Forced Displacement and State Building

State-led population movements, both voluntary and coerced, have long been a prominent tool of state building. There is much research on such movements, although the political context of resettlement varies considerably across cases. Internal displacements are often a result of traumatic events like civil wars,Footnote ⁷ territorial conflicts,Footnote ⁸ increasing security competition with a neighboring state,Footnote ⁹ or international agreements that conclude large multiparty wars.Footnote ¹⁰ Regardless of the specific context, most state-promoted resettlements aim to increase territorial control.

Governments often try to remove a minority population from a region via displacement. A common aim is to produce more ethnic homogeneity in the targeted region(s) by sorting people of different ethnicities, religions, or other relevant social characteristics into different geographic areas. For example, following the Turkish War of Independence, Greece and Turkey agreed to exchange populations between their newly delineated territories, where the goal was to eliminate sizable Christian populations in the new Turkish state and sizable Muslim populations in the Greek state. The transfer of people within Poland and Germany to accommodate the new (and significantly repositioned) post-World War II boundaries was also aimed at cultural homogeneity. State-led resettlements that are meant to homogenize different jurisdictions tend to be forced and aim to facilitate shared national identity and the central state’s territorial control.

While such attempts at greater homogeneity are consistent with common post-eighteenth-century state-building strategies, resettlement results in more heterogeneity when governments resettle their coethnics in regions populated by other groups. Such resettlement policies have the goal of increasing territorial control by placing putatively loyal coethnics among a local population seen as lacking strong ties to the state. They tend to target peripheral or border regions, and also commonly involve at least some voluntary migration.

The outbreak of civil conflict between the newcomers and the local population is documented in many such cases, as highlighted by the “sons of the soil” idea.Footnote ¹¹ Fearon and Laitin show that around a third of post-World War II ethnic civil wars are “sons of the soil” conflicts, where migrants ethnically tied to the state become embroiled in escalating conflicts with their new neighbors.Footnote ¹² They note that in many cases, the displaced people become staunchly loyal to the government because they depend on the state for their current land and also often enjoy resource transfers from the center. With a focus on Rwanda, McNamee makes this case for the resettlement of Hutus to border areas traditionally dominated by Tutsis, highlighting the importance of acquired land as an immobile asset.Footnote ¹³ In short, if the state with the displaced population’s ethnic kin at the helm loses power, the risk that the migrants lose their land significantly increases. This line of reasoning is consistent with that of the Iron Emir, who sent Pashtun kin to the strategically important north as a bulwark against a non-Pashtun population deemed “disloyal.”

Such population movements do not always result in severe intergroup conflict or civil war. While the “sons of the soil” cases involve placing loyal coethnics among putatively disloyal groups, states often promote the settlement of areas because they are seen as unpopulated or economically underdeveloped. An example is Indonesia’s transmigration program, which promoted the voluntary movement of around 2 million people between 1979 and 1988. Agricultural development and nation building were key objectives, with the latter being seen as urgent in a very diverse country where Indonesian, the official language of the central state, was spoken by a small minority of the population.Footnote ¹⁴ Bazzi and co-authors show that people relocated to diverse localities developed less exclusively local identities and more national identification, while those displaced to more homogeneous areas populated by their kin retain more exclusive identities and identify more with their ethnic group than with the national state.Footnote ¹⁵ This is consistent with research that has shown that people living in more diverse settings have more meaningful cross-ethnic tiesFootnote ¹⁶ and more tolerance of other groups.Footnote ¹⁷

The literature on border politics and border disputes also analyzes demographic engineering in border regions. The pathbreaking work of Huth highlights that the presence of cross-border ethnic minorities in contested border regions is associated with interstate territorial disputes.Footnote ¹⁸ Armed conflict is especially likely when these transnational ethnic groups are excluded from power.Footnote ¹⁹ Carter found that states with ethnic minorities with ties to the ethnic majority in a cross-border rival are associated with attempts at demographic engineering to consolidate territorial control,Footnote ²⁰ a dynamic that McNamee and Zhang show is at work in the China–USSR case.Footnote ²¹ However, the long-term implications of these demographic engineering efforts are not as clear, as most studies either examine the logic behind the strategyFootnote ²² or have a short-term or cross-national focus that provides mixed evidence on these policies’ efficacy.Footnote ²³ In one of the few exceptions, Zhang shows that China’s resettlement of ethnic Han to the west had the intended effect of quelling insurgency only at first, when the central government invested very heavily in these efforts; insurgency re-emerged after those investments waned.Footnote ²⁴ In sum, more research is needed to understand the effects of forced resettlement efforts in a contested border region, as the effects of these policies on local actors’ identities and behavior are often shaped over decades rather than months or years.Footnote ²⁵ We contend that the long-term difficulties for weak states to maintain the loyalties of settlers in the periphery over the long term are underemphasized in the literature.

State-Led Resettlement of Pashtuns in Nineteenth-Century Afghanistan

The resettlement of Pashtuns to the north from their historical base in the south and east began in the early 1880s under the transformative and brutal rule of Emir Abd al-Rahman Khan (1880–1901). Historians provide several motivations for the Iron Emir’s interest in the movement of Pashtuns to the northern provinces, and also detail how the policy was implemented in fits and starts from the early 1880s to 1896. This is a case of state-sponsored resettlement that involved a mix of forced and voluntary resettlement. One motivation was the need to consolidate territorial control in a contested border region, amid concerns about Russian encroachment.Footnote ²⁶ Abd al-Rahman viewed the settlement of his Pashtun kin along the frontier of Afghan Turkestan as essential.Footnote ²⁷

Another motivation was that the people in the north were viewed as having weak loyalty to the regime in Kabul. Thus, the emir planned to populate this strategically and economically important area with Pashtuns, who he presumed would be loyal to him and their fellow Pashtuns, particularly when surrounded by non-Pashtun populations.Footnote ²⁸ However, the emir did not rely solely on one strategy here; he also forcibly resettled Ghilzai tribes that he had conflictual relations with, and he obtained the voluntary resettlement of Durrani Pashtun tribes whose loyalty he was more confident in. In practice, these resettlements were quite haphazard and inconsistent in implementation, as detailed by Lee,Footnote ²⁹ although the emir did try to resettle Pashtuns such that tribal alliances in the south that did not favor him were disrupted. In this regard, he had the explicit intent of disrupting the social and political ties and identities of the migrants.

As that last point suggests, the primary political and social identities of Pashtun communities in the 1880s and 1890s in Afghanistan were with the kinship network and qawm. Footnote ³⁰ Thus, one goal was to use resettlement to shift these communities’ primary allegiances away from their qawms and kin and toward the central state.Footnote ³¹ In this regard the emir may have been successful, as close ties between these displaced Pashtuns and ethnic kin to the south were necessarily disrupted given the lack of infrastructure, with the western ranges of the Hindu Kush standing in the way.

While the resettlement of Pashtuns from the south and west to the north also likely had immediate effects on the Pashtuns who did not migrate, we expect these effects to be fleeting. The resettled Pashtuns had to endure the move to the north and adjust to living in a new place surrounded by non-coethnics, unlike those who did not migrate. Moreover, the Pashtuns who stayed in their traditional areas of settlement in the south and east were the dominant group at the time of displacement and remained so afterward. Thus, we argue that any effects on communities in the south and east would be muted and short lived. Accordingly, our theoretical focus is on how resettlement affects the identities of migrants, highlighting how the process and immediate consequences of resettlement interact with state policies to influence the networks and allegiances of resettled communities.

The Long-Term Effects of Resettlement

Of central interest are the conditions under which settlers will develop long-term attitudes consistent with the state-building goals of the central government. The state’s overarching objective is to develop a group of loyalists among the local population in the target region. A key obstacle to fostering pro-state loyalty are the traumas associated with the process and aftermath of resettlement. While this issue is starkest in cases of forced resettlement, it is also common in cases of voluntary state-sponsored resettlement, which is an under-appreciated point. The state can overcome the effects of trauma and foster a bastion of loyal subjects if it quickly and consistently establishes itself as a provider for and protector of resettled communities. But this is often hard because such resettlement schemes tend to target remote and difficult-to-reach border regions, where limits in capacity and large effective distances from the capital undercut sustained engagement and support. We outline our arguments with the details of the resettlement of Pashtuns to northern Afghanistan in the last two decades of the nineteenth century.Footnote ³²

The Trauma of Resettlement

Much research highlights the long-term effects of trauma and state repression on the political views and behaviors of the victims’ descendants. The intergenerational transmission of traumatic events is argued to lead victims’ descendants, decades later, to oppose political groups affiliated with the perpetratorsFootnote ³³ and to exhibit weaker identification with the state when it is culpable.Footnote ³⁴ While the idea that negative views of the state are a function of the trauma of resettlement is better recognized for forced resettlements, we argue that it is also fairly common in cases of voluntary resettlement, where the state incentivizes settlers to move.

Trauma for migrants and their families can derive from both the process of resettlement itself and hardships that migrants meet after resettlement. There is ample evidence that the resettlement of Pashtuns in the 1880s was traumatic and often included violence, hardship, and death among the settlers.Footnote ³⁵ Many of them, especially the Ghilzai, were prisoners “exiled” to the north, while as late as 1888 at least 14,000 Pashtuns were forcibly deported to areas in contemporary Herat, which was known then as Afghan Turkestan. In neighboring Badghis, both Lee and Tapper highlight the disastrous consequences of the voluntary arrival of thousands of Ishaqzai from the Durrani tribe in Badghis.Footnote ³⁶ This swift resettlement of thousands of Pashtuns from Pusht Rud in the south “wrought havoc on the local economy, and merely increased the misery and suffering that the indigenous population was already obliged to endure”; it also “proved to be a disaster for the maldars themselves.”Footnote ³⁷ These examples highlight the trauma faced by migrants after resettlement in the north. In sum, historical treatments of northern Afghanistan in the late nineteenth century lay bare the appalling brutality of the Iron Emir’s regime and the repression and hardship of resettlement in the 1880s and 1890s.

A growing literature clarifies how the trauma from state-led resettlement is transmitted across generations. Traumatic and disruptive shocks such as forced displacement can lead norms of cooperation to break down as people adopt shorter time horizons and steeper discount rates.Footnote ³⁸ Forced resettlement can also deplete wealth and assets, disrupt social networks, and undermine access to food stores and public goods,Footnote ³⁹ consequences that transmit intergenerationally via persistent social, economic, and political isolation.Footnote ⁴⁰ These events have strong influence on contemporary beliefs through mechanisms of memory-making, including oral histories passed down through generations, detailing harm and hardship. Lee effectively describes this mechanism in the Afghan case, highlighting how the trauma of forced resettlement and state repression has been transmitted across generations by descendants of the victims: “Whilst the Amir’s ‘achievements’ are constantly reworked and expanded by successive generations of academics and journalists, for the peoples of Afghanistan, it is the atrocities which dominate the folk memories which have survived from this era.”Footnote ⁴¹

A rich theoretical literature shows how the effects of disruptive historical events can be amplified and solidified over time, even in the face of contemporary incentives that reflect changes in context. Tabellini and Guiso, Sapienza, and Zingales show that the intergenerational transmission of political values from parents to children can easily lead to the long-term persistence of pessimistic ideas, as parents tend to pass down their own cultural values and want to protect their children from risk and harm.Footnote ⁴² Bin Oslan demonstrates that the long-term persistence of communities’ political differences need not rely on parental transmission but can also be sustained when intra-group interpersonal interactions are more common than inter-group interactions.Footnote ⁴³ The upshot here is that once a legacy of skepticism toward the state becomes entrenched in and passed down through communities, these narratives become increasingly difficult to alter. Accordingly, if the state hopes to produce a bastion of support in displaced communities, immediately counteracting the negative experiences of the resettlement process and its aftermath is essential.

The Ties That Bind: Resettlement and Political Identity

A possible counterbalance to these traumatic effects is a stream of support and resources from the state. Along these lines, the shorter-term loyalty of settlers in “sons of the soil” cases often hinges on state support of their communities. However, as noted, the state needs to provide support to counteract the trauma and hardship of resettlement because negative narratives are difficult to alter once they become entrenched in a community and passed down across generations.

For example, when in the 1960s Sinhalese people migrated to the Eastern Province of Sri Lanka, a region traditionally dominated by Tamils, the government swiftly set up large military installations such as the Air Force Farm to house the Sinhalese-dominated security forces. This show of force and the fact that the security forces sided with Sinhalese migrants in disputes both emboldened migrants and fostered stronger identification with the state,Footnote ⁴⁴ helping obviate the trauma of relocation to a new region. Without these investments in state security presence, the Sinhalese migrants would have likely experienced greater hardship and isolation in their new settlements. In short, this case highlights how support from the state after arrival can reduce the hostile views of the central government that often result when the aftermath of resettlement is difficult.

In our case, the resettled Pashtuns were not shielded from considerable hardship during or after moving to the north. The process of relocation, both in forced displacement and in more voluntary cases, is widely described as traumatic and poorly implemented. Moreover, the evidence suggests that resettled communities did not forge strong connections to the central government, let alone become dependent on Kabul for support. The central state’s weak administrative presence during and after Emir Abd al-Rahman’s rule was of course a significant impediment. Consequently, the displaced Pashtuns were not consistently provided with state resources that would “buy” their continued loyalty, as even the initial subsidies that some of the voluntary migrants received were small and short-lived.Footnote ⁴⁵ At the time of resettlement, subsidies were very selectively applied to specific groups, while prisoners or more coercively resettled groups usually went without. In general, the historical record suggests that after their haphazard migration process, the displaced Pashtuns did not receive significant or sustained support from the Afghan government.

The lack of central-government resources in the north did not indicate a lack of policy ideas on the part of the Iron Emir. Rather, he calculated that a sense of “Pashtunism” and ethnocentric loyalty would be fostered among the displaced, and that this would translate into greater loyalty to the central government. His reasoning hinged on two notions: that the resettled communities’ relations with their new neighbors would be conflictual; and that the migrants would view the central state as the protector of their resources and status in the targeted region.

Evidence is mixed as to whether relations between the resettled Pashtuns and their new neighbors were consistently hostile.Footnote ⁴⁶ However, that is beside the point in this case. The central government did not have a strong or consistent presence in the region; the migrants had no reason to see the Pashtun-controlled government in Kabul as a protector of their status and territorial possessions in the north. And given that the central government was not a reliable and essential protector, the displaced Pashtuns would likely only lower their opinion of and identification with the central state in the presence of significant inter-ethnic conflict.

Region and Identity in the Borderland

In cases where resettlement is traumatic and the government does not establish itself as an essential supporter of migrant communities, the local context shapes the development of settlers’ political attitudes. Regions of displacement are commonly border regions that the central government has struggled to consolidate control over. People living in border regions often exhibit political behaviors distinct from those more proximate to state power, especially where there are significant cross-border threats and limited state capacity.Footnote ⁴⁷ As Braun and Kienitz note, proximity to a border need not strengthen the “visibility of and identification with the state.”Footnote ⁴⁸ When resources from the state are scarce, communities near a border develop practices and engage with local institutions for subsistence and survival.Footnote ⁴⁹

We argue that resettled communities that are not dependent on support from the central government and that lack strong ties to their kin in traditional areas of settlement are likely to develop regional political identities centered on local institutions. Resettled people forge ties with their new neighbors and learn to work within and shape the local and regional political system. It is not necessary for relations with non-ethnic kin to be mostly harmonious or for interactions with local political actors to be uniformly positive. Rather, it is important that migrant communities forge meaningful ties with local populations and institutions, and that these ties are more important than ties to the national state.

Thrust into a new region bordering a foreign power, with ties to ethnic kin elsewhere in the country strained or severed, most resettled communities will find it beneficial to engage with local communities. It is uncommon for small communities to be fully self-sufficient economically, especially when they are new to a region and lack full knowledge of important issues, such as growing conditions or seasonal disruptions of activities such as grazing or hunting. Accordingly, resettled communities have strong incentives to engage in economic exchange with neighboring groups. Social ties between migrants and non-coethnics also become more common and meaningful, as resettled communities increasingly find it in their interest to engage with their new neighbors, and economic exchange leads to stronger social ties.Footnote ⁵⁰ This is consistent with prominent empirical research showing that exposure to ethnic heterogeneity is associated with more inter-ethnic social ties and denser social networks.Footnote ⁵¹

These economic and social ties to other ethnic groups, coupled with sparse ties to the central government and their kin in their traditional areas of settlement, significantly affect the migrants’ political behavior and identities over time. There is much evidence that individuals’ social ties affect their political attitudes and behavior.Footnote ⁵² Consistent with this, a growing literature in the social sciences finds that exposure to ethnic diversity leads to relatively positive views of out-groups.Footnote ⁵³ In cases of resettlement, exposure to different groups can also undercut narrower ethnic identities and promote broader ones.Footnote ⁵⁴

Marriage is one of the most significant social and political ties, and there is evidence in marriage patterns that resettled Pashtuns’ developed meaningful ties with non-Pashtuns.Footnote ⁵⁵ In most areas of resettlement in the north there were not enough kin to support traditional patterns of exclusive intra-ethnic marriage, which led to some Pashtun men marrying Uzbek or Tajik women. The practice initially preferred by the resettled Pashtuns was to find wives from their families in the south to marry, but this broke down due to the distance and difficulty of the move to the north. As Dupree notes in his influential treatment, “The Ghilzai who remained in the south did not want to see their daughters move away from the localized, extended family unit ... so the northern-based Ghilzai males had to look elsewhere for brides, and began to take daughters from the neighboring Uzbek and Tajik, and relations between the involved villages intensified.”Footnote ⁵⁶ In sum, this is one of numerous pieces of evidence in the historical record that the displaced Pashtuns forged meaningful connections with their non-Pashtun neighbors, which also led the strong ties to family and ethnic kin in the south, which had traditionally been bound by marriages, to dissipate over time.

The historical record also suggests that migrant communities’ interactions with regional and local governments were more substantial and positive than those with the central government. It is widely noted that Abd al-Rahman’s government struggled to exert influence over frontier areas during his reign.Footnote ⁵⁷ This is unsurprising given that Afghanistan’s northern borders were newly drawn in the late 1880s and were quite arbitrary relative to existing economic, social, and political networks. During this period one of the most significant episodes of local resistance to the emir’s repressive policies across Afghanistan came from the governor of Afghan Turkestan, Ishaq Khan. Khan had ignored some of the emir’s more draconian directives and in June of 1888 coordinated a rebellion against Kabul. Khan’s rebellion was brutally defeated but remained popular in local narratives, especially in light of the atrocities perpetrated by the central state following its suppression. The rebellion of Ishaq Khan underlines the Iron Emir’s lack of administrative capacity in the north, as the uprising was planned for almost two months without Abd al Rahman having news of it.Footnote ⁵⁸ This is notable for a regime known for its unprecedented network of informers, which inculcated a culture of fear in much of the country.Footnote ⁵⁹

In sum, resettled communities forged meaningful connections in their local environment, with little support from the central government to counteract the trauma of relocation. While relations with their new neighbors were not without conflict, they were not consistently protected by the central state, and to the degree the state did intervene, the historical record suggests it did so with a heavy hand and even brutality.Footnote ⁶⁰ We argue that this led to the development of a political and social identity centered on locality and region. This identity is more encompassing than a more exclusive identification with being Pashtun (a goal of the emir), or with the more traditional identification with one’s immediate kinship network, or qawm, such as Ishaqzai.

Data and Identification Strategy

To accurately analyze differences between the Pashtuns in Afghanistan’s northern displaced regions and those elsewhere, we need two key data sets. First, we need a detailed measurement strategy to pinpoint Pashtun village locations and to identify descendants of resettled populations. Since no data of which we are aware provide this, we develop our own strategy, which leverages a high-quality 1921 British ethnic map and an algorithm that uses detailed data on important geographic features to identify the expansion of these initial resettled Pashtun communities after the early twentieth century.

Second, we need extensive, geographically precise survey data, surpassing those used in earlier studies. Given the relatively small geographic pockets of territory with a Pashtun population in northern Afghanistan, any comparison of the political attitudes of displaced Pashtuns to those in the rest of the country requires a comprehensive survey of enough Pashtun respondents in the displaced region. Moreover, with Pashtun populations concentrated in specific northern villages, surveys retaining each village’s geolocation are crucial. In what follows, we describe data that meet these two requirements.

Expansion of Displaced Pashtuns in Northern Afghanistan

Understanding which Pashtuns in Afghanistan’s northern provinces descended from displaced groups (versus later emigrants) requires a relatively precise map. We use a detailed 1921 map produced by the British government (Figure 1) because at this point Pashtuns in the north were primarily those displaced by the regime or their descendants.

Figure 1. Initial displacement sites recorded in the 1921 British map

Note: Afghanistan Ethnic Groups, US Library of Congress, G7631.E1 2005 .U5, available at <https://www.loc.gov/resource/g7631e.ct002134/>.

To match contemporary villages to the displaced Pashtuns, we need to account for two processes of resettlement. First, the historical literature notes that the communities of displaced Pashtuns expanded their settlement areas over time, generally moving along valleys and mountain-runoff rivers that could sustain agriculture. Thus, to ensure that we correctly capture the descendants of displaced Pashtuns we need to account for this movement, which took place on foot and with animals. Second, it is documented that Pashtuns emigrated to the North after the 1930s, settling in new areas that previously did not have Pashtun settlements.

We devise a spatial method to distinguish between descendants of displaced Pashtuns and later emigrants by tracing the former’s locations and movements over a century. Specifically, we assess the suitability of land in northern Afghanistan for Pashtun expansion, assuming that Pashtuns will migrate to regions suitable for living. We divide northern AfghanistanFootnote ⁶¹ into 236,928 1 km $ \times $ 1 km grid-squares, and use a LASSO (least absolute shrinkage and selection operator) model to predict the settlement suitability of each square. The outcome variable $Y$ is distance to the nearest village (regardless of ethnicity), and the predictors $X$ are a set of variables that may affect people’s decisions to reside: elevation, relative elevation, slope degree, valley, longitude, latitude, distance to river, riverside, crop, rainfed agriculture, and seven key dimensions of soil quality (nutrient availability, nutrient retention rapacity, rooting conditions, oxygen availability to roots, excess salts, toxicity, and workability).Footnote ⁶² The advantage of LASSO over ordinary least squares is that LASSO penalizes overfitting and thus prioritizes the variables with higher predictive power.

Next, we calculate each grid-square’s likelihood of housing descendants of displaced Pashtuns. We use the pre-1921 map (Figure 1) to identify initial Pashtun settlements. The LASSO model takes into account several local features (Figure 2, panel A) and calculates a cost surface (panel B). Combining these two pieces of information, we get the cost distance for displaced Pashtuns to migrate from their core sites to any place in northern Afghanistan (panel C). Darker areas in panel C indicate a higher likelihood of hosting displaced Pashtuns, further categorized using the Jenks natural breaks optimization method, with probable areas shown in panel D.Footnote ⁶³

Figure 2. Methods used to trace the movement of displaced Pashtuns over time

Note: Each feature is calculated on a 1 km $ \times $ 1 km base. Southern Afghanistan is omitted.

We carefully validated the identified locations using extensive historical evidence from archival sources and the academic literature, drawing from what we believe are all the available resources. As detailed in the appendix, we carefully reviewed 1,374 references—narrowed down to a dozen pivotal works—that explicitly document the presence of displaced Pashtuns in specific locales. Notably, several sources describe migration patterns along major sites such as the Harirud and Kunduz Rivers, which are precisely captured by our algorithm. The historical accounts closely align with the spatial clusters we identified, providing strong support for the accuracy of our settlement measure and confirming that it reflects recorded movements and is not just model-driven.

Measurement Strategy

Having delineated the region of displacement, we can pinpoint displaced Pashtuns at an individual level via their self-identified ethnicity. According to the Foghorn survey, “Pashtun” is the predominant ethnicity, comprising just over 40 percent of all respondents, from a total of forty-four ethnicities. Out of the roughly 33,000 Pashtun respondents, about 1,800 are identified as descendants of the displaced, while a similar number of Pashtuns are not identified as descendants but reside in the same northern provinces. While most of these other potential ethnicities have limited representation, most reflect the population distribution in Afghanistan, identifying as Pashtun, Tajik, Uzbek, Turkmen, Hazara, Baloch, Nuristani, Aimak, or Arab. Figure 3 maps the geographic dispersion of Pashtuns and other ethnicities across sampled villages. Pashtuns are observed to populate the southern provinces continuously from the Iranian border in the west to the Pakistani border in the east, whereas the northern provinces bordering Turkmenistan, Uzbekistan, and Tajikistan have a mix of Pashtun settlements interspersed among non-Pashtun villages. For the MISTI survey, we adopt the same method to identify displaced Pashtuns, but only at the district level. A respondent is considered a displaced Pashtun if they self-identify as Pashtun and their district intersects with the displaced region.

Figure 3. Pashtun and non-Pashtun villages in Afghanistan

We estimate the individual-level effect of being a displaced Pashtun from the north, versus a Pashtun residing elsewhere, by regressing the outcome of interest on a series of mutually exclusive binary indicators that are the product of ethnicity and region: displaced pashtuns, non-pashtuns in displaced regions, and non-pashtuns in non-displaced regions. This makes non-displaced pashtuns the excluded category against which the estimated coefficients of these three included variables are compared. In the models that include respondents across all of Afghanistan, we include Pashtuns from northern provinces that we do not identify as descendants of the emir’s resettlement policy in this excluded category of non-displaced pashtuns. We also exclusively compare this set of non-displaced Pashtuns from the north to displaced pashtuns in analyses that include only northern provinces. We also include province fixed effects in all specifications, which condition out any fixed features of being from a specific province, northern or otherwise, in addition to including survey-wave fixed effects in all models.Footnote ⁶⁸

We also control for variables that might confound the effect of our theoretically central variables. In the Foghorn models, we first control for individual-level variables that are known to affect political attitudes: age, years of education, college-education status, employment status, community-elder status, marital status, and household income, all of which are self-reported in the survey. Second, we include an indicator of whether the individual lives in an urban area, like Kandahar or Kabul. The urban experience in Afghanistan is known to be very different from rural and village life, making this a vital control. Lastly, we adjust all our models using survey weights to correct for any sample unrepresentativeness.Footnote ⁶⁹ In the MISTI models, similarly, we control for gender, age, employment status, head-of-household status, marital status, and household income.

Armed with both the Foghorn data and our measure of displaced Pashtun communities, we can provide a systematic assessment of the common claim that displaced pashtuns are relatively well-off. This claim is often tied to the idea that displaced pashtuns are more supportive of the central government because they have economically benefited from their resettlement to the north over the long term. We infer each individual’s economic status with a measure that uses principal component analysis to combine responses on per capita household income, food security, and economically significant assets such as running water, electricity, and television. We find no significant differences between displaced pashtuns and their non-displaced coethnics (the excluded category) or their non-Pashtun neighbors (Figure 4). These results hardly support the idea that the Pashtuns who live in the areas of those displaced and resettled in the late nineteenth century are significantly better off than their non-Pashtun neighbors.

Figure 4. Pashtun and non-Pashtun economic status across Afghanistan

Notes: The panel head is the dependent variable derived from survey questions. We display only the estimated coefficients for the three population groups, compared to the baseline group, non-displaced pashtuns. The full tables encompassing all the control variables can be found in the Appendix. Robust standard errors are clustered by district and survey waves to be as conservative as possible. Survey weights are included to ensure representativeness of the survey sample to the underlying population.

Are the Central Government and Its Allies Viewed as Effective?

We now turn to analysis of individual respondents’ views of and engagement with key political actors and institutions. Our primary interest is to estimate the effects of being a Pashtun descendant of those resettled by the emir. Models presented here contrast Pashtuns in the displaced region, as well as non-Pashtuns, with Pashtuns who are not identified as descendants of those displaced—that is, mostly Pashtuns in the south and east. This approach enables us to assess whether displaced Pashtuns have significantly different attitudes relative to their Pashtun kin or non-Pashtuns in the same region. These surveys were fielded in 2017 and 2018, which means that all of these responses were recorded when the Ghani government was still in power and the Taliban’s territorial control in urban centers was limited.

Our first set of models focus on respondents’ views of the effectiveness of three key governing actors: the provincial council, the central government, and the Ulama councils—as well as their views of their communities’ relations with the local government. A higher value indicates higher perceived effectiveness or superior relations.Footnote ⁷¹ Figure 6 visualizes our results.

Figure 6. Perceived effectiveness of government and institutions

Notes: Panel heads are the dependent variables derived from survey questions. We display only the estimated coefficients for the three population groups, compared to the baseline group, non-displaced Pashtuns. The full tables encompassing all the control variables can be found in the Appendix. Robust standard errors are clustered by district and survey waves to be as conservative as possible. Survey weights are included to ensure representativeness of the survey sample to the underlying population.

Our findings suggest that whether Pashtun respondents are descendants of those resettled by the emir significantly shapes their perceptions of key governing bodies’ effectiveness. The association between displaced pashtuns and perceptions of central government effectiveness is insignificant, meaning that the Pashtuns we identify as descendants of the nineteenth-century migrants are not staunch supporters of the central government, in contrast to a key long-term state-building objective of resettlement policies like this one. On the other hand, we see a positive and statistically significant coefficient for displaced pashtuns with all three other dependent variables. Notably, compared to their ethnic kin who we do not identify as descendants of those displaced by the emir’s traumatic large-scale resettlement campaign, displaced pashtuns expressed a significantly more favorable view of community relations with local government. This is also reflected in displaced pashtuns’ perception of local governmental actors’ and institutions’ effectiveness. These estimated effects reflect meaningful shifts in attitudes; for instance, the mean of the excluded category is 1.585 for the provincial council outcome, which means the estimated effect of 0.175 corresponds to an 11 percent increase relative to the mean.

Furthermore, when comparing the perspectives of displaced pashtuns with other ethnic groups in the same localities, it becomes evident that there is similarity among these groups. The direction and significance of the estimates across all outcome variables are strikingly similar for both displaced pashtuns and non-pashtuns from the displaced region. This suggests that displaced Pashtuns are more similar, in their political attitudes, to their non-Pashtun neighbors in the north than to their coethnics in the rest of the country. This is notable, as these are groups that have been living in the same regions of the north for over a century.

We follow up on this point by replicating the analysis of Figure 6 focusing only on respondents living in northern provinces,Footnote ⁷² with the non-displaced Pashtuns (the relative newcomers) as the excluded category. While we include province fixed effects in all of our models of survey data to address geographic differences across Afghanistan, this approach is even more restrictive. Although the sample size decreases by over 70 percent, we recover very similar results (see the Appendix for details).Footnote ⁷³

Are the Central Government and Other Governing Actors Viewed as Legitimate?

Figure 7 presents the results on the perceived legitimacy of the central and provincial governments, the Wolesi Jirga (House of the People), and government courts. Although the displaced pashtuns’ views of the provincial government are not significantly different from those of Pashtuns who were not resettled by the Iron Emir, their views of the legitimacy of the central government are significantly more negative. The estimated effect of –0.221 corresponds to an 11 percent decrease relative to the mean. This observation aligns with previous findings regarding the perceived effectiveness of the central government, suggesting that displaced pashtuns view the central government as both ineffective and illegitimate. This flies in the face of the state-building objectives that drove the large-scale resettlement of Pashtuns to the frontier.

Figure 7. Perceived legitimacy of government and institutions

Notes: Panel heads are the dependent variables derived from survey questions. We display only the estimated coefficients for the three population groups, compared to the baseline group, non-displaced Pashtuns. The full tables encompassing all the control variables can be found in the Appendix. Robust standard errors are clustered by district and survey waves to be as conservative as possible. Survey weights are included to ensure representativeness of the survey sample to the underlying population.

We also note that displaced pashtuns’ negative views of the central government’s legitimacy are similar to those of non-Pashtuns. While the magnitude of the coefficient for displaced pashtuns is larger and not close to or overlapping with zero, the 95 percent confidence intervals for these groups indicate no statistical differences among them for all four outcomes. We further assess the differences between displaced pashtuns and other Pashtuns by again looking exclusively at Afghans in the northern provinces. Consistent with the plots in Figure 7, the only significant difference we find between displaced pashtuns and Afghans who do not live in the displaced region but do live in the north is in their views on the legitimacy of the central government (see the Appendix for these results). While the coefficient is significant at only the .10 level in this much smaller sample, the coefficient remains a meaningfully large –0.123, with p = 0.056. That is, displaced pashtuns have a consistently negative view of the central government and its legitimacy. The descendants of the Iron Emir’s resettlement policies have not become “strong supporters of the government in the north,”Footnote ⁷⁴ and they differ in their views from their ethnic kin elsewhere in Afghanistan.

Views on the Taliban

We now turn to respondents’ views of the Taliban. These responses were collected before the Taliban’s 2021 takeover of Afghanistan, and they offer insights into the subsequent support of the Afghan populace for their new regime. Initially, we find that displaced pashtuns are significantly less likely to agree that the Taliban has more influence than the government, compared to their ethnic counterparts in the south and east. In the first model, all coefficients for Pashtuns and non-Pashtuns are negative, which implies a stronger sense of governmental influence across these groups compared to Pashtuns in the south and east. This estimate indicates that displaced pashtuns have dim views of the central government even though it has more influence in their localities than the Taliban, its main competitor. If anything, this result further undermines the idea that the long-term effects of the Iron Emir’s resettlement policies were to create a group of Pashtuns loyal to and dependent on the central state.

In the second model, for the relation of local communities with the Taliban, all groups’ views are generally more negative than the baseline group, Pashtuns in the south and east. However, the estimates for neither Pashtuns nor non-Pashtuns in displaced regions reach statistical significance. In the third and fourth models, the coefficients for displaced pashtuns are both insignificant. Meanwhile, non-Pashtuns have generally less favorable views of the Taliban’s statements and judicial system. Across all categories, the coefficients for all three groups are either negative or insignificant when compared to non-displaced Pashtuns. This trend aligns with the widely held belief that Pashtuns, particularly those in the eastern provinces of Afghanistan, were more supportive of the Taliban. Finally, we note that the estimates are very similar when we look only at the northern provinces (see the Appendix). This suggests that the differences we uncover between displaced pashtuns and their kin elsewhere in Afghanistan are robust to comparing these communities to more recent Pashtun migrants in the north.

Identity in Resettled Communities

The results in Figures 6 to 8 demonstrate that the long-term effects of the large-scale resettlement of Pashtuns to the north cut against the central state-building objective of building a cadre of coethnics loyal to and supportive of the central government. These results support our main argument, that the long-term effects of this traumatic, large-scale, and poorly implemented resettlement over a century ago are communities that have stronger ties with and more positive views of their local and regional government and institutions. We also demonstrate across numerous outcomes that the views of displaced pashtuns are much more similar to non-Pashtuns who live in the same northern region of Afghanistan than to Pashtun kin elsewhere in the country. However, the analysis has not yet directly assessed our claim that displaced Pashtuns have developed political identities centered on their locality and region, rather than their ethnicity or Afghan identity.

Figure 8. Views of the Taliban and its institutions

Notes: Panel heads are the dependent variables derived from survey questions. We display only the estimated coefficients for the three population groups, compared to the baseline group, non-displaced Pashtuns. The full tables encompassing all the control variables can be found in the appendix. Robust standard errors are clustered by district and survey waves to be as conservative as possible. Survey weights are included to ensure representativeness of the survey sample to the underlying population.

Theoretically, people can have identities that combine different aspects of their life and experience, for example, their religion and their nationality. However, our focus here is on the one aspect of “who they are” that they identify as the most salient identity category when prompted to pick just one. We use a MISTI survey item that elicits this directly.Footnote ⁷⁵

Figure 9 reports the results of five regression models focused on this primary political identity. These models incorporate all survey waves that asked respondents about their primary political identity, covering 2012 to 2014. As with the Foghorn data, we included fixed effects for each survey wave and province to adjust for disparities across time and Afghanistan’s diverse regions. We also considered individual indicators for gender, age, employment status, household leadership, marital status, and household income.

Figure 9. Primary identity attachment: transnational, national, and ethnic identities

Notes: Panel heads are the dependent variables derived from survey questions. We display only the estimated coefficients for the three population groups, compared to the baseline group, non-displaced Pashtuns. The full tables encompassing all the control variables can be found in the appendix. Robust standard errors are clustered by district and survey waves to be as conservative as possible.

The results are striking and support our central claim that the long-term consequences of resettlement are Pashtun communities that identify with their locality and region rather than their ethnicity or nationality. On this front, we find pronounced differences among Pashtuns in different regions of Afghanistan, as displaced pashtuns express a stronger localized political identity—73 percent more than non-displaced pashtuns, the excluded category. Importantly, non-pashtuns from displaced provinces are also significantly more likely to prioritize their local identity than Pashtuns in the south and east. We did not observe a similar effect for family, ethnic, Afghan, or religious identity. The finding that displaced pashtuns are less likely to consider their Pashtun ethnicity or their Afghan nationality as their primary political identities, compared to their counterparts in the south and east, is a key piece of evidence for our central claim: resettled communities developed relatively local identities rather than the ethnocentric or national identities envisioned in the Iron Emir’s state-building project. Instead, displaced Pashtuns are significantly more inclined to prioritize the multi-ethnic local identity that they share with neighboring Tajiks and Uzbeks.

The results are very similar when we examine only respondents in northern provinces (see the Appendix). Thus, the local identity among displaced pashtuns is not simply an artifact of living in the north, as we recover the same effect when we look only at northern Afghans and the excluded category is Pashtuns that we do not identify as descendants of those resettled in the 1880s and 1890s (again, see the Appendix for these results).

It is worth noting that identifying one’s local identity as primary is very uncommon in the survey data: fewer than 8 percent of all respondents across survey waves choose this option, making it the second-least-common response. Ethnic and Afghan are the most common answers, at about 25 percent each. In sum, the results shown in Figure 9 provide strong support for our core arguments. In combination with the findings from Foghorn (as in Figure 6), they suggest that descendants of resettled Pashtuns have political views that are more similar to those of their non-Pashtun neighbors than their kin elsewhere in Afghanistan.

Displaced Pashtuns, Insurgency, and Battlefield Cooperation

The findings summarized thus far provide a wealth of evidence that displaced Pashtuns are neither relatively pro-central-government nor similar in political attitudes to their Pashtun kin. We supplement this detailed survey evidence with a salient and potentially costly behavior: the provision of tips to the central government in response to insurgent attacks. Do displaced Pashtuns’ distinct political attitudes translate into differences in behavior during the ongoing civil war? To answer this question, we use the military reports on civilian casualties and collaboration collected by coalition forces from 2006 to 2014.

We start by replicating and then building on an established result: that civilians respond to harm from insurgents by shifting the flow of sensitive information toward the counterinsurgency. We do this by studying tips on improvised explosive devices (IEDs) and other tactical threats at the district-week level. Our model specification follows Wright and co-authorsFootnote ⁷⁶ and includes district and week-of-year fixed effects, as well as trends in combat activity (direct fire, indirect fire, and IED explosions). These results are introduced in Table 1. In columns 1 and 3, we see a robust positive association between civilian harm by insurgents in the prior week and the flow of information in the current week. In columns 2 and 4, we introduce marginal effects, leveraging our district-specific measures of displaced Pashtuns. Note that while the interactions are positive, they are imprecise, suggesting that areas with a larger displaced Pashtun population are not more (or less) responsive to civilian harm than districts with no displaced Pashtuns (the baseline effect).

Table 1. Civilian harm shapes the flow of information

Notes: Outcomes of interest are noted in each column heading. Model parameters are described in the table footer. The outcome and measurement of harm are expressed in per capita terms (per 1,000), and population weights are included in each regression. Standard errors, clustered at the district level, in parentheses. ** $p \lt .05$ ; *** $p \lt .01$ .

However, it is possible that the information flow itself differs between areas, with the descendants of resettled Pashtuns less willing to share precise, actionable intelligence with the central government. Such a gap could have several mechanisms, each consistent with our theoretical argument and the survey-based findings presented thus far. The descendants of resettled Pashtun populations might be less willing to share tips with the government if they do not see it as particularly effective or legitimate. That is, sharing sensitive information with a central government that is unlikely to use it effectively is plausibly viewed as not worth the risk of being identified by insurgents as a collaborator. Of course, the counterinsurgents (that is, the central government) might also view information shared by these communities skeptically.

To check whether this flow of information shapes battlefield outcomes, we reproduce results from Sonin and Wright,Footnote ⁷⁷ allowing the influence of tips to vary using our measure of Pashtun displacement and focusing on an important set of counterinsurgent outcomes: IEDs found and cleared; weapons caches found and cleared; safe-house raids; and insurgent detentions (Table 2). Notice that the marginal effects of displacement are all negative, and the interaction coefficients in columns 2 and 4 are significant at the 95 percent level. These results suggest that while the flow of information overall does not change with displacement status, the impact of this information is substantively and, in the cases of weapons and detentions, significantly diminished. In sum, this is systematic evidence that Pashtuns who live in villages settled by those displaced over a century ago are generally less willing to provide actionable information to the central government in response to insurgent activity.

Table 2. Displacement associated with reduced battlefield effectivenes

Notes: Outcomes of interest are noted in each column heading. Model parameters are described in the table footer. The outcome and measurement of information flow are expressed in per capita terms (per 1,000), and population weights are included in each regression. Standard errors, clustered at the district level, in parentheses. ** $p \lt .05$ ; *** $p \lt .01$ .