Targeting Lynch Victims: Social Marginality or Status Transgressions?

Abstract

This paper presents the first evidence yielded by a newly-compiled database of known lynch victims. Using information from the original census enumerators’ manuscripts, we identify the individual- and household-level characteristics of more than 900 black males lynched in ten southern states between 1882 and 1930. First, we use the information gathered for successfully linked cases to present a profile of individual-level and household-level characteristics of a large sample of lynch victims. Second, we compare these characteristics to a randomly-generated sample of black men living in the counties where lynchings occurred. We use our findings from this comparative analysis to assess the empirical support for alternative theoretical perspectives on the selection of individuals as victims of southern mob violence. Third, we consider whether the individual-level risk factors for being targeted as a lynch victim varied substantially over time or across space. Our results demonstrate that victims were generally less embedded within the social and economic fabric of their communities than were other black men, suggesting that social marginality increased the likelihood of being targeted for lynching. These findings were generally consistent across decades, and within different socio-demographic contexts.

I. Introduction

There is no more powerful symbol of post-bellum southern racial oppression than the lynching of African Americans. At least 2,500 blacks are known to have been so murdered during the lynching era – a rate of roughly one mob killing every week for five decades (Tolnay and Beck 1995: 17).¹ Despite the wealth of information social scientists have uncovered regarding the contextual factors and institutional arrangements that made lynching more likely (Ames 1938; Dray 2003; Gullickson 2009; Hovland and Sears 1940; Perloff 2000; Raper 1969 [1933]; Soule 1992; Stovel 2001; Tolnay and Beck 1995; Vandiver 2006; Wasserman 1998), we know very little about the people who were victimized by vigilante mobs, aside from the fact that they were disproportionately black and male. Beyond race and gender, who was targeted by this lethal form of racial violence?

This paper presents the first substantive results from an innovative data collection effort using online genealogical resources to create a database of people known to have been lynched in the American South between 1882 and 1930. This newly-created database merges the most comprehensive, confirmed, inventory of known lynch victims (Beck-Tolnay 2004), covering ten southern states over a period of nearly 50 years², with individual and household records from historical enumerators’ manuscripts from the U.S. population census. The Beck-Tolnay inventory includes limited information on each victim — date and location of the lynching and, for most victims, their name, sex, and race, as well as their alleged crime or transgression. This information has been instrumental in identifying several contextual predictors of collective violence, but does not provide sufficient information to allow us to offer a profile of victims nor to investigate the individual characteristics that put people at greater risk of being lynched.

Census records, however, include extensive demographic data for each individual and all household members, including age, race, sex, marital status, literacy, and occupation. Census manuscripts also identify the relationships among household members and specify whether each family rented or owned their home.³ Using this information, we compare characteristics of black male lynch victims with those of other black men living in the counties in which lynchings took place, and evaluate whether significant differences existed between these groups. Gaining a better understanding of whether and how victims were distinguished from the broader population may grant additional leverage on the motivations behind this wave of racial violence, perhaps providing theoretical grounding for micro-level research on the targets of other ethnic violence campaigns, and related contemporary issues such as bias crimes.

Prior research has documented substantial variation in the intensity of southern mob violence over time and across space (Beck and Tolnay 1995; Brundage 1993; Corzine, Creech, and Corzine. 1983; Corzine, Huff-Corzine, and Creech, 1988; Olzak 1990; Pfeifer 2004; Phillips 1987; Reed 1972; Stovel 2001; Vandiver 2006). Temporally, it has been well documented that lynching peaked during the 1890s and eventually became rare events after 1930. Spatially, there is strong evidence that blacks were exposed to the greatest risk of being lynched in areas where their population was larger (absolutely and proportionately) and a larger proportion of the white population struggled economically. In light of these aggregate patterns in the frequency of lynching, we consider whether the personal characteristics that placed individuals at a heightened risk of victimization also varied by time period and location within the South.

II. Background

A. Brief History of Southern Lynching

Lynching in the United States was not an exclusively southern phenomenon, nor were blacks its only victims. Lynchings also occurred in the West, and to a much lesser extent, in the North (see e.g., Pfeifer 2004). In the Southwest, Mexicans and Native Americans were often the targets of mob violence, while Chinese were sometimes victimized in the far West and Mountain states. However, extralegal executions were substantially more frequent in the South and, within that region, blacks were considerably more likely than whites to be the victims. Southern lynching also fulfilled a uniquely important role in the subjugation of the African American population that was not shared by other regions (e.g., Brundage 1993; Tolnay and Beck 1995).

The frequency of lynching in the American South increased sharply after the Civil War and remained a relatively common occurrence through the onset of the Great Depression in the early 1930s. So distinctive was the surge in lynching activity that this period of southern history is sometimes referred to as “the lynching era.” Roughly 90% of southern lynch victims were African American and the vast majority of them were males. Most historians and social scientists explain the relative infrequency of lynching in the South prior to the Civil War as a result of economic interests. That is, it made little sense for the southern elite, especially planters, to tolerate extensive extralegal executions of blacks because slaves represented a significant proportion of their wealth. Therefore, the punishment of slaves, though often severe, typically was not lethal.⁴ The financial incentive for protecting southern blacks changed dramatically after the Civil War when four million slaves gained independence and enjoyed the privilege to compete with whites in the southern economy, especially within the agricultural sector (Ransom and Sutch 1977). This new status increased their vulnerability to discriminatory strategies of all kinds, including mob violence.

Between the 1880s and the early 1930s, lynching was only one tactic used by whites to subjugate southern blacks, socially, politically, and economically. Other elements in that powerful arsenal of oppression included Jim Crow laws requiring racial segregation in public places. At one extreme this included excluding blacks from entire towns and neighborhoods (Loewen 2005). Less extreme, but still clearly intent on defining blacks as an inferior caste, were rules dictating racial separation in virtually all public and private facilities. Grandfather clauses, poll taxes, and literacy tests, were used to disenfranchise the southern black population and to neutralize their potential political influence (Kousser 1974; Woodward 1966). Occupational discrimination restricted black workers to the least desirable and least remunerative jobs in the southern labor market. When those non-lethal strategies were inadequate to achieve their purposes, or when whites felt that an especially strong “message” needed to be delivered to the black population, they often resorted to violence.

The frequency of lynching declined sharply after 1930. Why this occurred is still the subject of much speculation and social scientific inquiry. One common explanation argues that lynching was replaced by a more efficient criminal justice system in which the likelihood of swift and severe punishment for serious crimes, including state-sanctioned executions, increased (e.g., Pfeifer 2004; Phillips 1987; Vandiver 2006; see, however, Tolnay and Beck 1995, chapter 4). A second explanation (Ames 1942; Brundage 1993; Dray 2003; Gordon 1937; White 1929 [1969]) attributes the decline in lynching to increased opposition to the practice by the southern press and to increased resistance to potential lynch mobs by southern law enforcement authorities (from sheriffs to governors). Still another explanation argues that the mass exodus of southern Blacks during the Great Migration depleted southern employers’ plentiful and inexpensive labor force and motivated them to take steps to reduce the level of violence aimed at blacks (Tolnay and Beck 1992, 1995: chapter 7; Wilkerson 2010). While lynchings continued to occur after 1930, including some that were especially horrific in their execution, their numbers paled in comparison to those suffered during the “lynching era,” and their centrality to both the social control of the black population and the maintenance of white domination was reduced.

B. The Context of Racial Violence

Prior research on lynching is generally bifurcated into two analytic approaches designed to document the “tragedy of lynching” and to help us gain a better understanding of the social, cultural, economic, and demographic factors that promoted it: qualitative case studies and comparative quantitative analyses. Case studies have made important contributions to our knowledge, marshaling extensive evidence to provide rich detail about single lynching incidents and the circumstances surrounding them (Dinnerstein 1968; Downey and Hyser 1991; Griffin 1993; McGovern 1982; Smead 1986; Wexler 2003). Although case studies often tell richly detailed stories about specific lynching episodes and individual victims, the extent to which the facts of a single event can be generalized to a larger set of lynchings is questionable.

Comparative quantitative studies, using limited information about a large number of lynchings, provide the bulk of our knowledge about the factors that created an environment hospitable to racial violence. By examining the distribution of incidents or victims over time and across space, and estimating covariation between temporal or spatial rates of lynching and theoretically-relevant social, economic, political or demographic factors, comparative quantitative work has enabled social scientists to test theories about the ecological correlates of lynching (see, e.g., Beck, Massey and Tolnay 1989; Beck and Tolnay 1990; Brundage 1993; Corzine, Creech, and Corzine 1983; Corzine, Huff-Corzine and Creech 1988; Gullickson 2009; Olzak 1990; Soule 1992; Tolnay and Beck 1992, 1995; Tolnay, Beck and Massey 1989; Tolnay, Deane and Beck 1996; Wright 1990). Although the relative importance of specific macrostructures remains a subject of debate, previous comparative studies indicate that southern blacks were most at risk of being lynched in environments where or when:

(1) the black population was proportionately larger, (2) the white population suffered economic hardship, (3) the Democratic party was stronger, (4) the white population was threatened economically or socially by the black population, (5) the level of black out-migration was lower, and (6) alternative legal sanctions for serious crimes were lacking (Bailey et al. 2008, p. 48).

The southern media (Ames, 1938; Perloff, 2000; Wasserman, 1998), law enforcement and government officials (Dray 2003; Senechal de la Roche, 1996), and broadly-defined cultural support for racist attitudes and behaviors (Berends, 2004; Brundage 1997; Oliver, 2001) also influenced lynching prevalence. However, empirical evaluations of the latter influences have not been conducted.⁵

C. Theoretical Perspectives on Vulnerability

Aside from the fact that lynch victims in the South were overwhelmingly black and male, we know little about the socio-demographic characteristics of individuals targeted by mob violence. While it is probably reasonable to assume that most victims were older adolescents or younger adults, in only a small proportion of lynching incidents is the age of the victim known. In even fewer cases do we know anything about the family status or socio-economic standing of the victims. Lacking more extensive, direct evidence we draw from prior research to propose two, alternative and contrasting, expectations regarding the selection of victims from among the black population. The first set of expectations is based heavily on the scholarship of Roberta Senechal de la Roche which claims that victims frequently were targeted because of their marginal status within the community (1996, 1997, 2001). The second set reflects a recurrent theme in the literature on lynching suggesting that victims were targeted because of their distinctively high levels of human and social capital compared to other blacks within their communities (e.g., Bailey 1914; Dray 2003; Tolnay and Beck 1995).

The Marginalization Perspective

Senechal de la Roche (1996, 1997, 2001) expands on Black’s (1976) work on relational distance and partisanship to propose that those who were less rooted in the local community were more likely to be victimized by lynch mobs. She argues that strangers are at greater risk than are “locals” because they lack a web of protective social relations within the community. She finds (2001) that violence is more likely to become collective when the victims and perpetrators belong to groups with dramatically unequal status, and the perpetrator is able to marshal the support of third parties. Regarding the last point, it is possible that the targeting of more successful blacks by lynch mobs would have prompted greater outrage and condemnation from local whites or even a violent reaction from the local black population. In the context of ethnic riots, Horowitz writes (2001:135), “Fear of retribution tends to inhibit violence. This basic behavioral propensity has led some psychologists and others to suggest that aggression is likely to be directed at helpless targets, because of their inability to retaliate or to persuade state authorities to retaliate in their behalf.” Aggressive or violent responses from the black community to lynchings were not common, but they did occur (Shapiro 1988; Dray 2003). Although southern white communities were generally supportive of lynch mob activity, in many cases potential lynchings were averted through the intervention of local white citizens or authorities. In far fewer cases, arrests were made (Brundage 1993; Dray 2003). It is unlikely that negative responses to lynching from the white community, and violent reactions from the black community, were random occurrences. Rather, they likely were more frequent when the victim was a higher status individual with strong local ties.

To our knowledge, only one study has empirically assessed Senechal de la Roche’s predictions about the characteristics of lynch victims. Beck and Clark (2002) rely on newspaper accounts of Georgia lynching incidents to determine the characteristics of victims. They show that, in most decades, socially marginal individuals represented a minority of lynch victims. Only during the 1890s, a period of economic stress and rapid population growth in the “wiregrass” region of south Georgia, did marginalized African Americans represent the majority of victims. Beck and Clark argue that a definitive test of the proposition that strangers were more likely than non-strangers to be lynched requires knowledge of the proportion of the community that are strangers, and these data do not exist. In lieu of such information, their study, based on indirect estimation for the state of Georgia, is the best evidence we have thus far.

In the absence of more direct tests of Senechal de la Roche’s “marginalization” hypothesis regarding the targeting of southern lynch victims, perhaps evidence from studies of more contemporary hate, or bias, crimes can provide some guidance in anticipating the characteristics of historical southern lynch victims. Although intergroup hatred is an important component of both forms of violence, we are cautious about equating modern hate crimes with southern lynching. While there may be instrumental motives behind modern hate crimes (see for example, Lyons 2007), the structural circumstances and institutional support for historic southern lynchings and modern hate crimes are substantially different. Most importantly, hate crimes are just that – crimes. In most southern places, lynching was viewed differently from conventional criminal behavior, and mob members enjoyed virtual immunity from prosecution. Also, lynching had greater popular support in the historic South than do today’s hate crimes.

Research on modern hate crimes is a relatively new endeavor, reflecting the recency of such crimes as a unique category of criminal offense, and has engendered as much speculation about the offenders as about the characteristics of their victims (e.g., Messner, McHugh and Felson 2004). Furthermore, to some extent, the characteristics of hate crime victims are directly determined by the specific type of crime, itself. For example, in racially or ethnically motivated hate crimes, victims, by definition, will be members of a particular race or ethnic group; likewise, victims of hate crimes with sexual orientation as a motivation are likely to be, or perceived as, gays, lesbians, or transgendered individuals. With those limitations in mind it is possible to glean some useful generalizations from this literature. Especially useful in this regard is research conducted by the Bureau of Justice Statistics (BJS) using information from the National Crime Victimization Survey (NCVS) which is based on reports from the victims of crime. According to the BJS’s annual rates of victimization for hate crimes (Harlow 2005: Table 8), the relative risk was greater for those who were: male, under 20 years of age, single (both never-married and separated/divorced), high school dropouts, with annual incomes below $25,000.

Other research on the victims of modern bias crimes also suggests that targets are disproportionately younger and less embedded within the community. For example, Lauritsen and Schaum (2004) identify that violence against women tends to be perpetrated against those who are younger, have children, and have recently moved into a new community. Younger gay and bisexual men appear to be at elevated risk for anti-gay violence (Lampinen et. al. 2008). By comparing the characteristics of victims of hate crimes and non-hate crimes, Messner, McHugh and Felson (2004:602) find a higher risk of victimization among younger people (ages 16–25) for non-racial hate crimes (e.g., based on gender, sexual orientation, or religion), but not for racially motivated hate crimes. To some extent, this evidence for recent bias crimes comports well with Senechal de la Roche’s argument that lynch victims were more marginal or vulnerable members of society.

The Status Transgression Perspective

It is possible that southern lynch mobs felt provoked by well-established, higher-status, African Americans within their communities. The success of such individuals threatened the racial caste line that implicitly guaranteed a superior social position to any white person, relative to all black persons. This perspective was expressed clearly by Thomas Bailey in his book published in 1914, Race Orthodoxy in the South, and Other Aspects of the Negro Question, “Every prosperous negro who shows his prosperity in a way to be seen by the whites is a focus for hatred on the part of the ‘lower’ whites” (1914:30, emphasis in original). This type of racially-defined status threat is consistent with evidence from aggregate analyses of lynching frequency within areas and across time. For example, prior research suggests that blacks, in general, were more vulnerable to mob violence in settings where, and during times when, blacks appear to have been in a position to threaten white dominance, for example, when they represented a relatively large percentage of the total population. Although suggestive of the possibility that individual blacks may have provoked white lynch mobs through their economic and political striving, the existing evidence supporting such an interpretation is based on macro-level, cross-sectional and temporal patterns and not on the actual characteristics of individual victims. Past researchers have been careful to avoid the “ecological fallacy,” by drawing inferences about individual behavior from aggregate relationships.

The record of known lynch victims certainly includes many examples of successful blacks being targeted by mobs. For example, Eli Hilson, who was lynched in Mississippi in December of 1903, was a prosperous farm owner who had been warned by whitecappers to leave the area. After refusing their demands, Hilson’s house was riddled with bullets and he was killed (Memphis Commercial Appeal 1903). In 1892 a Louisiana mob targeted Dennis Cobb who was described by the New Orleans Times Democrat (1892) as a “well-to-do and highly respected colored man.” Although these incidents are consistent with a “status transgression” perspective on the targeting of lynch victims, we are reluctant to embrace fully the argument that successful blacks were disproportionately targeted by southern lynch mobs based solely on such anecdotal evidence. Rather, it remains a plausible hypothesis, and a counterpoint to the marginalization hypothesis, that requires empirical testing.⁶

D. General Objectives and Research Hypotheses

This paper seeks to determine the extent to which the characteristics of lynch victims, as recorded in the U.S. population census enumerators’ manuscripts, support the competing perspectives on vulnerability discussed above. Census records from the late nineteenth and early twentieth centuries include measures of individual-level and household-level characteristics that would help to determine whether lynch victims were selected disproportionately from among the more successful, higher-status, segments of the African American population, from those with lower or average status, or from those who were more marginal to the community. For example, commonly occurring characteristics included in the population census are: literacy, occupation, marital status, relationship to the household head, state of birth, home ownership, and “mulatto” versus “Negro/Black” racial classification. In contrast, income, wealth, religion, political party membership, and voting behavior are not available in census records for the lynching era.

We first take advantage of the new information revealed by matching lynch victims to their census records to present a “profile” of black males who were lynched between 1882 and 1930. For our purposes, we focus on those individual and household characteristics mentioned in the research hypotheses stated below, specifically: age, marital status, state of birth, literacy, home ownership, household headship, racial status (i.e., mulatto versus black/negro), and occupation.

Our second objective is to compare lynch victims’ characteristics with those of the general black population. These comparisons will identify significant differences between the two groups that might be useful for further illuminating perpetrators’ motivations, or for adjudicating among competing expectations regarding victims’ characteristics. Our analyses will test a set of hypotheses in an attempt to determine whether the characteristics of lynching victims provide consistent support for one, or more, of the perspectives discussed above.

The Marginalization Perspective

Senechal de la Roche’s theory (1996, 1997, 2001) assumes that lynchings targeted individuals with more marginal status in the community, and a greater degree of social and relational distance to members of the white mob. If this theory holds, we would anticipate:

Hypothesis 1: Compared to the general black male population, lynch victims were (a) less likely to be married and to have been enumerated as heads of households, (b) more likely to have born out-of-state, and, (c) less likely to reside in owned than rented dwellings.

The Status Transgression Perspective

Conversely, lynching may have been used to punish those blacks who, through their social and economic success, challenged the prevailing racial hierarchy. As such, we would expect more economically successful, upwardly-mobile blacks to be disproportionately represented among lynch victims. Thus, according to the status transgression perspective, we expect to find support for the following:

Hypothesis 2: Compared to the general black male population, lynch victims were more likely to be (a) literate, (b) homeowners, (c) working in higher status occupations requiring greater skill, and (d) enumerated as “mulatto” rather than as “Negro” or “black.”

After discerning which, if any, individual- or household-level characteristics are related to the likelihood of being targeted for lynching, we consider whether the importance of those risk factors varied across the decades of the “lynching era” or across areas that differed along selected socio-demographic characteristics that have been tied to variation in the overall frequency of lynching. Our consideration of possible temporal and spatial variation in the risk factors for being lynched is largely an exploratory effort that is motivated by prior research which has demonstrated that the contextual conditions conducive to the “volume” of lynching varied over time and across space. One might expect, for example, that during time periods or in settings in which whites perceived greater economic threat or competition from the local African American population that victims were selected more for their economic success, rather than their social marginality.

III. Data, Variables, and Methods

Data

We use a newly-created data source⁷ which includes individual-level and household-level data for lynch victims. Our creation of a new database of lynch victims is noteworthy because of its geographic and temporal scope and the variety of documents we incorporate.⁸ Our data collection efforts occur entirely online, using a genealogy website subscription⁹ and its searchable web-based interface to access high-quality .jpg images of historic census manuscripts and other historic documents. We search for each victim in the census records immediately prior to his lynching – a backward search. For example, if someone was lynched in 1902, we search in the 1900 census, beginning within the county of lynching, and expanding outward to contiguous and nearby counties. We also conduct a forward search, for example, looking in the 1910 census for candidate matches we identified in the 1900 census, to eliminate “false-positives.” Unfortunately for our purposes, most of the original enumerators’ manuscripts for the 1890 census were destroyed in a fire. Therefore, it is impossible to search for victims lynched during the 1890s in the 1890 census records. Rather than sacrifice all lynchings that occurred during the 1890s, however, for victims lynched between 1890 and 1895 we search in the census enumerators’ manuscripts for the 1880 census. This solution is less than ideal, given the time elapsed between 1880 and the dates of these lynchings, but it is the only opportunity to retain at least some of the victims from the 1890s in our study. We have chosen to not include lynchings that occurred between 1896 and 1899 in our linked data base (i.e., linked to the 1880 census). A longer interval between the time of lynching and date of census enumeration decreases the likelihood of successful matches, primarily as a result of inter-county migration by the victim prior to his death, and increases the likelihood of “false positives” through the in-migration of individuals with the same or very similar name as the victim. Furthermore, linking victims to a census that is too far in the past results in less useful information for our purposes. For example, a 20 year old victim in 1898 would have been only two years old in 1880 and we would not have information on key characteristics such as literacy, marital status, and occupation.

While our research primarily relies on census records, we incorporate additional online sources to help verify matches and to locate additional information on lynch victims. Chief among these are World War I draft registration cards, available for more than 24 million men in 1917 and 1918, and including each person’s name, race, date of birth, employer, occupation, marital status, and often next-of-kin. We also occasionally rely on death records to verify facts about particular victims, or to adjudicate between multiple possible matches, although temporal and geographic coverage is uneven. Within the American South, few records exist before the early 20^th century, and implementation of vital records registries varied both within and between states. We also utilize historical newspapers, since articles about lynchings sometimes include details about the victim – for example they may refer to the victim’s “youth”, or report his occupation, or the names of family members – which can help to narrow our field of match candidates.

Although the database includes information on all victims, regardless of their race or gender, the current analysis is restricted to presenting a profile of black male victims, and to identifying differences between the victims and a randomly-generated sample of black men.¹⁰ For the majority of victims for whom we were able to locate appropriate matches in the historical census records, we identified only a single potential match. These single matches were carefully evaluated for the likelihood that they were, in fact, the same individual that was lynched. For each case an “estimated subjective probability” was assigned, reflecting the level of confidence that the research team had in the linkage between the lynching records and the census manuscripts. For potential single matches, only those cases with an estimated subjective probability of at least .75 are included in the following analyses. In cases with more than one possible match identified for a specific victim, estimated subjective probabilities were assigned to all potential matches, up to a maximum of three. Only those multiple matches receiving a probability score of at least 0.50 are retained for the analytic sample.¹¹

Using these eligibility criteria, we successfully matched 45.1% of the victims from our inventory with their census records. Table 1 presents the record of matching victims to the original census enumerators’ manuscripts, by time period.¹² No objective criterion is available for judging this level of successful matches. However, it is possible to compare it to the success rates obtained by other record linkage projects. Guest (1987) describes a 39.4% linkage rate for the National Panel Study that attempted to link the census records for adult white males between 1880 and 1900. Steckel’s (1988) effort to link families in the 1860 census with their records in the 1850 census in order to study migration yielded a successful linkage rate of 59%. Rosenwaike and colleagues (1998) achieved a 56.3% success rate in their effort to link death certificate records from 1980 through 1985 for African Americans to their earlier census records for 1900, 1910, and 1920. Ferrie (1996) estimates a 21% success rate for his project that attempts to link males enumerated in the 1850 census to their records in the 1860 census. Given the high levels of census underenumeration for black males during this time period (Coale and Rives 1973), the substantial levels of inter-county migration for southern blacks (Tolnay 1999), and that we began our searches with many fewer individual-level characteristics than were available for the cited previous record linkage efforts, we are reassured by the record of successful matches reported in Table 1.

Table 1.

Record of Matching Named Black Male Lynch Victims to Their Original Census Enumerators’ Manuscripts, by Decade. Ten Southeastern States.

Years	Cases	No Match	Single Match	Multiple Matchesa
1882 – 1889	338	46.2%	42.6%	11.2%
1890 – 1895	436	59.9%	31.2%	8.9%
1900 – 1909	522	58.8%	30.5%	10.7%
1910 – 1919	385	56.7%	33.2%	10.1%
1920 – 1929	192	44.3%	48.4%	7.3%
Total	1,873	54.9%	35.2%	9.9%

^aAt least one match attains 50% or greater estimated subjective probability.

Variables and Method

Using information for the 923 matches¹³ in which the lynch victim and his linked census records satisfy the eligibility criteria described above, we first present a profile of southern lynch victims for the period 1882–1930. This profile offers the first glimpse that social science researchers have had of the personal characteristics of the thousands of black men who were murdered by lynch mobs during the “lynching era.”

Second, we compare the personal and household characteristics of lynch victims with those of the general population of black males. To facilitate this comparison we have generated a comparison sample for 1880, 1900, 1910, and 1920, using machine-readable Public Use Microdata Samples, or PUMS files (Ruggles et. al. 2008)¹⁴ for each county that experienced the lynching of a black male in that decade. That is, we compare the characteristics of lynch victims to the characteristics of black male non-victims residing in the county in which the lynching occurred, rather than non-victims residing in the county in which the victim was located in the census manuscripts.¹⁵ Our comparative analytic strategy proceeds as follows. First, we assess the unadjusted, bivariate differences between victims and non-victims for selected characteristics. The following characteristics are available for all census years (i.e., 1880, 1900, 1910, and 1920):

• Age refers to the victim’s age at the time he was lynched;

• Married identifies those who were married at the time of census enumeration;

• Head of household is based on the individual’s reported relationship to the household head;

• Born out of state, as reflected in reported birthplace;

• Literate, indicates an ability to read and write;

• Skilled worker identifies those individuals who were recorded as having occupations that required some level of skill.¹⁶

We are able to distinguish “mulatto” from “black/Negro” only for the 1880, 1910, and 1920 censuses. Likewise, information about home ownership is available only in 1900, 1910, and 1920, and, for our purposes, is restricted to victims enumerated as heads of households.¹⁷

For our second comparative strategy we estimate adjusted differences between victims and non-victims, using a logistic regression model. This baseline model uses victim status as the dependent variable and predictor variables age, age-squared and a series of dichotomous variables identifying decade. Incorporating decade dummy variables accounts for temporal variation in the incidence of lynching. Using both age and age-squared controls for the likelihood that age distributions of victims and nonvictims are materially different, and that risk of victimization declines among older men. We then estimate a series of logistic regression models that include additional sets of predictor variables, linked to social status and social marginalization.

Our final analytic objective is to determine whether any of the personal risk factors identified in our comparison of victims and non-victims varied significantly across time periods or according to selected socio-demographic characteristics of locations. Toward that end we first estimate logistic regression models of victimization separately for each decade and compare the effects of individual-level predictors across time. We next consider the possibility that the predictors of victimization varied by context. Drawing from previous research on the macro-level correlates of the intensity of lynching (Tolnay and Beck 1995) we distinguish counties by (1) the proportionate size of their African American populations (i.e., % Black), (2) decadal change in the size of their African American populations, and (3) the proportion of farmers who were tenants.¹⁸ Size and growth of the African American population have been used extensively in the literature to measure the level of racial threat to the local white population (see e.g., Beck et al. 1989; Corzine, Creech, and Corzine. 1983; Corzine, Huff-Corzine, and Creech, 1988; Reed 1972; Tolnay and Beck 1995). White farm tenancy is intended to represent economic stress within the local population (Corzine, Huff-Corzine, and Creech 1988; Tolnay and Beck 1995).¹⁹ We construct quartiles for each of these contextual variables and assign individuals (victims and non-victims) to the appropriate quartile, based on the characteristics of their counties. By including appropriate multiplicative interaction terms in our logistic regression models, we are able to determine whether the influence of individual-level risk factors for victimization differed across contexts.²⁰

Caveats

Like all record linkage efforts, ours encountered a variety of challenges, in addition to those mentioned above (i.e., census underenumeration, geographic mobility, and the destruction of the 1890 census manuscripts). First, for a relatively modest proportion (9.3%) of the cases in the original Beck-Tolnay inventory of lynch victims, the name of the victim is not known. Without a name it is impossible to link a victim to a census record. Second, the Beck-Tolnay inventory contains only incidents in which a mob was successful in killing their targeted victim. There also were many cases of potential lynchings that were averted, most often through the efforts of local citizens and/or authorities. Although a comprehensive inventory of averted or prevented lynchings is not available, it is likely that they were more common during later decades. A case could be made that, ideally, a study of the selection of victims by southern mobs should include those cases in which an individual was targeted by the mob, but was spared execution through the intervention of actors not involved in the selection. Third, many of the victims included in the Beck-Tolnay inventory could not be linked to their census records. To the extent that these three sources of sample truncation were random with respect to the personal characteristics of lynch victims, they do not pose a serious threat to our investigation. However, given the lack of appropriate information, we are unable to demonstrate such randomness. Rather, we propose what seem to us to be some reasonable a priori assumptions.

1. Victims without names were likely to be drawn from the more socially marginal population of southern black males.

2. A potential lynching was more likely to be averted when the intended victim was known within the local community and, therefore, more likely to be of higher socioeconomic status and not socially marginal.

3. While the socially marginal, of lower socioeconomic status, were probably more likely to be underenumerated in the census, this source of truncation also applied to the comparison sample of non-victims that is derived from the census manuscripts.

It is important to keep these caveats, as well as our a priori assumptions, in mind when interpreting the empirical results that follow.

IV. Results

A. A Statistical Profile of Lynch Victims and Comparison with Non-Victims

Table 2 presents selected characteristics of southern lynch victims, aggregated over decades and across states. Beyond the detailed information available in a handful of case studies of individual lynchings, Table 2 is the first glimpse that social scientists have had of the profile of the victims of southern mob violence. The bulk of victims were in their twenties when they were killed, with 59% falling below age 30. Approximately one in every six victims was born in a state that was different from the state in which they were lynched. Four-in-ten adult victims were recorded as the head of their household and about 15 percent of the household heads owned the dwelling in which their family resided. A slight majority (55%) of the adult victims were currently married. Just under half (45%) of all victims were able to read and write. Only 3.5% of victims were employed in skilled occupations. About one-in-nine victims was enumerated as a “mulatto,” rather than as “black” or “negro.”

Table 2.

Selected Characteristics of Black Male Lynch Victims and Non-Victims, Ten Southeastern States, 1882 – 1929

	Victims	Non-Victims
	Proportion	Proportion
Age
< 20 years	0.204***	0.341
20 – 29 years	0.381***	0.257
30 – 39 years	0.208***	0.154
40 – 49 years	0.109	0.115
50 years and older	0.098**	0.133
N of Cases	923	21,136
Marrieda	0.551***	0.625
N of Cases	650	16,460
Head of Householda	0.443***	0.656
N of Cases	650	16,460
Born out of State	0.177	0.197
N of Cases	923	21,136
Literateb	0.449	0.452
N of Cases	817	21,136
Homeownerc	0.145	0.120
N of Cases	166	10,810
Racial Identificationd
Black	0.890*	0.860
Mulatto	0.111*	0.140
N of Cases	688	15,316
Skilled Occupatione	0.035***	0.069
N of Cases	718	17,851

Notes:

^***p ≤ 0.001;

^**p ≤ 0.01;

^*p ≤ 0.05.

^a)Sample restricted to those aged 16 and older;

^b)Sample restricted to those aged 10 and older;

^c)Sample restricted to those enumerated as heads of household;

^d)In 1900, census records do not differentiate between blacks and mixed-race people, so this sample includes only victims and a corresponding comparison sample from 1 the following decades: 880, 1890, 1910, and 1920;

^e)Sample restricted to those aged 14 and older.

The statistical profile of black male lynch victims becomes even more meaningful when we have an appropriate standard for comparison. Also included in Table 2 are the characteristics of black males who lived in the same counties in which lynchings occurred, and who were within the same age range of the victims. Overall, the bivariate evidence offers support for the marginalization hypothesis. Consistent with the argument that southern lynch mobs selected victims who were less rooted in the community, on average lynch victims were significantly less likely than non-victims to be married or enumerated as heads of households.

In contrast, the differences between victims and non-victims provide little support for the status transgression hypothesis. Victims were no more likely than non-victims to be literate or to be homeowners. Furthermore, victims were significantly less likely to have skilled occupations or to be enumerated as “mulatto,” in contradiction to the status transgression hypothesis.

In sum, the bivariate contrasts between lynch victims and non-victims suggest that the victims were both more marginal to their communities and of generally lower social standing than non-victims.

B. Modeling the Likelihood of Victimization

Because some of the characteristics described in Table 2 are not independent of each other, it is useful to also combine them in models predicting the likelihood of lynching in order to gain a better understanding of their unique influence on the risk of victimization. The baseline model reported in Model 1 of Table 3 describes (1) temporal variation in the likelihood of being lynched, in the form of coefficients for dummy variables representing the time periods (with the 1880s serving as the reference); and (2) variation by the age of the individual, modeled as a quadratic.²¹ Our findings comport well with previous aggregate-level evidence showing that lynching peaked in the 1880s and 1890s and then declined in later decades (see e.g., Tolnay and Beck 1995). The variable risk of being lynched by age, of course, has never been described because the required data were not available. Nonetheless, the non-linear relationship suggested by the results for Model 1 is intuitively reasonable. That is, as shown in Figure 1, the likelihood of being lynched rose with age until reaching a peak of roughly 34 years and then declined. The age pattern for the risk of lynching is consistent with the frequent finding that young males are more likely to be the victims of violent crimes, more generally.

Table 3.

Results from Logistic Regression Analyses Predicting Likelihood of Being Lynched, Black Males Age 10 – 70, 1882 – 1930

	Model 1	Model 2a	Model 3a	Model 4b	Model 5a
Age	0.136*** (0.013)	0.251*** (0.021)	0.104*** (0.022)	0.191*** (0.018)	0.239*** (0.014)
Age Squared	−0.002*** (0.000)	−0.003*** (0.000)	−0.001*** (0.000)	−0.002*** (0.000)	−0.003*** (0.000)
1890 – 1895	0.019 (0.104)	−0.504*** (0.144)		−0.406** (0.126)	0.001 (0.107)
1900 – 1909	−0.386*** (0.099)	−0.341** (0.116)			−0.427*** (0.102)
1910 – 1919	−0.419*** (0.105)	−0.467*** (0.124)	−0.130 (0.115)	−0.451*** (0.118)	−0.521*** (0.110)
1920 – 1929	−0.706*** (0.121)	−0.644*** (0.136)	−0.287* (0.129)	−0.736*** (0.136)	−0.811*** (0.128)
Married, Head of Household		−1.233*** (0.099)			−1.602*** (0.091)
Not Married, Head of Household		−0.786*** (0.127)			−0.402*** (0.122)
Born Out of State		−0.239* (0.101)			−0.342*** (0.093)
Literate			0.120 (0.105)	0.185 (0.098)	0.009 (0.075)
Skilled Worker			−0.922*** (0.260)	−0.638** (0.239)	−1.056*** (0.205)
Homeowner			−0.111 (0.130)
Mulatto				−0.268 (0.139)
Intercept	−4.846*** (0.203)	−6.903*** (0.365)	−4.952*** (0.398)	−6.020*** (0.312)	−6.058*** (0.222)
Pseudo R2	0.030	0.080	0.014	0.042	0.098
N=	22,053	17,110	11,751	14,714	22,053

Note: Standard errors in parentheses.

^***p ≤ 0.001;

^**p ≤ 0.01;

^*p ≤ 0.05.

^a)Sample restricted to black males who were 16 and older at time of census enumeration;

^b)Sample restricted to black males age 12 and older at time of census enumeration.

FIGURE 1.

Variation in Probability of Lynching by Age, Black Male Victims in Ten Southeastern States, 1882–1929

Note: Calculated using predicted values derived coefficients reported for Model 1, Table 3, holding regional distribution constant.

Model 2 offers an initial, multivariate, test of the marginalization hypothesis.²² The results are uniformly consistent with the bivariate evidence presented in Table 2. Being the head of a household – and especially being married – was associated with significantly lower risks of being lynched.²³ Converting the logit coefficients for these two variables to odds ratios reveals that the odds of being lynched for married heads of household were only 29% of the corresponding odds for single men who were not heads of household (e^−1.233 = 0.29). Similarly, the odds of being lynched for unmarried heads of households were only 46% of those for unmarried men who were not heads of household (e^−0.786 = 0.455). Being born out-of-state is significantly negatively related to the likelihood of being lynched, net of the other independent variables included in the equation, contrary to the prediction of the marginalization hypothesis. As is true of the results for all subsequent models, the differences by marital status, household headship, and place of birth in Model 2 are net of the influences of age and time period.

Model 3 and Model 4 offer separate tests of the status transgression hypothesis by alternately including home ownership and mulatto status on the right-hand side of the equation, along with occupation. Separate models were warranted because the two characteristics are available in different combinations of decennial censuses – 1900, 1910, and 1920 for home ownership; 1880, 1910, and 1920 for mulatto status. Both models omit the indicators that were used in Model 2 to test the marginalization hypothesis.²⁴ Consistent with the bivariate evidence, the results provided by Model 3 and Model 4 provide little empirical support for the status transgression hypothesis. Model 3 reports no relationship between home ownership and the likelihood of lynching, while the odds of skilled workers being lynched were only 40% of those for other workers (e^−.922 = .40). Model 4 reveals that the likelihood of being lynched was actually lower for those who were classified as “mulatto” in the census (although the relationship is not statistically significant by conventional standards, the coefficient for mulatto status is 1.98 times larger than its standard error). In addition, the negative relationship between holding a skilled occupation and being lynched persists in Model 4. Both of these findings are inconsistent with the status transgression hypothesis. The positive coefficient for literacy in Model 4 fails to reach statistical significance at the p < .05 level but is 1.89 times larger than its standard error.

In light of the non-significance of homeownership in Model 3 and borderline negative effect of mulatto status in Model 4, which contradicts the status transgression hypothesis, we have omitted both variables in our analysis that combines indicators for the marginalization and status transgression hypotheses in the same equation, allowing us to retain data from all four decades. The results from that model estimation are reported in Model 5. For the most part, the findings from the comprehensive model reinforce those from the separate analyses. That is, the significantly lower likelihood of being lynched for married men and heads of households suggests lower risk for individuals who were more rooted in their communities and, therefore, less marginal within local society. Further, the significantly lower probability of being lynched for skilled workers, as well as the non-significant influence of literacy, contradict the predictions of the status transgression hypothesis. In contrast, the negative effect of being born in a different state is inconsistent with the social marginalization hypothesis. We have no good explanation for why those who were born outside of the state of lynching should have been exposed to a lower likelihood of being lynched. It is possible that this effect partially reflects a stronger influence of inter-state migration on our ability to successfully match lynch victims with their census records, than on census enumeration, which is related to inclusion in our comparison sample of non-victims obtained from the historical PUMS files. Alternatively, it could reflect the operation of positive selection in the inter-state migration of black males which is not entirely accounted for by the other predictor variables included in the models.

C. Temporal and Spatial Variation in the Risk Factors for Victimization

Table 4 reports, by time period, the coefficients and standard errors for the individual-level predictors of victimization included in Model 5 of Table 3. We have chosen to replicate Model 5 in our consideration of temporal and spatial variation in risk factors in order to retain information for all census years. Caution must be exercised when interpreting the time-specific results in Table 4 because the number of victims within decades is reduced. While the coefficients representing the effects of individual-level characteristics do vary across decades, in some cases significantly so, no obvious pattern of variation in coefficients emerges and, most importantly, in no decade do we find support for the status transgression perspective. That is, being able to read and write, or holding a higher status occupation, never put black males at a significantly greater risk of being targeted for lynching. In contrast, being a married head of household appears to have been a protective factor against being lynched during all time periods. Similarly, skilled workers enjoyed a consistently lower risk of being targeted in all time periods, reaching statistical significance in 1890–95 and 1900–09.

Table 4.

Effects of Individual-Level Characteristics on Lynching Probability by Decade

	1882–89	1890–95	1900–09	1910–19	1920–29
Age	0.266*** (0.032)	0.386*** (0.038)	0.193*** (0.028)	0.712*** (0.030)	0.184*** (0.041)
Age Squared	−0.003*** (0.000)	−0.004*** (0.001)	−0.002*** (0.000)	−0.002*** (0.000)	−0.002*** (0.001)
Married, Head Household	−1.559*** (0.120)	−2.427*** (0.216)	−0.889*** (0.170)	−1.157*** (0.194)	−0.527* (0.249)
Married, Not HH Head	−0.380 (0.354)	−2.125*** (0.612)	−0.427 (0.410)	−0.541 (0.440)	0.489 (0.371)
Born Out of State	−0.508* (0.215)	−0.702** (0.231)	0.189 (0.176)	−0.023 (0.205)	−0.158 (0.242)
Literate	−0.058 (0.179)	−0.157 (0.190)	0.115 (0.139)	0.019 (0.159)	0.127 (0.216)
Skilled Worker	−0.268 (0.311)	−2.457* (1.019)	−1.457*** (0.392)	−0.621 (0.467)	−0.219 (0.477)
N=	3,701	3,413	6,053	5,000	3,886
Adj. R-Sqr.	0.123	0.235	0.092	0.059	0.128

Note: Coefficients are presented first. Standard errors are in parentheses.

^***Significant at the p ≤ 0.001 level;

^*significant at the p ≤ 0.05 level.

Our consideration of spatial variation in the effects of individual-level risk factors for lynching is summarized in Table 5 which reports the coefficients for individual-level predictors within quartiles of three contextual variables, as well as symbols to indicate significant differences in coefficients across quartiles. Like the evidence for temporal variation in risk factors, Table 5 shows a number of cases of statistically significant differences across quartiles. However, those differences reveal no clearly discernible pattern that would alter the general conclusions that we drew from the general evidence in Table 3. That is, under no local conditions do we observe support for the status transgression perspective. The effect of literacy on the risk of victimization fails to achieve statistical significance within any local context. And, when the effect of occupation is statistically significant (in 9 out of 12 cases), more skilled workers have a lower likelihood of being lynched. In contrast, married heads of households are less likely to be singled out by lynch mobs within all socio-demographic environments, consistent with the evidence in Table 3. Again, the occasionally negative effect of being born out-of-state (8 out of 12 cases) is consistent with the more general findings in Table 3, and represents the strongest and most consistent empirical evidence that contradicts the marginalization hypothesis.

Table 5.

Variation in Individual-Level Effects on Risk of Lynching by Local Context

Percent Black	1st Quartile	2nd Quartile	3rd Quartile	4th Quartile
Age	0.242*** ♦ (0.017)	0.228*** (0.015)	0.247*** ♦ (0.016)	0.253*** ♦ (0.015)
Married Household Head	−0.970*** (0.222)	−1.077*** (0.175)	−1.145*** (0.167)	−1.285*** (0.122)
Born Out of State	−0.567* (0.271)	−0.136 (0.214)	−0.470* (0.218)	−0.274 (0.144)
Literacy	−0.181 (0.210)	−0.101 (0.163)	−0.123 (0.159)	−0.075 (0.114)
Skilled Occupation	−0.415 ♦ (0.634)	−1.531** (0.471)	−0.751 ♦ (0.460)	−0.438 ♦ (0.324)
Change in Pct. Black	1st Quartile	2nd Quartile	3rd Quartile	4th Quartile
Age	0.222*** (0.016)	0.231*** (0.017)	0.223*** (0.016)	0.232*** (0.016)
Married Household Head	−1.447*** (0.241)	−1.420*** (0.231)	−1.390*** (0.240)	−1.120*** (0.120)
Born Out of State	−0.642* (0.300)	−0.406 (0.306)	−0.334 (0.310)	−0.541* (0.212)
Literacy	0.191 ♠ (0.229)	0.337 ♠ (0.219)	0.282 ♠ (0.230)	−0.171 (0.187)
Skilled Occupation	−1.617* (0.648)	−1.469* (0.669)	−1.327 (0.700)	−1.396** (0.541)
Farm Tenancy	1st Quartile	2nd Quartile	3rd Quartile	4th Quartile
Age	0.238*** (0.016)	0.225*** (0.015)	0.239*** (0.016)	0.237*** ♦ (0.015
Married Household Head	−1.251*** (0.187)	−1.504*** (0.192)	−1.279*** (0.192)	−1.317*** (0.130)
Born in State	−0.625** ♣ (0.203)	−0.556* ♣ (0.249)	−0.282 (0.233)	−0.519 ** ♣ (0.162)
Literacy	−0.125 (0.174)	−0.147 (0.181)	−0.135 (0.181)	−0.086 (0.122)
Skilled Occupation	−1.252** (0.449)	−1.253** (0.465)	−0.991* (0.445)	−0.822* (0.376)

Note:

^***significant at the p ≤ 0.001 level;

^**significant at the p ≤ 0.01 level;

^*significant at the p ≤ 0.05 level.

^♠significantly different from 4^th quartile at the p ≤ 0.05 level;

^♣significantly different from 3^rd quartile at the p ≤ 0.05 level;

^♦significantly different from 2^nd quartile at the p ≤ 0.05 level.

In sum, our exploratory consideration of temporal and spatial variation in the effects of individual-level risk factors for lynching reveal some differences over time and across space, but do not significantly alter our inference of greater empirical support for the importance of social marginality versus status transgressions.

V. Conclusion

Social scientists have long relied on aggregate patterns, or case studies, to better understand the history of racial violence in the American South. From those efforts we have learned a great deal about the institutional forces that created climates conducive to mob violence. Until now, however, researchers have lacked sufficient information about individual victims to draw inferences about the characteristics that shaped the risk of being singled out by lynch mobs. The evidence presented in this paper is the first of its kind, based on a newly created database that includes extensive individual- and household-level information for the victims of southern lynching. The availability of that information has allowed us to present a statistical profile of the victims. Coupled with a comparison sample of non-victims, it has further allowed us to assess the role of selected characteristics in determining the relative risk of victimization.

Our findings suggest that the risk of lynching was greater for those black males who were less rooted in their communities, and therefore more marginal to local society. In contrast, we find no evidence to suggest that black males were singled out for their social or economic success, either educational (as measured by literacy) or occupational. Our consideration of temporal and spatial variation in the effects of individual-level risk factors revealed some differences across decades and between local areas with different socio-demographic profiles, but no empirical evidence to challenge our conclusion that socially marginal, not socially successful, men were more likely to be targeted by lynch mobs.

What specific mechanisms might account for the patterns we have described? To answer that question, of course, requires us to speculate beyond the statistical evidence. Perhaps southern lynch mobs disproportionately targeted individuals who could expect less protection from their neighbors -- white and black. Black males who provided more skilled services to the local economy may have been viewed as more essential by the white elite. In a similar vein, it is possible that lynch mobs and their supporters were reluctant to risk provoking strong reactions from the local black community, which would have been more likely for victims who were deeply rooted within the black community or were respected for their occupational status. Therefore, when they selected their victims, mobs disproportionately targeted the more socially and economically marginal members of their communities. Finally, local blacks may have been less likely to cooperate with potential lynch mobs by sharing information regarding the location or identity of their intended victim when the targeted individual was a higher status, well-integrated, member of the community.²⁵

An alternative interpretation of our findings might argue that individuals who are socially marginal, and less rooted within the local community, are more likely to violate behavioral norms (minor or serious) and, therefore, put themselves at greater risk of being lynched by provoking lynch mobs with their behavior. According to this argument, in the context of the historical South, marginalized black men were more likely to be lynched because they were also more likely to commit crimes and engage in racially-charged behavioral transgressions than were black men who were more attached to the “conventional order.” The logic of this argument, drawn loosely from the “Social Control Theory” (Hirschi 1969, 1977) of non-normative behavior, has enjoyed support for a variety of social settings. However, we are reluctant to embrace it as the primary interpretation of our findings. First, it “blames the victim” without adequate support for the inference (1) that lynch victims were guilty of any offense, and (2) that unmarried, non-heads of households, with less skilled occupations, in the South during this time period, were more likely to engage in non-normative behavior. Second, it shares an intellectual affinity with the “Popular Justice” explanation for southern lynching (i.e., that lynching substituted for absent or ineffective law enforcement) which, when examined statistically, has been found wanting (Tolnay and Beck 1995: Chapter 4).

However, it is possible that communities varied in the extent to which relatively minor, caste-driven, behavioral expectations for black men were enforced, for example yielding to whites on the sidewalk or addressing whites respectfully. If this were the case, we would expect to find that victims who were accused of minor crimes or of non-criminal social transgressions would be less similar to other black men in their communities than was true for men accused of more serious crimes. In supplementary analyses,²⁶ we disaggregated the victims into three groups based on the nature and severity of the offense they were accused of committing. We separately identified men who were accused of serious nonsexual crimes that inflicted or could have inflicted bodily harm – specifically murder, assault, robbery and arson. The second category of victims included black men accused of sex crimes in which physical contact was alleged to have occurred, or could have been inferred based on the reported circumstances. The final subset of victims included men who were accused of minor crimes like theft, as well as those whose actions threatened the social order. When compared with appropriate samples of non-victims, drawn from the same counties in which lynchings of the three types occurred, the basic findings reported in Table 2 persist.²⁷ That is, for all three groups of offenses, more marginal black men were more likely to be targeted by lynch mobs. Thus, we find no support for the notion that men accused of committing minor crimes or social transgressions were more marginal within their communities than were men who were accused of more serious sexual or nonsexual crimes.

A final mechanism could have operated at a very early stage in the potential lynching incident. That is, given the stimulus for a possible lynching (i.e., an alleged crime or offense), it is possible that black men with stronger ties to the local community were more likely to be dealt with by the official legal establishment, while less socially rooted men were more likely to be dealt with informally. This possible explanation for the greater vulnerability of socially marginal black men could be evaluated systematically if a comparable census-linked database of targeted individuals in averted lynchings were available. Given the limitations of existing data, the question regarding the precise mechanisms accounting for the greater vulnerability of socially marginal black men must remain for future scholars to answer.

It is, of course, impossible for us to adjudicate, definitively, among these alternative interpretations from the statistical evidence we have presented in this paper. To do so would require knowing what was going through the minds of the members of the lynch mob. However, it is possible that future research, building from the comparative evidence offered here, will be able to move us even closer to a fuller understanding of why more marginal black men were disproportionately targeted by such mobs.

While our findings offer little support for the status transgression hypothesis, it is important to remember that the statistical evidence allows us to make probabilistic, rather than deterministic, statements about the relative risk of being lynched. The historical record, and the database we have constructed, include many cases in which successful black men, who were leaders within their communities, were victimized by southern lynch mobs. Two additional examples help to emphasize this point. Louis Rice, a 65-year-old physician and farm owner, was lynched in Lauderdale County, Tennessee on March 22, 1900. Rice’s alleged offense was his advocacy within the court system on behalf of another black man who was accused of murdering a white man (testifying and performing research for the defendant). However, it is possible that his elevated status within the community contributed to his vulnerability. S.S. Mincey, a leader in the Republican Party, was lynched on July 29, 1930 in Montgomery County, Georgia. Mincey was kidnapped and beaten to death because he refused to give up his post as local Republican Party chair. In Mincey’s case, his prominent position in local politics directly led to his death at the hands of a lynch mob. Thus, while social marginality appears to have increased the likelihood of being targeted by southern lynch mobs, an elevated social standing certainly did not provide immunity.

It is important, we believe, to emphasize the important distinction between (1) the macro-level, institutional, societal forces that created environments within which lynching was more likely, and (2) the micro-level, individual-level, characteristics that determined the relative risk of victimization. Thus, our failure to infer support for the status transgression hypothesis should not be construed as evidence that racial competition, and efforts to control the black population, did not operate at a higher level of aggregation. That is a different question which requires a different type of evidence.

We recognize that our data and analyses have a number of limitations. First, we cannot claim that our sample of victims who were successfully linked to their census records is representative of all victims lynched in these ten states during these fifty years. Second, while the census records provide vastly more information about the victims and their families than was previously available, they are limited to the information collected by the Census Bureau. Third, we could not include in our study comparable information for the large number of potential lynch victims who were spared through the intervention of local citizens or legal authorities. Finally, our consideration of temporal and spatial variation in the targeting of lynch victims has been exploratory, and deserves a more comprehensive investigation by future researchers. Still, the evidence we have presented raises a figurative curtain that has obscured the individual characteristics of the vast majority of southern lynch victims since the lynching era ended. And, they provide important leverage for gaining a better understanding of the kinds of individuals who faced a greater risk of being targeted by lynch mobs.