Social Transformation and Violence: Evidence from U.S. Reconstruction

Abstract

How do political actors create and institutionalize revolutionary social transformation, and what are the consequences of their efforts? In this paper, we provide a framework for understanding the conditions under which revolutionary social transformation unfolds and becomes institutionalized over time. We argue that a direct consequence of social transformation and the institutionalization thereof, however, is violence against the revolution’s beneficiaries which can likewise endure over the long-term. We test our arguments using historical, county-level data on post-U.S. Civil War Reconstruction and we supply both quantitative and qualitative evidence for our mechanisms. We ultimately demonstrate that social transformation and violence are often causally linked, not mutually exclusive outcomes, thereby expanding our understanding of how social orders are created and maintained.

Keywords

civil war conflict processes social movements subnational politics

How is revolutionary social transformation institutionalized, and what are the consequences thereof? Revolutionary social and political change occurs by destroying social and political orders then institutionalizing new ones (Huntington, 2006, p. 266; Mondlane, 1983, p. 163; Skocpol 1979, pp. 163–164). Yet, enduring transformations of social and political orders are rare (Tilly, 1977, p. 220) and frequently transpire over the course of decades if not centuries (North et al., 2009, p. 27). Even among the actors who do realize their transformative, revolutionary ambitions, the process of consolidating such change is hardly simple or straightforward: during the Chinese Civil War, for example, Mao Tse-Tung and the People’s Liberation Army (PLA) implemented their revolutionary land redistribution schemes in southern China, but the changes wrought as a result of their programs triggered a violent backlash against the PLA so intense, it caused the PLA’s catastrophic defeat and retreat from the south of China (Opper, 2018).

In this paper, we argue that both success toward institutionalizing revolutionary social change and reactionary violence are causally intertwined. We conceive of revolutions as bundles of interlocking political, social, and economic programs. The objective of these programs is to alter pre-existing social hierarchies across a broad population, typically to mitigate if not eliminate status hierarchies that exist between (subjugated and oppressed) social out-groups and (privileged and dominant) social in-groups. We identify three conditions that, when present together, make social transformation more likely to occur: revolutionary actors introduce programs that transfer physical (wealth and assets) or ideational (skills, knowledge, and practices) goods to social out-groups, revolutionaries undertake this transfer of physical or informational assets within a protected environment (supportive legal milieu or secure space), and social out-group members who typically stand to benefit from these programs participate in them. Social out-groups immediately capitalize upon the goods and information they receive from revolutionaries and put them into use and practice, thereby disrupting the existing distribution of power between social groups and incrementally reducing crystallized status hierarchies.

Furthermore, social out-group behavior in response to these initial conditions becomes the impetus for institutionalizing social transformation over the long-term. By repeatedly putting assets gleaned from revolutionary programs into practice, social out-groups embed knowledge, skills, norms, and behaviors within a broader community, such that these practices and knowledge are learned and reproduced over generations. Within protected spaces created by revolutionaries during the implementation of programs, social out-group members establish informal and formal communal networks and organizations. These networks serve to promote communal interests, coordinate communal collective action, preserve assets from revolutionaries and expand access to these assets within the community. Both practices and networks are extremely resilient to pressure and difficult to eradicate. As a result, even in the absence of protected spaces and revolutionary programs that transfer assets to out-groups and make the initiation of social transformation more likely, social out-group members continue to institutionalize revolutionary programs.

Yet, the improvement in the lives of social out-groups and corresponding reduction (but not necessarily elimination) of social inequities provokes resentment and hostility among privileged in-groups toward the gainful out-groups (Brewer, 1999; Petersen, 2002; Quillian, 1995; Smångs, 2016b; Wells, 2012). In places where reductions in social inequities are greater, this resentment is more likely to manifest as political violence from privileged in-groups who seek to maintain or reclaim their dominant status (Smångs, 2016b; Wells, 2012). Privileged in-group members are more likely to resort to violence against social out-groups because social out-groups may be easier targets than armed and militarized revolutionary actors, and because during the revolutionary period, in-groups may lack access to alternative, non-directly violent tools for exerting social control and preserving the hierarchy over out-groups they once had. Furthermore, because social out-groups work to institutionalize and preserve benefits over the long-term, the resentment and violence against social out-groups correspondingly endures over the long-term until social in-groups are wholly demobilized (sometimes annihilated) by revolutionaries, or until social in-groups reclaim pre-revolutionary levels of social control.

Thus, social transformation and violence are frequently intimately intertwined. Furthermore, both social changes as well as resentful violence can persist over time. Indeed, Vladimir Lenin, one of history’s infamous initiators of widespread social change, claimed that even when transformative schemes “commence peacefully,” they nevertheless “end in furious wars” (Lenin, 1919).

We test our theory using sub-national data from Reconstruction after the U.S. Civil War (see, e.g., Ruef, 2014).¹ We select the U.S. case for three reasons. First, Reconstruction is an example of a political actor’s attempt to construct a new social order on the scale of a social revolution (Du Bois, 2013; Foner, 2011; Moore, 1993; Skocpol, 1995): indeed, some of Reconstruction’s leaders understood its objectives to be revolutionary (Shortreed, 1959, p. 77) and even communist rebel groups intent on achieving radical social transformation recognize Reconstruction as America’s true revolution (Connell, 1993, p. 71). Second, our case selection suggests that violence is unlikely to emerge after conflict due to a clear wartime victor (Licklider, 1995, pp. 684–685), as well as the presence of strong (Fearon & Laitin, 2003) and accountable state institutions (Walter, 2015). Third, pre-Reconstruction (pre-treatment) measures for social transformation (out-group literacy) and a specific form of counter-revolutionary violence (white-on-Black lynching) are extremely low if not non-existent in most places, thereby facilitating causal inference: though not all Black persons in the South were enslaved, most were, and almost all enslaved persons could not receive an education (Davis, 2011), while lynching was not a common form of symbolic racial violence until after the Civil War (Pfeifer, 2011).

What we find supports our claims. Counties where former enslaved persons (social out-groups) had access to revolutionaries that transferred physical and ideational assets within protected spaces (Union troops) were more likely to see improved literacy and decreased social inequities between Black and white communities. These relative social improvements persist into the 1920s, almost 50 years after Reconstruction ended. Yet whites were also more likely to lynch Black persons in places where Black-white social inequities were reduced, and this violence also persisted decades after Reconstruction ended.

Our work ultimately specifies the conditions under which social transformation and the institutionalization thereof is more likely to occur, and the conditions under which retributive violence could follow. Although we are not the first to find that efforts to construct a new social order sometimes have violence-producing consequences (Lenin, 1919; Opper, 2018), our contribution is to instead highlight the causal linkage between social transformation and reactionary violence.

Conceptualizing Revolutionary Change and Its Challenges

In this paper, we conceptualize states and societies as composed of social groups. Social groups are people who share a common identity: class, nation, linguistic group, race, ethnicity, gender, religion, etc. (Tajfel & Turner, 1979). These social groups are often arranged hierarchically vis-a-vis each other with some groups dominating (social in-groups) and other groups subordinated (social out-groups) (Tajfel & Turner, 1979). To preserve and maintain their relational position and exert control over other social groups, in-groups construct social orders. Social orders refer to the formal and informal institutions and organizations, procedures, rules, and norms that structure access to social, political, and economic power, and dictate interactions between social groups.² Social orders thus reinforce and make meaningful the hierarchies and status differentials between social groups.

Revolutionaries emerge within this social landscape and strive for social transformation by replacing the existing social order. If social orders mirror and reinforce social hierarchies between social groups, then replacing the social order corresponds to replacing the nature of hierarchies between social groups, thereby achieving social transformation.³ In most cases of revolution, the objective of transformative schemes is to make relational positions between social groups more egalitarian and less hierarchic (Moore, 1993; Skocpol, 1979).⁴

To replace social orders and thus corresponding social hierarchies, revolutionary actors implement multi-dimensional and inter-connected projects and institutions (Skocpol, 1979) that far exceed the scope of reforms. Rather than simply re-allocating resources within an existing social order, revolutionary projects entail creating new institutions and organizations through which power is redistributed between social groups.⁵ Because of the herculean task of revolutionaries, however, the current scholarship is almost uniformly pessimistic about realizing and institutionalizing social transformation (Johnson, 1982, p. 1; Tilly, 1977, p. 220).

The failure to institutionalize social change typically occurs for one of two reasons: first, the institutions that preserve a social order resist change (Tilly, 1977, p. 220) or second, the entire project collapses into violence (see, e.g., Lenin, 1919; Opper, 2018). Institutions are sticky, path dependent, and created for the express purpose of enduring overtime (Pierson, 2000, p. 262), even in the face of severe shocks (Acemoglu & Robinson, 2006). Any changes to such institutions are typically incremental, transpiring over decades (North et al., 2009, p. 27). Even if revolutionaries construct new institutions intent on pursuing certain objectives, these institutions, and the people who populate them, may fall short in achieving such objectives (Lieberman, 2002).

Alternatively, social transformation may not be institutionalized because revolutionary projects descend into counter-revolutionary violence that prevents any change from occurring. Locals may actively dislike revolutionary efforts, forming bands of resistance (see, e.g., Lenin, 1919; Opper, 2018). Counter-revolutionaries even learn from one another to maximize and eradicate potential challengers (Weld, 2019). At the same time, the institutions revolutionaries construct to change the social order may cause catastrophic problems that collapse revolutionary endeavors (Scott, 1998, p. 4).

Thus, existing works suggests revolutionaries typically fail to institutionalize social change, either because ancien régime social orders are so entrenched or because revolutionary projects descend into violence before change can be institutionalized. Instead, in the next section, we argue that rather than conceive of social transformation and violence as mutually exclusive outcomes, the two are frequently interrelated.

Theory: Social Transformation and Violence

The realization of revolutionary social change cannot occur with a single policy, nor does it occur instantaneously (North et al., 2009, p. 27). Rather, we conceive of revolutionary social transformation as unfolding in two stages: an initiation stage and then an institutionalization period.

The initiation of social transformation is made more likely by the simultaneous presence of three elements.⁶ The first element involves the implementation of revolutionary programs wherein physical and ideational assets are transferred to a target population, typically members of an oppressed social out-group.⁷ Physical assets often involve land and wealth, both of which have been central to redistributive schemes from nearly two centuries of revolution: French, Russian, Chinese, Mexican and Eritrean, for example (Houtart 1980; Huizer 1968; Mathiez 1920; and Opper 2018). Likewise, since the Protestant Reformation, ideational assets like literacy programs have been key components of revolutionary projects (Arnove & Graff, 1987, pp. 1–2). These ideational projects “embrace[s] a new type of person in a qualitatively difference society” (Arnove & Graff, 1987, p. 5; see also Zhang & Lee, 2019; Freire, 2005, pp. 135–138), and typically involve basic literacy and arithmetic, while also conveying information about new political and legal rights, as well as social expectations and cultural practices.

When simultaneously combined with these programs, the second element that makes the initiation of social transformation more likely is a protected space to implement the transfer assets to beneficiaries (see, e.g, Chacon & Jensen, 2017; Downs, 2015). This space could be a supportive legal milieu, a physical building or a militarily defended area. Within such spaces, revolutionaries can introduce and enforce their programs with little resistance. For instance, in the absence of a conducive legal environment, out-group members may be unable to exercise new rights and privileges. In-group members may be unwilling to voluntarily relinquish their wealth and property to social out-groups, and so the beneficiaries of redistributive programs need a space to claim the physical assets to which they are entitled without fear of in-group members refusing to relinquish their goods, or they could face legal challenges to redistributed wealth and property. Likewise, social out-group members often need a designated place wherein they can obtain informational assets from revolutionaries, such as a school.⁸

The transfer of these physical and ideational resources even within a protected space, however, is less likely to lead to social transformation without a third, critical element. The third element that makes the initiation of social revolutionary change more likely is the participation of social out-groups in revolutionary programs and their putting to use the assets they receive from revolutionary actors.⁹ When social out-groups receive and exercise the assets transferred to them by revolutionaries within protected spaces, they begin to reduce power imbalances along pre-existing social hierarchies. For instance, when social out-groups safely accept revolutionaries’ transfer of assets, the sources of power begin to shift from social in-groups to out-groups (see, e.g., Jones, 1991; Moise, 2017). During the Chinese Civil War, the redistribution of land quickly altered power relations centered on land-ownership and renting (see, e.g., Opper, 2018), while for a short period during the early years of the Soviet Union, workers claimed ownership shares in factories, thereby seizing the means of production (Siegelbaum, 2015).

Beyond merely shifting the bases of power, the confluence of the three factors enlivens the new social order. When social out-groups receive ideational assets from revolutionaries about new laws, rights, and entitlements then act on this information within supported spaces, they begin to populate the institutions that make up a new social order and through which political, social, and economic powers are expressed. In other words, the new social order is put into practice. During the Taiping Rebellion, in contrast to imperial institutions, the rebel organization introduced quasi-democratic institutions that enabled persons who were once unable to participate in political processes to influence the creation of new social orders and exert social control over previously dominant social groups (Chien, 1973, pp. 139–141, 200, 257, 439).

Although the joint presence of all three initial elements makes the initiation of social transformation more likely, the actions of social out-groups members specifically institutionalize social transformation over the long-term. By putting revolutionary programs into practice, new rights, privileges, norms, knowledge, and skills become part of the communal routine, increasingly woven into and embedded in everyday life of at least social out-group members (see, e.g., Adler, 2019, pp. 122–127). When practices become routinized, they are reproduced and recreated over generations, emerging as part of the cultural framework of a community, and becoming so deeply embedded that they become taken for granted norms that motivate behavior through a logic of appropriateness (Adler, 2019, pp. 122–127; March & Olsen, 2004, pp. 478–479).

Additionally, when revolutionaries are able to implement programs in secured spaces, social out-group members converge upon these focal points. Once there and participating in revolutionary programs, out-group members are able to establish formal and informal networks, organizations and institutions. Through these formal and informal channels, out-group members can transfer their skills, knowledge and resources throughout their community: both to social out-group members who may not have access to revolutionary actors and programs, as well as within communities across time (see, e.g., Ingham, 2003; Montgomery, 1995; Ruef & Grigoryeva, 2018). These informal networks preserve existing levels of social transformation and carry forward this transformation over time.

Even under extreme pressure, these practices, networks, and institutions can continue to operate and function despite attempts to erode them across time. As examples, despite the Soviet Union’s attempts to erode national identities in non-Russian territories it controlled, language, culture, and shared national myths learned during pre-communist schooling projects were reproduced informally across generations, and could not be eradicated (Darden & Grzymala-Busse, 2006). Similarly, Jewish survival strategies during the Holocaust were shaped by pre-World War II governance experiences that were not eradicated by Nazi repression: Jewish communities could rely on networks, language, skills, and institutional knowledge gained through pre-World War II transformative assimilatory education programs to evade Nazi occupation, while Jewish communities who experienced selective repression by the pre-World War II state were able to rely on pre-Holocaust communal networks, organizations, and mobilization techniques to resist Nazi occupation (Finkel, 2017, pp. 10–12). Thus, even under the extreme pressure of Nazi occupation and the Holocaust, skills and languages cannot be unlearned, and communal institutions and networks continue to persist and support out-group communities.

Taken together, this means that when revolutionaries implement programs that transfer physical and ideational assets within a safe space to social out-group communities who put these assets into practice, social transformation begins, but social out-groups’ behaviors then institutionalize social transformation by routinizing revolutionary programs or forming intra-communal networks to preserve and expand assets from revolutionary programs. These routines, knowledge, and communal networks can endure over time, even under extreme pressure. Thus, we hypothesize that:

H1: When revolutionaries transfer material and ideational assets within secure spaces to participatory social out-groups, reductions in hierarchies between social groups are more likely, and this reduction in social group hierarchies is more likely to endure over the long-term.

Yet, the social, economic, and political gains made by social out-group communities must be contextualized within society as a whole and therefore must be understood vis-a-vis the dominant social in-group that actively oppressed them (Brewer, 1999; Hewstone et al., 2002; Quillian, 1995). By changing or threatening to change the distribution and allocation of political and economic resources across social groups and eliminating or limiting in-group’s ability to control other social groups, revolutionaries and out-group members threaten or reduce in-group members’ relative status (Brewer 1999; Quillian 1995). The loss of status or the threat thereof in turn fuels privileged community sentiments of revenge and resentment toward the upwardly mobile social out-group (Quillian, 1995; Smångs, 2016a, 2016b; Wells, 2012). Feelings of resentment and revenge are sufficient psychological motivators for political violence that can range from low-level inter-personal intimidation or bodily harm, such as hate crimes, or full scale insurgencies that culminate in ethnic cleansing or genocide (Acharya et al., 2016; Petersen, 2002).

Beyond being psychologically primed for violence, social in-groups may view violence as the only way to maintain or recoup their socially dominant position and ability to exert control over other social groups. Revolutionaries may block social in-groups from participating in the institutions and organizations that make up new social orders, so social in-group members cannot wield social control through these institutions and organizations. Even if revolutionaries allow in-groups to participate, their participation may be limited, reducing in-group’s ability to exert any influence within these institutions. For instance, in Eritrea, the revolutionary Eritrean People’s Liberation Front created new quasi-democratic institutions with representation based on class, so wealthier persons participated but had almost no influence in these institutions (Leonard, 1988, p. 118). Finally, revolutionary actors may also confiscate and redistribute in-group wealth, thereby limiting in-group members’ ability to convert economic power to political and social influence. As a result, in-group members may perceive violence as the only way to maintain or recoup their dominant position.

With limited options and a psychological state primed for violence, in-group members may be especially keen to wield violence against out-group members in particular, as opposed to revolutionaries. Compared to revolutionaries who may be militarized or control armed organizations, civilian out-groups are comparatively “less risky” targets. Social out-group members are also the impetus for the institutionalization of revolutionary change. By targeting out-group members, in-groups may forestall or hamper the institutionalization of a new social order. Taken together, as social transformation begins and becomes institutionalized, in-group members become primed for violence psychologically, they perceive violence as the only means available to them to recoup or maintain their status, and they recognize civilian social out-group members as comparatively “easy” targets of violence. Social transformation therefore causes in-group violence against out-group members to be more likely.

Finally, as discussed above, after the initial implementation of revolutionary programs, social out-groups can institutionalize social transformation over the long-term, even in the absence of revolutionary actors. Yet, because social out-groups institutionalize and preserve gains from social transformation over time, in-group sentiments of resentment and revenge will similarly endure over the long-term. As a result, direct violence by social in-groups against social out-groups is also likely to endure over the long-term. This direct violence will endure until one of two conditions are met. First, social in-group members are able to recoup some or most of their dominant social status, in the process regaining access to mechanisms of social control that do not rely on direct, interpersonal violence. Under these conditions, sentiments of resentment diminish and social in-groups need not only turn to violence to preserve their status, both of which make violence less likely. By contrast, if revolutionary actors or an otherwise militarily equipped force eliminate social in-groups who wield violence against social out-groups,¹⁰ then violence against social out-groups will also diminish.

Several examples illustrate the powerful effect of in-group resentment and their proclivity for violence against out-groups. The Vendée Uprising erupted in concert with ancien régime in-groups directly responding to the institutionalization of the Revolutionary French Regime (Ozouf et al., 1989, pp. 165–166). In Eastern Europe, resentment at the loss of status in the social hierarchy of ethnic groups motivated political violence (Petersen, 2002, p. 25). As part of the social change ushered in by the Mexican Revolution, the Mexican government undertook one of the most comprehensive land reform programs in the world, but peasants who benefited from land redistribution often found themselves to be the targets of violence by powerful and vengeful landowners whose land and relative power was threatened (Huizer, 1968, pp. 118, 125).

Therefore, social transformation makes violence by social in-groups toward the out-group members more likely. Because social out-groups institutionalize social transformation over the long-term, in-group resentment toward the out-group also persists over the long-term. As a result, we hypothesize that:

H2: Reducing social hierarchies between salient social out-groups and in-groups creates resentment among the in-group leading to violence against the out-group, both of which are more likely to endure over time.

To summarize, in this section, we provide a framework for understanding how social transformation begins then endures as well as the short- and long-term violence-producing implications of such transformation. As a result, unlike previous works, rather than conceive of violence and social transformation as mutually exclusive, we argue that social transformation and violence are intimately and causally linked.

Research Design: Case Study of U.S. Reconstruction

We test our theory using the case of Reconstruction after the U.S. Civil War. Our analysis proceeds across several stages. First we begin by justifying our case selection. Then we provide an overview of the case and supply qualitative evidence of our theorized causal processes.

We ultimately test our theory quantitatively across three stages. In the first stage, we assess non-random selection in the distribution of revolutionary programs that transferred material or ideational assets within secured spaces (see, e.g., Fortna, 2004; Mazumder, 2018).

Given these results, in the second stage, we test whether the implementation of revolutionary programs that transferred these assets within a secured space achieved social transformation and whether social transformation persisted across time (H1). To do so, we use a number of different specifications including a difference-in-difference model that addresses non-random selection in the first stage.

In the third stage, we test whether reductions in hierarchies between social groups also causes in-group violence against out-groups, and if this violence persists over time (H2). We replicate the same modeling strategy in the second stage and use a number of estimation techniques including a difference-in-difference approach to account for non-random selection.

Finally, we provide a brief overview of additional robustness checks.

Case Justification

The case of U.S. Reconstruction is particularly instructive and we selected it to test our theory for a number of reasons. First, like most revolutions, U.S. Reconstruction is an example of an attempt to erase an existing social order and replace it with a new, more egalitarian social order favorable to the central state (Bensel, 1990, pp. 1–2). Using the language of King and Smith (2005, p. 75), Reconstruction corresponds to a fundamental change in the social order of the American South from a “white supremacist” order to more of an “transformative egalitarian” social order. Likewise, Skocpol (1995, p. 86) writes that the Republican party during and immediately after the American Civil War promoted “regulatory and redistributive policies” that, if they had been implemented to their entirety “would have added up to a ‘virtual revolution from above’.” Moore (1993, pp. 111–156) includes the U.S. Civil War as a case of bourgeois-capitalist revolution. Thaddeus Stevens one of the chief architects of U.S. Reconstruction even described the project “as a radical revolution” and that the U.S. should “remodel our institutions” (Shortreed, 1959, p. 77). Downs (2015, p. 5) describes Reconstruction as a “revolutionary experiment.” Both Du Bois (2013) and Foner (2011) argue that the U.S. attempted revolutionary change, even if such change was never fully realized. Even leftist, independence-seeking rebel groups operating in Africa in the mid-1970s recognize American Reconstruction as a revolutionary, transformative project, particularly in comparison to the American Revolution (Connell, 1993, p. 71). Thus, U.S. Reconstruction can be understood as a case of political actors introducing revolutionary projects and we explore their success (or lack thereof) in the following sections.¹¹

Second, by selecting the U.S. case, we are able to account for the factors identified by existing research as most likely to trigger post-conflict violence, and thus isolate the effects of revolutionary projects themselves. One of the strongest predictors associated with no return to violence is a clear military victory (Licklider, 1995, pp. 684–685), something the Union army undoubtedly secured. Weak post-conflict institutions are not to blame for post-war violence in the U.S. (Fearon & Laitin, 2003;Walter, 2015): national-level political and economic institutions remained strong in the war’s immediate aftermath, and the federal government invested heavily in the southern economy, schools, transit, and infrastructure (see, e.g., Foner, 2011; Du Bois, 2013). During Reconstruction, elections took place and a large number of Black politicians were elected to national and state offices (Logan, 2018), and in stages before but in all cases by 1884, loyalty oaths for holding office or voting for Confederates were eliminated (Hyman, 1959, p. 265).

Third, in the U.S. context, the social in-groups and out-groups are salient and clearly defined, with political preferences for social change (or not) generally shared across these lines. The out-group community, Black persons and especially former enslaved persons, sought to improve their social, economic, and political livelihoods from the condition of enslavement, and to reduce long-standing social inequities between Black persons and white persons. The in-group, (wealthy) Southern whites, sought to maintain antebellum social hierarchies and keep labor immobile and inexpensive.

Finally, the conditions of enslavement allow us to more accurately assess the effect of revolutionary projects. We operationalize Black social advancement through literacy rates, and literacy rates in comparison to whites. Because enslaved persons were legally forbidden from education everywhere but Maryland and Kentucky (Davis, 2011, p. 106), pre-war, pre-treatment values for these variables are very low and close to zero in some places (Margo, 1990). Likewise, although some mob executions of enslaved persons took place before Reconstruction (Pfeifer, 2011, p. 34),¹² the number of white-on-Black lynchings over the same period of time is almost 40 times higher after Reconstruction (Auut, 2017) indicating that increases in literacy or white-on-Black lynchings are not simply continuations of pre-existing trends.¹³

Overview of the Case of U.S. Reconstruction and Qualitative Evidence in Favor of Theory

From 1861 to 1865, the U.S. military engaged in hundreds of battles across multiple fronts during the Civil War, finally securing victory over Robert E. Lee’s Confederate army on April 9, 1865. At the time of surrender, the U.S. government not only sought to prevent a counter-insurgency from forming, but also sought to ensure that enslavement had ended and some voting and civil rights for freedpersons would be protected (Downs, 2015, pp. 13–14). To that end, after a brief, post-surrender period of time in which whites reasserted control over the South and introduced nefarious Black Codes that perpetuated enslavement in all but name, by 1866, “radical” U.S. Congresspersons gained control over Reconstruction from President Andrew Johnson (Donald et al., 2016, pp. 944–945) and passed a series of intensive reforms with the explicit aim of preventing a reactionary counterinsurgency in the South while improving the livelihoods of the formerly enslaved.¹⁴

For instance, Congress passed the 13th, 14th, and 15th Amendments abolishing slavery, extending equal protection under law and enshrining suffrage for men of all races (Frohnen, 2008). Congress invested heavily in the Bureau of Refugees, Freedmen, and Abandoned Lands, or the Freedmen’s Bureau, which was responsible for a number of directives to mitigate the consequences of war, including providing food and necessities to needy white persons, but to also aid enslaved persons negotiate contracts, reconnect with family members, purchase property, secure medical support, and gain an education (Donald et al., 2016, pp. 951–955). Congress also passed Civil Rights Act of 1866 abolishing the nefarious Black Codes, then passed the Reconstruction Acts in 1867 which created five military districts throughout the South and put forth the requirements of readmission to the Union as states (Frohnen, 2008).

The control of these districts fell to generals, some of whom retained this position from the Civil War (Downs, 2015, pp. 181–182). Their decisions would be enforced by 60,000 troops deployed across the non-Texas South in 1866, and an additional 25,000 troops deployed later in Reconstruction (Downs, 2015, p. 91). These generals had a degree of freedom to interpret their mandate. As a result, the nature, presence and expansiveness of Union troop presence often depended upon these commanders’ interpretation of the law and objectives for Reconstruction. Some commanders were more moderate (George Meade) and specifically stationed troops in cities and railroad hubs (Downs, 2015, pp. 205–206), while other commanders were far more expansive in the implementation of Reconstruction and deployment of troops (Philip Sheridan) (Downs, 2015, p. 184). Although more conservative President Andrew Johnson could and did remove military commanders (Downs, 2015, p. 185), Johnson’s ability to control the army and regulate occupation was limited by Congress and General Ulysses Grant (Downs, 2015, p. 183), and even if Johnson replaced generals, their replacements were often just as radical their predecessors (Downs, 2015, p. 181). Ultimately, what resulted is that Union troops unevenly deployed across these military districts (Downs, 2015, pp. 25–26) to counties that varied in terms of how much they relied on pre-existing infrastructure and state capacity: some troops were stationed in military barracks, other encamped on the edge of town, while others occupied run-down “hovels” and abandoned buildings (Downs, 2015, p. 191).

Regardless of where they were placed, both during the Civil War and especially after, Union troops created safe spaces for the implementation of revolutionary programs that transferred material and information assets to social out-groups. For instance, during the Civil War, Special Field Order 15 issued by Union General William Tecumseh Sherman during his March from Atlanta to the Georgia coast enabled the resettlement of enslaved persons living on confiscated and abandoned properties (Sherman, 1865/2020). In so doing, Sherman implemented a revolutionary program that transferred wealth and assets while simultaneously creating a protected space for this transfer of resources. A consequence of these orders was “an extraordinary experiment in [B]lack self-rule” whereby Black persons created a government, schools, and a militia, the result of which was contingent upon the “space wartime created” alongside “Sherman’s field order” (Downs, 2015, p. 31), an incidence that illustrates social transformation at the intersection of space, asset transfer, and out-group participation.

After war, Union troops created safe spaces from vengeful and resentful whites and then helped with the implementation of revolutionary programs among former enslaved persons (Downs, 2015, pp. 6, 52; Foner, 2011, pp. 80–81; Hahn, 2003, p. 173). Indeed, in the context of U.S. Reconstruction, places with concentrated revolutionary resources like Union troops offered “protection from the violence so pervasive in much of the rural south” and a space to create “black social institutions–schools, churches, and fraternal societies” (Foner, 2011, p. 81). Downs (2015, p. 6) describes the presence of Union troops as creating a “narrow but precious space for freedpeople to organize economically, socially, and politically.” Black Union troops were instrumental in the social and economic mobility of former enslaved persons and they “helped construct schools, churches and orphanages, organizing debating societies, and held political gatherings” (Foner, 2011, p. 80). Having faced discrimination and having fought for pay equity, Black troops encouraged “freedpeople to press more expansive definitions of freedom” and “provided both models and access to government for people” (Downs, 2015, p. 52), accelerating former enslaved people’s efforts to mobilize, organize, and ensure social progress (Hahn, 2003, see, e.g., p. 173). Thus, Union troops were uniquely suited to both create a safe space and favorable structural conditions for the implementation of revolutionary programs, then implement these programs themselves.

While Union troops created a protected space for, and then implemented revolutionary programs, Du Bois (2013, e.g., p. 596), Hahn (2003, pp. 164–165) and Foner (2011, p. 97) locate the impetus for the institutionalization of social transformation with freedpersons themselves. Through Reconstruction’s revolutionary projects to build “public schools and private colleges, and by organizing the Negro church,” former enslaved persons had “acquired enough leadership and knowledge to thwart the worst designs of the new slave drivers” and without these institutions, organizations and networks born out of reconstruction “the Negro would, to all intents and purposes, have been driven back to slavery” (Du Bois, 2013, p. 596). Thus, Black communities learned information and practices from revolutionary programs and reproduced that knowledge over generations.

Likewise, extended families re-connected through revolutionary programs formed the basis of “societies,” “associations,” and “joint-stock companies” (Hahn, 2003, p. 167), and these cooperative networks would expand material assets gained during Reconstruction by investing in land and property for its members. For example, “[w]hen the Freedmen’s Bureau assistant commissioner in Louisiana made over 50,000 acres of land on 68 abandoned plantations available to freed people in September 1865, the applicants included 73 cooperative groups composed of 584 men, 458 women and 644 children. Black families in Wilkes County, Georgia managed to raise seven thousand dollars by November of 1865 with the intention of acquiring land in the still thinly populated southwestern part of the state” (Hahn, 2003, p. 167). Similarly, after gaining informational assets from Union troops, Black political leaders like Robert Smalls “utilized the spaces available in Military Reconstruction to build powerful political machines” (Downs, 2015, p. 209), thereby creating communal institutions and networks that would perpetuate freedpersons’ gains from Reconstruction.

Black churches also represent theorized communal institutions and networks that preserved revolutionary programs over the long-term. In the segregated South, church attendance was primarily based on race, so Black churches served Black congregants. These churches became a focal point for preserving and expanding gains made from revolutionary programs. Indeed, when Reconstruction ended, Black churches:

“. . . mustered the personnel and furnished some of the money that were required to operate schools. Educated clergymen or their wives and members of their congregations served as teachers. Church buildings functioned as schoolhouses. Congregations raised funds to pay teachers’ salaries, purchase books, or rent additional space for classrooms to accommodate teachers” (Montgomery, 1995, p. 148).

By the end of Reconstruction, Black persons had also begun to establish trade and technical schools to more directly improve economic advancement (Montgomery, 1995, p. 234).

But revolutionary programs were unable to change white southerners’ desire subjugate Black persons and southern whites’ wrath was soon apparent. In the eyes of white Southerners, the era of Reconstruction represented the “years the lowest elements of society ‘took control of the government’” and that Black suffrage would “‘involve total and absolute ruin to the South, and infinite and irreparable loss to the whole country’” (Cox Richardson, 2009, p. 202). The progress former enslaved persons achieved by leveraging Reconstruction’s resources became a clear source of racial resentment and a trigger of violence for years. Union generals pointed to the emancipation of former enslaved persons, as well as “uncertainty surrounding [southern white’s] political rights and financial values” as a cause of violence against Black persons (United States Congress, 1872, p. 19). They also reflected that “very many of the crimes which have been committed have no political bearing,” but rather, “most of the numerous outrages upon freedmen result from hostility to the race induced by their enfranchisement” (United States Congress, 1872, p. 20).

One of the largest perpetrators of violence was the Ku Klux Klan (KKK). The founder of the KKK, Nathan Bedford Forrest, reported that when Reconstruction began, he received “50 to 100 letters a day . . . from all over the Southern States: men complaining, being dissatisfied, persons whose friends had been killed or their families insulted, and they were writing to me to know what thy ought to do” (United States Congress, 1872, p. 11). His response was to organize the KKK “as a means of subduing the undue bumptiousness and nocturnal prowlings of some of those who seemed incapable of using their new-found freedom” (Horn, 1969, p. 18). To the Klan, “the most ‘offensive’ [B]lacks of all seemed to be those who achieved a modicum of economic success, for, as a white Mississippi farmer commented, the Klan ‘do not like to see the negro go ahead’” (Foner, 2011, p. 429). In one of many incidences of violence, the KKK burned 230 Mississippi schools in response to white Klansman’s feelings of political marginalization (Lyons, 2014, p. 132).

Scholarly work on this period also supports our proposed causal mechanism. Wells (2012), a victim of attempted lynching herself, concluded that “the whole matter [of lynching] is explained by the well-known opposition growing out of slavery to the progress of the race. This is crystallized in the oft-repeated slogan: ‘This is a white man’s country and the white man must rule’” (Wells, 2012, p. 33). She further noted that “The mob spirit has grown with the increasing intelligence of the Afro-American” (Wells, 2012, p. 38). Writing on what caused white-on-Black riots in Memphis, Foner (2011, p. 262) notes that it was “difficult to say which proved more threatening to local whites. . . the large numbers of impoverished rural freedman who thronged the streets in search of employment, or the considerable group that managed to achieve modest economic success” as many Black victims of white mobs were robbed of cash, watches, tools, and furniture. When whites saw their economic status threatened, they frequently lynched relatively successful Black farmers to vent their resentment (Smångs, 2016a, 2016b; Tolnay & Beck, 1995) point to status competition between racial groups as a trigger for whites lynching Blacks. Likewise, Cox Richardson (2009, p. 218) attributes the sources of lynching as “rooted in social and economic tensions” because “[w]hat Southerners feared, after all, was ‘Negro Supremacy’.” Finally, Hahn (2003, p. 427) writes that “lynch mobs sought to reestablish the boundary that they believed were being traversed, to crush the attacks and violations they could associate with the weakening of their own leverage.” Though Reconstruction ultimately ended in 1877, the violence perpetrated by whites toward Black persons in this period, however, would endure for decades (Auut, 2017).

Key Measures

Given the case summary discussed above, in this section, we review our key measures and data. As an overview, Table 1 summarizes our theoretical concepts, how they apply to the U.S. case, how we operationalize each concept and our predictions for each measure. Below, we discuss our operationalization of concepts in greater detail.

Table 1.

Summary of Key Concepts and Measures.

Stage 1: selection
Concept	U.S. case	Measures	Predicted relationship
Implementing revolutionary programs under favorable conditions	Union troops	(1) Union troops as % of total county population (1865–1872)	NA (used as explanatory variable in subsequent stages)
Stage 2: social transformation
Concept	U.S. case	Measures	Predicted relationship
Out-group social transformation	Black literacy	(1) Black literacy in 1870 (placebo), 1900, 1910, 1920	Increase
		(2) Literacy differences relative to whites in 1870 (placebo), 1900, 1910, and 1920	Decrease
		(3) Difference-in-difference: black literacy in 1850 (pre-treatment) and 1900, 1910, and 1920 (post-treatment)	Increase
Stage 3: violence
Concept	U.S. case	Measures	Predicted relationship
In-group resentment toward out-group	White-on-Black lynching	1) Count of whites lynching black persons (1911–1920 and 1921–1930)	Increase
		(2) Difference-in-difference: white-on-black lynching 10-years post-reconstruction (1877–1886), 20-years post-reconstruction (1877–1896), and 50-years post-reconstruction (1877–1926)	Increase

Implementing revolutionary programs

To measure the implementation of revolutionary programs, we primarily focus on Union troop size as a percentage of county population in 1860 (Troops as % Population) obtained from Mapping Occupation dataset (Downs & Nesbit, 2015). We focus on this measure because Union troops both created a safe space for the implementation of revolutionary programs and also implemented revolutionary programs by transferring ideational and material assets to social out-groups, as discussed above. Furthermore, because other post-conflict reconstruction initiatives—like the 13th, 14th and 15th Amendments—ostensibly applied to all men equally, Union troops as a percentage of county population provides sub-national variation we can exploit. Finally, because we are interested in the interaction between revolutionaries, revolutionary programs and people, we are especially interested in Union troops as a percent of the total county population, as opposed to Union troops over space, but we still consider the latter measure as a robustness check (see Supplemental Appendix Tables A1 and A2). We consider counties with more troops as a percent of the total population to be associated with a more robust implementation of revolutionary programs.

Figure 1 shows the percent of average troops deployed to a county by population of the county.¹⁵ In the first stage of analysis that measures selection, Union troops as a percent of total county population is our outcome measure, but in the second and third stage of our analysis, Union troops as a percent of the total county population is one of our key explanatory variables.

Figure 1.

Implementation of revolutionary programs, by county.

Measuring social transformation in response to revolutionary programs

To measure social transformation we use historical census data on Black literacy rates in 1870, 1900, 1910, and 1920, the years for which literacy by race is available (Minnesota Population Center, 2016). Increases in literacy correspond to minimal social transformation. We also estimate differences in literacy between Black persons and whites to test if the confluence of revolutionary programs, safe spaces and social out-group participation reduced social inequities between social groups. Because data on literacy rates are not available by both race and age, we assess the effects of Union troops on literacy rates of the entire Black population, rather than the school-aged Black population. Table 2 summarizes our education variables.

Table 2.

Census Education Variables.

Percent	Black		White		Difference
Percent	Mean	N	Mean	N	Difference
Can write, 1870	42.8	1036	81.8	1036	−39.0
Literate, 1900	67.8	1033	91.2	1036	−23.4
Literate, 1910	76.9	1034	93.7	1036	−16.8
Literate, 1920	75.1	1022	93.6	1035	−18.5

Measuring violence in response to social transformation

We also argued that reduced hierarchies between social in-groups and out-groups arising from the implementation of revolutionary projects would stoke violent resentment among social in-groups resulting in violence against social out-groups. To measure violent white resentment in the U.S. context, following Acharya et al. (2016) we estimate the count of white-on-Black lynchings in a county over certain periods of time ranging from 10 years after Reconstruction to 50 years after reconstruction ends.¹⁶ Data on white-on-Black lynchings are drawn from Auut (2017), which synthesizes over two dozen lynching datasets by academics and researchers, while eliminating duplicates. Figure 2 plots the annual count of lynchings by race.

Figure 2.

Annual lynchings. The figure displays lynchings over time by race. The black vertical line corresponds to reconstruction’s end.

Data Structure, Sample, and Unit of Analysis

For our analysis, we create two data-sets: a cross-sectional dataset and a panel dataset for our difference-in-difference analyses. Because censuses do not always collect the same information each year or they fail to do so by race, it was not possible to use a panel dataset for all analyses, so we primarily use the cross-sectional dataset. In both datasets, our geographic unit is county. We selected the county level because historical census data are available at this level.¹⁷ Counties also provide a distinct boundary that we can define. Because county boundaries have changed over the years, we use historical county boundaries as they were defined in 1870 to match our data using the spatial boundaries created by Siczewicz (2011). In the panel data set, the temporal component falls across treatment/pre-treatment and post-treatment periods based on availability of data.

Our samples for both datasets includes the 1,040 counties for which we have data in all enslaving states.¹⁸ In the panel dataset, we use almost the same sample of counties but because of the temporal dimension of the panel dataset, it contains 2,076 observations.¹⁹

Temporally, we limit our analysis to the years after Reconstruction (after 1877) and the height of the Great Migration in the 1920s and 1930s. The Great Migration prompted Southern white elites to make structural changes prevent further Black out-migration. Although Black migration to the North began shortly after the Civil War, it was not until 1910 that the percent of Black persons living in the south dropped below 90% (Gibson & Jung, 2002). After the Great Migration, however, comparisons to measures from the era under study are thus difficult given high levels of movement (Tolnay & Beck, 1995, pp. 213; 219–233).

Stage 1: Estimating Non-Random Selection in the Implementation of Revolutionary Programs

We begin by assessing variation in the robustness of the implementation revolutionary programs within secure spaces (operationalized as Union troops as percent of county population) to identify any factors that might bias estimates about the effects of these programs (Fortna, 2004; Mazumder, 2018). Understanding strategic selection is essential for identification: for instance, if places with more troops experienced higher literacy rates, this could be the result of Union troops’ efforts, or it could be because troops deployed to areas where education was already high. By modeling strategic selection in the implementation of revolutionary programs within secure spaces, we can be sure that second- and third-stage results about the effects of such programs are either unbiased by strategic selection, or we can mitigate these issues through estimation techniques.

Model specification

We first use our cross-sectional dataset of all southern counties to estimate the determinants of variation in county-level Union troop deployments by county population. We are particularly focused on five sets of factors that might explain variation in the implementation of revolutionary projects: county-level destruction during the U.S. Civil War, pre-war ease of transit to county, pre-war economic capacity, pre-war institutionalized racial resentment, and pre-war county demographics.

To measure Civil War destruction, we account for Union Battle Deaths and Confederate Battle Deaths from National Park Service (2011). More troops might be sent to counties where battle casualties were high, or where Union soldiers were victorious. Second, because troops might be more likely to be stationed in counties where it is easier to move goods and personnel, we include measures for Rail Access or Water Access measured in 1860 before the Civil War (Acharya et al., 2016). Third, to account for county capacity, we use Farm Value Per Capita, Total Cash Value of Farms, Total Improved Acreage (Acharya et al., 2016) and Terrain Ruggedness (Hornbeck & Naidu, 2014).²⁰ Fourth, we account for population demographics like population size by race and class using the variables Log of Population, Proportion Enslaved Blacks, Proportion of Free Blacks, Proportion of Small Farms, Inequality of Farm Holdings measured in 1860 in a county (Acharya et al., 2016). Lastly, troops might be more likely to be stationed in counties that are heavily economically reliant on enslavement because it was the institution of slavery that precipitated the Civil War in the first place. To measure enslaving institutions in a county, we include a measure of Cotton Suitability measured in 1860 (Acharya et al., 2016).

In all models, we also account for unobservable state-level policies with state fixed effects and report robust standard errors. We estimate our models using OLS regression. As a robustness check, in Supplemental Appendix Table A1 we use alternative measures of the distribution of revolutionary programs (any troop presence, troops per square mile, duration of troop presence, and Freedmen’s Bureau presence) to be sure there is nothing unique about our choice of measure.

Results

Table 3 reports the full results. To interpret the findings, non-significant coefficients indicate that the covariates are not systematically associated with the distribution of resources (no strategic selection). Of particular concern are counties with climates more suitable to cotton and places with more enslaved persons: according to Acharya et al. (2016), these locations are more likely to have institutionalized racial resentment which makes social transformation less likely and white-on-Black lynchings more likely.

Table 3.

Determinants of Union Troop Presence.

	(1)
	Troops as % pop
Union deaths	0.000 (0.000)
Confederate deaths	0.000 (0.000)
Log of population	−4.210 (2.176)
Prop. of enslaved	−4.736 (1.939)
Prop. free black	43.767 (38.140)
Prop. of small farms	−3.038 (3.996)
Total Improved acreage	0.000 (0.000)
Inequality of farm holdings	−11.050 (3.492)
Total cash value of farm	0.000 (0.000)
Farm value per capita	−0.009 (0.006)
Rail access	2.028 (0.543)
Waterway access	1.231 (0.448)
Cotton suitability	−5.240 (3.826)
Terrain ruggedness	−0.012 (0.005)
Constant	48.799 (23.148)
State fixed effects	X
Observations	1039

OLS coefficients are reported. All independent variables except union deaths and confederate deaths are from 1860. State fixed effects are included and robust standard errors are reported.

What we find is that Union troops as a percent of the total population were deployed to more accessible locations, but not counties with higher rates of racial resentment, nor counties that have higher pre-war levels of economic capacity. We find a negative relationship with the between the proportion of enslaved persons and Union troops as percent of population. These results suggest that in the following analyses, we are able to estimate the effect of military resources without concerns of selecting into “hard” or “easy” counties, more or less resistant to change or more or less prone to violence.²¹

Because we are concerned with causal identification, we take steps to mitigate non-random selection by using a difference-in-difference estimation strategy in the following stages.

Stage 2: The Effects of Revolutionary Programs on Social Transformation

We now turn to the socially transformative consequences of revolutionary projects implemented within protected spaces. Here, the relevant counterfactual compares the degree of social transformation in a county where revolutionaries implemented projects under more favorable conditions to a place where these conditions were absent or not present to the same degree. In other words, Union troops are not the sole vehicles through which other social transformative projects could have unfolded: public schools opened, the southern economy partially industrialized, charitable and religious societies worked to provide services throughout the South, and the Federal Government introduced programs like the Freedmen’s Bureau, discussed above. We argue that Union troops cause an additional degree of social transformation in addition to these factors.

Model specification

In this analysis, we use our cross-sectional county-level dataset to evaluate the socially transformative effect of revolutionary programs administered within a protected space—measured as the presence of Union troops as a percent of county populations—on social transformation—measured as Black literacy and Black-white differentials in literacy.²²

In all models, we replicate almost the same analysis conducted in Stage 1 by including almost all the same covariates (discussed in detail above) to address county-level variation in experiences with Reconstruction (Duration of Troop Presence from Downs and Nesbit (2015) and Freedmen’s Bureau from Google, n.d.), population demographics and size by race and class measured in 1860,²³ county-level capacity measured in 1860,²⁴ ease of transit and accessibility in 1860,²⁵ and the extent of institutionalized racial resentment measured in 1860.²⁶ Across all models, we use OLS regression with state fixed effects and robust standard errors. As a robustness check, in Supplemental Appendix Table A2 we replicate the same analysis described above but replace Union troops as a percent of the total county population with Union troops per square mile.

Results

The full results for the effects revolutionaries transferring assets within secure spaces (Union troops as percent population) on social transformation (black education) are reported in Table 4. We begin with a placebo test in column 1, which measures the effects of political and military resources on Black literacy rates in 1870. We selected this census year because Reconstruction had not yet concluded, so we would not expect to see the full effects of Reconstruction by this time. This test indicates that revolutionary resources were not simply amplifying trends that began before Reconstruction as Union troops as a percent of the total county populations had no effect on literacy rates in 1870. In 1900, 1910, and 1920, however, we find that Black literacy rates were lower in counties with more Union troops. Figure 3a presents the effects of troop size as percent of the population on Black literacy rates in 1900, 1910, and 1920.

Table 4.

Black Literacy Rates (Social Transformation).

	(1)	(2)	(3)	(4)
	1870	1900	1910	1920
Union troop presence	−0.482 (0.373)
Troops as % population		0.076 (0.029)	0.064 (0.014)	0.093 (0.029)
Months of presence		0.003 (0.009)	0.013 (0.003)	0.018 (0.005)
Freedmen’s bureau	0.360 (0.717)	0.361 (0.355)	0.438 (0.288)	1.156 (0.282)
Log of population	−0.833 (0.609)	2.972 (0.805)	0.839 (0.644)	0.477 (0.878)
Prop. of slave	−9.797 (3.604)	−15.625 (3.492)	−10.667 (1.908)	−14.739 (3.360)
Prop. free black	−19.452 (10.826)	−1.500 (8.580)	−14.245 (8.108)	−37.006 (11.959)
Prop. of small farms	8.105 (3.081)	1.275 (4.123)	−1.206 (1.626)	1.888 (3.843)
Total improved acreage	0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)
Inequality of farm holdings	−13.392 (5.170)	4.801 (7.163)	3.741 (2.206)	4.792 (4.730)
Total cash value of farm	−0.000 (0.000)	−0.000 (0.000)	−0.000 (0.000)	0.000 (0.000)
Farm value per capita	0.006 (0.006)	0.012 (0.003)	−0.000 (0.003)	−0.001 (0.003)
Rail access	0.977 (0.664)	0.941 (0.339)	1.033 (0.323)	1.132 (0.331)
Waterway access	−0.919 (0.498)	1.542 (0.478)	0.879 (0.340)	1.077 (0.417)
Cotton suitability	4.818 (3.729)	6.492 (4.601)	3.469 (2.108)	7.185 (3.794)
Terrain ruggedness	0.006 (0.019)	0.016 (0.005)	−0.003 (0.009)	0.002 (0.009)
Constant	52.282 (6.133)	37.289 (8.467)	69.260 (6.375)	67.783 (9.366)
State fixed effects	X	X	X	X
Observations	1036	1034	1035	1023

OLS regression coefficients are reported. The dependent variable is the literacy rate of the black population for that time period. Other than occupation variables, independent variables are from 1860. State fixed effects are included and robust standard errors are reported.

Figure 3.

Revolutionary programs increase literacy. (a) Black literacy rates. (b) Differences in black-white literacy. The figures report the marginal effect of union troops as percent of the population (a) on county-level black literacy rates for 1900, 1910, and 1920. Dashed lines represent 95% confidence intervals. The positive relationship supports our hypothesis that more resources increases black literacy. Figure (b) presents the marginal effects of Union troops as percent of the population on differences between white and black literacy rates for 1900, 1910, and 1920. Dashed lines represent 95% confidence intervals. The negative relationship supports our hypothesis that the implementation of revolutionary programs decreases differences between Black and white literacy rates.

We also compare rates of Black education to their white counterparts in the same county, to determine whether revolutionary projects successfully reduced inequities between races. The dependent variables are constructed as differences in literacy rates of whites in comparison to Blacks. Table 5 displays the full results. A negative coefficient in columns 1 to 4 of Table 5 corresponds to a larger percentage of the Black population being literate in comparison to whites. Figure 3b graphically presents these findings. We find that revolutionary programs implemented within a protected space statistically significantly reduced literacy inequities between Black and white populations over the long-term. Together, the results suggests that the implementation of revolutionary projects improved absolute levels of literacy rates among Blacks, and diminished inequities between whites and Blacks.

Table 5.

Black-White Differences in Literacy Rates (Reductions in Social Hierarchies).

	(1)	(2)	(3)	(4)
	1870	1900	1910	1920
Troop presence	1.286 (0.725)
Troops as % population		−0.101 (0.026)	−0.081 (0.018)	−0.108 (0.021)
Months of presence		0.007 (0.018)	−0.004 (0.008)	−0.009 (0.009)
Freedmen’s bureau	0.489 (0.517)	−0.010 (0.544)	−0.242 (0.331)	−1.160 (0.396)
Log of population	2.781 (1.699)	−2.653 (0.606)	−0.804 (0.484)	−0.766 (0.736)
Prop. of slave	15.095 (4.922)	20.436 (3.965)	13.708 (2.181)	17.705 (3.670)
Prop. free black	2.268 (20.701)	−8.502 (17.167)	7.140 (9.189)	31.417 (11.659)
Prop. of small farms	−18.489 (6.160)	−6.077 (4.845)	−2.348 (1.744)	−4.901 (3.100)
Total improved acreage	−0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
Inequality of farm holdings	11.720 (11.159)	−4.110 (9.486)	−2.725 (3.483)	−3.432 (5.823)
Total cash value of farm	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
Farm value per capita	−0.001 (0.009)	−0.015 (0.007)	−0.003 (0.002)	−0.005 (0.005)
Rail access	1.342 (1.047)	−0.206 (0.271)	−0.735 (0.322)	−0.807 (0.319)
Waterway access	1.533 (1.212)	−1.599 (1.030)	−0.687 (0.562)	−0.879 (0.501)
Cotton suitability	13.082 (4.147)	0.782 (5.434)	2.833 (2.446)	−0.176 (4.775)
Terrain ruggedness	0.000 (0.019)	−0.024 (0.007)	−0.003 (0.009)	−0.010 (0.010)
Constant	5.000 (13.571)	48.208 (6.775)	21.729 (4.805)	25.689 (7.870)
State fixed effects	X	X	X	X
Observations	1036	1034	1035	1023

The dependent variable is the difference in literacy rates between white population and the black population for that time period. Other than occupation variables, independent variables are from 1860. State fixed effects are included and robust standard errors are reported.

Difference-in-difference

To compliment the primary analysis above and address concerns of non-random selection of revolutionary programs, we also estimate the effects of the implementation of revolutionary programs on our literacy response variables using difference-in-difference models with our panel dataset.

Difference-in-difference models require pre-treatment measures of our dependent variables, meaning that in addition to data on literacy rates from 1900, 1910, and 1920 described above, we also need data on literacy rates prior to the Civil War, and so we use data on literacy rates from the 1850s. Importantly, the 1850 census only measures literacy rates of free Blacks, and does not take into account the literacy of enslaved Blacks. Because education was legally denied to enslaved persons in almost all enslaving states (Davis, 2011, p. 106), we therefore assume that literacy among enslaved populations was zero. While it is possible that the actual percent is higher, Margo (1990) notes that the overwhelming majority of enslaved persons were illiterate.²⁷ We calculate total 1850 Black literacy as the literacy rate for free Black persons listed in the census, weighted by the 1850 proportion of the free Black population in a county, and the assumed literacy of 0% for enslaved Blacks, weighted by the proportion of the Black population who were enslaved in a county in 1850. Our computed average literacy rate was 4.5% for all Black persons in the South (compared to 91% for whites, 81.3% for free Blacks). The literacy rate we calculate is actually higher than other sources with sub-national data on all Black literacy rates before the Civil War: Du Bois’ data on Georgia suggest that total Black literacy (free plus enslaved) in 1860 was closer to 1% (Du Bois, 1900).²⁸

To ease interpretation of our results, we use a binary measure of Union Troop Presence coded as a “1” if any Union troops were stationed in a county during Reconstruction, and a “0” otherwise. We estimate our models using OLS regression. To facilitate comparisons across all models, we include the same covariates included in Tables 4 and 5: county-level variation in experiences with Reconstruction,²⁹ population demographics and size by race and class measured in 1860,³⁰ county-level capacity measured in 1860,³¹ ease of transit and accessibility in 1860,³² and extent of institutionalized racial resentment measured in 1860.³³

In terms of interpretation, difference-in-difference approaches compare four observations: treated observations at time t (pre-treatment), untreated observations at time t (pre-treatment), treated observations at time t+1 (post-treatment), untreated observations at time t+1 (post-treatment). Thus, we first compare the 1850 literacy rates in counties that would eventually have Union troops to those that would not to assess systematic differences between them. We then look within treated counties over time to assess the effect of the presence of Union troops on literacy rates in the same counties in 1900, 1910, and 1920 (after Reconstruction), and we also compare these effects with untreated counties over time. The difference-in-difference approach allows us to control for differences between counties as well as changes within counties over time.

Supplemental Tables A3, A4, and A5 in the Appendix report the full results of the analysis. The interaction term and lower-order constituent terms must be interpreted appropriately. The Time coefficient captures the change in literacy rates in untreated counties, meaning counties without Union troops, over time (untreated observations at time t + 1). The coefficient for Troop Presence refers to the pre-treatment literacy rates of counties in 1850 (treated observations at time t). The interaction of Time × Troop Presence refer to the implementation of revolutionary programs on treated observations in 1900, 1910, and 1920, or treated observations at time t+1. The baseline literacy rate of untreated observations in 1850 (time t) must be calculated independently, when values for the variables of Time and Troop Presence variables are set to zero.

The large, positive and statistically significant coefficient of Time × Troop Presence indicates that the presence of Union troops increased Black literacy rates in 1900, 1910, and 1920. At the same time, white literacy rates at these times are unaffected by Union Troop presence.

Like an interaction term, these coefficients cannot be interpreted independently. To facilitate interpretation, we set all control variables to their means, and calculate the predicted effect of revolutionary resources over time. Figure 4 graphically presents the predicted effect of literacy rates in a county, as well as the confidence intervals for these predictions with all other variables set to their means. We also report whether the predicted differences in literacy rates between treated observations post-treatment and untreated observations post-treatments were statistically significantly different from one another. Supplemental Appendix Table A6 also presents this information in another format.

Figure 4.

Predicted literacy rates from difference-in-difference models. (a) We reject the null hypothesis that the post-reconstruction predicted black literacy rates in counties with troops are the same as predicted black literacy rates in counties without troops in the 1850–1910 and 1850–1920 models (p-value_1850–1900: .24; p-value_1850–1920: .07; p-value_1850–1900: .005). (b) We cannot reject the null hypothesis that the post-reconstruction predicted white literacy rates in counties with troops are the same as predicted white literacy rates in counties without troops across models (p-value_1850–1900: .90; p-value_1850–1920: .41; p-value_1850–1900: .46).

What we find is that literacy rates across both treated and untreated observations increase precipitously from 1850 to 1900, 1910, and 1920. But independent of this positive upward trend, the presence of Union troops also exerts a statistically significant effect on increasing literacy rates in most models, consistent with expectations, showing an average increase in Black literacy of at least one to two percentage points per county, and this difference is statistically significant. As a placebo test, we also estimate the same model using white literacy rates as our response variable to ensure that Union troop presence improved Black literacy specifically and not literacy generally. Our placebo test confirms that Union troop presence improved the literacy rates of Black persons specifically.

Substantive interpretation of estimates

According to the results presented in Tables 4 and 5, a 10% increase in the size of Union Troop presence in a county relative to its population corresponds to a 1% increase in Black literacy and a 1% reduction in differences in Black literacy rates relative to white literacy rates. Similarly, our difference-in-difference results reported in Supplemental Table A6 indicate that any Union troop presence was associated with about a 1% to 1.5% increase in Black literacy rates in a given county. While a 1% to 1.5% increase may seem small, from 1900 to 1920, the total Black population across all counties with any Union troop presence ranged in size from 4.6 million to 4.9 million. A 1% to 1.5% increase in Black literacy rates in these counties would therefore mean that between 46,000 and 74,000 additional Black persons acquired literacy as a result of Union troop presence. Thus, when contextualized in terms of human livelihoods, the effects of Union troop presence on social advancement are not insignificant.

Stage 3: Social Transformation Provokes Violence

In the previous stage, we found that the implementation of revolutionary programs caused social transformation in the form of improvements in Black literacy rates and also reduced social inequities between Black persons and white persons. Our theory predicts that a change in status hierarchies between social in-groups and out-groups makes in-group violence against upwardly mobile social out-groups more likely. We test this element of our theory here.

Model specification

To directly test whether social transformation causes violence, we use census data from our cross-sectional, county-level dataset to create four explanatory variables that capture social transformation: change in Black literacy from 1900 to 1910 and from 1910 to 1920, and differences in the over time change in literacy between Black persons and white persons from 1900 to 1910 and from 1910 to 1920. These variables tell us the rate at which Black literacy increased, and rate in which Black literacy increased in comparison to whites, respectively. We expect that larger increases in Black literacy rates, or when Black literacy increases at a faster rate than white literacy, that resentment among whites toward Black will be most acute, therefore triggering greater violence by whites against Blacks.

To evaluate the relationship between social transformation and violence, we first test whether an increase in Black literacy rates is associated with an increase in the count of white-on-Black lynchings in a county over the 10-year period immediately following the year literacy was last measured. Data for white-on-Black lynchings are from Auut (2017). For example, we estimate the effect of the difference in the rate of change between Black and white literacy rates from 1900 to 1910 on the number of white-on-Black lynchings from 1911 through 1921. We anticipate a positive relationship between change in Black literacy and the count of white-on-Black lynchings.

Next, we test the effect of differences in the rates of change in Black and white literacy over a 10- or 20-year period on lynching over the 10-year period literacy was last measured. In this case, we anticipate that places where Black literacy rose at a faster rate than white literacy rates will have more white-on-Black lynchings. Negative coefficients indicate that Black literacy rose at a faster rate than white literacy and produced a greater number of white-on-Black lynchings in a county in the subsequent period of time.

To maximize comparability across models, we use the same covariates in our previous two Stages presented above (county-level variation in experiences with Reconstruction,³⁴ population demographics and size by race and class measured in 1860,³⁵ county-level capacity measured in 1860,³⁶ ease of accessibility in 1860,³⁷ extent of institutionalized racial resentment measured in 1860³⁸). Given the count construction of our dependent variable, we use negative binomial regression with state fixed effects and robust standard errors in all models.

As a robustness check, in Supplemental Appendix Table A7, we also includes measures for electoral competition to account for the fact that rates of white-on-Black lynching could accelerate when under intense electoral competition. Results remain unchanged. In Supplemental Appendix Table A8 and A9, we re-analyze the data using OLS regression. Though not as strong as the results presented here, they are consistent in direction of the negative binomial results.

Results

Columns 1 and 2 of Table 6 report the effects of changes in Black literacy in a 10-year period on the count of white-on-Black lynching in the following 10-year period. The positive coefficients support our hypothesis: increases in Black literacy rates triggered white resentment and whites lynched more Blacks. Columns 3 and 4 of Table 6 report the effect of changes in white literacy in comparison to Black literacy in a 10-year period on the count of white-on-Black lynching in the following 10-year period. The negative coefficient shows that in places where Black literacy increased faster than white literacy, there is a positive association with lynching. Figure 5 presents the results graphically.

Table 6.

White-on-Black (In-Group Violence in Reaction to Social Transformation).

	(1)	(2)	(3)	(4)
	1911–1920	1921–1930	1911–1920	1921–1930
∆ in black lit. 1910–1900	0.020 (0.008)
∆ in black lit. 1920–1910		0.026 (0.013)
Diff ∆ in black and white lit. 1910–1900			−0.017 (0.008)
Diff ∆ in black and white lit. 1920–1910				−0.027 (0.013)
Log of population	0.452 (0.169)	−0.081 (0.198)	0.441 (0.166)	−0.085 (0.198)
Prop. of slave	0.497 (0.748)	1.535 (0.886)	0.461 (0.757)	1.534 (0.887)
Prop. free black	−14.580 (8.001)	−9.161 (7.524)	−14.528 (7.943)	−9.265 (7.563)
Prop. of small farms	−0.749 (0.848)	0.497 (0.879)	−0.775 (0.851)	0.473 (0.885)
Total improved acreage	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
Inequality of farm holdings	−0.486 (1.372)	1.735 (1.505)	−0.419 (1.380)	1.670 (1.505)
Total cash value of farm	−0.000 (0.000)	0.000 (0.000)	−0.000 (0.000)	0.000 (0.000)
Farm value per capita	0.002 (0.001)	0.001 (0.001)	0.002 (0.001)	0.001 (0.001)
Rail access	−0.224 (0.172)	0.423 (0.202)	−0.228 (0.173)	0.417 (0.202)
Waterway access	0.122 (0.159)	0.179 (0.195)	0.122 (0.160)	0.178 (0.194)
Cotton suitability	2.599 (0.722)	2.365 (0.938)	2.568 (0.719)	2.383 (0.940)
Terrain ruggedness	−0.010 (0.005)	−0.004 (0.005)	−0.010 (0.005)	−0.004 (0.005)
Constant	−5.310 (1.798)	−4.855 (2.005)	−5.099 (1.761)	−4.738 (1.996)
lnalpha	0.243 (0.192)	−0.819 (0.583)	0.246 (0.192)	−0.817 (0.580)
State fixed effects	X	X	X	X
Observations	1032	1023	1032	1023

Negative binomial regression coefficients are reported. The dependent variable is the count of white-on-black lynchings in a county during that time period. We estimate the effect of changes in black literacy on white-on-black lynchings. State fixed effects are included in all models. Robust standard errors are reported.

Figure 5.

Social transformation provokes violence. (a) Change in black literacy. (b) Difference in change in black-white literacy. Figure (a) reports the predicted probability of an additional lynching in a county when the county-level change in black literacy rates is at the 10th and 90th percentile, but all other covariates are held at their means, in 1911–1920 and 1921–1930. The positive relationship supports our hypothesis that increases in black literacy cause higher rates of violence by in-groups toward out-groups. Figure (b) presents the predicted probability of an additional lynching in a county when the county-level change in Black literacy rates relative to whites is at the 10th and 90th percentile, but all other covariates are held at their means, in 1911–1920 and 1921–1930. The negative relationship supports our hypothesis that the greater the reduction in social status hierarchies, the greater the likelihood of violence by in-groups toward out-groups. In both figures, vertical lines represent 95% confidence intervals.

Difference-in-difference

To once again account for potential non-random selection in the implementation of revolutionary programs, we replicate the same difference-in-difference estimation strategy used in Stage 2 of our analysis to estimate the effects of Union troop presence on the count of white-on-Black lynchings in a county over a given period of time. We include all the same covariates, state fixed effects, robust standard errors and we report negative binomial regression estimates.

Supplemental Appendix Table A10 reports the full results of the difference-in-difference analysis of revolutionary on lynching. We find that over time, white-on-Black lynchings in both treated and untreated counties increases. However, Union troop presence exerts an independent, positive effect on white-on-Black lynchings, compared to counties without Union troop presence. Once again, the strong, positive and statistically significant coefficient for Time × Troop Presence indicates that the presence of Union troops increased white-on-Black lynchings 10, 20, and 50 years after Reconstruction ended. For simplicity in interpretation, Supplemental Table A11 presents the predicted effect of Union troops on white-on-Black lynchings over time, with all other variables set to their means. We again include confidence intervals and report the probability that the two predicted values are the same. Figure 6 presents these same results graphically. The results indicate that 10, 20, and 50 years after Reconstruction, the expected count of white-on-Black lynching was higher in counties that experienced Union Troop presence compared to those that did not, and these differences are statistically significant. The results of the analyses support our hypothesis and theoretical expectations. These results are also reported in a tabular format in Supplemental Appendix Table A11.

Figure 6.

Estimated counts of lynching per county. We reject the null hypothesis that the post-reconstruction expected counts of white-on-black lynchings in counties with troops are the same as the expected counts of white-on-black lynchings in counties without troops across all models (p-value_{Reconstruction+10 years}: 0.03; p-value_{Reconstruction+20 years}: 0.005; p-value_{Reconstruction+50 years}: 0.06).

Substantive interpretation of estimates

Although the predicted effect sizes may once again seem small, when contextualized in terms of human lives across the entire South, these effects are not insubstantial. For instance, as demonstrated in Figure 6, our estimates indicate that the presence of Union troops is associated with about a 20% to 75% increase in the expected count of lynchings in a county, depending on the model and time period. Ten years after Reconstruction, counties that experienced Union troop presence are estimated to have 75% more white-on-Black lynchings than counties that never experienced Union troops presence. According to data from Auut (2017), from 1877 to 1886, there were 154 white-on-Black lynchings in counties without Union troops. Had untreated counties experienced Union troop presence, we would expect to observe a 75% increase in expected counts of white-on-Black lynchings in these counties, meaning an additional 116 white-on-Black lynchings in just 10 years. Our estimates also suggest that over the 50-year period after Reconstruction ended, the expected count of white-on-Black lynchings in counties with Union troops is about 20% higher than in counties with no Union troop presence. According to our data from Auut (2017), from 1877 to 1926, there were 1,206 white-on-Black lynchings in counties without Union troop presence. If these untreated counties experienced Union troop presence, we would observe an additional 241 Black people lynched by whites across these counties over a 50-year time period.

Likewise, with all covariates at their means but when Change in Black Literacy Rate moves from the 90th (a large increase in Black literacy rates) to the 10th percentile (a smaller increase in Black literacy rates), the expected count of white-on-Black lynchings in a county decreases by about 25% in 1911 to 1920 (Table 6, Model 1), and by about 30% in 1921 to 1930 (Table 6, Model 2). Contextualized substantively, from 1911 to 1920, there were 43 white-on-Black lynchings in the counties where the increase in Black literacy rates was at or above the 90th percentile. If Black literacy rates in these counties had increased at the same, slower pace of counties where the change in Black literacy rates was at or below the 10th percentile, we would expect a 25% decrease in the expected counts of white-on-Black lynching, meaning we would expect about 11 fewer Black persons lynched by whites across this subset of counties from 1911 to1920. When Black social advancement was limited, whites lynched fewer Black persons in comparison to places where Black social advancement was more pronounced and reduced Black-white social inequities.

Robustness Checks

In our theory, we argued that out-group communal institutions sustained social transformation, but as a result, could also be a target of violence. In the appendix, we test this implication using a measure of Black churches. We find that counties with more Black churches have both higher literacy rates and higher rates of white-on-Black lynching, even when accounting for Black population size and other county-level covariates in our previous stages (see Supplemental Appendix Tables A12 to A14). Finally, we focus on literacy to capture social transformation. Though data on economic factors by race are not widely available during our period of study (Margo, 1984, 1990), we also explore the relationship between Union troops and Black wages in 1940 (Supplemental Appendix Table A15). Results are consistent with expectations.

External Generalizability

While we examine an historical case of attempted social transformation, our findings nevertheless have contemporary relevance. Work by Huntington (2006, p. 266) and Skocpol (1979, pp. 163–164) refer to the process of changing social orders as statebuilding (also referred to in these texts as state-building or state building). While many might conceptualize statebuilding exclusively as a foreign process, Lake (2016, p. 1) notes that both contemporary foreign statebuilding and internal statebuilding both entail the creation of a new social order (see also, Barma, 2017).

We also see many of the same patterns identified here in contemporary cases of foreign statebuilding. Foreign statebuilding interventions are sometimes socially and politically transformative in their objectives (Barma, 2017)—such as moving from a one-party dictatorship to a democracy,³⁹ or the imposed requirement of women’s nearly equal participation in economic, social, and political life.⁴⁰ In Afghanistan, for example, statebuilding attempts to reduce gender inequities succeeded in fostering women’s political representation (Beath et al., 2013) and women’s literacy improved to almost 20% after women were almost totally excluded from educational opportunities. (Kurtz et al., 2018, p. 8). But beliefs about women’s role in the family and private life stayed unchanged (Beath et al., 2013), and women and girls continued to face extreme, often lethal violence by simply attending school.⁴¹

Similarly, though reforms are less extreme than the revolutionary projects discussed here, research finds that when reforms make progress toward socio-economic equity, they trigger less extreme resentment among privileged populations. For instance, when welfare—political resources aimed to reduce social inequities—is tied to race, it typically faces a backlash from whites (Brown, 2013). Such findings therefore support our contentions here that altering social hierarchies triggers resentment among social in-groups.

Conclusion

In this paper, we argued that when revolutionaries succeed in reducing social hierarchies, they are also more likely to provoke violence. As social out-group institutionalize social transformation over the long-term, violence too persists over time. We tested our claims and proposed mechanisms with historical subnational data on U.S. Reconstruction. Our quantitative results, illustrative anecdotes, and scholarly works on Reconstruction and lynching all support our theory. Moving forward, new research could apply our framework to understanding the consequences of revolutionary schemes undertaken by other political actors, such as rebels or empires.

Ultimately, this paper speaks to fundamental questions about how political actors consolidate order over people and space, and the challenges and tradeoffs they face in doing so. We demonstrate that patterns of social transformation and violence occur together, and that social and political engineering in pursuit of equality between social groups is frequently matched by in-groups’ violent efforts to preserve their social standing. In so doing, we identify one mechanism that explains why social transformation is often so challenging: social transformation and violence are intimately and causally intertwined.

Supplemental Material

sj-pdf-1-cps-10.1177_0010414021997164 – Supplemental material for Social Transformation and Violence: Evidence from U.S. Reconstruction

Supplemental material, sj-pdf-1-cps-10.1177_0010414021997164 for Social Transformation and Violence: Evidence from U.S. Reconstruction by Megan A. Stewart and Karin E. Kitchens in Comparative Political Studies

Footnotes

Acknowledgements

We thank Ahsan Butt, James Igoe Walsh, Laia Balcells, Elizabeth Saunders, Kathleen McNamara, David Edelstein, Jordan Tama, Emily Ritter, Adi Dasgupta, Casey Dorff, Bridget Coggins, Joseph Young, Susanna Campbell, Cathy Schneider, Boaz Atzili, Daniel Esser, Yang Zhang, Timothy Luke, Anjali Dayal, Alexandra Stark, Rebecca Davis Gibbons, Haillie Lee, Jacob Shapiro, Carl Müller-Crepon, and participants of the DC-Area International Relations Workshop, American University’s Political Violence Research Working Group, the Online Peace Science Colloquium, participants at ESOC lab meetings, the editors at CPS and three anonymous reviewers.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Megan A. Stewart

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Megan A. Stewart is an Assistant Professor at the School of International Service at American University. Her book has been published with Cambridge University Press and her articles have appeared in International Organization, Journal of Politics, Journal of Conflict Resolution, among other outlets.

Karin E. Kitchens is an Assistant Professor in the Political Science Department at Virginia Tech. Her articles have been published at Journal of Politics, State Politics & Policy Quarterly, Politics & Gender, among other outlets.

References

Acemoglu

Robinson

J. A.

(2006). De facto political power and institutional persistence. The American Economic Review, 96(2), 325–330.

Acharya

Blackwell

Sen

(2016). The political legacy of American slavery. The Journal of Politics, 78(3), 621–641.

Adler

(2019). World ordering: A social theory of cognitive evolution. Cambridge University Press.

Arnove

R. F.

Graff

H. J.

(1987). National literacy campaigns: Historical and comparative perspectives. Springer Science & Business Media.

Auut

(2017). Monroe work today compiled dataset (version 0.9) [Database]. http://www.monroeworktoday.org/metadata.html

Barma

(2017). The peacebuilding puzzle: Political order in post-conflict states. Cambridge University Press.

Beath

Christia

Enikolopov

(2013). Empowering women through development aid: Evidence from a field experiment in Afghanistan. American Political Science Review, 107(3), 540–557.

Bensel

R. F.

(1990). Yankee Leviathan: The origins of central state authority in America, 1859-1877. Cambridge University Press.

Brewer

M. B.

(1999). The psychology of prejudice: In group love and outgroup hate? Journal of Social Issues, 55(3), 429–444.

10.

Brown

H. E.

(2013). Racialized conflict and policy spillover effects: The role of race in the contemporary US welfare state 1. American Journal of Sociology, 119(2), 394–443.

11.

Chacon

Jensen

(2017). De facto power, democracy, and taxation: Evidence from military occupation during reconstruction. Cambridge University Press.

12.

Chien

Y.-W.

(1973). The Taiping revolutionary movement. Yale University Press.

13.

Connell

(1993). Against all odds: A chronicle of the Eritrean revolution: With a new foreword on the postwar transition. The Red Sea Press.

14.

Cox Richardson

. (2009). The death of reconstruction. Harvard University Press.

15.

Darden

Grzymala-Busse

(2006). The great divide: Literacy, nationalism, and the communist collapse. World Politics, 59(1), 83–115.

16.

Davis

A. Y.

(2011). Women, race, & class. Vintage.

17.

Donald

D. H.

Baker

J. H.

Holt

M. F.

(2016). The civil war and reconstruction. Golden Springs Publishing. Kindle Edition.

18.

Downs

G. P.

(2015). After appomattox. Harvard University Press.

19.

Downs

G. P.

Nesbit

(2015). Mapping occupation: Force, freedom, and the army in reconstruction. http://mappingoccupation.org

20.

Du Bois

W. E. B

. (1900). [The Georgia Negro] illiteracy. According to the Library of Congress: “Chart prepared by Du Bois for the Negro Exhibit of the American Section at the Paris Exposition Universelle in 1900 to show the economic and social progress of African Americans since emancipation.” Retrieved March 27, 2020, from https://www.loc.gov/resource/ppmsca.33876/

21.

Du Bois

W. E. B

. (2013). Black reconstruction in America: Toward a history of the part which black folk played int he attempt to reconstruct democracy in America, 1860-1880. Transaction Publishers.

22.

Fearon

J. D.

Laitin

D. D.

(2003). Ethnicity, insurgency, and civil war. American Political Science Review, 97(1), 75–90.

23.

Finkel

(2017). Ordinary Jews: Choice and survival during the holocaust. Princeton University Press.

24.

Foner

(2011). Reconstruction: America’s unfinished revolution, 1863-1877. Harper Collins.

25.

Fortna

V. P.

(2004). Does peacekeeping keep peace? International intervention and the duration of peace after civil war. International Studies Quarterly, 48(2), 269–292.

26.

Freire

(2005). Pedagogy of the oppressed. Bloomsbury Publishing.

27.

Frohnen

(2008). The American nation: Primary sources. Indianapolis: Liberty fund. Retrieved May 11, 2017, from http://oll.libertyfund.org/titles/2282

28.

George

A. L.

Bennett

(2005). Case studies and theory development in the social sciences. MIT Press.

29.

Gibson

Jung

(2002). Historical census statistics on population totals by race, 1790 to 1990, and by Hispanic origin, 1790 to 1990, for the United States, regions, divisions, and states. US Census Bureau.

30.

Google. (n.d.). Freedmen’s bureau field offices [Database]. Records based on National Archive Microfilm. https://www.google.com/maps/d/u/0/viewer?mid=11kGBhLi6gJ9jvXqF9I2X47uHAgk&hl=en_US&ll=33.960461351713974%2C-86.24256812499999&z=6

31.

Hahn

(2003). A nation under our feet: Black political struggles in the rural South from slavery to the great migration. Belknap Press of Harvard University Press.

32.

Haines

M. R.

, & ICSPR. (2010). Historical, demographic, economic, and social data: The United States, 1790-2002. Dataset manual. Inter-university Consortium for Political and Social Research.

33.

Hewstone

Rubin

Willis

(2002). Intergroup bias. Annual Review of Psychology, 53(1), 575–604.

34.

Horn

S. F.

(1969). Invisible empire: The story of the Ku Klux Klan, 1866-1871 (Vol. 81). Patterson Smith Publishing Corporation.

35.

Hornbeck

Naidu

(2014). When the levee breaks: Black migration and economic development in the American South. American Economic Review, 104(3), 963–990.

36.

Houtart

(1980). The social revolution in Eritrea. In: Davidson

Cliffe

Selassie

B. H.

(Eds.), Behind the war in Eritrea. Spokesman Books.

37.

Huizer

(1968). Peasant organization in the process of agrarian reform in Mexico. Studies in Comparative International Development (SCID), 4(6), 115–145.

38.

Huntington

S. P.

(2006). Political order in changing societies. Yale University Press.

39.

Hyman

H. M.

(1959). To try men’s souls: Loyalty tests in American history. University of California Press.

40.

Ingham

J. N.

(2003). Building businesses, creating communities: Residential segregation and the growth of African American business in southern cities, 1880–1915. Business History Review, 77(4), 639–665.

41.

Johnson

C. A.

(1982). Revolutionary change (Vol. 47). Stanford University Press.

42.

Jones

P. M.

(1991). The Agrarian law: Schemes for land redistribution during the French revolution. Past & Present, 133, 96–133.

43.

King

D. S.

Smith

R. M.

(2005). Racial orders in American political development. American Political Science Review, 99(1), 75–92.

44.

Kurtz

Tesfaye

Wolfe

(2018). Impacts of vocational training and cash transfers on youth support for political violence in Afghanistan. Mercy Corps.

45.

Lake

D. A.

(2016). The statebuilder’s dilemma: On the limits of foreign intervention. Cornell University Press.

46.

Lenin

. (1919). First all-Russia congress on adult education. Published in the pamphlet: N. Lenin, Two speeches at the first all-Russia congress on adult education, Moscow, 1919; Published according to the pamphlet. Retrieved November 29, 2018, from https://www.marxists.org/archive/lenin/works/1919/may/06.htm

47.

Leonard

(1988). Popular participation in liberation and revolution. In: Cliffe

Davidson

(Eds.), The long struggle of Eritrea for independence and constructive peace (pp. 105–135). Red Sea Press.

48.

Licklider

(1995). The consequences of negotiated settlements in civil wars, 1945–1993. American Political Science Review, 89(3), 681–690.

49.

Lieberman

R. C.

(2002). Ideas, institutions, and political order: Explaining political change. American Political Science Review, 96(4), 697–712.

50.

Logan

T. D.

(2018). Do black politicians matter? Technical report. National Bureau of Economic Research.

51.

Lyons

P. B.

(2014). Statesmanship and reconstruction: Moderate versus radical republicans on restoring the union after the civil war. Lexington Books.

52.

March

J. G.

Olsen

J. P.

(2004). The Oxford handbook of political science. Oxford Handbook Online.

53.

Margo

R. A.

(1984). Accumulation of property by southern blacks before world war I: Comment and further evidence. The American Economic Review, 74(4), 768–776.

54.

Margo

R. A.

(1990). Race and schooling in the South, 1880-1950: An economic history. University of Chicago Press.

55.

Mathiez

(1920). Le Bolchevisme et le Jacobinisme [Bolshevism and Jacobinism]. Librairie du Parti Socialiste et de l?HumanitÃľ Trans. Mitchell Abidor. Retrieved March 22, 2018, from https://www.marxists.org/history/france/revolution/mathiez/1920/bolshevismjacobinism.htm

56.

Mazumder

(2018). The persistent effect of the U.S. civil rights movement on political attitudes. American Journal of Political Science, 62(4), 922–935. https://doi.org/10.1111/ajps.12384

57.

Minnesota Population Center. (2016). National historical geographic information system: Version 11.0 [Database]. https://www.nhgis.org/

58.

Moise

E. E.

(2017). Land reform in China and North Vietnam: Consolidating the revolution at the village level. UNC Press Books.

59.

Mondlane

(1983). The struggle for Mozambique. Zed Books.

60.

Montgomery

W. E.

(1995). Under their own vine and fig tree: The African-American church in the South, 1865-1900. LSU Press.

61.

Moore

(1993). Social origins of dictatorship and democracy: Lord and peasant in the making of the modern world (Vol. 268). Beacon Press.

62.

National Park Service. (2011). Civil war battle summaries of major battles [Database]. http://www.archive.org/details/CivilWarBattleSummariesOfMajorBattles

63.

North

D. C.

Wallis

J. J.

Weingast

B. R.

(2009). Violence and social orders: A conceptual framework for interpreting recorded human history. Cambridge University Press.

64.

Opper

(2018). Revolution defeated: The collapse of the Chinese Soviet Republic. Twentieth-Century China, 43(1), 45–66.

65.

Ozouf

(1989). A critical dictionary of the French revolution. Harvard University Press.

66.

Petersen

R. D.

(2002). Understanding ethnic violence: Fear, hatred, and resentment in twentieth-century Eastern Europe. Cambridge University Press.

67.

Pfeifer

M. J.

(2011). The roots of rough justice: Origins of American lynching. University of Illinois Press.

68.

Pierson

(2000). Increasing returns, path dependence, and the study of politics. American Political Science Review, 94(2), 251–267.

69.

Quillian

(1995). Prejudice as a response to perceived group threat: Population composition and anti-immigrant and racial prejudice in Europe. American Sociological Review, 60(4), 586–611.

70.

Ruef

(2014). Between slavery and capitalism: The legacy of emancipation in the American South. Princeton University Press.

71.

Ruef

Grigoryeva

(2018). Jim Crow, ethnic enclaves, and status attainment: Occupational mobility among US Blacks, 1880–1940. American Journal of Sociology, 124(3), 814–859.

72.

Scott

J. C.

(1998). Seeing like a state: How certain schemes to improve the human condition have failed. Yale University Press.

73.

Sherman

W. T.

(1865/2020). Order by the Commander of the Military Division of the Mississippi [Special Field Orders, No. 15.]. First issued 16 January 1865 by William Tecumseh Sherman. Updated 14 February 2020 by Steven F. Miller. Retrieved April 10, 2020, from http://www.freedmen.umd.edu/sfo15.htm

74.

Shortreed

(1959). The antislavery radicals: From crusade to revolution 1840-1868. Past & Present, 16, 65–87.

75.

Siczewicz

(2011). U.S. historical counties [Spatial Data]. In: Long

J. H.

(Ed.), Atlas of historical county boundaries. Emily Kelley. http://publications.newberry.org/ahcbp

76.

Siegelbaum

(2015). Workers organization. From the Seventeen moments in soviet history project. Retrieved August 22, 2019, from http://soviethistory.msu.edu/1917-2/workers-organization/

77.

Skocpol

(1979). States and social revolutions: A comparative analysis of France, Russia and China. Cambridge University Press.

78.

Skocpol

(1995). Protecting soldiers and mothers. Harvard University Press.

79.

Smångs

(2016a). Doing violence, making race: Southern lynching and white racial group formation. American Journal of Sociology, 121(5), 1329–1374.

80.

Smångs

(2016b). Interracial status competition and southern lynching, 1882–1930. Ethnic and Racial Studies, 39(10), 1849–1868.

81.

Stewart

M.A.

(2021). Governing for Revolution: Social Transformation in Civil War. Cambridge University Press.

82.

Stewart

M. A.

Kitchens

K. E.

(2020). Replication data for: Social transformation and violence: Evidence from U.S. Reconstruction. https://doi.org/10.7910/DVN/CCHS9C

83.

Tajfel

Turner

J. C.

(1979). An integrative theory of intergroup conflict. The Social Psychology of Intergroup Relations, 33(47), 74.

84.

Tilly

(1977). From mobilization to evolution. Addison-Wesley.

85.

Tolnay

S. E.

Beck

E. M.

(1995). A festival of violence: An analysis of southern lynchings, 1882-1930. University of Illinois Press.

86.

United States Congress. (1872). Report of the Joint Select Committee to inquire into the condition of affairs in the late insurrectionary states, made to the two houses of congress February 19, 1872. Retrieved May 6, 2019, from https://archive.org/details/reportofjointsel01unit/page/10

87.

Viola

(2000). The role of the OGPU in dekulakization, mass deportation, and resettlement in 1930. The Carl Beck Papers in Russian and East European Studies, 1406, 55.

88.

Walter

B. F.

(2015). Why bad governance leads to repeat civil war. Journal of Conflict Resolution, 59(7), 1242–1272.

89.

Weld

(2019). The other door: Spain and the Guatemalan counter-revolution, 1944–54. Journal of Latin American Studies, 51(2), 307–331.

90.

Wells

I. B.

(2012). Southern horrors: Lynch law in all its phases. Apple Books Edition.

91.

Williamson

O. E.

(2000). The new institutional economics: Taking stock, looking ahead. Journal of Economic Literature, 38(3), 595–613.

92.

Zhang

Lee

M. M.

(2019). Speaking like a state: Language and state-society interactions in 19th century France. American Journal of Political Science, 64(4), 1001–1016.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.45 MB