Practicing What They Preach? Lynching and Religion in the American South, 1890 – 1929

Abstract

This project employs a moral solidarity framework to explore the relationship between organized religion and lynching in the American South. We ask whether a county’s religious composition impacted its rate of lynching, net of demographic and economic controls. We find evidence for the solidarity thesis using three religious metrics. First, our findings show that counties with greater religious diversity experienced more lynching, supporting the notion that a pluralistic religious marketplace with competing religious denominations weakened the bonds of a cohesive moral community and might have enhanced white racial solidarity. Second, counties in which a larger share of the black population worshipped in churches controlled by blacks experienced higher levels of racial violence, indicating a threat to the prevailing moral community or inter-group racially based solidarity. Finally, we find a lower incidence of lynching in counties where a larger share of church members belonged to denominations with racially mixed denominations, suggesting that cross-racial solidarity served to reduce racial violence.

During the five-decade period between the fall of Reconstruction and the onset of the Great Depression, American communities roiled in religious fervor. The Social Gospel movement propelled Progressive policies in Northern cities and within U.S. federal government. Millennialist belief spawned new religious movements, and newly-emancipated blacks created their own religious institutions. This same half-century witnessed such a high rate of lynchings in the American South – roughly one each week – that it is commonly referred to as the “lynching era.”

Prior work has explored the relationship between lynching and institutional features of southern life, including economic, political, legal, and cultural conditions (Finnegan 1997; Hovland and Sears 1940; Soule 1992; Tolnay and Beck 1995), the southern media (Ames 1938; Perloff 2000; Wasserman, 1998;), and cultural support for racist attitudes and behaviors (Berends 2004; Brundage, 1997; Oliver 2001). This inquiry seeks to identify what role, if any, religious organizations played in racial violence, by examining the link between lynching and the local religious landscape. In light of the power of other social institutions to shape the contours of racial violence, a comprehensive investigation into religion’s influence on lynching seems overdue, particularly given the impact that Christian organizations had on southern social life and local power structures, and the racialized nature of southern denominations (Brundage 1993: 218; Hill et. al. 1972; Schweiger and Matthews 2004; Wilson 1980).

INTRODUCTION

Lynching is defined by the NAACP as an extra-legal killing perpetrated by three or more individuals, who claimed their murderous actions were intended to uphold justice or tradition (Tolnay and Beck 1995: 260). While lynchings have occurred throughout U.S. history, across geographic regions, and have targeted individuals from various racial and ethnic backgrounds (Pfeiffer 2004), certainly the most widely-known incarnation, and the one we focus on here, is the lynching of African Americans in the Jim Crow-era south. Whites were occasionally lynched in southern states, typically in sparsely-populated, predominantly white counties, and targeting individuals accused of murder, rape, or both, supporting a “popular justice” explanation (Beck and Tolnay 1997). Scholars generally agree, however, that the campaign of terror waged against American blacks was intended to suppress political participation and economic independence.

This work is the first systematic exploration into the role that the characteristics of the local religious marketplace may have played in the incidence of racial violence. We adopt aspects of Tolnay and Beck’s (1995) groundbreaking work, which linked lynching to cotton production and economic stress, to control for well-known community-level contributors to racial violence. In this way, we build on and extend Tolnay and Beck’s (1995) research by incorporating religious measures into a religious economy model while including their key explanatory variables as well as additional important demographic controls (Finke 1989). Thus, we are able to assess the role of religion within a rich model that is comparable to findings from prior research. The main thrust of our analyses, however, focus on how the kinds of religious institutions wielding influence at the local level may have affected race relations and lynching incidence, net of economic and demographic factors. Religious structures and cultural restrictions limiting contact between dominant and subordinate groups were critical to the (white) Southern “way of life” (Hill et. al. 1972). But to what extent might religious institutions have served to under-gird or dampen the racial caste system and racial violence?

This paper employs a moral solidarity theoretical perspective to explain why a county’s religious environment might have mattered for its trajectory of lynching activity. Religious institutions may be viewed as producing a “moral community” (Durkheim 1984; 2001), enhancing solidarity within groups through shared beliefs, symbols, practices, and experiences. We test for evidence of solidarity in several ways, reflecting different dimensions of social cohesiveness. First, we examine the socially integrative effects of religious organizations and their potential to create – or reinforce – moral community and impact racial violence. In addition, we assess religion as a sign of solidarity based on racial divisions or cross-racial solidarity and ask whether it became a marker for racial and regional identities, intensifying racial conflict. Integrating Beck and Tolnay’s (2004) inventory of lynch victims with county-level census data on local economic, demographic, and religious conditions, we ask whether a county’s religious composition influenced its rate of lynching, net of a variety of demographic and economic controls. Our primary unit of analysis is the number of lynching events by county¹ in ten Southern states (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, and Tennessee) across four decades. We use negative binomial regression to measure religious influence on county lynching events during four separate decades and combine data for a pooled analysis to explore the impact of religious variables over time.

LYNCHING, RELIGION AND RACE

The link between lynching and race is unquestionable, particularly within the American South. At least 2,500 blacks are known to have been so murdered in former Confederate States during the lynching era – a rate of roughly one mob killing every week for five decades (Tolnay and Beck 1995: 17). While lynchings occurred throughout the United States in the late 19^th and early 20^th centuries, and targeted people from a variety of racial and ethnic backgrounds (Pfeifer 2004), the practice was substantially more frequent in the South and 90% of Southern victims were African American (Citation X). Of the victims in Tolnay and Beck’s (1995) lynching inventory, 94% died at the hands of white lynch mobs². A small proportion of cases were perpetrated by mixed-race mobs, or were black-on-black lynchings but this was not typical. Thus, we are primarily interested in lynching of blacks in the American South in the period from 1890–1929, and the role it played in the system of racial stratification. Southern lynching fulfilled a uniquely important role in suppressing the African American population that was not shared by other regions (e.g., Brundage 1993; Tolnay and Beck 1995), although it was only one among a panoply of tactics employed. While we acknowledge that lynchings did occur outside of these parameters – time, location, and racial composition – we argue that different dynamics might be used to explain other lynching events.³

Research on lynching has focused on other institutional or cultural factors and largely overlooked the role of religion and churches in episodes of racial violence. The studies that have explored the influence of religion on lynching mostly use case studies or anecdotal evidence to focus on the role of religious leaders or parishioners as opposed to religion as an institution. Evidence on the role of individual church members and ministers in or around lynchings is mixed. Many authors implicate clergy and congregants in particular lynching events (Brundage 1997; McGovern 1982:67; Raper 1969 [1933]). For example, Reinhold Neibhur’s, an American theologian, stated that “If there were a drunken orgy somewhere, I would bet ten to one a church member was not in it… But if there were a lynching, I would bet ten to one a church member was in it” (quoted in Grant 1975: 87). Indeed, church leaders often provided justification for specific lynchings (Miller 1957). Others report that it was ministers and lay members who were at the fore of the anti-lynching movement (Brundage 1993; Hall 1993), and that they attempted to intervene to prevent some lynchings (Miller 1957).

White religious institutions in the post-bellum South were intimately involved in maintaining the existing racial hierarchy. Indeed, the predominant Christian organizations disseminated a theology rooted in white superiority, and providing ideological justification for separation of the races became a primary function of white religious groups (Bailey 2005; Berends 2004; Harvey 2004; Matthews 2004; Snay 1993; Spain 1961). Christianity was used to support slavery based on biblical and evangelistic reasons (Emerson and Smith 2000). Most local and national religious organizations were racially segregated, and to the extent that blacks were incorporated into white-controlled denominations, they were typically relegated to separate congregations, or separate areas of the church in mixed-race congregations.

The historical regional divergence of American religious institutions predates the Civil War. Because the plantation was the first southern institution, and all other institutions developed within its sphere, other social structures, including the church, were forced to accommodate its needs (Hill et. al. 1972: chapter 3). Southern white churches resisted institutional changes that would alter the social arrangements of the day – both racial hierarchy and the political and economic domination enjoyed by the landed elite (Spain 1961; Wilson 1980), and most failed to challenge the racialized status quo (Brundage 1993; Raper 1969 [1933]). Many national white Protestant denominations, including Methodists (Farish 1969), Baptists (Spain 1961) and Presbyterians (United States Bureau of the Census 1906), split over the issue of slavery in the decades prior to the Civil War, with the separate Northern and Southern organizations spawned by these denominational splits maintained well into the twentieth century.

White Southern religious groups voiced unanimous, enthusiastic support for maintaining racial segregation, a theological perspective at odds with mainstream present-day interpretations of Christian doctrine. It requires an extraordinary logical leap, however, to move from advocating racial segregation and racial hierarchy to encouraging racial violence. It is conceivable that churches could promote segregationist views while simultaneously rejecting racial violence. We do not contend that religion per se caused racial violence, nor that religious conviction suppressed impulses toward collective violence. Rather we focus on ways that religion, as an important social institution, influenced the shape of local racial conflict and thus played a mitigating or exacerbating role in the incidence of lynching.

National bodies of both northern and southern denominations adopted platforms objecting to lynch law, although the groundswell of opposition to racial violence did not compel these stances until the prevalence of lynching was on the wane in the 1920s (Miller 1957). In fact by the early 1930’s, no organized church failed to officially condemn all lynchings (Grant 1975). Brundage (1993) also identifies the use of Christian morality in condemning lynching and cites church organizations combating mob violence in Virginia in the late 1890s and early 1900s. Raper highlights the disjuncture between national religious bodies’ condemnation of mob violence and the lack of outcry from local organizations (1933: 21). Despite individual ministers or church members occasionally intervening, both local churches and broader religious organizations failed to take a consistent public position against racial mob-violence.

LYNCHING, RELIGION AND MORAL COMMUNITY

Religious organizations’ public positions on social matters such as lynching have the potential to impact their adherents’ moral beliefs and behaviors. Religious institutions may be viewed as producing a “moral community” (Durkheim 1984; 2001), enhancing solidarity within groups through shared beliefs, symbols, practices, and experiences. A moral community can be defined as a group drawn together by an interest in living according to a common moral philosophy. While moral communities are often associated with benevolent ideals, they need not be, nor need they be associated with specifically religious philosophies. As Durkheim observed, moral communities can also take shape around a secular philosophy, such as nationalism.

As repugnant and immoral as the ideology of white supremacy may appear to contemporary sociologists, those drawn together by it into a philosophy of life understood themselves in moral terms. This is consistent with Smith’s (2003) research on morality and social action that argues violations of the moral order are not behaviorally consistent due to the complexity of moral systems. For example, in the South a Christian could be compassionate toward the poor while acquiescing to – or even participating in –episodes of racial violence. While these can be regarded as inconsistent behaviors within in a Christian moral system, the coexistence of such divergent moral behaviors was central to the structure of Southern life and race relations because both would be in support of the “good.” Sociologically, to say that a group has moral community is not to say that those values are benevolent or universal but rather that they produce cohesiveness among group members. In the culture of Southern white supremacy, lynching produced the solidarity that helped to bolster racial hierarchy.

Religion frequently reinforces the moral order and is a key factor in social control (Stark and Bainbridge 1997), operating through social integration (attachments and social networks) and institutions (churches) to inculcate moral conformity (Stark et al. 1983). Beliefs and rituals express the ideals and norms of a moral community. It is out of shared experiences of the sacred in a ritually activated group that moral communities form and reproduce themselves over time. In this sense, the culture of white supremacy was no different from what Durkheim observed in other settings. Central to white supremacy was the notion that the demands of Christian morality and good citizenship necessitated racial separation and white social and political dominance. Miscegenation and “race-mixing” of all kinds, black criminality, and defiance of the racial order were all constituted as grave threats to the “moral community” of Southern society. In the case of the American South, the connection between religion and society was strong and not unlike what Durkheim (1984, p. 119) described of earlier historical periods where “everything social was religious.”⁴ In this way religion’s embeddedness in people’s collective conscience, norms and customs strengthened the social solidarity that was produced through religious community.

In the enactment of white supremacist moral community, lynchings played a vital role. They were fully ritualized, solidarity-enhancing events in the Durkheimian sense (Collins 2004; Durkheim 1984; 2001). In many of the communities during our period of study, lynching was not a furtive act nor generally regarded as shameful affair. Rather, lynching served as a public spectacle that often took on carnival-like characteristics – “festivals of violence”, as Tolnay and Beck (1995) aptly put it. Not only did large crowds often gather to murder the victims but also assembled to view the corpse, to mangle further or disfigure it in highly ritualized fashion, to pose for photographs, and to enjoy musical performances and orations. The emotional resonance of lynching was extended beyond those immediately present through a lively trade in picture-postcards and souvenirs taken from lynching victims (Allen et. al. 2000).⁵

These patterns are indicative of a ritual process premised on reinforcing racial solidarity among whites and expressing a “moral” lesson to blacks. Following the Civil War, white Southern Protestantism played a central part in merging the Lost (Confederate) Cause and the assertion of white supremacy (Berends 2004; Matthews 2004; Wilson 1980), enabling simultaneous membership in “two moral communities, church and South” (Hill et. al. 1972: 46). Public ceremonies venerating the Confederacy and (white) southern “way of life” (Wilson 1980: 34) were major ecumenical events. Religious leaders were deeply entangled with Southern political and legal authorities that officially sanctioned Jim Crow segregation (Ayers 1995; Feldman 2005). The overlapping of racial/ethnic and religious cleavages often intensifies inter-group struggles by merging religious and racial identities (Hechter 1999; Kunovich and Hodson 1999; Williams 1994). In this context, traumatic social change and economic threat -- as was experienced in the post-Reconstruction American South – are further prone to heighten ethnic intolerance among religious adherents (Hodson et. al. 1994).

Durkheim’s assertion that religion produces moral conformity is rooted in his assumption that mechanically integrated societies have a moral consensus based on common membership in the same church. The effect might be expected to attenuate when community members are divided into several competing religious organizations. In other words, a common Christianity might be expected to produce moral consensus – in this case in support of white supremacy – but if Christian denominations with different doctrines and rituals coexist, this may heighten competition within that locality and weaken the bonds of moral community (Finke and Stark 2005; Inverarity 1976).

Ordinarily, we might expect that if these religious cleavages overlap with other socially-meaningful identities, the local religious landscape may predict the erosion of racial consensus, if not intra-group conflict over the issue (Williams 1994). However, in the Jim Crow South, competition among religious organizations for members may have led organizations to sharpen their racial positions in order to differentiate themselves (Emerson and Smith 2000), potentially breeding extremism in attempts to capture a niche in the religious market. New and insurgent religious organizations eager to demonstrate legitimacy may have most eagerly embraced prevailing cultural norms and popular ideology, thereby reinforcing norms of racial solidarity (Thomas 1989).

Paradoxically, Christian organizational fragmentation might have enhanced white racial solidarity as preachers and religious entrepreneurs strove to demonstrate their racist credentials and expressly confirm community morals. Thus, a greater diversity of ideas in the local religious marketplace (Finke and Stark 1992, 2005; Stark and Bainbridge 1987; Stark and Finke 2000) may impact racial violence. If religious competition occurred without putting white supremacist ideology into question but rather reinforcing it through a general endorsement of the principal, than the effect of denominational heterogeneity among Southern whites may have been to reinforce the salience of this basic element of white Christian moral community rather than undermine it.⁶

Most scholars of Southern lynching utilize competition/threat frameworks (e.g. Olzak 1990, Soule 1992, Tolnay and Beck 1995). Alternatively, some investigators have examined eruptions of collective violence from Durkheimain perspectives of anomie and moral breakdown (Evans 2009; Godoy 2006). These perspectives are not necessarily inconsistent. Indeed, we incorporate elements of the competition/threat framework into a moral solidarity thesis. First we acknowledge that blacks could be perceived as threatening in spheres of economic competition as has been previously found to have direct effects on lynching prevalence. Furthermore, the religious moral community in the South that we posit was premised on white supremacy. Religious competition and fragmentation were thus also threatening to the consensus underlying the racial order. Some churches and sects progressed toward “Northern ideals” and “racial integration” while others held tight to the ideals of racial purity and white moral superiority. Thus, we hypothesize that the level of religious diversity would signal a threat to Southern racial consensus. In a context where the morality of white supremacy might be perceived at under threat, lynching may have served to reaffirm or restore the moral community – however perverse.

Stemming from a moral solidarity perspective, Erikson’s seminal work (1966) focuses on the integrative role that such acts play in re-establishing community boundaries and values. Godoy (2006) similarly identifies that collective violence served to unify communities in which moral community was eroding. Some scholars interrogate the related “legal breakdown” and “popular justice” claims (such as Inverarity 1976), or Durkheim’s laws of quantitative and qualitative change (Massey and Myers 1989), but the results have typically been weak.⁷ Senechal de la Roche’s work, while employing Black’s paradigm of pure sociology,⁸ rather than Durkheim, focuses on micro-level processes linked to conditions of community embeddedness, and particularly the differences in level of community integration between the victim and perpetrators. She hypothesizes (1997) that “functional interdependence” – a concept reminiscent of organic solidarity – should dampen racial violence by increasing the likelihood that more powerful individuals would intervene on behalf of the intended victim.

Evans (2009) directly embeds his comparative analysis of racial violence in South Africa and the historical American South within a Durkheimian perspective, and argues that racial violence served a symbolic purpose, as a ritual means of maintaining consensus among whites (p. 245), linking religious feeling and group identity to the Southern practice of lynching (p. 249). Godoy (2006), conversely, frames lynching in contemporary Latin America as “desperate” acts undertaken by communities bereft of social capital and unable to otherwise integrate members into a common moral community. She terms this type of collective violence “the lowest common denominator of social control” (p. 104). Within this context, if denominational splintering signaled the erosion of moral community, some groups may have reacted with violence in attempts to regain collective cohesion. Despite these excellent recent works on the role of collective violence in enhancing moral community, the bulk of the research on lynching focuses on competition over directly quantifiable valued resources – typically in the economic or political realm.

Religion and Solidarity Hypotheses

We suspect the incidence of lynching varies across ecological units to the degree that moral integration varies within those units. As moral community is not directly measurable – a non-material social fact – we use indicators from the religious sphere. In the case of the American South, churches might effectively integrate members but not necessarily suppress lynching behavior, since, in the context of the moral community rooted in the prevailing racial hierarchy, this form of racial violence would not necessarily be regarded as deviant behavior nor offensive to the collective conscience (Durkheim 1984). If few important religious groups questioned white racial supremacy and most strove to reaffirm it through loud and visible endorsements (Bailey 2005; Berends 2004; Harvey 2004; Matthews 2004; Snay 1993; Spain 1961), then the competition born of diversity may have provided ongoing occasion to remind citizens of their moral community. We argue that religious community in this period in the American South reflects a Durkheimian notion of organic solidarity. Consequently, our first hypothesis predicts that higher levels of religious pluralism in a county will increase the county’s level of lynching.

If we conceptualize religion as producing solidarity among white Southern Protestants, the relative strength of religious groups that challenge prevailing social arrangements might influence a community’s level of racial violence. The kinds of denominations that were present in the local religious marketplace – and the level of support they provided to the existing power structures – might create greater impetus for economically and martially humiliated whites to defend their status. In the Jim Crow South, where racially-segregated religious organizations used theology to support or subvert the prevailing racial caste system (Hill et. al. 1972; Matthews 2004; Schweiger 2004; Wheeler 1986), racial violence may have been influenced by the kinds of religious organizations wielding local influence.

Strength of Black Churches

Religious separatism among oppressed minority groups can be a form of protest, and may have served as both an emancipatory structure for blacks and a catalyst for racial violence. Religion may be used by subordinate groups to assert group identity, aggravating ethnic tension (Hechter 1999). Black churches, the only institution blacks controlled, promoted economic self-sufficiency, literacy, and political and civil rights for African Americans during the late nineteenth and early twentieth centuries (Anderson 1988; Dollard [1937] 1988; Frazier 1964; Hill 1980; Mays and Nicholson 1933; Montgomery 1992; Spain 1961; Wheeler 1986), directly challenging white privilege (Brundage 1993:110), and serving as a “catch all” node of religious, economic, social and political strength (Brundage 1993; Montgomery 1992). The strength or presence of black churches might reflect a threatening racially-based religious solidarity that served as an institutional base for challenging prevailing economic and political power structures, and may also serve as an institutional indicator for elevated levels of racial conflict. Thus, our second hypothesis predicts a positive relationship between lynching and the percent of a county’s black population that belongs to black-controlled denominations.⁹

Mixed Race Denominations

While the North was becoming more religiously diverse during the late 19^th century with growing secularization of public institutions as a consequence (Smith 2003; Stark 2001), Southern religion became more homogenous and racially segregated, and Southern culture was increasingly dominated by conservative Protestantism (Hill 1980; Spain 1961; Wilson 1980). “Mixed race” denominations might have reflected a move toward cross-racial solidarity, or a willingness among whites to work cooperatively with blacks. This in turn might lead to the decay of a white-dominated moral community. Shared religious experiences and organizational resources may have reduced inter-group racial conflict.¹⁰ To the extent that there was a breakdown in the collective conscience based on race there might be more cross-racial solidarity resulting in less racial violence. It is also possible that the presence of interracial religious groups may have lowered the perception of racial threat since whites probably continued to control institutional resources and doctrinal messages in these organizations.¹¹ Our third hypothesis, then, suggests that we will find a lower incidence of lynching in counties where a larger percent of religious adherents belong to organizations with mixed-race memberships.

All hypothesized relationships between religion and lynching are assumed to be net of cross-sectional variation in economic distress, cotton dependence, and demographic composition (Finke 1989; Tolnay and Beck 1995). This paper examines religion as an institutional arena in which indicators of moral solidarity are measured. In this way, examining the role of religion, a key social institution, may offer new insights into the phenomenon of southern mob violence.

DATA AND METHODS

Data for tis paper come from a variety of sources. The first, compiled by Beck and Tolnay (2004), documents 2,805 lynchings in 10 southern states between 1882 and 1930, and includes the date, county, and state of each lynching.¹² We also employ the Censuses of Religious Bodies (CRB), a series of surveys of religious organizations compiled by the U.S. Bureau of the Census, providing county-level membership statistics for American religious organizations in 1906, 1916, and 1926. In addition, the CRB provides historical narratives for each denomination, including data on racial membership composition. Religious data for 1890 are taken from the U.S. Census Report on the Statistics of Churches and include the same information available in the Censuses of Religious Bodies. The final datasets are from the Inter-University Consortium for Political and Social Research’s Historical, Demographic, Economic, and Social Data: The United States, 1790–1970, constructed from United States Census Bureau statistics and providing information on the socio-demographic composition and economic/agricultural features of U.S. counties. We use these sources to construct decade-specific data sets for 1890, 1900, 1910 and 1920.

Unit of Analysis

Our analytic unit in most cases is the county, and we use the term “county” to discuss our measures and results. However, a small minority of observations in each decade are actually clusters of counties. For our analyses using a pooled sample to examine change over time, the geographic unit of analysis is the smallest indivisible unit with consistent boundaries between 1890 and 1930 (Horan and Hargis 1995). For decade-specific analyses, in instances where county boundaries were altered over the course of a decade – due to the creation of new counties from pre-existing geographic units, or because contiguous counties adjusted their shared boundary – counties were “bundled” to include the smallest indivisible geographic unit that was consistent throughout the decade.

Dependent Variable

Our dependent variable is the number of lynching events with black victims per county per decade. While we have annual counts of lynching events, temporal collapsing of the data is desirable for several reasons. First, lynchings were relatively rare events, and the number of lynchings varied substantially – both geographically and temporally – over the forty-year period under investigation. Using decadal, rather than annual, measures provides more statistical leverage on the data. Additionally, data on counties’ demographic, economic, and religious characteristics provide decennial enumerations rather than annual counts, rendering annual analyses of the data less meaningful.

We estimate the number of lynching events with black victims – rather than the number of black victims – in each county within each decade for both empirical and theoretical reasons. Theoretically, we are most interested in the structural factors that enable the collective impulse for violent outbursts to develop into a completed lynching. It seems plausible that the number of victims killed by a lynch mob may be more a function of the number of people the mob encounters and is able to physically overtake than of a difference in structural support for single- versus multiple-victim lynchings. Measuring lynching based on a count of events rather than of victims reduces the degree of difference between counties and should provide a more accurate picture of structures that enabled racial hostilities to express themselves through collective violence. Because we are most interested in capturing the incidence rather than the magnitude of lynching, it is most appropriate to use a count of lynching events with at least one black victim.¹³ Additionally, 86% of lynching incidents involved a single black victim, and trends in number of victims track quite closely with trends in number of lynching events (Tolnay and Beck, 1995: 31). Note that incidents which simultaneously targeted black and white victims are included.

Independent Variables

We predict lynching events using a combination of religious, economic/agricultural, historical, and demographic factors.

Religious Variables

Our measures of the county’s religious environment include:

Religious Diversity

We use the complement of Herfindahl’s (1950) index of market concentration, which accounts for the number of competing firms – in this case, religious groups – and their relative sizes, producing lower diversity scores in economies that are dominated by a monopoly or oligarchy of groups. Religious diversity is calculated as 1−Σp_i² (Voas et. al. 2002), where p is the percent of a religious group’s members among members of all religious groups¹⁴.

Religious Composition

We calculate two compositional measures. We first measure the relative strength of denominations organized and led by blacks and operating outside the auspices of white religious organizations (e.g., African Methodist Episcopal Church) by dividing the number of members in these black churches by the total black population.¹⁵ Second, we include a measure of the percent of religious adherents who belong to interracial denominations by dividing the number of religious adherents in mixed race (black and white) denominations by the total number of religious adherents.¹⁶

Economic Variables

We incorporate three measures of local economic conditions (Tolnay and Beck 1995). To control for local dependence on the cotton economy we include the percent of the county’s agricultural land dedicated to cotton production. We also calculate the percent of farmers who are tenants¹⁷. Finally, we incorporate a measure of the overall value of the county’s annual agricultural products, to control for the greater concentration of racial violence in more rural areas.

Prior Violence against Blacks

To control for possible temporal dependence in racial violence, we include the cumulative number of lynching events with black victims occurring between 1882 (the first year for which data are available in the Beck-Tolnay inventory) and the start of the decade under analysis.¹⁸

Demographic Variables

The strength of a community’s religious organizations is often linked to population stability and demographic composition (Finke 1989). We include a county’s sex ratio, the proportion of its residents who are under age 21, and the percent population growth in the prior decade. These demographic variables control for the destabilizing effects of migration, the presence of women (who are more likely than are men to be church members), and the possibility that natural population increase is linked to higher levels of church participation. We also include the proportion of each county’s residents who are black, to control for the effects of perceived racial threat, and a quadratic term to capture possible non-linearity (Corzine, Creech and Corzine, 1983; Raper [1933] 1969). Finally, we control for the total county population, indicating the number of potential targets for mob violence.

Detailed descriptive statistics for all variables included in our analyses appear in Table 1. They demonstrate broad variation in all measures, and a decline in lynching events over time. Of particular note for the purposes of this inquiry is the degree of variation in religious measures, both as distributions within a decade and as religious trends changed the landscape of the local religious economy across decades.

Table 1.

Descriptive Statistics for County-Level Religious, Economic, and Demographic Characteristics, 1890 – 1929

1890	Number Lynchings This Decade	Index of Religious Diversity	Black Church Strength	Percent Members in Mixed-Race Denoms.	Percent Black	Percent Farmland in Cotton	Farm Tenancy Rate	Total Value Agricultural Products	N Prior Black Lynchings	Percent Population Growth	Sex Ratio	Percent Population Under Age 21
Mean	0.81	0.72	31.63	45.87	33.80	20.89	23.26	721,666	0.40	21.58	1.03	56.81
25th Percentile	0.00	0.68	15.09	28.42	11.55	0.01	12.97	325,430	0.00	5.49	0.99	54.69
Median	0.00	0.75	26.97	42.84	30.69	21.16	20.96	571,130	0.00	15.17	1.02	57.87
75th Percentile	1.00	0.80	42.14	61.87	53.14	36.90	31.50	937,060	1.00	29.83	1.06	59.86
Minimum Value	0.00	0.08	0.00	0.00	0.06	0.00	0.00	8,900	0.00	−41.71	0.85	36.56
Maximum Value	9.00	0.91	100.00	100.00	94.00	86.04	81.28	5,801,930	6.00	282.08	1.59	66.87
1900
Mean	0.62	0.64	27.41	16.18	33.92	18.00	44.78	955,317	1.23	20.45	1.03	54.36
25th Percentile	0.00	0.56	18.90	3.25	11.31	0.00	29.29	486,422	0.00	8.11	0.99	52.22
Median	0.00	0.66	27.29	9.32	30.24	13.75	40.90	816,760	1.00	15.92	1.02	55.66
75th Percentile	1.00	0.75	35.18	20.01	53.38	33.63	61.26	1,241,106	2.00	26.22	1.06	57.62
Minimum Value	0.00	0.08	0.00	0.00	0.01	0.00	3.26	15,496	0.00	−48.28	0.00	0.00
Maximum Value	9.00	1.00	96.40	96.85	94.02	82.61	95.50	7,430,477	11.00	462.25	1.74	64.13
1910
Mean	0.47	0.71	36.73	22.09	32.77	20.21	47.23	1,680,478	1.87	12.93	1.03	52.57
25th Percentile	0.00	0.68	24.44	6.96	10.02	0.01	31.11	810,796	0.00	0.15	1.00	50.43
Median	0.00	0.75	36.05	15.39	30.44	19.92	45.07	1,352,218	1.00	10.42	1.02	54.03
75th Percentile	1.00	0.80	47.55	30.65	51.69	35.64	63.60	2,085,958	3.00	21.67	1.06	55.88
Minimum Value	0.00	0.01	0.00	0.00	0.00	0.00	3.74	8,739	0.00	−47.47	0.00	0.00
Maximum Value	10.00	1.00	100.00	100.00	94.19	67.22	95.40	20,236,108	30.00	227.00	1.28	62.54
1920
Mean	0.23	0.71	37.11	36.03	31.20	19.36	46.98	4,252,052	2.25	6.69	1.02	51.41
25th Percentile	0.00	0.68	23.14	22.27	9.58	0.00	29.51	1,870,234	0.00	−4.25	0.99	49.01
Median	0.00	0.75	35.98	31.52	28.92	19.13	44.98	3,157,381	1.00	4.25	1.02	52.88
75th Percentile	0.00	0.80	49.79	46.12	48.23	32.42	64.33	5,397,681	3.00	13.08	1.05	55.02
Minimum Value	0.00	0.05	0.00	0.86	0.00	0.00	0.00	6,959	0.00	−53.32	0.00	31.40
Maximum Value	4.00	1.00	100.00	97.92	90.77	75.22	95.43	61,713,338	28.00	258.28	1.95	61.51

Model Estimation

We estimate a series of nested regression models using Negative Binomial or “compound” Poisson regression to deal with overdispersion and the possibility that observed events are not independent (Long 1997).¹⁹ Our analysis proceeds in three stages. First, in bivariate models, presented in Table 2, we estimate a series of decade-specific relationships, regressing three religious variables on the number of lynching events in each county during each decade, as well as pooling the from all four decades. Second, we estimate baseline multivariate models to control for well-known economic and demographic factors that may be related to both the local religious marketplace Finke (1989) and to the level of racial violence. Finally, we include three religious variables -- religious diversity, the share of local blacks belonging to black-controlled religious organizations, and the percent of all church members who belong to mixed-race denominations – singly as well as together in a “religious economy” model.²⁰. The full model is represented below:

• Log μ_i (Number of Lynching Events with Black Victims) = β₀ + β_{RELIGIOUS DIVERSITY} + β_{BLACK CHURHCES} + β_{MIXED CHURCHES} + β_{PERCENT BLACK} + β_{PERCENT BLACK}² + β_{PERCENT POPULATION GROWTH} + β_{SEX RATIO} + β_{PERCENT CHILDREN} + β_{TOTAL POPULATION} + β_{COTTON DOMINANCE} + β_{AGRICULTURAL VALUE} + β_{FARM TENANCY} + β_{PRIOR VIOLENCE} + β_STATE + ε_i

Table 2.

Reduced-Form Regression Results Using Religious Variables to Predict Lynching, 1890 – 1929

	1890–1920	1890	1900	1910	1920
Index of Religious Diversity	1.973***	1.773***	0.877†	1.718**	3.552***
	(0.304)	(0.480)	(0.465)	(0.626)	(0.965)
Percent Blacks in Black Churches	−0.007***	−0.003	0.015**	0.012**	0.006
	(0.002)	(0.003)	(0.006)	(0.004)	(0.004)
Percent Members in Mixed-Race Denominations	−0.008***	−0.015***	−0.014***	−0.021***	−0.012*
	(0.002)	(0.003)	(0.004)	(0.005)	(0.006)
Number of observations:	2672	783	779	771	803

Note: Estimation equations use negative binomial regression with robust standard errors. State dummy variables are included in the model. Coefficients appear first. Standard errors are in parentheses.

^***p ≤ 0.001;

^**p ≤ 0.01;

^*p ≤ 0.05;

^†p ≤ 0.10.

In the initial analyses, we pool the data from all decades using robust standard errors to account for possible autocorrelation or heteroscedasticity, and cluster by county. The pooled models also include a time sequence variable ranging from 1 to 4, reflecting the four decades, to capture the linear time relationship. We estimate a final set of models separately by decade, again using negative binomial regression equations with robust standard errors and state fixed effects.²¹

RESULTS

Table 2 presents bivariate regression results using religious variables for each decade and for the pooled analyses. Our bivariate regression equations sequentially use three measures of the local religious marketplace to predict lynching incidence across the four decade time span and in each of four decades (1890–1899, 1900–1909, 1910–1919, and 1920–1929). Using the pooled data, religious diversity is positively associated with racial violence, indicating that communities with greater religious competition are likely to experience a larger number of lynchings. Counties in which a larger share of black residents belonged to organizations controlled by blacks, and those in which higher proportions of church members belong to racially-mixed organizations, appear to have lower levels of racial violence. In each decade, at least two religious variables are significantly related to racial violence. Without controlling for economic or demographic characteristics, or a county’s history of racial violence, a more diverse local religious marketplace was associated with higher lynching incidence in every decade. The relative strength of black-controlled denominations increased the incidence of lynching in three of the four decades. Larger proportions of church members who belong to racially-mixed denominations appear to suppress racial violence in all four decades. In supplemental analysis (results available upon request) we also found that the county’s share of all religious adherents who belonged to non-Southern denominations was associated with lower levels of lynching.

Our primary set of regression equations use pooled data for all four decades, employing a negative binomial approach with robust standard errors that will be replicated in the single-decade models. The pooled models also incorporate a linear time variable. Results from the pooled models are presented in Table 3. Multiple regression results are presented separately by decade in tables 4 (1890–1899 and 1900–1909) and 5 (1910–1919 and 1929–1929). We also estimated a second set of regression equations restricted to the 1900–1929 time period that incorporate measures of white farm tenancy, which are not available for 1890. As the results are substantively the same in both sets of analyses, we present the results that utilize data from all four decades.²²

Table 3.

Effects of Religious, Economic and Demographic Variables on Lynching Incidence Over Time, 1890 – 1929

	Baseline Model	Religious Economy	Religious Diversity	Black Churches	Mixed-Race Denominations
Index of Religious		1.034***	0.962***
Diversity		(0.281)	(0.280)
Strength of Black-Controlled		0.005*		0.005**
Denominations		(0.002)		(0.002)
Strength of Mixed-Race		−0.004†			−0.005**
Denominations		(0.002)			(0.002)
Percent Black	0.072***	0.068***	0.069***	0.072***	0.071***
	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)
Percent Black Squared	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Percent Population Growth	0.009***	0.009***	0.009***	0.008***	0.009***
	(0.002)	(0.002)	(0.001)	(0.002)	(0.002)
Sex Ratio	1.771**	1.946**	2.121**	1.788**	1.486*
	(0.637)	(0.668)	(0.689)	(0.626)	(0.622)
Percent Population	−0.008	−0.006	−0.004	−0.008	−0.012
Under Age 21	(0.009)	(0.009)	(0.009)	(0.009)	(0.009)
Population (in 1000's)	0.005***	0.005***	0.005***	0.005***	0.005***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Percent Farmland	0.012***	0.008**	0.009**	0.012***	0.011***
Devoted to Cotton	(0.003)	(0.003)	(0.003)	(0.003)	(0.003)
Total Value Agricultural	−0.000**	−0.000**	−0.000**	−0.000**	−0.000*
Products (in $10,000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Rate of Farm Tenancy	0.001	−0.002	0.003	0.001	0.000
	(0.002)	(0.003)	(0.003)	(0.002)	(0.002)
Prior Lynching Events	0.090***	0.084***	0.087***	0.089***	0.087***
	(0.012)	(0.011)	(0.012)	(0.012)	(0.012)
Time Sequence	−0.488***	−0.551***	−0.498***	−0.533***	−0.504***
	(0.050)	(0.054)	(0.049)	(0.054)	(0.052)
Intercept	−3.017**	3.873***	−4.256***	−3.129**	−2.251*
	(1.004)	(1.161)	(1.113)	(0.994)	(1.037)
Pseudo-Log Likelihood	−2281.192	−2270.456	−2275.854	−2278.301	−2277.858

Note: N=2,672. State dummy variables are included in the model. Coefficients presented first. Standard errors in parentheses.

^***p ≤ .001;

^**p ≤.01;

^*p ≤.05;

^†p ≤ .10.

Table 4.

County-Level Predictors of Lynching Incidence by Decade, Considering Religious Measures 1890 –1899 and 1900 – 1909

	1890 – 1899					1900 – 1909
	Baseline Model	Religious Economy	Religious Diversity	Black Churches	Mixed- Race	Baseline Model	Religious Economy	Religious Diversity	Black Churches	Mixed- Race
Index of Religious		0.608	0.811†				1.013*	1.129*
Diversity		(0.483)	(0.475)				(0.510)	(0.479)
Strength of Black		−0.000		0.002			0.000		0.003
Churches		(0.004)		(0.003)			(0.006)		(0.006)
Strength of Mixed-		−0.006			−0.007*		−0.011*			−0.012**
Race Groups		(0.004)			(0.004)		(0.005)			(0.004)
Percent Black	0.056***	0.051***	0.053***	0.056***	0.052***	0.070***	0.067***	0.067***	0.069***	0.069***
	(0.009)	(0.010)	(0.009)	(0.009)	(0.009)	(0.011)	(0.011)	(0.011)	(0.011)	(0.010)
Percent Black	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***
Squared	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Percent Population	−0.002	−0.002	−0.002	−0.002	−0.002	0.011**	0.011**	0.011**	0.011**	0.010**
Growth	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)
Sex Ratio	0.903	0.793	1.036	0.935	0.662	2.666**	2.609**	3.048**	2.707**	2.266*
	(0.928)	(0.957)	(0.931)	(0.934)	(0.939)	(0.960)	(1.015)	(0.998)	(0.970)	(0.979)
Percent Population	−0.016	−0.015	−0.012	−0.015	−0.017	0.002	−0.003	0.008	0.000	−0.009
Under Age 21	(0.015)	(0.016)	(0.016)	(0.015)	(0.015)	(0.018)	(0.017)	(0.018)	(0.018)	(0.017)
Population	0.002	0.002	0.001	0.002	0.002	−0.001	−0.001	−0.002	−0.001	−0.000
(in 1,000's)	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)	(0.003)	(0.003)	(0.003)	(0.003)	(0.003)
Percent Farmland	0.014**	0.011*	0.012**	0.014**	0.012**	−0.002	−0.008	−0.003	−0.003	−0.008
Planted Cotton	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
Value Agricultural	0.003***	0.003***	0.003***	0.003***	0.003***	0.002***	0.003***	0.003***	0.002***	0.003***
Products	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)
Farm Tenancy	−0.016**	−0.016**	−0.016**	−0.015**	−0.016**	0.006	0.007	0.006	0.006	0.007
	(0.005)	(0.005)	(0.005)	(0.005)	(0.014)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)
Number of Prior	0.168***	0.157***	0.163***	0.168***	0.159***	0.133**	0.113**	0.129***	0.131***	0.116**
Lynchings	(0.037)	(0.038)	(0.036)	(0.037)	(0.037)	(0.040)	(0.038)	(0.040)	(0.040)	(0.038)
Intercept	−1.406	−1.364	−2.261	−1.538	−0.630	−5. 156***	−5.584**	−6.990***	−5.580**	−4.243**
	(1.688)	(1.982)	(1.810)	(1.724)	(1.759)	(1.727)	(1.872)	(1.822)	(1.737)	(1.730)
Log Pseudo Likelihood	−842.962	−840.220	−841.543	−842.828	−840.957	−701.085	−695.116	−698.518	−700.941	−697.222

Note: State dummy variables included. Coefficients presented first, standard errors in parentheses.

^***p ≤0.001;

^**p ≤0.01;

^*p ≤0.05.

In the bivariate pooled analyses, the measure of religious diversity, the percentage of all adherents belonging to mixed race denominations and the percentage of adherents belonging to black denominations, had significant effects on the number of lynching events each county experienced. All of three variables maintained their influence on a county’s level of racial violence when controlling for demographic composition and variables related to the cotton economy. Both the index of religious diversity and the percent of the black population belonging to black churches had a positive and significant effect on the incidence of lynching. Counties with higher proportions of church members who belong to mixed-race denominations typically experience fewer lynching events. We also find strong evidence of a decline in the likelihood of lethal mob violence over time, net of all other covariates in the model. This finding supports empirical evidence of an overall decline in the rate of lynching throughout the 20^th century.

In the models including religious variables separately, the same effects obtain, although the coefficient for mixed-race denominations is more highly significant. The coefficients and standard errors of the religious variables are relatively unchanged compared to the pooled model, suggesting that these factors operated independently. We hypothesized that these religious variables captured different aspects of the potential relationship between religious institutions and racial violence – conceptualizing religious diversity as signaling the breakdown of moral community, the strength of mixed-race organizations as linked to the degree of fusion of regional, racial, and religious identities, and the strength of black churches as indicative of racial solidarity that threatened prevailing social arrangements. The results from the first model – including all three religious variables – imply that these effects were not mutually exclusive, and indeed, appear to have operated simultaneously.

Because the coefficients produced by negative binomial regression can be difficult to interpret, we have exponentiated the coefficients, creating a percentage or factor change in the incidence of lynchings. These factor-change coefficients suggest that religious diversity is the strongest religious determinant of racial violence. For a county with a diversity index of 1.0 compared to one with a diversity index of 0.0 (complete diversity vs. no diversity), we would expect to find that the incidence of lynching increased by a factor of 2.57 per decade when the other religious variables are excluded from the model, meaning we would expect to observe 2.57 times as many lynchings. Religious diversity increases the incidence of lynching by a factor of 2.77, on average, when all religious measures are controlled. Conversely, while stronger black churches appear to significantly increase—and racially mixed denominations to significantly depress—the incidence of lynching, the size of these effects are smaller. Comparing otherwise identical counties in which one county has no religious adherents in black-controlled denominations, and in the other all black residents belong to black-controlled denominations, we would expect the number of lynchings in a decade to increase by a factor of 1.65 when this measure is examined in isolation, or a 65% increase in the incidence of lynching. Counties in which all black residents belong to black-controlled denominations would expect to see their incidence of lynching increase by a factor of 1.82, meaning the incidence would be 82% higher than in counties where no blacks belonged to black-controlled denominations, when the black church measure is used in concert with other religious variables. The average effects of the strength of mixed-race denominations in the full religious economies model suggest that incidence of lynching would only be 61% as high in counties in which all religious members belonged to mixed-race organizations, compared to counties with no church members in mixed-race denominations. When strength of mixed-race denominations is used alone, the factor change is .67.

Our decade-specific analyses clearly demonstrate that the effects of these religious measures vary over time. In 1890 (table 4), none of the religious variables attain statistical significance when all three are included in the model. When religious variables are entered separately, we observe that religious diversity has a weak positive effect, and the strength of mixed-race denominations has a moderate negative effect on racial violence. In simple terms, comparing counties with no diversity to those with complete diversity, we would expect the incidence of lynching to be 2.25 times as high during the decade (at P ≤ .10), using no additional religious measures. Comparing a county in which all religious adherents belonged to mixed-race denominations with an otherwise-identical county where no church members belonged to a mixed-race organization, we would expect the number of lynchings to be 50% lower. Adding religious measures does not appreciably alter the strength or direction of the economic or demographic measures, nor does it reflect a substantial change in the goodness of fit of the model overall.

In 1900 (table 4), the effect of religious contexts on racial violence begins to emerge. In the model incorporating all three religious measures, religious diversity appears to increase the incidence of lynching, and the percentage of religious adherents in mixed-race denominations exhibits a negative relationship with racial violence. This suggests that counties with a greater number of adherents in mixed-raced churches experienced less racial violence, and counties with higher levels of religious diversity were more likely to experience lynching events. In this decade, we expect the incidence of lynching to be 3.09 times higher in a county with complete religious diversity compared to one with no diversity, if only the religious measure of diversity is included, and 2.75 times higher if the strength of black churches and mixed-race denominations is controlled. Considering the effects of mixed-race denominations in the full religious economy model, we would expect the incidence of lynching in a county in which 100% of church members belonged to mixed-race denominations would be only one-third of that experienced by an otherwise-identical county where all religious adherents worshipped in racially segregated institutions. When other religious measures are excluded from the equation, the ratio is .30. The percentage of local blacks belonging to black-controlled religious organizations was not significant in either the reduced form or the full model. The full model including all three religious variables reflects a significant change in the goodness of fit (log-likelihood x² = 11.94, P < .01).

The influence of the local religious marketplace weakens in the decade beginning with 1910, as shown in table 5. When all three religious measures are included in the 1910 model, only the negative effect of mixed race organizations attains statistical significance—and only weakly so. This effect strengthens when other religious variables are excluded from the model. Counties with a greater share of religious adherents in “outsider” institutions—those that were racially mixed—experienced lower levels of racial violence. The expected incidence of lynching in counties where the churches were fully integrated was only 41% as high as in counties with total segregation. Put another way, in communities in which the largest share of church members belonged to groups that strictly policed racial borders, the incidence of lynching was higher. Results for 1920 are also presented in table 5. In this decade, none of the religious variables are significant in the full religious economy model,²⁴ and when each religious measure is included separately, again, only the strength of racially mixed religious groups has a somewhat anemic negative effect.²⁵ When no other religious measures are controlled, we would expect that, on average, moving from a county with complete racial-religious segregation to one of full integration would result in an incidence of lynching that was only 27% as high during the 1920s, all else equal.

Table 5.

County-Level Religious, Economic, and Demographic Predictors of Lynching Incidence, 1910 – 1919 and 1920 – 1929

	1910 – 1919					1920 – 1929
	Baseline Model	Religious Economy	Religious Diversity	Black Churches	Mixed-Race	Baseline Model	Religious Economy	Religious Diversity	Black Churches	Mixed- Race
Index of Religious		0.179	0.520				1.635	1.146
Diversity		(0.747)	(0.674)				(1.042)	(0.938)
Strength of Black		−0.001		0.000			0.005		0.008
Churches		(0.004)		(0.004)			(0.006)		(0.005)
Strength of Mixed-		−0.010†			−0.009*		−0.012			−0.013†
Race Groups		(0.005)			(0.005)		(0.009)			(0.007)
Percent Black	0.722***	0.071***	0.069***	0.072***	0.071*	0.094***	0.082***	0.089***	0.092***	0.092***
	(0.013)	(0.014)	(0.013)	(0.013)	(0.013)	(0.018)	(0.020)	(0.19)	(0.018)	(0.018)
Percent Black	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001***	−0.001**	−0.001***	−0.001***	−0.001***
Squared	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Percent Population	0.008*	0.007†	0.008*	0.008*	0.007†	0.005*	0.005†	0.005*	0.005*	0.005†
Growth	(0.004)	(0.004)	(0.004)	(0.004)	(0.004)	(0.003)	(0.003)	(0.003)	(0.003)	(0.003)
Sex Ratio	−0.220	−0.398	0.074	−0.213	−0.466	0.019	0.070	0.182	0.193	−0.320
	(1.353)	(1.436)	(1.362)	(1.360)	(1.358)	(1.040)	(1.079)	(1.021)	(1.054)	(1.090)
Percent Population	−0.009	−0.017	−0.009	−0.009	−0.017	−0.013	−0.016	−0.008	−0.010	−0.025
Under Age 21	(0.019)	(0.020)	(0.018)	(0.019)	(0.020)	(0.019)	(0.021)	(0.019)	(0.019)	(0.020)
Total Population	0.002	0.003	0.002	0.002	0.003	0.000	0.000	0.000	0.001	0.000
(in 1,000's)	(0.002)	(0.002)	(0.002)	(0.002)	(0.002)	(0.003)	(0.003)	(0.003)	(0.003)	(0.003)
Percent Farmland	0.017*	0.014†	0.017*	0.017*	0.014†	−0.005	−0.008	−0.006	−0.006	−0.007
Planted Cotton	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)	(0.008)
Value Agricultural	0.001	0.001	0.001	0.001	0.001	0.000	0.000	0.000	0.000	0.000
Products ($10K)	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.000)	(0.000)	(0.001)	(0.000)	(0.000)
Farm Tenancy	−0.000	0.000	0.000	−0.000	0.000	−0.000	0.001	−0.000	0.001	0.000
	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)	(0.006)
Number of Prior	0.112***	0.102***	0.108***	0.112***	0.103***	0.070**	0.065**	0.067**	0.070**	0.068*
Lynchings	(0.021)	(0.021)	(0.021)	(0.021)	(0.020)	(0.027)	(0.025)	(0.026)	(0.027)	(0.027)
Intercept	−2.642	−1.814	−3.315	−2.671	−1.698	−4.037*	−4.824†	−5.189**	−4.844**	−2.582
	(2.069)	(2.513)	(2.098)	(2.128)	(2.202)	(1.840)	(2.571)	(2.001)	(1.883)	(2.025)
Log Pseudo Likelihood	−564.244	−562.403	−563.945	−564.241	−562.490	−378.162	−374.949	−377.451	−377.059	−376.476

Note: State dummy variables included in the model. Coefficients presented first, standard errors in parentheses.

^***p≤0.001;

^**p≤0.01;

^*p≤0.05.

Overall, findings from the pooled and the decade-specific analyses, which replicate Tolnay and Beck’s (1995) work while including demographic measures (Finke 1989) that are linked to religious participation, perform as expected in all decades. The economic variables are in the direction consistent with both competition/threat perspectives and a moral community framework. Consistent with prior research relying on competition/threat frameworks, counties with a higher level of reliance on the cotton economy experienced a higher incidence of lynching. Similarly, we find a curvilinear relationship between the incidence of lynching and the percent of the population that was black. Building from this account we include religious variables that are associated with incidence of lynching reflective of a moral community perspective. Thus the effects that we identify for the religious variables are net of contextual influences of economic distress, cotton dependence, prior racial violence and demographic factors.

DISCUSSION

This paper examined whether local religious institutional structures influenced the incidence of lethal racial violence in the American South. Specifically, we explore a solidarity thesis assessing the relationship between moral community and racial violence. Our analyses provide support for the existence of a relationship between local religious structures and lynching regimes. We find support for a solidarity perspective. First, our findings show that counties with greater religious diversity experienced more lynching, supporting the notion that a pluralistic religious marketplace with competing religions weakened the bonds of a cohesive moral community and might have enhanced the need to affirm white racial solidarity. Second, we identify that communities in which larger shares of the black population belonged to black-controlled religious groups experienced a higher level of racial violence, suggesting that racially-organized institutional challenges to the prevailing racial hierarchy may have sharpened white racial solidarity. Finally, we find a lower incidence of lynching in counties where a larger share of church members belonged to denominations with racially mixed denominations, suggesting that cross-racial solidarity served to reduce racial violence.²³

We identify that communities with greater religious diversity also experience higher levels of racial mob violence. This suggests that competing religious denominations within the same county fuelled stress among and between racial groups adding to the climate of racial violence and, encouraging lynchings as a way to cement racial solidarity among whites. A more competitive religious marketplace might have signaled a redoubling of efforts to maintain the racial hierarchy. In this way, religion served as a conduit for lynching. Religious pluralism, which in many instances weakens solidarity, actually produced within group solidarity among Southern whites. From a Durkheimian perspective, the solidarity produced that was associated with higher lynching rates is a possible latent consequence of pluralism.

We also find evidence that the composition of the local religious marketplace may have influenced a county’s lynching trajectory. Counties in which higher shares of local blacks were members of black-controlled organizations experienced a greater number of lynchings, net of demographic and economic characteristics. This suggests that institutionally-based racial challenges to the prevailing social hierarchy may have served to increase white racial solidarity and exacerbate levels of racial violence. Places where a greater proportion of adherents belonged to mixed-race denominations typically had lower incidence of lynching than did counties with similar economic and demographic profiles, and similar lynching histories. That a stronger presence of mixed-race denominations reduced the level of racial violence would appear to support a different aspect of our solidarity perspective, namely that cross-racial institutional structures might have reduced violence by increasing both social interaction and racial tolerance. In some contexts religious identities might have become somewhat detached from regional and racial identities thus creating a different form of solidarity that was less associated with racially-motivated violence. Counties in which the dominant religious institutions rigidly policed boundaries between the races, and buttressed the existing political and economic power structures, were significantly more likely to erupt in lethal mob violence than were counties with less overlap between religious, regional, and racial structures.²⁴

In their research, Stark and collaborators (1983) identify two elements of moral community: social integration and the importance of churches as institutions in cultivating behavioral conformity. First, the lack of integration might be directly related to the presence of several competing religions. Moreover, racial divisions in the church might limit the churches’ ability to inculcate value conformity. Both serve to weaken the sense of moral community and reduce the social control function that Stark and others document. Moreover, in their study of crime, Stark and collaborators (1983) report that not all types of crime are equally related to moral community and argue that impulsive deviance (e.g., murder) may be less tied to moral community as they represent “momentary acts of nonconformity” (p.16) of which deterrence is limited. As such religion’s power to impede extreme acts of violence and spontaneous acts – those not part of a pattern of criminality – may be harder to determine.

We do not believe that white parishioners would have objected to lynching for reasons of social justice or racial equality (Bailey 2005; Berends 2004; Farish 1969; Feldman 2005; Harvey 2004; Hill et. al., 1972; Matthews 2005; Snay 1993; Wilson 1980). It is more likely that in Southern communities, where only small minorities of religious adherents belonged to organizations that welcomed both blacks and whites, churches largely buttressed the prevailing doctrine of white supremacy. The degree to which these predominantly regional religious organizations successfully reinforced the racial caste system may have impacted white Southerners’ impulses to resort to extralegal violence against blacks.

To be clear, these findings obtain for the particular conjuncture of religious and racial institutions that prevailed in the post-Reconstruction south. The sharp focus on racial hierarchy buttressed by all social institutions, including religious ones, guaranteed that during this particular historic and political moment, the prevailing regional doctrine supported white supremacy and social and spatial segregation of the races. Although Southern Christian theology was often used to support this racial ideology, it is noteworthy that the places where this message was most effectively imbued into local culture, as measured by lack of religious diversity, were actually the safest in some ways for blacks. It appears however, that in communities where religious organizations with more tolerant racial practices were able to take hold, the racial environment improved. Perhaps it was in those places where the monolith had only begun to crumble and diversify – producing a climate of social, ideological and theological change – that the vulnerable were in the gravest danger.

It is also worthy of note that while these relationships obtain in the pooled models, changes over time in the relationship between specific religious measures and the level of racial violence appear to reflect changes in the religious landscape. During the 1890s the religious arena did not yet fully reflect the intense levels of racial and regional polarization that prevailed by the onset of the Great Depression. During this decade, a larger number of religious denominations were viable within the South, and a larger share of believers filled the pews of churches headquartered outside the South. It is noteworthy that the link between the constellation of local religious options and racial violence had not yet emerged. As Southern religion retrenched to reflect Jim Crow segregation that was being imposed on the broader Southern society, and commitment to regional identities and regional denominations strengthened – reflected, for example, in the splintering of white and black Baptist organizations, and the declining popularity of non-Southern denominations – the link between the local religious arena and racial violence first emerged, in the decade between 1900 – 1909, and then began to subside, as lynching became a less-common cultural practice and revivalistic spasms launched new, millennialist movements.

Limitations

The main limitations we have identified relate to measurement of religious variables, particularly problems associated with statistics on churches and religious membership. Possible issues with religious data from the US Census include varying definitions of membership, falsification of membership statistics, bad record-keeping by churches, inadequate coverage by the census, and errors in tabulation (Christiano 1987). Most of these potential inconsistencies will be small and difficult to estimate, and appear to be concentrated among data collected before and after the period under examination here.

However, variations in the definition of church membership, both over time and between organizations, are troubling. Organizations vary in their threshold requirements for membership, ranging from attending church on four consecutive Sundays to a unanimous vote for an offer of membership by all current members. Stark and Bainbridge (1981), using the 1926 data, report that varying definitions of membership do not affect overall measures of church membership greatly. The most problematic case seems to be measures of Jewish membership, which ranged from head of household counts only in 1906 – a clear undercount – to overcounting Jews by using the “cultural population” rather than religious membership in 1926. Despite these limitations, there is no evidence to suggest that the weaknesses in the data are systemically related to the outcome of interest – the incidence of lynching. As such, although we recognize there are some inherent problems with using religious data of the kind used in this study, we argue that their impact in the present analysis is minimal.

An additional problem lies with our reliance on Census Bureau reports for coding of religious bodies as monoracial or multiracial, since the census reports quantify racial diversity by identifying black denominations and black congregations within white-controlled denominations. This approach would mean that churches that embraced the radical notion of mixed-race congregations (as with the nascent Pentecostal/Holiness movements in the latter decades of our analysis) would have been overlooked. This may represent a small but important omission from our analysis. Similarly, our decision to include only those organizations that were organized and controlled by blacks in our measure of the strength of black churches, undoubtedly glosses over the variety of autonomy exercised by black congregations that were auxiliary to white churches.

There are also a few analytic limitations in the paper worth noting. The analysis presented here focuses on explaining country-variations in lynchings of blacks. Also, the analysis does not differentiate by the racial composition of the lynch mobs. The focus on black victims and the unspecified race of lynch mobs is unproblematic for several reasons. First, the overwhelming majority of lynchings targeted black victims. Second, blacks were nearly always lynched by whites. Third, studying the small number of lynchings of Southern whites would be limited in size and scope, and reflects a different underlying social process than is presented here for the lynchings of Southern blacks. In the few cases of white lynchings, the members of the lynch mob were also predominantly if not solely white. In fact, there were no cases of lynchings in the Beck and Tolnay (1995) lynching inventory used in this analysis in which blacks were the primary instigators of lynching white victims.²⁵

The operationalization of time in these analyses may present another limitation. Decades are used as the dividing points due to both the ability to link to census data collection and to the relative infrequency of lynching. Although lynching data is available on an annual basis, because lynching events were relatively infrequent in any given county, some sort of temporal collapsing is needed for statistical reasons. Decades provide a parsimonious and logical approach, since key controls and religious explanatory measures are available in ten-year increments. Alternative strategies, such as empirically determining temporal divides, were appealing but would require that religious measures be estimated for those new cut points, introducing the potential for measurement error.

Other analyses, such as Beck and Tolnay’s (1990) that used 3-year moving averages, suggest that there are some temporal effects in the relationship between economic variables and lynchings. However, the results are in the same direction across the entire time period with more significant effects before 1900. This suggests that looking at data before and after 1900 is a key distinction which is captured by a decade specific analysis. Moreover, we have no reason to believe that the underlying social processes we find are driven by decade dividing points.

Our inability to parse out the influence of the economic stresses experienced by white tenant farmers in 1890 is also of concern, given the strength of this measure in earlier research (see, e.g., Tolnay and Beck 1993). However, in preliminary analyses that replicated the Beck-Tolnay model, we find no meaningful differences between the size and significance of the race-neutral and white-restricted tenancy measures in decades for which both measures are available, nor are there distinctions in the effects of religious measures in models based on which tenancy measure was used. This fact, coupled with the profound changes occurring in the Southern religious landscape between 1890 and 1906, lead us to believe that we are, indeed, accurately identifying a change in the relationship between the role played by local institutional structures and the prevalence of mob violence in Southern communities.

It is possible that communities’ receptivity to divergent types of religious organizations may have been so constrained by their racial climate that the relationship we have identified is associational rather than causal. Whether the strength of these kinds of religious organizations was caused by or was the result of more racially tolerant communities is something we are unable to comment on. For example, Olzak and West’s “salience theory” asserts that elevated rates of ethnic conflict will “increase demand for protective ethnic organizations” (1991: 459), suggesting that perhaps the strength of black churches in counties with high levels of racial violence is a symptom rather than an outcome of local incidence of lynching. However, it is clearly an area worthy of additional investigation. Future research will also need to explore the potential diffusion or deterrent effects for contiguous counties with different combinations of denominational strength, as well as the possibility that dynamic change in denominational options over the course of the study period (expansion or contraction of available denominational affiliations) has an independent effect on the level of racial violence. Additionally, preliminary analysis of the data indicates that the shift in membership between denominations was quite strong in many counties between some decades. Identifying patterns of change in denominational allegiance and potential relationships to prevalence of lynching could help to further explain the relationship between these two types of collective behavior.

CONCLUSION

Religious organization and expression in the post-Reconstruction era American South manifested in ways that directly – and indirectly – influenced the racial landscape. Southern Christianity largely rejected the Social Gospel that swept the north, choosing instead to focus on saving individual souls (Spain 1961). White Southerners have historically used religious expression and structures as barriers against the tides of social change (Feldman 2005), and non-religious civil society developed more slowly in the South than in other regions of the U.S. (Young 2002). Adopting progressive social agendas or actively working for racial justice as Christians would have also challenged their regionalist identity. Even in instances in which southern communities embraced social activism, they were unwilling to directly confront the prevailing racial hierarchy (Young 2002: 668). We believe our results support the idea that the trends in social change towards greater racial tolerance in the South were linked to the religious landscape. The strength of the effects of religious diversity suggests that competition between religious denominations – which was not necessarily antagonist – and the meaningful theological differences that separated them were linked to more racial violence. However, in those Southern communities that were receptive to religious organizations that welcomed both black and white members, the level of racial violence was suppressed.

Our analyses focus on a specific time and region of the country, and indeed revolve around the claim that white Christian expression in the Jim Crow South, and its unique fusion of regional, religious, and racial identities, served to build white racial solidarity and foster racial violence. This may seem a discrete question with little importance to contemporary American society. However, the geographic dispersion of the Southern diaspora and subsequent influence on the broader national religious landscape (Gregory 2005) implies that these linkages between religious organization and race relations may reverberate within contemporary communities throughout the United States.

Indeed, a number of sociologists and social historians have examined whether religious contexts reinforce contemporary racism or condone the use of violence. For example, Messner and colleagues (2005) use county-level Presbyterian membership in 1850 as a proxy for the local salience of honor codes, and unsuccessfully attempt to use this measure to predict contemporary homicide rates. Jacobs and colleagues (2005) find a positive relationship between the share of local religious adherents who are fundamentalist Christian and the imposition of the death penalty, but fail to find a racial effect.²⁶ Emerson and Smith (2000) analyze contemporary evangelicals’ understandings of the causes, consequences, and potential solutions to racial inequality in the United States, and conclude that competing organizations which cater to popular views as a tactic for increasing membership simultaneously reinforce social distinctions by segmenting the religious marketplace along racial lines.

Our results support Emerson and Smith’s assertion that an unintended consequence of an active and racially divided religious marketplace is the reinforcement of boundaries between groups and an increase in social distance between the races. The ability of congregations to isolate similar people within closed social communities tends to solidify boundaries between social groups. It is possible that this factor grounds the positive influence of religious diversity on the incidence of lynching. This is also consistent with Emerson and Smith’s (2000) assertion that evangelicals are not typically “countercultural” and they often do not heed the call for social change or challenge underlying social structures. This seems to be the case with religious organizations and lynchings.

This paper suggests that the community-level contours of religious membership may reverberate in areas of society far outside the strictly religious realm, and in both expected and unexpected ways. The bottom line is this: the doctrinal bent of local religious organizations, and the social integration they are able to foster, have real implications for collective behavior. Through this analysis, we have added another macro-level social measure to the battery used to explain racial violence. Religion represents a key social institution; this was particularly salient in the American South during this time period, with potential for enhancing moral community, in both positive and negative manifestations. Including religious-based measures adds to our understanding of the patterns of racial violence. Implications for contemporary American society – in which 11:00 on Sunday morning remains the hour in which our nation is segregated along racial, economic and political lines – remain to be seen.

Appendix A.

Classification by Decade, Religious Organizations Reporting Members in Ten Southern States, 1890 – 1926

	Primary Racial Group				Mixed Race				Headquarters Location
Denomination	1890	1906	1916	1926	1890	1906	1916	1926	1890	1906	1916	1926
Adventist Bodies (Including Christian and 7th Day)	W	‡	W	W	N	‡	Y	Y	N	‡	N	N
Assemblies of God	---	---	W	‡	---	---	Y	‡	---	---	N	‡
American Baptist	---	---	---	W	---	---	---	Y	---	---	---	N
Colored, National and Negro Baptist	B	‡	B	B	N	‡	N	N	S	‡	S	S
Duck River Baptist	---	W	W	‡	---	N	N	‡	---	S	S	‡
Free Baptist	---	W	‡	‡	---	Y	‡	‡	---	N	‡	‡
Freewill Baptist	W	W	W	W	Y	N	N	N	N	S	S	S
United Freewill Colored/ Colored Freewill/ United Amer. Free Baptist	---	B	B	B	---	N	N	N	---	S	S	S
General Baptist	W	W	W	W	N	Y	N	N	N	N	N	N
Primitive Baptist	W	W	W	W	Y	N	N	N	S	S	S	S
Colored Primitive Baptist	---	B	B	B	---	N	N	N	---	S	S	S
Regular (Southern)/ Southern and National/ Southern Baptist	W	W	W	W	Y	Y	Y	Y	S	S	S	S
Separate Baptist	---	---	W	‡	---	---	N	‡	---	---	S	‡
Two Seed in the Spirit Baptist	W	‡	‡	‡	Y	‡	‡	‡	S	‡	‡	‡
United Baptist	---	W	W	W	---	N	N	N	---	S	S	S
Roman Catholic	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Christian Church/Christian Connection	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Christian Science	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Independent Christian	W	‡	‡	‡	N	‡	‡	‡	S	‡	‡	‡
Church of Christ	---	W	W	W	---	N	Y	Y	---	S	S	S
Church of God	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Church of God (General Assembly)	---	---	B	B	---	---	N	Y	---	---	S	S
Church of God in Christ	---	---	---	B	---	---	---	N	---	---	---	S
Church of the Living God	---	B	B	‡	---	N	N	‡	---	S	S	‡
Swedenborgian	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Congregationalist	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Disciples of Christ	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Conservative Dunker	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Evangelical Association	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Free Christian Zion (Colored)	---	B	B	‡	---	N	N	‡		S	S	‡
Orthodox Friends	W	W	W		N	Y	Y		N	N	N
(German) Evangelical Synod	W	W	W	---	N	N	N	---	N	N	N	---
(Eastern) Greek Orthodox	---	W	W	---	---	N	N	---	---	N	N	---
Jewish Congregations (Reformed and Orthodox)	W	W	W	W	N	N	N	N	N	N	N	N
Latter Day Saints (Reformed and Christian)	W	W	W	‡	N	N	N	‡	N	N	N	‡
Lutheran General Council	W	‡	‡	*	N	‡	‡	*	N	‡	‡	*
Lutheran General Synod	W	W	W	*	N	N	N	*	N	N	N	*
Lutheran: Misouri Synod	---	---	---	W	---	---	---	N	---	---	---	Y
Lutheran: Ohio Synod	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Lutheran Synod of the South	W	W	W	*	Y	N	N	*	S	S	S	*
Lutheran Synodical Conference	W	W	W	‡	Y	Y	Y	‡	N	N	N	‡
Lutheran United Church	---	---	---	W	---	---	---	Y	---	---	---	N
Mennonite	W	‡	W	‡	N	‡	N	‡	N	‡	N	‡
Amish Mennonite	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
African Methodist $	---	B	‡	‡	---	N	‡	‡	---	N	‡	‡
African Methodist Episcopal	B	B	B	B	N	N	N	N	N	N	N	N
African Methodist Episcopal Zion	B	B	B	B	N	N	N	N	N	N	N	N
Colored Methodist Episcopal	B	B	B	B	N	N	N	N	S	S	S	S
Free Methodist	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Methodist Congregational	---	---	W	‡	---	---	N	‡	---	---	S	‡
Methodist Episcopal	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Methodist Episcopal, South	W	W	W	W	N	N	N	N	S	S	S	S
Methodist Protestant	W	W	W	W	Y	Y	Y	Y	S	N	N	N
Wesleyan Methodist	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Moravian Church	W	W	W	‡	N	N	N	‡	N	N	N	‡
Pentecostal	---	---	W	‡	---	---	Y	‡	---	---	S	‡
Presbyterian Associate (Reformed)	---	W	W	W	---	N	Y	N	---	S	S	S
Presbyterian Church of the USA	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Cumberland Presbyterian	W	W	W	W	N	Y	Y	N	S	S	S	S
Colored Cumberland Presbyterian	---	---	B	B	---	---	N	N	---	---	S	S
Colored Presbyterian	B	‡	‡	‡	N	‡	‡	‡	S	‡	‡	‡
United Presbyterian	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Presbyterian Church of the US, South	W	W	W	W	Y	N	Y	Y	S	S	S	S
Protestant Episcopal	W	W	W	W	Y	Y	Y	Y	N	N	N	N
Reformed Church of the US/Reformed Episcopal	W	W	W	W	Y	N	Y	Y	N	N	N	N
Salvation Army	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Spiritual Church	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
United Brethren of Christ	W	W	‡	‡	N	Y	‡	‡	N	N	‡	‡
Unitarian Church	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡
Universalist Church	W	‡	‡	‡	N	‡	‡	‡	N	‡	‡	‡

Notes: --- indicates first reporting of statistics for this organization has not yet occurred;

^‡indicates separate statistics for this organization not reported in 10 states that are the subject of this inquiry;

^*indicates denomination ceased to exist because merged into Lutheran United Church;

^$In some jurisdictions, black Methodist denominations were reported separately in 1906, while in others, a “catch-all” measure of “African Methodists” was reported.