD.⁴¹ “Treatment” is defined as eligibility for the road conscription program, which is determined by a community’s location vis-à-vis provincial borders. Because labor was conscripted from only the provinces where the road was built, clusters of municipalities within a given province were assigned to treatment (i.e., subject to the Leguía draft) or control. I use the Qhapaq Ñan route—rather than the number of conscripts or Leguía’s road—to code treated provinces for two reasons. First, all communities in provinces with a portion of the Qhapaq Ñan—that is, which Leguía used as his guide for the road—were eligible for conscription.⁴² Because most communities that were eligible for conscription ultimately worked, the effect of eligibility should be similar to and perhaps a conservative estimate of the effect of experiencing conscription. Second, data on road conscription in Peru are only partially available. I thus conduct an intent-to-treat analysis, which provides an estimate of the effect of eligibility for labor conscription.⁴³ I define the study group as provinces located in the mountainous Andean region that neighbor or are themselves a Qhapaq Ñan province (Figure 3).⁴⁴

Figure 3. Qhapaq Ñan Provinces Study Group (1940 Borders)

Note: The map depicts the central sierra route of the Qhapaq Ñan along with adjacent “control” provinces in the mountainous sierra. White provinces are not included in the study group.

The running variable is a municipality’s distance to a border dividing a Qhapaq Ñan province from a non-Qhapaq Ñan province—not distance to the road itself.⁴⁵ As a result, although the effects capture conscription, they do not necessarily indicate broader effects of exposure to the Qhapaq Ñan or later roads.⁴⁶ Often, in fact, communities located near a treat-control border were quite distant from the road, even those located in Qhapaq Ñan provinces.⁴⁷

Outcomes

The main dependent variable in my analysis is the omnibus measure of accommodation described in the first section of the paper. I provide summary statistics for the omnibus measure and its component items in Table 2.⁴⁸ I also analyze the four component items of the omnibus measure as separate outcomes in the Supplementary Material.

Table 2. Summary Statistics: Accommodation Measures (Study Group Only)

To analyze the near-term mechanism, I examine Indigenous mobilizations using data from Kammann (1982), Kapsoli (1982), and Kapsoli and Reátegui (1987). I supplement these secondary sources with primary data from bulletins issued by the Section on Indigenous Affairs (1922–30), available in the Peruvian National Library. This dataset thus includes all documented Indigenous movements between 1920 and 1930—the time frame for which data are available.⁴⁹ Using this dataset, I link community names with their municipalities—as of 1920—and use as an outcome the number of Indigenous mobilizations in a given municipality.⁵⁰ I include any mobilizational event by Indigenous communities that explicitly targeted abuses by state officials during the period of conscription. I examine the enduring mechanisms of reproduction using qualitative evidence from historical primary and secondary sources.

Identifying Assumptions

The key identifying assumption under the RDD approach requires the continuity of potential outcomes across provincial borders separating treated units from control units. In other words, any observed or unobserved confounders should not in expectation vary discontinuously across the treat-control boundary. While not directly testable, case knowledge can help evaluate the plausibility of this assumption.⁵¹ For any single pair of neighboring provinces, there may have been pretreatment differences between Indigenous communities on either side of the provincial border resulting from how the specific border was drawn or because of historical differences in provincial-level administration.⁵² However, this presents a threat to inference only if these pretreatment differences are systematically correlated with the presence of the Qhapaq Ñan. There is no reason to expect this to be the case because the location of the Qhapaq Ñan was not well known at the time that provincial borders were assigned.⁵³ Provincial boundaries were fluid until 1850, nearly three centuries after much of the Qhapaq Ñan was destroyed through wars between the Spanish and Indigenous groups; the road was never rebuilt and the parts that had not been destroyed quickly fell into disuse (Esquivel 2013, 34–5). Furthermore, the administrative boundaries drawn during Spanish colonial rule, some of which became the borders of modern provinces, were “imprecise” and “vague” with borders running “along the peaks of mountains, in stream beds, or in relatively unpopulated stretches of the high puna” (Cook 2003, 415–6).⁵⁴ Thus, communities on either side of a treat-control border should not—in expectation—exhibit discontinuous changes in key baseline covariates.⁵⁵

To test this assumption, I perform a series of balance tests. I analyze six municipal-level covariates for which pretreatment data are available. These include measures of ethnic and economic composition, taken from an 1876 census: number of haciendas, total population, number of Indigenous communities, rural population, and Indigenous population. I also analyze the percentage of residents of each municipality with primary education, which is taken from a 1902 education census. Balance tests show that municipalities on either side of the border did not exhibit significant pretreatment differences on potentially important covariates.⁵⁶ Sorting may also pose a threat to inference if municipalities select into or out of Qhapaq Ñan provinces, but there is no evidence of this from a conditional density (McCrary) test ( $ p=0.2 $ ).⁵⁷

A second assumption requires that a municipality’s potential outcomes depend only on whether it is in a Qhapaq Ñan province and not whether other municipalities are “assigned” to a Qhapaq Ñan province. The particular concern in this case is that one municipality’s exposure to labor conscription may generate mobilization that spills over to other communities that are not exposed to labor conscription. As I discuss in the next section, I find no evidence of this.⁵⁸ Individual movement out of communities would not represent a concern for spillovers as individuals must be born in a community to achieve membership. Any movement out of a community is thus part of the treatment but not a threat to the noninterference assumption as individuals cannot easily relocate from one community to another.⁵⁹

A further concern involves bundling of the treatment; those provincial officials exposed to the treatment—and that have access to Leguía’s roads—may have adopted different policies toward Indigenous communities than governments in control provinces. However, in the Peruvian case, provincial governments control neither recognition nor Indigenous policy more generally, making this unlikely to account for the differences I observe for accommodation and assimilation.⁶⁰

THE EFFECTS OF LABOR CONSCRIPTION ON ACCOMMODATION

In this section, I use both the RDD and qualitative evidence to trace the long-term effects of labor conscription in Peru. I begin with an analysis of how labor conscription shaped Indigenous leaders’ ability to mobilize collective resistance against the state. I then use a process-tracing approach to examine whether labor conscription increased the likelihood that communities achieved accommodation.

The historical record suggests that Leguía’s road-building program posed a clear threat to Indigenous communities and, as such, it provoked a defensive response from Indigenous community leaders. Heilman (2010b) documents how Indigenous leaders in Ayacucho developed quasi-military organizations to violently resist conscription. Throughout central and southern Peru, Indigenous leaders organized uprisings to oppose the road conscription program (Hirsch 2010, 265; Mayer 1995, 280). When Leguía was ousted in 1930, “hatred of the [road conscription] law erupted into violence” as Indigenous communities throughout Peru “sacked government offices and burned conscription files” (Davies 1974, 85).⁶¹ Between 1922 and 1930, there were 837 documented Indigenous movements, 213 of which can be directly linked to abuses by local authorities and the road conscription program (Kapsoli 1982, 61).⁶²

To test for a causal relationship between conscription and increased Indigenous mobilization, I use the RDD and a database of municipal-level Indigenous movements against local officials that occurred between 1920 and 1930.⁶³ Importantly, for noninterference concerns, there is no evidence of cross-provincial mobilization in response to conscription.⁶⁴

Figure 4 demonstrates that labor conscription increased the likelihood of a municipality having an Indigenous movement by about 30 percentage points ( $ \approx $ 0.75 standard deviations, or sds).⁶⁵ Importantly, because these data are collected from the period during which conscription occurred, it also suggests that conscription was indeed felt by eligible communities.⁶⁶ In fact, 31 Indigenous movements during this period targeted explicitly the Road Conscription Program (Kapsoli 1982, 61).⁶⁷

Figure 4. The Effect of Labor Conscription on Community Mobilization against State (1920–30)

Note: Point estimates taken from a local-linear regression-discontinuity analysis. Ninety percent and ninety-five percent confidence intervals plotted. The dependent variable is whether a municipality experienced Indigenous mobilization against local officials between 1920 and 1930. Running variable is the municipality’s distance from a border dividing a treated (i.e., Qhapaq Ñan) province from a control one. SEs clustered at the province level. P-values adjusted for multiple comparisons using a Benjamini–Hochberg procedure (two outcome variables). Bias-corrected estimates include robust confidence intervals. $ N=607 $ . Supplementary Tables S7 and S8 contain further information. Source: Kammann (1982), Kapsoli (1982), Kapsoli and Reátegui (1987), and Boletínes de Asuntos Indígenas (1922–1930). $ {}^{*}p<0.1 $ ; $ {}^{**}p<0.05 $ ; $ {}^{***}p<0.01 $ .

The act of mobilizing to resist conscription fundamentally redefined power relations within Peruvian Indigenous communities, increasing leaders’ ability to organize collective action among their members. Traditionally, community leaders occupied a position of relative equality with other community members. Resistance to road conscription, however, allowed these leaders to take on new and more powerful roles. In 1920s Peru, some Indigenous community leaders reclaimed precolonial titles of Inca, the Quechua word for “nobleman” or “king,” whereas others adopted the military title of colonel (Heilman 2010b; Kapsoli and Reátegui 1987). Indigenous peasants eagerly mobilized behind these leaders “because of their profound anger at official government authorities and their agents” (Heilman 2010b, 501). Remarking on Indigenous communities in Arequipa during the period of labor conscription, Salcedo (1921) observes that members were “willing to fight $ \dots $ against enemies of the community at the orders of leaders” (50). Indigenous leaders capitalized on this willingness of their members to act collectively. They formed military branches to defend their communities against road conscription and established “brigades and local cell groups to foment public disorder” (Mayer 1995, 305).⁶⁸ Leaders deployed this collective mobilization to burn villages, assassinate junta vial officials, and seize local government offices.⁶⁹

Conscription—and the mobilization against it—also led leaders to join Indigenous organizations, which promised to aid communities in opposing conscription. The most notable Indigenous organization in the 1920s was the national-level Tahuantinsuyo Pro-Indigenous Rights Committee (CPIT). Subcommittees of this national organization emerged throughout Peru and organized community resistance to road conscription along with making demands for the government to respect Indigenous institutions (de la Cadena 2000, 82–99).⁷⁰ Of the 50 subcommittees established in Peru between 1922 and 1927, over 80% were located in conscription-eligible provinces.⁷¹ Many of these efforts were locally initiated; de la Cadena (2000) documents a contemporaneous account of the spread of these subcommittees: “The best men among the Indians are organizing and leading Indigenous societies (Sociedades Indigenas) in spite of their relative lack of literacy” (94).⁷² While these organizations were almost fully disbanded and repressed by Leguía in 1927, they provided important socialization experiences for Indigenous leaders, who subsequently became key officials within the influential Peruvian Communist Party and the populist Aprista party (Arroyo 2004, 206). Within these non-Indigenous organizations, leaders were not silenced but rather continued to advocate for community recognition. Wilson (2018) observes, “[CPIT] was put down by presidential decree. However, in the Andean provinces, popular opposition to the traditional political order did not disappear but rather increased its reach and brought about a surge of $ \dots $ Indigeneity” (156–7, translation mine).

This leads to the final mechanism of interest: collective memory. After the brutal repression of CPIT in 1927, community members feared discussing openly their experiences with the Committee and resistance to conscription (Heilman 2010a, 59). Indigenous leaders, however, took advantage of their newly acquired literacy, which had been advocated by CPIT, to write down their experiences (Wilson 2018). As the political system opened, stories of conscription and resistance were passed down through stories, song, dance, and pilgrimages to work sites.⁷³ The sustained memory of conscription and of a community’s resistance to it became markers of local identity for some communities, facilitating enduring collective mobilization, especially against the state.⁷⁴

The presence of these mechanisms aided communities in overcoming barriers to achieving accommodation. Communal land, for example, was available to all Indigenous communities from 1920 onward, but several barriers existed to achieving titles to collectively held land. Community leaders first had to convince their members to engage in the long and costly process of obtaining a title. After mobilizing the support of their community members, community leaders needed to secure bureaucratic, legal, and financial resources. Lawyers and researchers had to be hired to prove a community’s residence on its land from time immemorial, surveyors had to be contracted to define the borders of a community, fees had to be paid to the government, and in some cases, translators were needed to help community leaders navigate Spanish-language legal documents. Even after all documents had been submitted to the government, communities frequently had to pressure government agencies and officials to process their title; delays of years and decades were common. Finally, communities often had to mobilize collectively to ensure that governments enforced protections for their communal land titles; this could involve protracted court cases, visits to government agencies in Lima, and collective marches or demonstrations to demand government protection of communal lands.

The legacies arising from mobilization against conscription increased the likelihood that communities achieved accommodation. Leaders who had assumed more power in the wake of conscription were more likely to convince their members to pursue a communal land title. In some cases, this collective action was mobilized by Indigenous leaders to recover communal lands lost to large estates, especially through land invasions.⁷⁵ Ties to Indigenous and labor organizations, as well as Left parties, then provided Indigenous communities the resources to navigate bureaucratic and administrative procedures required to obtain a title to this land. These organizations “provided strategy and tactics for the seizure of the land, created networks $ \dots $ and developed an effective repertoire of political activism in the countryside” (Paredes 2014, 230, translation mine).⁷⁶ Finally, the collective memory of conscription facilitated later mobilization and lobbying to pressure governments to process and approve community applications for a title. Paredes (2014) argues that “the fight for land was part of the collective memory of grievances and the political culture of Indigenous communities” (231, translation mine) and that mobilization for land rights was especially common among “Indigenous peasants that had a very vivid memory of traditional exploitation” (228–9, fn. 21, translation mine).

Returning to the RDD, communities that faced road conscription have been, in fact, more likely to achieve accommodation. Using the omnibus measure, I find strong, significant effects of labor conscription, which are in the hypothesized direction (Figure 5). Communities that were eligible for labor conscription achieve higher scores ( $ \approx $ 0.3 items, or $ \approx $ 0.4 sds) on the omnibus measure of accommodation ( $ p<0.001 $ ). Analyses of the individual components of the index also suggest strong, significant, and positive results. Conscription for Leguía’s road increases the number of Indigenous institutions a community reports preserving by about 0.3 items on the seven-point scale ( $ \approx $ 0.25 sds, $ p<0.01 $ ).⁷⁷ Exposure to road conscription under Leguía generates a 12 percentage point increase in the likelihood communities have a communal land title ( $ \approx $ 0.3 sds, $ p<0.01 $ ) and a 9 percentage point ( $ \approx $ 0.3 sds, $ p<0.01 $ ) increase in the likelihood communities are registered with the government.⁷⁸ Conscription does not increase bilingualism, suggesting that its effects operate primarily on the domains over which Indigenous leaders have more direct control: community recognition, communal land titles, and traditional institutions.⁷⁹

Figure 5. The Effect of Labor Conscription on Accommodation (Omnibus Measure)

Note: Ninety percent and ninety-five percent confidence intervals plotted. The dependent variable is an omnibus measure of accommodation that considers whether a community (1) maintains Indigenous institutions, (2) is bilingual, (3) has recognition, and (4) has a completed communal land title. Values range from 0 to 4. See Figure 4 for specifications. $ N=2,583 $ . Supplementary Tables S7 and S8 contain further information. Source: INEI (2014). $ {}^{*}p<0.1 $ ; $ {}^{**}p<0.05 $ ; $ {}^{***}p<0.01 $ .

In tracing the effects of labor conscription, I specified and evaluated two alternative mechanisms that could account for the observed positive relationship between conscription and accommodation. I first explored whether Indigenous leaders collaborated with the government to mobilize workers for conscription and then leveraged these ties to achieve later accommodation. I found no evidence of this. In fact, Leguía abolished the position of the community leader in 1921 in an effort to delegitimize certain traditional authorities (Davies 1974, 72). I also investigated whether the Tahuantinsuyo Committee, which Leguía supported early in his administration, was used to establish a paternalistic relationship between the Peruvian government and Indigenous communities. If this were true, cooptation, rather than collective mobilization, might account for why conscripted communities achieved accommodation. Contrary to this proposition, the Committee appears to have functioned autonomously; it openly opposed the government after 1924 and was ultimately banned and repressed by Leguía (Arroyo 2004).

I analyzed three further explanations that might account for the positive relationship between labor conscription and accommodation. First, governments could have offered Indigenous communities accommodation in exchange for their participation in road conscription—or to atone for past wrongs, absent community pressure. If this were the case, affected communities should have been the first to receive recognition. However, the data instead suggest that road conscription led to much later community recognition.⁸⁰ Second, communities eligible to work on the road may have also been more likely to have been exposed to the road network itself and thus have experienced greater market integration; connections to external actors—and increased wealth—could have facilitated accommodation. However, there is no evidence of this effect.⁸¹ In fact, a key criticism of the conscription law during debates over its adoption was that Indigenous groups would be responsible for building the road but—because of their isolation—would generally not benefit from it (de la Cadena 2000, 96). Finally, communities located in a Qhapaq Ñan province may have been located closer to the Inca Road and thus had stronger Indigenous institutions and identities prior to conscription.⁸² I find that the communities for which an effect is identified—that is, those close to a treat-control provincial border—are, on average, quite distant from the Qhapaq Ñan. The average straight-line distance between a treat-control provincial border—using the point on the border closest to the Qhapaq Ñan—and the Qhapaq Ñan is 28 km, a significant distance in the mountainous terrain of the Andes.⁸³

Where conscription did not occur, Indigenous institutions were less likely to survive. By way of example, Carhuanca, a municipality in the department of Ayacucho, did not experience labor conscription.⁸⁴ Carhuanca’s traditional leaders have experienced a reduction in their authority and community members’ memories of these officials are largely negative (Heilman 2010a, 103). Carhuaquinos are now more likely to make demands “in class rather than ethnic terms,” suggesting—consistent with my theory—a low likelihood of accommodation and a greater probability of assimilation (Heilman 2010a, 70).

DISCUSSION

Moving beyond Peru, Indigenous communities in other cases in Andean South America likewise met labor conscription with collective resistance. In the Bolivian province of Cinti, for example, the president was forced to deploy national police to “enforce the conscription levies” and quell the “strident resistance by Indians to $ \dots $ [forced] road construction” (Kohl 2020, 16). In the wake of particularly severe labor exactions in Chayanta, Bolivia, for example, an Indigenous leader “took the title of heir of the Incas, stirring up the Indians with promises to $ \dots $ reestablish the rules of his race in all branches of public administration” (Platt 1987, 309). Similarly, in late nineteenth-century Ecuador, “Indigenous communities rebelled against $ \dots $ intolerable state demands” the most common of which was labor conscription (O’Connor 2007, 79).⁸⁵ The largest such uprising occurred in Chimborazo, where “thousands of Indian men and their wives rose up against white-mestizo society,” killing local officials and burning down villages (O’Connor 2007, 79). Indigenous rebellion in these cases generally did not achieve its intended goal of stopping conscription. Those convicted of participating in the Chimborazo rebellion, for example, were sentenced to labor on public roads, which “promoted [President] García Moreno’s agenda for building up national infrastructure while punishing [rebels] with one of the burdens against which they had rebelled” (O’Connor 2007, 81).

Organizing these acts of collective resistance, however, empowered Indigenous elites in new ways that endured beyond the period of conscription. As in Peru, the process of organizing collective resistance led Bolivian Indigenous leaders to reestablish a precolonial institution—the cacique (chief), an enduring, community-level political office with substantial authority (Gotkowitz 2008, 87; Irurozqui 2000, 105). These Indigenous leaders also used conscription to establish linkages with nascent Indigenous organizations like the Oldest Autonomous Mayors (AMP), a group that argued that “Indians could no longer be subjected to ‘forced’ labor” (Ari 2014, 102). As in Peru, these Indigenous organizations enabled communities to make demands for “specific ethnic programs” (Platt 1987, 304). Irurozqui (2000) analyzes the lasting “impact the rebellions [of the early twentieth century] had on Indian national and political consciousness,” creating an Indigenous identity that facilitated long-term mobilization for ethnic rights (87). Areas of Bolivia that were deemed exempt from labor conscription, such as the Cochabamba region, have been less likely to demand accommodation than those in the neighboring departments of La Paz and Potosí, which experienced much higher levels of labor conscription.⁸⁶

Similar effects of labor conscription can be observed outside the Andean region. Smith (2020) demonstrates that in several Mexican states,⁸⁷ “Opposition to the use of communal labor for road building was extremely widespread” (289). In some cases, Indigenous leaders used this resistance to achieve government recognition of deliberative assemblies that mirrored Indigenous institutions (Smith 2020, 291–2).

In addition to conscripting Indigenous labor to build infrastructure, Latin American states also compelled Indigenous men to serve in the military with notably similar effects. In Peru, members of native communities constituted the vast majority of servicemen at the turn of the twentieth century (Forment 2013, 174–5). With soldiers often failing to understand why they were fighting and returning home with severe injuries, traditional Indigenous leaders mobilized their communities to resist military conscription (Hunefeldt 2018, 377–84). Even though this mobilization was violently repressed, it generated enduring collective action (Thurner 1997, 92). Communities that resisted military conscription became future “centers of rebel activity” (Mallon 1995, 238). In Bolivia, Indigenous soldiers used their status as veterans to make demands on the state, “claim[ing] credibility and authority based on their personal participation in the war and the collective contribution of the many Indigenous men who had fought on the front lines” (Shesko 2020, 128). These leaders received attention from national politicians and the press. Shesko (2020) observes, “Unlike in the prewar period, the mainstream press in 1945 was willing to present Chipana Ramos [an Indigenous veteran] as speaking with authority and making claims for Indians’ place in the nation as ‘sons of the same soil’” (126). Like road conscription, military service—albeit through perhaps distinct mechanisms—appears to have reinforced Indigenous leaders’ ability to demand rights for their communities.

CONCLUSION

In 1926, José Mariátegui, a noted Peruvian intellectual, wrote, “[T]he resistance to road conscription leaves no room for doubt about public sentiment regarding this service” (Mariátegui 1926, translation mine). Using as-if random variation in exposure to the road conscription program, this paper has demonstrated that—consistent with Mariátegui’s claim—conscription triggered collective resistance by Indigenous communities. Expanding beyond the 1920s, this paper has also shown that these investments in collective mobilization endured in ways that later enabled communities to achieve accommodation of their long-standing institutions.

Labor conscription is, however, just one form of extraction. In the Supplementary Material, I develop a typology of extractive activity in colonial and postindependence Latin America.⁸⁸ I consider Indigenous head taxes, debt peonage, and the notorious mining mita.

Labor conscription differs from these other forms of extraction in two key ways, which make it particularly likely to generate the effects observed in this paper. First, it provided a substantial disruption to Indigenous communities, but did not destroy them. In other words, it constituted a clear harm to communities’ social and economic well-being, increasing the likelihood of collective resistance.⁸⁹ Second, labor conscription did not coopt Indigenous elites and thus differed from the Indian tribute (Platt 1982, 43), the notorious mining mita (Dell 2010, 1877), and many instances of rural elite extraction (e.g., debt peonage, the encomienda system). By failing to incorporate community leaders, the central state ultimately encouraged a sustained resistance by these leaders, who used their renewed collective action capacity to lobby for protections for the traditional institutions over which they presided.

Achieving this accommodation has had important implications for communities in contemporary Peru.⁹⁰ In 2016, I visited the province of Abancay and spoke with a former president of a local community. The man complained about a mining company, which had operated near the community “for about 30 years” and had encroached on the community’s water supply. He told me, “Our leaders in the past failed us. They did not oppose the company $ \dots $ What can we do now? Join the mining company [as employees] or leave the community or both.” When I asked the man to clarify which leaders had failed his community, he replied, “El gobierno [the government].”⁹¹ This former president highlights how accommodation—meaningfully enforced land titles and other government protection (e.g., prior consultation)—can improve community welfare by reducing resource-related conflicts.⁹²

To be sure, accommodation has not offset the cost Indigenous communities incurred from conscription, which included a decline in material wealth, loss of human life, and enduring trauma. The findings, however, highlight how groups can respond to government exploitation and abuse by investing in their collective action capacity—what Finkel (2015) labels the “phoenix effect of state repression.”

From the vantage point of the state, the phoenix effect of labor conscription provides a cautionary tale. In their efforts to bypass and weaken Indigenous authorities, central state incumbents unintentionally sowed the seeds of future ethnic mobilization. Far from solidifying the state’s monopoly of control over peripheral areas, labor conscription encouraged communities to reinvest in Indigenous authority and strengthen their long-standing—and resilient—political, economic, and social institutions.