Modernization and Lynching in the New South

This article evaluates an emerging body of historical scholarship that challenges prevailing views of the primacy of rural conditions in southern lynching by positing that it was symbiotically associated with the processes of modernization underway in the region in the decades around 1900. Statistical analyses of lynching data that differentiate among events according to communal participation, support, and ceremony in Georgia and Louisiana from 1882 to 1930 and local-level indices of modernization (urbanization, rural depopulation, industrialization, agricultural commercialization, and dissolution of traditional family roles) yield results that both support and contradict such a modernization thesis of lynching. The findings imply that the consequences of the social transformation in the South coinciding with the lynching era were not uniform throughout the region with regard to racial conflict and violence and that broad arguments proposing an intrinsic connection between modernization and lynchings therefore are overstated.

T HE lynching of African Americans by white mobs around the turn of the nine- teenth century has garnered scholarly attention for over a century, resulting in a consensus view of its fundamentally rural character.While early observers saw it as an expression of primordial southern backwardness and racism (Cutler 1905;Raper 1933;White 1929), later observers viewed it as grounded in the late nineteenth and early twentieth centuries' relations of agricultural production, characterized by widespread landlessness and farm tenancy among whites and blacks alike (Beck and Tolnay 1990;Corzine, Huff-Corzine, and Creech 1988;Tolnay and Beck 1995).
An increasing number of historical studies approach lynching from a different angle, viewing it within the context of the modernization processes coinciding with the "lynching era" from 1880 to 1930, whereby the South experienced a transition away from small-scale local communities into more of a large-scale mass society.Taking issue with earlier arguments that modernization broke southern backwardness and thereby eradicated the root causes of lynching, this emerging literature exhorts us to broaden our understanding of lynching and reconsider its role in a critical transformative period of U.S. society in general and of its race relations in particular.Rather than seeing modernization as a progressive force dampening racist attitudes and racial violence, it argues that the processes of modernization visited upon the South in the decades around 1900 were, because of their destabilizing effects on traditional white communal values and relation as well as on relations between whites and blacks, intrinsically conducive to lynching (e.g., Feimster 2009;Goldsby 2006;Pfeifer 2004;Wood 2009).
This departure from conventional wisdom reflects to a certain extent the academic division of labor characterizing the contemporary lynching literature.Sociological contributions to this scholarship consist mainly of quantitative studies using statistical methods to analyze exhaustive inventories containing hundreds or even thousands of lynching events.In contrast, what I call the "modernization thesis of lynching" emerges from historical scholarship, either in the form of case studies of particular lynchings, including their textual or visual representations, or studies using comprehensive event catalogues but unaccompanied by rigorous methods of analysis.From the standpoint of the sociological strand of lynching research, the conclusions of the historical strand therefore rest upon weak methodological and empirical foundations, remaining more suggestive than conclusive and requiring further systematic assessment.
These limitations are no reason for dismissing out of hand claims proffering a connection between modernization and lynching because they offer a set of working hypotheses amenable to empirical testing.The modernization thesis of lynching, furthermore, compels attention because it carries considerable face plausibility in view of well-established sociological theory and research into social movements and collective action.In an attempt at interdisciplinary dialogue and knowledge accumulation in an otherwise fragmented research field, this article accordingly engages and evaluates the modernization thesis of lynching by, for one thing, illuminating its consistency with well-established sociological theory and research and, for another, testing a number of hypotheses derived from its core claims using systematic methods of data collection and analysis.
The Modernization Thesis of Lynching: Theoretical Foundations, Historical Background, and Empirical Arguments The explanatory principles of the modernization thesis of lynching invoke theoreticalconceptual models casting intergroup mobilization as a response to broad structural social changes.The effect of structural transformations can cut two analytically different ways depending upon the mobilizing goals and claims-"competitive" and "reactive" mobilization (Tilly 1970(Tilly , 1978)).Competitive mobilization refers to actions by members of one group claiming priority access to valued scarce resources also claimed by members of another rivaling group.This type of mobilization may not manifest itself primarily in collective action but rather in "uncoordinated though similar individual acts" (Blalock 1967:160).Reactive mobilization refers to actions attempting to reestablish certain moral and social principles revolving around intra-as well as intergroup values and relations perceived as challenged or lost.Unlike competitive mobilization, which may very well mainly take the form of individual actions, reactive mobilization thus primarily manifests itself in collective action.Insofar, reactive mobilization represents "collective responses to collective experiences" (Tilly 1970:163).

Modernization and Lynching as Competitive Mobilization
Quantitative sociological lynching studies deriving their theoretical impetus from perspectives of intergroup competition to explain lynching demonstrate how it was more common where and when blacks threatened whites' agricultural economic interests.In such circumstances, lynching furthered the interests of whites across classes by, on the one hand, serving white landlords by controlling the large black rural labor force and, on the other hand, serving smallholder and landless whites by reducing the competition from African Americans for land, farm tenant, and labor opportunities in the southern agricultural economy (Beck and Tolnay 1990;Corzine et al. 1988;Tolnay and Beck 1995).
This type of competition-based explanation also informs research arguing that modernization processes sparked lynching.Thus, in parallel with studies focusing on lynching as a means to reduce black threat to white economic interests linked to the southern agricultural economy, some authors maintain that increasing interracial contact and competition under urbanizing and industrializing conditions found expression in mob violence as well.Pfeifer argues that "The conglomeration of rural African Americans and country whites in developing urban and industrial spaces provoked substantial tensions . . .which were sometimes channeled into large working-class mobs" (Pfeifer 2004:141f).Wood (2009:13f) likewise asserts that "If urban life had threatened white authority by bringing whites and blacks together . . .lynchings performed on city streets . . .reclaimed urban, public spaces as decidedly white spaces." Almost no studies in the quantitative sociology lynching literature explore these issues and the few exceptions have yielded mixed results.In a cross-county study of lynching in Georgia, 1890-1900, Soule (1992) ) argues that industrial growth increased interracial economic competition, which, in turn, led to lynchings.In contrast, in a time-series analysis spanning the years 1882-1914, Olzak (1992) ) finds no evidence of a relationship between the level of urbanization of African Americans and lynching in the South.

Modernization and Lynching as Reactive Mobilization
The modernization thesis of lynching amounts to more than a competition-based explanation though and most importantly invokes mob violence as a form of reactive white collective action resisting the modernizing forces undermining the traditional values and relations at the foundation of white communities.Thus, rather than understanding lynching as promoting comparatively limited instrumental economic interests, authors arguing that it was part and parcel of southern modernization cast it as expressive of broader white moral and social concerns.What moral values and social relations did modernization threaten, then, and how did racial violence against blacks supposedly restore them?The South emerging from the upheavals of the Civil War and the abolition of slavery remained predominantly an agricultural region, but it was not the same agricultural region as earlier.The shift from primarily self-sufficient subsistence farming to commercial cash crop production, above all cotton but also tobacco and rice, dramatically changed the conditions of southern life because it eclipsed not only yeomen farmers-constituting the majority of the antebellum white population-but also disrupted a distinctive way of community life.As antebellum yeomen were embedded in local interpersonal networks (including planters in planation areas) involving the reciprocal exchange of goods and services, their self-sufficiency did not operate on the individual but on the collective community level.In other words, in antebellum times, ideals and relations of communal autonomy, sovereignty, and solidarity dominated people's lives (Ayers 1992;Redding 1992).
Thus, while New South leaders advocated for the necessity of regional modernization through urban, industrial, and commercial expansion, the yeoman tradition remained a strong ideal against which common whites of the postbellum South measured society and themselves.Modernizing developments therefore threatened, in their eyes, a highly prized social and moral order.The post-Civil War expansion of large-scale markets in land and farm products transformed freeholders from independent producers into commercial farmers at the mercy of national and international agricultural markets.A crucial consequence of the transition to cash crop production was that farmers had to obtain food and other supplies from merchants on credit using liens on their future crop as well as their land as security.But because of declining farm commodity prices and an appreciating dollar, such crop lien credit trapped a great many white farmers in ever-deepening indebtedness, leading them to default on their loans and lose their farms.In this way, many rural whites fell into farm tenancy or wage labor or had to pack up to find livelihood in growing urban and industrial areas.In consequence, the white industrial labor force and urban population of the South in the late nineteenth and early twentieth century consisted to a large extent of people with strong rural roots (Ayers 1992).
The transformation of the southern agricultural landscape directed, in turn, developments further tearing apart the fabric of the rural South because the migration of whites away from the countryside not only uprooted individuals from their rural communities but also furthered the inroad of modernization into those very communities.This was because the region's postbellum industrial and urban developments were fundamentally shaped by the dominance of its agricultural sector.The majority of southern manufacturers were located in labor-surplus rural areas and functioned as an extension of the agricultural economy, as indicated by its concentration in cotton and cottonseed mills, lumber mills, fertilizer factories, and turpentine camps (Cobb 1984).
An important line of argument pursued by proponents of the modernization thesis of lynching is that modernizing developments were conducive to lynching not only because they disrupted traditional white communal values and relationships but also familial ones.In the antebellum South, most white men had for generations shared and realized the tenets of Jeffersonian republicanism, which portrayed them as independent agrarian producers and citizens and relegated white women and children to a position of dependence.The ideology and practice of white male social, political, and economic primacy were symbiotically coupled with a tradition of chivalry, wherein a white man's honor and authority was contingent upon his ability to protect and provide for the dependents (that is, females and children) in his household.Modernizing developments had important implications for such traditional family and gender roles.The increasing indebtedness and landlessness, often accompanied by urbanization and proletarization, among rural whites caused by the commercialization of agriculture not only put severe strains on the ideal of white male economic independence but also implied a loss of social status and honor.Likewise, undermining the traditional familial dominion of white males as heads of farm households was the increasing number of white, especially young and unmarried, women who joined the ranks of wage earners outside the household along with the modernization of the South (Ayers 1992).
Before moving on, it may at this point be worth emphasizing what type of argument the modernization thesis of lynching does not make.As the modernization thesis of lynching argues that certain antebellum moral and social values were not suspended while new ones were being formed but still very much in play in the postbellum period, it does not entertain a Durkheimian perspective that explains collective mobilization as a response by isolated and wayward individuals to conditions of moral ambiguity and social disorganization caused by structural change.Instead, the modernization thesis of lynching views lynching as a form of reactive collective action-a form of "reactionary violence" (Tilly 1970(Tilly , 1978))-growing out of collective white experiences revolving around values and relations among and between southern whites and blacks.In other words, the modernization thesis of lynching posits lynching as a collective act of resistance by whites perceiving the moral and social foundations of their traditional communal life imperiled.
In this, the modernization thesis of lynching parallels research viewing the agrarian protest movement sweeping the late nineteenth-century South in the form of the Farmers' Alliance and the People's Party as a reaction to the erosion of traditional rural values and relations by the emerging urban-industrial-commercial society.Scholars agree that this movement expressed rural traditionalism and provided a channel for popular rural discontent through which whites who felt that they were losing ground attempted to collectively reclaim their place and voice in southern society.With its programs of joint marketing and credit, wherein more well-to-do members would help out their less-fortunate fellows, the Alliance, for example, meant to lessen farmers' dependence of crop-lien credit and grant them greater economic self-determination and stability, thereby recapturing the "cooperative commonwealth" among the tillers of the soil (Goodwyn 1976:xx)."Farmer activism," Redding (1992:345) thus observes, "was not just a response to the economic crisis, but also to the loss of independence and community." Along similar lines of argument, the modernization thesis of lynching argues that lynching represented an adaptation of traditional principles and practices of localistic popular autonomy, sovereignty, and solidarity to new circumstances.In this view, lynching was a form of reactionary collective mobilization that offered, on the one hand, whites in rural communities destabilized by demographic shifts and agricultural commercialization and, on the other hand, whites dislodged from traditional rural communities in urban and industrial circumstances a means to resist disruptive modernizing forces.Furthermore, some scholars argue that lynching grew out of the transformation of traditional white familial roles in general and the decline of white male power and authority in particular, including, among other things, the participation of young, unmarried white women in the wage earning labor force (Feimster 2009;Pfeifer 2004;Wood 2009).

The Modernization Thesis of Lynching and Different Forms of Lynchings
The modernization thesis of lynching does not simply posit a positive association between modernization and lynching in general but makes certain conjectures regarding the specific nature of the white-on-black mob violence spurred by such developments.In order to gain conceptual purchase on this issue, I distinguish between "private" and "public" lynchings.This distinction draws on the fourfold lynching typology developed by Brundage (1993), consisting of private, terrorist, posse, and mass mobs, with "private" lynchings comprising the former two and "public" the latter two types.The reason for collapsing Brundage's fourfold typology into a twofold typology is that the significant overlap in size, composition, and behavior makes it difficult to conceptually and practically distinguish private from terrorist mobs on the one hand and posses from mass mobs on the other (cf.Brundage 1993;Pfeifer 2004).
Private lynchings were a form of interpersonal violence typically concerned more with "secrecy than ceremony" (Brundage 1993:30), perpetrated by relatively small groups united by bonds of kinship or friendship to pay retribution for (alleged) harm done to self, family, or friends.Note that while mobs of this type occasionally claimed to defend community values by responding to blacks' social propriety violations, they, like mobs openly settling personal scores, were often carried out under disguise as well as constituted by family and friends of an offended white party and commonly represented the culmination of long-standing interpersonal conflicts (Brundage 1993;Pfeifer 2004).
Public lynchings signified collective violence because they marshaled broadbased community participation and support.While the legitimacy of private mobs could be questioned, public mobs were hardly ever challenged but rather glorified as expressing the public will and contributing to the public good.As Wood (2009:43) observes, "Making a lynching public and spectacular rendered it more legitimate than an act of vigilante violence performed secretly."The "spectacular" character of public lynchings that Wood alludes to refers to their occasional ceremonial elements, including excessive brutality.Furthermore, groups of white men ("posses") that in their search for African Americans suspected of serious crimes against whites ended up killing rather than apprehending the suspects were similarly a highly emotionally charged form of interracial violence enjoying wide communal participation and approval among whites.Although lacking explicit ceremonial elements, these killings brought perpetrators and audiences together in collective prekilling frenzies and postkilling celebrations (Brundage 1993;Pfeifer 2004).
By conceiving lynching as a collective white revolt against modernizing developments in the New South, the modernization thesis of lynching thus argues that the collective violence of public lynchings worked to enact traditional communal and familial values and relations among southern whites as well as to preserve their dominance over blacks during a period of social transformation.One of the main proponents of the modernization thesis of lynching, Amy Louise Wood (2009:9, 14), accordingly argues that, "[Public] lynching spectacles revealed the threads with which white southerners stitched together an idealized white community . . .The spectacle of [public] lynching, in these ways, erupted and thrived along that fault line where modernity and tradition collided."

Modernization and Lynching Hypotheses
The modernization thesis of lynching echoes, as discussed above, theoretical perspectives viewing collective experiences and perceptions as key mechanisms in generating reactionary collective mobilization to effect the restoration of traditional communal values and relations (Tilly 1970(Tilly , 1978)).Granting the modernization thesis of lynching as much benefit of the doubt as possible, it may therefore be considered valid insofar as modernizing developments and conditions were associated with public but not necessarily with private lynchings.This is so for two reasons.One, positing that white mob violence against blacks resulted from interracial competition for scarce resources under modernizing developments and conditions does not have clear implications for whether it should have assumed the form of public or private lynchings.Two, positing, like the modernization thesis of lynching, that white mob violence against blacks resulted from collective white experiences and perceptions revolving around the disruption of traditional white communal values and relations by modernizing developments and conditions does have clear implications for whether it should have assumed the form of public or private lynchings.In case white mob violence against blacks represented a form of collective white reactionary mobilization growing out of collective white experiences and perceptions of communal imperilment, modernizing developments and conditions should specifically have been associated with the collective violence of public lynchings because it allowed white southerners to unite around and enact traditional community values and relations in ways that the interpersonal violence of private lynchings could not.
The extent to which white mob violence against blacks arose where modernization and tradition clashed in such a fashion thus suggests a number of hypotheses about the association between lynchings and such modernizing developments and conditions as shifts in the balance from the countryside to town and city, from local economies to national and international markets, and from the primary agricultural to the secondary industrial economic sector, as well as transformations of white family roles.
Urbanization.One straightforward way to conceptualize conditions favoring town and city at the expense of the countryside is in terms of the relative size of the urban population.The modernization thesis of lynching holds that urban circumstances were conducive to white violence against blacks because they, for one thing, increased interracial resource competition and, for another, generated collective experiences of communal disruption and displacement among whites with strong rural roots.As discussed above, while the former conjecture does not clearly imply whether we should expect urban circumstances to be associated with the interpersonal violence of private or the collective violence of public lynchings, the latter clearly implies that such circumstances were associated with public lynchings.This leads to the following hypotheses: • Hypothesis 1a: Localities with larger relative urban populations had higher rates of public lynchings.
• Hypothesis 1b: Localities with larger relative urban populations had higher rates of private lynchings.
Rural white population decline.Another dimension of shifts in the balance from the countryside to town and city is declining rural populations.The modernization thesis of lynching suggests in that regard that rural whites in communities destabilized by declining populations sought to prevent the erosion of traditional communal vales and relations by collective violence against blacks.As the modernization thesis of lynching does not have clear implications for how declining rural populations would affect rates of interpersonal racial violence, I formulate the following hypothesis: • Hypothesis 2: Localities with higher levels of rural white population decline had higher rates of public lynchings.
Commercialization of agriculture.The postbellum commercialization of southern agriculture transformed the majority of rural whites from primarily self-sufficient subsistence farmers to cash crop producers.Such developments not only disrupted traditional reciprocal community exchange networks but also made white farmers dependent on national and international markets for farm products, leading many of them down the path of downward social mobility of landlessness and indebtedness.In view of this, the received wisdom of quantitative sociological lynching scholarship in recent decades has it that whites used lynching as a means to reduce the competition from African Americans for opportunities in the southern agricultural economy.The modernization thesis of lynching goes beyond such an intergroup economic competition perspective to suggest that white communities disaffected by the commercialization of southern agriculture enacted collective violence against blacks in order to restore values and relations of communal autonomy, sovereignty, and solidarity.Accordingly, we should anticipate the following: • Hypothesis 3: Localities with higher levels of agricultural commercialization had higher rates of public lynchings.
Industrialization.Like urbanization, industrialization destabilized rural white communities and uprooted white individuals from such communities as well as threw whites into contact and competition with blacks under new and uncertain conditions.The modernization thesis of lynching argues that racial violence against blacks followed industrialization like it followed urbanization, as whites sought to transplant, on the one hand, traditional principles of communal autonomy, sovereignty, and solidarity on the collective level and, on the other hand, white racial authority on the interpersonal level to modernizing conditions.This leads, like hypotheses 1a and 1b above, to the following predictions: • Hypothesis 4a: Localities with higher levels of industrialization had higher rates of public lynchings.
• Hypothesis 4b: Localities with higher levels of industrialization had higher rates of private lynchings.
Transformation of familial gender roles.One important aspect of modernization is the breaking up of traditional family and gender roles.The modernization thesis of lynching suggests that such developments undermined traditional white masculine ideals and that racial mob violence against blacks served to enact chivalric principles casting white men, as a group as well as individuals, as providers and protectors of white females and children, which leads to the following hypotheses: • Hypothesis 5a: Localities with higher levels of traditional family and gender role dissolution had higher rates of public lynchings.
• Hypothesis 5b: Localities with higher levels of traditional family and gender role dissolution had higher rates of private lynchings.

Lynching Data
In order to evaluate these hypotheses, I draw upon a data set of lynchings in Georgia and Louisiana in the period 1882-1930 that contains 598 lynching events and differentiates among events according to levels of popular participation, support, and ceremony.These data augment the Georgia and Louisiana parts of the Tolnay and Beck (1995) lynching inventory for the period 1882-1930 by including 39 white-on-black lynchings in Georgia discovered or confirmed (as far as lynching inventories go) either by myself or E. M. Beck subsequent to the construction of the original inventory.I have also disconfirmed a couple of events in the Tolnay-Beck inventory that in light of newly discovered information should not be considered lynchings. 1 The necessity of disaggregating different forms of lynchings to properly and fairly evaluate the modernization thesis of lynching guided for reasons of data availability the choice of states.The online, full-text ProQuest Historical Newspapers database offers easy access to contemporary newspaper reports on lynchings in the Atlanta Constitution-the most important and widely read newspaper in Georgia at the time.Pfeifer's Rough Justice: Lynching and American Society 1874-1947 (2004) lists lynchings in Louisiana according to the same lynch mob typology upon which the present coding of Georgia lynchings at root rests. 2  The geographical scope of the study is also justified on substantive grounds.While no states can be seen as representative of the South as a whole, Georgia and Louisiana spanned all the major subregions of the South at the time in terms of demographic, economic, historical, and physiographic characteristics (Ayers 1992).There is thus little reason to suppose that the geographical scope of the study to Georgia and Louisiana drives its results.More than this, Georgia has been seen by many contemporary as well as later observers as the cradle of the ethos and developments of the New South (Ayers 1992).And with regard to Louisiana, Pfeifer (2004) uses a comprehensive inventory of lynchings in that state to argue that there was a link between lynching and modernizing developments in the post-Reconstruction South.The inclusion of Georgia and Louisiana in this study thus makes for a thorough as well as fair evaluation of the modernization thesis of lynching.
In view of the earlier conceptual discussion of different lynch mob types, mobs reported as composed by less than 50 participants, a cutoff point consistent with Brundage's (1993) operationalization, were classified as private lynchings (using 30 participants as the cutoff does not affect the results reported here).Smaller mobs wearing disguises or carrying out their deed without salient ceremony in or nearby the lynch victim's home or in a secluded place (for example, a forest or swamp) were also coded as private lynchings.All mobs composed of more than 50 participants were coded as public lynchings, as were mobs drawing participants from beyond the family and friends of the victim of the (real or imagined) precipitating black crime or offense, indicated by such descriptors of the mob as constituted by the "community," "citizens," or "people" of a locality.In some instances, such mobs formed in order to capture blacks accused of serious crimes against whites and killings of African Americans pursued by such bands of whites were considered public lynchings.The occasional ceremonial elements of racial violence at this time allowed for popular white participation-either directly as perpetrator or vicariously as observer-and events involving conspicuous ritualism or symbolism (for example, torture, mutilation, burning, collective shooting of the victim, or signs left nearby the killing site) were coded as public lynchings.Events in which the lynch victim was brought to and killed at the scene of his alleged crime, brought before the victim of his alleged crime for public identification, or given the opportunity to publicly confess his alleged crime were likewise coded as public lynchings (for further details on the coding of lynching events, see Smångs 2016).
Note that although mob size is an important factor in coding lynchings as private or public, it is not the only or necessarily the most important factor.Mob composition (family and friends of a faulted white party or broader community) and behavior (conspicuous ceremonial elements) are also considered.The fact that contemporary newspapers did not always report mob sizes and that their reports when doing so should be treated as approximations rather than precise estimations does therefore not invalidate the coding procedure.Neither do complications arising from lynching events displaying inconsistent mob type features-for example, mobs larger than 50 participants wearing disguises (coded as public mobs) or mobs with less than 50 participants drawing members from the broader white community (coded as public mobs)-because they are too few to influence the results.
The listing and classification of Georgia lynchings in Brundage (1993) and Louisiana lynchings in Pfeifer (2004) allow for assessing the reliability of the coding procedure.After transforming the information in these two publications into the twofold lynching typology applied here, Cohen's kappa (κ) measure of intercoder reliability was calculated separately for the two states, yielding scores that in terms of conventional standards correspond to nearly perfect intercoder agreement (Landis and Koch 1977).The Kappa scores for Georgia lynchings, calculated using 250 events classified both by Brundage and myself, is κ = 0.84 (p > 0.00) and for Louisiana lynchings, based upon Pfeifer's coding of 146 events that I for the sake of this kind of evaluation also classified based upon newspaper articles in The Times-Picayune (collected through the database Early American Newspapers, Series III), is κ = 0.89 (p > 0.00).

Measuring Modernization
The analyses incorporate five different county-level measures of modernizing processes and conditions.The first of these measures is a county's level of urbanization in terms of the percent of its population living in places of at least 2,500 inhabitants (which follows the U.S. Census definition of urbanized areas).The second measure captures the destabilization of white rural communities in terms of the percentage decline in the absolute white population of a rural county over a five-year period.While the available data contain information on how many people in a county lived in rural and urban areas, they do not contain information on the racial composition of the population in each such area.In other words, if a county contained urban areas, the available data do not allow for calculating how many whites lived in urban and rural areas.The present measure of white rural population decline was therefore calculated only for counties with no urban population because only for such counties do the data allow for determining the total absolute rural white population and whether it declined over time.Note that this measure does not indicate overall population change but specifically captures decline in absolute rural white population and that it assumes higher values the higher that decline over a five-year period.As illustration, consider the Georgia county Habersham.In 1900, it had a white population of 11,812.Five years later, it had a white population (estimated by linear interpolation) of 10,617.5.As Habersham had no population living in urban areas in these years, its percentage decline in absolute rural white population from 1900 to 1905 was 10.11 percent.As further illustration, consider Appling County in Georgia, which had a white population in 1900 of 8,823 and in 1905 of 9,139 and no population living in urban areas.As Appling saw an increase in its white rural population in these five years, it assumes the value zero for the year 1905 on the measure for percentage decline in the absolute rural white over a five-year period.
The third measure captures a county's level of industrialization in terms of total manufacturing wages per capita. 3This indicator of industrialization was chosen over other alternatives such as the number of manufacturing establishments or manufacturing capitalization for two reasons.For one thing, census practices changed over time in ways invalidating for present purposes the number of manufacturing establishments as a measure of industrialization.Until 1900 (inclusive), the census included so-called neighborhood and hand industries, which used little or no power machinery, among manufacturing establishments, whereas it thereafter only included establishments operating on a factory basis.While this had a large effect on the number of establishments counted as manufacturers, it likely did not impact estimates of wages paid in manufacturing or number of wage earners to the same extent.This is because most wage earners in manufacturing were employed in factories rather than neighborhood or hand industries (U.S. Bureau of the Census 1932b).For another thing, manufacturing capitalization data are not available for the entire study period, and the estimates that are available are not reliable (Census of Manufactures 1919).
The fourth measure captures agricultural commercialization in terms of the percent of a county's improved farmland acreage devoted to the dominant southern cash crops: cotton, tobacco, and rice (Redding 1992).Sugar cane was another important cash crop at this time, especially in Louisiana but also in some parts of Georgia, but it is not included in the measure of commercialization of agriculture because its labor system was markedly different from other cash crops, relying almost exclusively on closely supervised black wage laborers worked in gang-style.Although racial repression governed the labor system on sugar plantations, it has been argued that it was not conducive to mob violence against blacks because its centralized wage work organization depended less on physical intimidation for labor force control and was less prone to violent confrontations between white landlords and black farm tenants (Pfeifer 2004).In order to take this bifurcation of commercial agricultural into account, the analyses include, along with the abovementioned measure of percent improved farmland acreage devoted to cash crop production, a measure of percent improved farmland acreage devoted to sugar cane production as a control variable.All measures discussed so far (as well as all the control variables described below) are based upon information in "Historical, Demographic, Economic, and Social Data: The United States, 1790-2000" (Haines 2006) and from the U.S. Bureau of the Census (1913, 1922, 1932a, 1932b).
The fifth measure captures the extent of traditional family and gender role dissolution in terms of the percent of unmarried white women aged 16-30 years in the nonagricultural labor force, and it is based on information from the following samples of the Integrated Public Use Microdata Series (IPUMS) (Ruggles et al. 2015): 1880 (10 percent sample), 1900 (5 percent sample), 1910 (1.4 percent sample with oversamples for, among other groups, African Americans), 1920 (1 percent sample), and 1930 (5 percent sample).This measure was calculated by aggregating individual-level IPUMS information on a person's sex, age, marital status, labor force status, and industry of employment up to the county level of analysis using sample person weights that indicate how many persons were represented by a given person in an IPUMS sample.This procedure yields, for example, an estimate of 380 unmarried white women aged 16-30 in Burke County, Georgia, in 1900 and an estimate that 20 of them were employed in the nonagricultural labor force.The value for Burke County on the measure in question in 1900 was thus 5.26 percent.In order to achieve better model fit, model estimations presented below include cubicroot transformed values of this measure of dissolution of traditional family and gender roles.This transformation was used instead of natural-log transformation because a number of observations take the value zero for this variable and because it provides better model fit than squared-root transformed values.
Values in intercensal years for the independent variables described above as well as the control variables described below were calculated by linear interpolation. 4

Control Variables
In order to avoid spurious results, the analyses include a number of control variables that prior scholarship suggests were related to white mob violence against blacks: linear and quadratic terms of black percentage of the population, absolute black population size (in 1,000s and natural-log transformed), average acreage farm size, and percent of farms owned.As the control variables are not the focus of present analyses, I neither report nor comment on them in the result section but present for interested readers their estimated model coefficients in Table A2 in the online supplement.Finally, in order to control for unobserved heterogeneity between Georgia and Louisiana, the analyses include a state dummy variable, which, as it for present purposes lacks theoretical as well substantive relevance, is left unreported.Table A1 in the online supplement gives descriptive statistics for the independent and control variables in this study.

Statistical Modeling Technique
In order to model the association between different measures of modernization and different types of lynching, I estimate continuous-time Cox event history models using counties as the unit of analysis.Cox models, formally given by Equation ( 1), express the dependent variable in terms of a hazard rate h i (t), interpretable as the "risk" of a given subject i experiencing the event of interest at any time t during the study period, and estimate how it varies with different covariate values.In Equation ( 1), β is accordingly a vector of unknown (but to be estimated) regression parameters and X i is a vector of covariate measurements for the ith subject.Based upon partial likelihood methods, a Cox model makes no assumptions regarding the underlying distributional form of actual time durations until events occur, leaving the so-called baseline hazard function h 0 (t) unspecified.
The specification and estimation of Cox models are restricted by the "proportional hazards" assumption (Box-Steffensmeier and Jones 2004), which requires that covariate coefficients remain constant across time, and it is violated in case the coefficient of a covariate increases or decreases as time passes. 5All models reported below were accordingly tested for proportional hazard violations, and evidence thereof was remedied by estimation of a subsequent expanded model including an interaction term between the offending covariate(s) and the logarithm of time elapsed from the beginning of the study period.The total coefficient of a nonproportional covariate is thus a combination of its time-independent (β 1 ) and time-dependent (β 2 ln[t]) coefficients, and it may vary either or as well as in size and sign from one time point to another.The standard error of the total coefficient is, furthermore, a function of the standard errors of the time-independent and time-dependent coefficients, their covariance, and the logarithm of time.The statistical significance of the total coefficient may in other words also vary over time (for further technical details, see Box-Steffensmeier and Zorn 2001;Golub and Steunenberg 2007).The analyses reported below are based upon two data sets-one including public lynchings and one including private lynchings-with multiple observations per county.An observation contains relevant covariate information for a given county at each date a lynching of the particular type occurred in Georgia or Louisiana from 1882 to 1930 as well as a binary (outcome) variable indicating whether the county in question was the county in which the lynching actually took place.As counties never ceased being at risk of experiencing lynchings, each county has in addition a (right-censored) record for December 31, 1930, serving to make sure that counties with as well as without lynchings of a particular type are properly included in model estimations until the end of the study period.In order to correct for the nonindependence of observations resulting from multiple records per unit of analysis, model estimations use robust standard errors clustered on county, and in order to avoid simultaneity problems between event occurrence and temporal changes in independent variables, they use one-year lagged rather than current-year values of continuous covariates.The models, furthermore, deal with the matter of units of analysis experiencing multiple events by stratifying estimations by event rank-order (that is, first event, second event, and so on) while still producing a single set of cross-strata parameter coefficients.Without taking into account the actual order of multiple events within units of analysis, model estimations proceed as if a county was at risk for all lynchings that it experienced during the span of the investigation at all times.That is, a model that ignores the actual order of events is estimated as if later events could have happened before earlier events, whereas estimation by rank-order stratification preserves the actual event sequence (Box-Steffensmeier and Jones 2004;Box-Steffensmeier and Zorn 2002).
Most counties enter the data at the beginning of the study period (i.e., January 1, 1882), but some counties were created at a later point, and they do not enter into model estimations until their creation.Covariate values of new counties were calculated as follows.First, as model estimations use one-year lagged covariate values, a county created after the beginning of the study period was assigned the mean value of each covariate-with the exception of absolute black population size as described below-across its counties of origin in the year before creation, and values in years between the year before creation and the next census were then approximated by linear interpolation.The absolute size of the black population in newly created counties was calculated in a slightly different manner.Instead of assigning a new county a black population size equal to the mean absolute black population size in its counties of origin, it was assigned an absolute black population size equaling the total black population in its counties of origin divided by the number of counties of origin plus one, and values in years between the year before creation and the next census were then approximated by linear interpolation.The addition of one in the denominator is necessary because the total black population in the counties involved in the creation of the new county would otherwise be larger after than before the creation of the new county.

Results
Table 1 reports the results of two Cox event history models evaluating the modernization thesis of lynching.As specified in hypotheses 1a and 1b, the modernization thesis of lynching predicts that urbanization was positively associated with racial mob violence against blacks in the form of public as well as private lynchings.Neither model 1, which predicts public lynchings, nor model 2, which predicts private lynchings, support these hypotheses.In fact, the results directly contradict the modernization thesis of lynching on this point, as the coefficient for the urbanization measure is statistically significantly negative in both models.
The urbanization covariate coefficient in model 1 is, however, not stable over time, as the model includes a term of its interaction with the logarithm of time elapsed since the beginning of the period of investigation.That the coefficient for this interaction term is positive indicates that the association between urbanization and public lynchings increased as time passed.As explained above, such time-dependent covariate coefficients may temporally differ in size, sign, and statistical significance and must therefore be evaluated at different points in time.To that end, Table 2 shows the temporal development of the covariate coefficient of urbanization in model 1 predicting public lynchings at various time points in the period 1882-1930.As we can see in Table 2, although the coefficient of urbanization increases over time, it remains negative and statistically significant until the end of the nineteenth century and does not attain substantive or statistical significant at any point during the period of investigation.We can therefore conclude that hypotheses 1a and 1b are not only unsupported but partly contradicted by the empirical findings.
But what do these results tell us substantively?How much less likely were public and private lynchings, respectively, to occur in more urbanized areas?In general terms, a k-unit difference in a covariate in an event history Cox model implies a 100[(exp(β*k)−1)] percent higher or lower risk, depending on whether the coefficient estimate (β) for the covariate is positive or negative, of a subject experiencing the event in question.In the present case, the results of model 2 suggest that a county with a one standard deviation (16.81 in model 2) higher value on the urbanization measure relative another county was about 19 percent less likely to experience a private lynching.Unlike urbanization's association with private lynchings, the results of model 1 indicate that its association with public lynchings was not stable over time.Thus, taking this time-dependent character, displayed in Table 2, into account, we can see that a public lynching was 64 percent less likely to occur in 1882 in a county with a one standard deviation (16.97 in model 1) higher value on the urbanization measure relative another county.The corresponding figures were 43 percent in 1885, 29 percent in 1890, 21 percent in 1895, and 9 percent in 1900, whereafter the estimated coefficients for urbanization in model 1 become too imprecise to be informative.
Hypotheses 2, predicting that that white population decline in rural areas was conducive to public lynchings, as whites used such collective violence to contest the erosion of their communal values and relations, fares a fate similar to hypotheses 1a and 1b.Again, the results presented in model 1 contradict the prediction of the hypothesis in that white rural population decline shows a negative and statistically significant association with public lynchings.Although the modernization thesis of lynching does not have clear implications for the association between white population decline in rural areas and private lynchings, and no explicit hypothesis was therefore stated above in that regard, it is worth noting that the results of model 2 suggests that white rural population decline likewise had a negative and statistically significant association with private lynchings.In substantive terms, a public lynching was about 21 percent and a private lynching about 27 percent less likely to take place in a county with a one standard deviation (2.68 in model 1 and 2.52 in model 2) higher value on the measure of rural white population decline compared to another county.
The third hypothesis predicts that the commercialization of southern agriculture led whites disaffected by their dependence on and their communities' disruption by national and international market for farm products as well as grinding crop-lien debt to enact and restore values and relations of communal autonomy, sovereignty, and solidarity by means of the collective violence of public lynchings.Yet again, the results fail to bear out the hypothesized prediction, as model 1 yielded a small, though positive, but highly statistically insignificant coefficient for the measure of the percent of a county's improved farmland devoted to commercial agriculture.In consequence, the estimated coefficient for the association between agricultural commercialization and public lynching in model 1 is too imprecise to be substantively informative.
The interpretation of the association between manufacturing wage per capita and public lynchings in model 1 is complicated by the fact that it is nonlinear as well as nonproportional.Let me therefore first address the association between manufacturing wage per capita and private lynchings in model 2 (as preliminary exploratory analyses showed no indication of nonlinearity or nonproportionality for manufacturing wage per capita; the model includes it only as a linear predictor).Simply and shortly, the results of model 2 do not lend any support for hypothesis 4b, which stated that higher levels of industrialization were positively associated with higher levels of private lynchings; the coefficient for manufacturing wage per capita is negative, though quite small, and far from conventional levels of statistical significance.
Moving to model 1, the results appear at first glance to straightforwardly support the prediction of hypothesis 4a that localities with higher levels of industrialization had higher levels of public lynchings, as the coefficient for manufacturing wage per capita is positive and statistically significant, though nonlinear.But these results are complicated by the fact that both the linear and quadratic terms of manufacturing wage per capita demonstrate time invariance, indicated by the inclusion of terms for their interaction with the logarithm of elapsed time since the onset of the study period.In general, the size, sign, and statistical significance of linear and quadratic coefficients cannot be interpreted separately as unconditional covariate coefficients but must be evaluated jointly by the calculation of "marginal effect" coefficients at relevant values of the covariate in question (Brambor, Clark, and Golder 2006).In order to properly evaluate the association between industrialization and public lynchings, we must thus calculate its marginal effects at different values of manufacturing wage per capita at different time points from 1882 to 1930.
Figure 1 graphically displays the marginal coefficient effect of manufacturing wage per capita at low (the 25th percentile of its distribution), intermediate (its median), and high (the 75th percentile of its distribution) levels in selected years during the period of investigation.Figure 1 shows, generally speaking, that the marginal coefficient effect of industrialization for public lynchings was positive and statistically significant at all levels of industrialization until the early twentieth century, whereafter it grew exceedingly small as well as statistically insignificant.More specifically, it shows that until the early twentieth century, the positive marginal effect coefficient of manufacturing wage per capita was strongest and most statistically significant at lower and intermediate levels of manufacturing wage per capita.The results support insofar the conjecture of the modernization thesis of lynching specified in hypothesis 4a that whites came together in the collective violence of public lynchings in contexts where industrialization emerged and disrupted the values and relations of their traditional communal way of life.Like manufacturing wage per capita, the (cubic root transformed) percent of unmarried white women aged 16-30 years in the nonagricultural labor force, measuring the dissolution of traditional family and gender roles, has a time-invariant covariate coefficient in the model predicting public lynchings but not in the model predicting private lynchings.I therefore again consider model 2 before model 1.While the coefficient for this measure is in the expected direction in model 2, its standard error is of such a magnitude as to render it far from conventional levels of statistical significance.In other words, because of the variability and imprecision of the coefficient estimate for the traditional family and gender role dissolution measure, the results fail to support hypothesis 5b.
Model 1 shows that the dissolution of traditional family and gender roles measured by the (cubic root transformed) percent unmarried white women aged 16-30 years in the nonagricultural labor force was positively associated with public lynchings.This association, however, grew weaker as time passed, as indicated by the negative coefficient for the interaction between the covariate and the natural log of time.Returning to Table 2, we see results partly supporting hypothesis 5a in that the association between the participation of unmarried young white women in the nonagricultural labor force and public lynchings was positive and statistically significant in the last couple of decades of the nineteenth century but that it thereafter became too small and too imprecise to permit clear conclusions.
Because of its nonlinearity, the interpretation of the association between traditional familial gender role dissolution and public lynching from the early 1880s to the late 1890s is facilitated by graphic display.Figure 2 shows how the likelihood of a public lynching, expressed in terms of the hazard ratio, differed at different levels of participation in the nonagricultural labor force by unmarried white women aged 16-30 in 1882, 1885, 1890, and 1895.Figure 2 shows, for one thing, that the likelihood of public lynchings, especially in the early 1880s, rose sharply with higher levels of traditional family and gender role dissolution, particularly in places where the value of this variable was low (that is, in the 0 to 20 percent range).The figure shows, for another thing, that the bearing of traditional family and gender role dissolution on the likelihood of public lynchings was much weaker in the mid-1890s than in the early 1880s.For example, a public lynching was in 1882 almost 40 times more likely to occur in a county in which 80 percent of unmarried white women aged 16-30 years participated in the nonagricultural labor force compared to a county in which no such women did so.In contrast, in 1895, a public lynching was about twice as likely to take place in the former compared to the latter type of county.

Concluding Discussion
The purpose of this article has been to evaluate arguments put forth in a number of historical studies that the lynching of African American by white mobs at the turn of the nineteenth century resulted from tensions and disjunctures between a declining agrarian society and an emerging modernizing society.The results of empirical analyses using systematic statistical methods and detailed lynching data in two southern states during nearly half a century both contradict and support such a modernization thesis of lynching.With regard to white rural populations' decline and urbanization, the results show, for one thing, that public lynchings were negatively associated with the developments toward the former during the entire study period and toward the latter development from the early 1880s to the late 1890s, whereas private lynchings were negatively associated with both types of developments throughout the span of the study.For another thing, the results yield no evidence of an association between agricultural commercialization and either lynching type.For a third thing, public lynchings were positively associated with industrializing developments from the early 1880s to the early 1900s and with developments disrupting traditional familial gender roles from the early 1880s to the late 1890s, whereas private lynchings do not appear to have been clearly associated with such developments.
This attempt at interdisciplinary cross-fertilization does thus bear mixed fruit in that while providing systematic evidence confirming some arguments proposing a link between modernization processes and lynchings, it also provides evidence directly contradicting or otherwise failing to support such a modernization thesis of lynching.The lesson implied in these results is twofold.The first lesson is methodological, speaking particularly to the liability of relying upon case studies of particular lynchings or comprehensive lynching inventories but using empirical examples as illustrative pointers rather than as systematic evidence subjected to rigorous methods of analysis.It seems that these methodologies led some proponents of the modernization hypothesis of lynching to the fallacy of "working backward."That is to say, "The analyst locates an effect that calls for explanation, then searches in the immediate or more distant past for evidence of the posited explainers" (Rule 1988:188).For example, it may very well be the case that "The most spectacular lynchings took place not in the countryside but in . . .newly urbanizing places" (Wood 2009:5;cf. Pfeifer 2004), but that does not constitute systematic evidence allowing for generalizations about conditions conducive to lynching.
The second lesson is substantive in highlighting that the patterns and consequences of the modernizing transformations of the South coinciding with the lynching era were not uniform across the region and that racial conflicts in some modernizing environments assumed different forms than in other such environments.In consequence, broad statements that lynching followed the "fault lines" of modernization (Wood 2009) and that the "parallels" between lynching and modernization cannot be ignored (Goldsby 2006) appear, in light of the results presented here, overstated.Let me stress that I do not mean to suggest that the results presented here imply that modernizing developments and conditions that showed a negative or no discernable association with lynchings in any form was free of racial tension or repression.That present results bore out a negative association between urbanization and public as well as private lynchings can, for example, be explained by the fact that violent racial conflict in towns and cities was handled by white policemen rather than bands of white citizens (Ayers 1992;Rabinowitz 1994).
While extended discussion and analysis of spatial diffusion processes are beyond the scope of this article, let me further discuss the point that the consequences of modernizing transformations were not uniform throughout the South with regard to racial conflict and violence and briefly address some implications thereof.Tolnay, Deane, and Beck (1996) have demonstrated a negative spatial effect of lynching in the South, in the sense that lynchings in one county decreased the lynching rate in geographically proximate counties.They (Tolnay et al. 1996:811) explain this finding by arguing that whites may have enjoyed lynchings vicariously; that is to say that "[W]hites in counties that did not lynch or lynched less frequently participated in vicarious violence through the activities of white mobs in other [especially nearby] areas."The negative association between urbanization as well as white rural population decline and especially public lynching may in part reflect such negative spatial diffusion processes.This is so because it is not unlikely that whites in localities undergoing such demographic transitions may have lacked the social cohesion necessary for staging public lynchings and therefore vicariously enjoyed such lynchings in nearby places to such an extent as to further reduce the likelihood of their undertaking such collective violence themselves.But I have to leave it to future research to systematically evaluate that conjecture.
In closing, then, in addressing crucial questions raised by historical lynching research, this article sheds new light on the role of the emblematic form of racial violence in the United States in developments bringing forth the society we still live in.In doing so, it also highlights the necessity of methodological rigor as well as of further research in order to advance and accumulate knowledge in this interdisciplinary field of scholarship.

Figure 1 :
Figure 1: Marginal effect coefficient of manufacturing wages per capita at low (its 25th percentile), intermediate (its median), and high (its 75th percentile) levels in different years from 1882 to 1930.

Figure 2 :
Figure 2: Estimated hazard ratios at different levels of family and gender role dissolution, 1882-1895.

Table 1 :
Event history model estimations of the association between modernization and different lynching types.