Explaining Lawless Violence

A Review Essay of Stewart E. Tolnay and E.M. Beck, A Festival of Violence: An Analysis of
Southern Lynchings, 1882-1930 (forthcoming, Historical Methods)

When, more than 30 years ago, I was writing my second graduate research paper, I was
strongly advised by the professor in the course, John Morton Blum, to give up my effort to weigh
the factors that I hypothesized might have caused the phenomenon that I was trying to explain.
Just list all the causes for which there was any credible evidence, I was told; don’t try even to
rank them and certainly don’t waste your time attempting to reject any. It’s not the historian’s
job, and it’s probably not possible, anyway. Tell a good story, with interesting characters and
active verbs. If you must, explain, but above all, entertain — that was the Blumian credo. I
largely ignored the adjuration, reinforcing the then-department chairman’s view of me as a rebel
with too few causes.
More recently, things have gotten even worse in history. Thus, the March, 1997 Journal
of American History spotlighted a shockingly incomplete, grossly exaggerated, and relentlessly
personalized string of musings by Joel Williamson on lynching in the post-Reconstruction South,
along with 6 referee reports on the paper. Citing none of the sociological studies of lynching
that have proliferated in recent years, but nonetheless dismissive of any social scientific
approach, Williamson’s article proclaimed that “Anytime, anyplace, white people in large
crowds might suddenly fall into a frenzy and lynch a black person . . .” Since on this view,
lynching was utterly unpredictable, Williamson made no effort at systematic examination.
Instead of clearly stating and forthrightly testing a set of explanations, he merely asserted an
interpretation whose core he borrowed from Jacquelyn Dowd Hall: “Lynching was done by all
classes of whites; it was done as a public ritual; and it was a tool to control not only blacks but
whites as well, and especially white women and, most especially, white women in relation to
sexual matters.” Historians, he went on, should cease “categorizing, measuring meticulously,
and pondering the oppressions that we all know about . . . it is, perhaps, time to think more about

1
the moral wars within.” Piling speculation on impulse, Williamson suggested that “the real war,
the essence of the conflict, concerned gender, not race, and . . . lynching and even
disfranchisement, segregation, and proscription had more to do with relations between white
men and white women than with relations between blacks and whites . . .” (Williamson 1997,
1228, 1247, 1251, 1253; Hall 1979, 153-57) Henceforth, if Williamson’s line were followed,
American history would revert to the old all-white story, blacks not even deserving a role in their
own oppression — insignificant bystander-victims in the only important struggle, the white
gender war.1

With the discipline of history increasingly thrashing in such reactionary, unproductive

spasms, those who seek logical thinking and the systematic analysis of evidence must turn more
and more to historical social scientists. Thus, it was no historian who refined, deepened, and
extended the existing data sets on lynching, but two sociologists, Tolnay and Beck.2 Instead of
contemplating their inner feelings to discover truths about lynching, they gathered evidence,
crafted explicit, falsifiable hypotheses, and tested them statistically. While there may be no
productive response to such self-indulgent pieces as Williamson’s, but only anger, frustration,
and condemnation (or, depending on one’s taste, joy, appreciation, and celebration),3 Festival of
Violence conclusively undermines some theories of lynching and strongly supports others.
Although their analysis of the connection of lynching with politics and law enforcement seems
to me inadequate, and although they might have mined their data more fully, Tolnay and Beck’s
solid advances and laudable example have brought us closer than we have ever been before to
the truth (to use a word currently scorned by many historians) about lynching.
Take, for example, Williamson’s notion that the reason white men lynched African-
Americans was to demonstrate to white women that they must defer to white men if they wanted
to be protected from black beast/rapists, as well as that any white woman who breached the
sexual color line would only be condemning her black lover to torture and death. Eschewing

2
speculation, Tolnay and Beck test the view that lynching was inextricably bound to rape or other
sexual behavior by simply counting the number of times that various reasons were publicly
offered as justifications for specific lynchings. The mind-numbing amount of combing through
sources that no doubt went into the chief summary table was well worth the effort (48).
Accusations of sexual assault or other violation of sexual norms were made in only a third of the
lynchings of blacks by whites in the South from 1882 to 1930. By contrast, murder was cited in
37 percent of these lynchings, and other non-sexual assaults, in another 10 percent . In fact,
allegations of sexual assault were involved in a larger percentage of the 148 cases in the Deep
South in which blacks were lynched by black or integrated mobs than in the 2314 instances
where blacks were lynched by white mobs. (97) If Williamson had looked at the available data
in Tolnay and Beck’s book or any previous systematic account of lynching, instead of putting a
trendy hyper-feminist twist on the propaganda of the turn of the century racist radicals, he could
never conscientiously have offered his conjecture as a possible hypothesis. This data by itself
shows conclusively that lynching was primarily about black-white relations, not white gender
relations.
Nor was lynching a substitute for more formal trials and punishment, for when other
important control variables are introduced into time-series and cross-sectional Poisson
regressions, there is no statistically significant relationship between legal executions and
lynchings of blacks. Lynching was not rough popular justice for increasingly unruly blacks
unshackled from the social controls of slavery, as late nineteenth century southern whites often
charged. Such a theory, which Williamson sometimes appears to give some credit to, is
disproven by Tolnay and Beck’s sophisticated statistical analysis. (104-11)
Before he put forth his gender war interpretation of lynching, Williamson, with equally
little evidence, proposed an economic/psychological/scapegoat theory. White small farmers,
threatened by the depression of the 1890s with loss of control of their property and therefore
control of their lives and families, punished the symbol of uncontrolled passion, the black rapist,

3
as a means of fictively slaying the demon within themselves. (Williamson 1984, 305-23) In this
Williamson view, lynching served as a means of reassuring economically shaken white men that
they were still in charge at home and in society, rather than as a method for disciplining white
women directly.4
Tolnay and Beck’s data support the economic underpinning, at least, of Williamson’s
loose theory. Although several previous scholars asserted that there was a negative correlation
between cotton prices and lynching, no one until now has run a sufficiently sophisticated
statistical procedure to guarantee that the apparent correlation was real, and not the product of
other variables. When cotton prices were low, there were more lynchings, especially before
1906, and the cotton belt was prime lynching territory, even controlling for the fact that there
were larger numbers of African-Americans there. Lynching was also seasonal, peaking in the
summer, when the demand for farm labor was especially heavy, and dropping markedly in the
winter, when fields lay fallow.
Tolnay and Beck’s gloss on these facts is socioeconomic, not psychosexual. What they
describe as the legacy of the slave plantation -- white racism and a one-crop agriculture
dependent on black labor -- weighed most heavily in cotton regions. 5 Racial conflict over labor
control became fiercer when profit margins shrank, and the pervasive racism of the Black Belt
enabled whites to use violence as a means of social control, with even less restraint than in more
temperate regions. (157-60)
Is there a way to choose between these two glosses on the cotton price/lynching
correlation? Although Tolnay and Beck do not directly address this question, their data
potentially allows one. For the purposes of their larger analysis, the sociologists consider a
lynching a lynching. That is, they do not subdivide lynchings into those purportedly the result of
rape, those supposed to have been responses to murder, etc. But their data set is rich enough to
allow a test between the psychosexual and social control hypotheses. If labor discipline was the
predominant motive for lynchings, and such discipline could be conveyed by a lynching, no
matter what the alleged offense, then one would expect to see similar temporal patterns of

4
lynching for sexual and non-sexual “offenses.” For instance, the proportion of lynchings for
rape to lynchings for murder should have been roughly the same in the depression of the 1890s
as throughout the period from 1882 to 1930. The seasonal pattern, however, should have been
different. If the labor control theory is correct, then lynchings for trivial offenses should have
been concentrated in the summer, for such “reasons” as “injuring livestock” or “unruly remarks”
might have been flimsy excuses for providing an example to keep African-American laborers in
line when they were most needed on the plantation. Williamson’s psychosexual thesis, on the
other hand, would predict a concentration of lynchings for rape during the agricultural
depression, and a random pattern of reasons over the course of each year. Without such data, the
choice between such hypotheses would be purely a matter of taste.
Like most historians today, Williamson scorns politics; Tolnay and Beck just disconnect
politics from policy. Thus, their chapter on politics and lynching suffers less from small errors
of fact6 than from a fundamental failure to consider how politics may have affected law
enforcement.7 It is quite puzzling that the authors of a book about spatial and temporal
variations in lawless violence never seriously consider that officials in some areas and at some
times might have been more willing to suppress lynching than police at other times and places.8
In particular, most Republican and Populist officials, whose election was usually at least partly
dependent on African-American votes, no doubt tried harder to protect blacks from lynching and
overtly political violence than Democratic officials in areas of similar social composition did.
Moreover, blacks were less of a threat to whites after disfranchisement (which also virtually
eliminated political parties other than the Democrats), but lacking the franchise, they were also
more vulnerable, because sheriffs then had few or no black votes to lose if they allowed lynchers
free rein.
These considerations undermine Tolnay and Beck’s dismissal of the “political threat”
hypothesis. Blacks were not only threats before disfranchisement; they were also possible allies
of white politicians. When Republicans, Independents, or Populists posed a major danger in a
county, Democrats typically responded by race-baiting, intimidating voters or opposition

5
politicians, and using fraud or violence. Opposition failure was often accompanied or succeeded
by anti-black violence. On the other hand, when the opposition succeeded in electing local
officials, and those officials managed to stay in office, blacks were presumably safer. Thus,
instead of testing linear models relating political competition to lynching, Tolnay and Beck
should have tested well-thought-out models containing thresholds, which would predict that
lynching would increase as party competition heated up, and increase further if, after a close
election, the Democrats won, but that it would decrease after an opposition victory.9
Furthermore, while the overall expected correlation between disfranchisement and lynching is
unclear, the removal of African-Americans as a political force ought to have increased the
correlation between socioeconomic determinants and lynching. Before disfranchisement,
politics was perhaps an intervening variable in a developmental sequence; afterwards, there was
no political prism between social forces and lynching, potentially diffracting those forces in
complicated ways. Such considerations suggest testing a model containing both political and
socioeconomic factors, properly specified, for the period before disfranchisement in each state,
and comparing the coefficients with those for the period after disfranchisement. If the
hypothesis is correct, then the relationships between social factors and lynching should be
stronger after disfranchisement than before.
Law enforcement concerns also underline a fundamental difficulty in any data set that
records events only if they occurred: we do not know the probability of the event unless it was
100 percent. In particular, Tolnay and Beck do not examine potential lynchings that were
prevented by law enforcement officials. Although such information would be particularly
difficult to obtain, there is at least one time series, unfortunately beginning only in 1914, that
contains evidence of lynchings prevented, which passed the number of persons lynched in 1920
and continued to exceed it thereafter.10 (See Hall 1979, 235) If it were possible to extend this
data set back into the period before disfranchisement, it would allow a particularly good test of
what might be called the “political control” hypothesis. The thesis would predict that
Republican or Populist or Independent sheriffs would be much more likely than Democratic

6
sheriffs in areas of similar social composition to quash possible lynchings. Another possible
test of the political control thesis would be whether lynchings for non-capital offenses
(everything except rape and murder during that period) were concentrated in areas controlled by
Democrats and in the period after disfranchisement, the thought being that opposition politicians
dependent on black votes would be particularly likely to suppress violence against blacks for
relatively trivial “offenses.” It is instructive that four mob murders of African-Americans that
Tolnay and Beck (21) describe in detail and which were in retribution for “insolence,” arguing,
using sarcasm, and failing to pay 10 cents in interest on a debt, all took place after
disfranchisement or in the midst of a disfranchisement campaign.
It is possible, of course, that the political threat and political control theses will not be
borne out by the data in properly specified models. In fact, post-1880 lynchings were not
generally attributed to political causes. Any connections during that period between politics and
lynching, like the ones between economics and lynching that Tolnay and Beck prefer, were only
indirect. Thus, the political threat hypothesis suggests that racial strife in politics created a
climate in which anti-black violence was more likely, or that politics was the latent, but not the
immediate cause of a symbolic, cautionary murder, while the political control hypothesis states
that the relationship between law enforcement officers and the black electorate encouraged or
discouraged lynching. In neither case is politics considered the possible direct cause of
lynching.
It is far different with the much greater violence of Reconstruction, which was overtly
and unquestionably the result of politics. In a six-month period in 1868 in Louisiana, Democrats
slaughtered nearly half as many black and white Republicans as were lynched in the whole
South in the fifty years after 1882. During and shortly after the 1874 election campaign in
Vicksburg, Mississippi, Democrats killed three times as many African-Americans as in the
South during the 1920s. Over Easter weekend, 1873, Democrats massacred more blacks in the
tiny hamlet of Colfax, Louisiana, than were lynched in the South in any year after 1881.
(Trelease 1971, 135; Foner 1988, 558; Tunnell 1984, 192; Tolnay and Beck 1995, 271-72) Why

7
is such violence less well-known to the public and less studied by scholars than the lynching
violence?11
I suggest that there are four reasons: First, twentieth century Americans don’t like to
believe that partisan politics and racism are intertwined, and nineteenth century southern white
Democrats had a strong interest in downplaying that connection, for fear of strengthening
sentiment for federal intervention. Thus, to the extent that Reconstruction violence is
mentioned, it is attributed to the Ku Klux Klan, the Redshirts, the Knights of the White Camelia,
or similar groups, which were, in fact, only the Democratic party in costume. Second,
Reconstruction violence is no puzzle, and the sources are not so good as those about lynching. It
is not difficult to discern the motives and sometimes, the identities, of the Reconstruction
violators, but it is often impossible to discover how many people they killed. As a scholarly
topic, Reconstruction violence is thus both too easy and too hard. Third, violence probably
seemed less shocking to the nation during Reconstruction than it did later because the mass
carnage of the Civil War was so close. Every northern and southern town had lost sons, often
many of them. A few hundred obscure murders of unknown men in places far away, not well
reported, may not have seemed particularly horrific, especially since Reconstruction murders
were generally businesslike, private, and quick, not the huge public spectacles, the dramatized
tortures of some later lynchings.12 Fourth, sex sells. Reconstruction violence was nearly
asexual, and while post-Reconstruction lynchings had much less to do with sexual matters than
contemporary apologists for lynching claimed or than Williamson and other historians like to
imagine, allegations of rape or other sexual impropriety were involved in several hundred
lynchings. And America’s fascination with sex is as well known as her puritanism. Thus, post-
1880 lynching may not have been very political – we don’t really know yet – but in sheer
numbers, as well as in public significance, 19th century violence against African-Americans was
predominantly due to political causes.
Lynching began to decline after 1900 and did so especially precipitously after 1920.13
Why? Like the muckraking historians and social scientists of the early twentieth century who

8
derided the role of ideas and institutions, finding crude self-interest behind every action or trend,
but unlike the“new institutionalist” sociologists (Evans, Rueschmeyer, and Skocpol 1984),
Tolnay and Beck dismiss anti-lynching reformers’ agitation and the passage of laws
guaranteeing protection for prisoners as epiphenomena, and increased black resistance as
ineffective. They do not test such explanations statistically. Instead, they settle on the response
to black migration as the only important cause of the fraying of the rope. According to this
view, white elites, fearful that more of their African-American labor forces would join the Great
Migration northward, and believing that lynching drove some to leave, frowned on the practice,
and it slowed markedly in the 20s. How the white elite effected its new goal, the sociologists do
not say, commenting only that they were “somehow able to reduce the number of black lynch
victims.” (230) As support for their migration hypothesis, Tolnay and Beck cite little direct
qualitative evidence, relying on statements by NAACP Executive Secretary and lynching
detective Walter White. Principally, they offer two cross-sectional regressions, one for the
1910s and one for the 1920s.
These regressions necessarily exclude explanations that vary across time, but not from
county to county, or at least those whose variation from area to area is difficult to measure.
Thus, county-level regressions cannot test whether southern white officials in the 1920s quelled
lynching in order to fend off the very real threat of federal anti-lynching legislation; or whether
statewide laws, forceful law and order governors, or efforts by such groups as the NAACP or the
Commission on Interracial Cooperation helped to bank the lynch-fires; or whether an
increasingly “modern,” bureaucratized southern governmental system gradually suppressed
irregular, extra-legal violence. While it is possible that notable state-level efforts, such as those
by Virginia and Kentucky governors in the 1890s, could be represented by “dummy” variables
with the same value for every county in a state, or that urban percentages or variations in county
expenditures might be taken as proxies for “modern” law enforcement attitudes, such solutions
are not altogether satisfactory. The point here is that particular methods may rule out certain
classes of explanations by fiat.

9
 Tolnay and Beck’s operationalization of the county-level regressions on the decline of
lynching is also questionable. Although they speak of explaining “reductions in the level of
racial violence” from one period to the next (224), in fact they only measure variations from
county to county in the levels of lynching in the 1910s and 1920s. Thus, they must assume that
the propensity for violence was the same in every county, if one controls for the variables that
they included in their equations. But if they had used changes in lynching from earlier decades
as the dependent variables of interest, they would have automatically controlled for every
county-level influence on the lynching rate except those, such as migration, that changed from
decade to decade. This would have allowed a cleaner test of their hypothesis, for it seems more
logical to predict the rate of change in lynching in each county from the rate of change in the
county’s black population, rather than the level of lynching in each county from its rate of
change in population. If Tolnay and Beck had acted here a bit more like historians, who are
particularly attuned to explaining change over time, and a bit less like social scientists, who tend
to seek timeless general rules, then their analysis of this matter might have been more
convincing.

Arguments with or between such cultural historians as Joel Williamson are often futile,
because such historians do not seem to prize clarity, to believe in fairly evaluating all
explanations, or to risk their claims against the evidence. By contrast, one can argue with
historical social scientists such as Tolnay and Beck over exactly how a model should be
specified or operationalized, and how data should be interpreted. Misunderstandings can be
eliminated, improvements suggested, disagreements narrowed or terminated, and the number of
causes limited. Increasingly, it is up to painstaking, thoughtful, methodologically sophisticated
social scientists like Stewart Tolnay and E.M. Beck to save history from its disciplinary vagaries.

10
 References

Brundage, W.F. 1993. Lynching in the New South: Georgia and Virginia, 1880-1930. Urbana
and Chicago: University of Illinois Press.

Dollard, J. 1937. Caste and Class in a Southern Town. New Haven, Connecticut: Yale
University Press.

Evans, P.B., D. Rueschemeyer, and T. Scocpol, eds. 1984. Bringing the State Back In.
Cambridge, Eng.: Cambridge University Press.

Foner, E. 1988. Reconstruction: America’s Unfinished Revolution, 1863-1877. New York:

Harper and Row.

Hall, J.D. 1979. Revolt Against Chivalry: Jessie Daniel Ames and the Women’s Campaign
Against Lynching. New York: Columbia University Press.

Hovland, C.I., and R.R. Sears. 1940. “Minor Studies of Aggression: Correlations of Economic
Indices with Lynchings.” Journal of Psychology: 301-10.

Trelease, A.W. 1971. White Terror: The Ku Klux Klan Conspiracy and Southern
Reconstruction. New York: Harper and Row.

Tunnell, T. 1984. Crucible of Reconstruction: War, Radicalism and Race in Louisiana, 1862-

11
1977. Baton Rouge and London: Louisiana State University Press.

Williamson, J. 1984. The Crucible of Race: Black-White Relations in the American South Since
Emancipation. New York: Oxford University Press.

- - - - - - . 1997. “Wounds Not Scars: Lynching, the National Conscience, and the American
Historian.” Journal of American History. 83: 1221-53.

Wright, G.C. 1990. Racial Violence in Kentucky, 1865-1940: Lynchings, Mob Rule, and “Legal
Lynchings”. Baton Rouge and London: Louisiana State University Press.

Endnotes

1. Hall actually treats lynching as primarily a racial phenomenon, she shows a careful
appreciation for social scientific studies of it, and she stoutly rejects as myth the view that
lynching was always concerned with actual rapes of white women by black men. (Hall 1979,
129-57)

2.Even the most careful empirical work on lynching by historians (Wright 1990; Brundage 1993)
lacks the rigorous hypothesis testing that sociologists, more statistically competent, can perform.

3.It is notable that, to be generous, only two of the 21 published letters to the editor about
Williamson’s article contained any criticism of the substance of his piece, and none mentioned
any social scientific works on the subject. See Journal of American History, 84 (1997), 748-65.
.

4.In fact, although he nowhere acknowledged it, Williamson was here adopting something very
close to Hovland, Sears, and Dollard’s “frustration-aggression” theory (Hovland and Sears 1940;
Dollard 1937, 315-63). Thus, whether he knew it or not, Williamson was a captive of now-
discarded social scientific theories.
Williamson also joined this economic with a more purely psychoanalytical view. Victorian
southern white men, he hypothesized, were guilt-stricken about having sex with white women,

12
whom they believed should never be lowered from their pedestals. African-Americans, white
men thought, suffered from no such prudish misgivings; in particular, black men felt no
inhibitions about having sex with white women. In a mixture of envy and a desire to punish
aberations from “normal” moral feelings, therefore, white men were driven to lynch black men.
Such a theory, if testable at all, requires different documents than Tolnay and Beck examine.

5.They might also have noted the legacy of one-party politics, aristocratic extremism, and a
particularly pronounced socioeconomic gap between the races.

6.For instance, Henry Cabot Lodge was a member of the House, not the Senate, when he
introduced the “Force Bill” in 1889; that bill affected registration and fraud, not voting per se;
and the bill’s timing had little direct effect on the Populists, who held their founding convention
after its defeat. (178)

7.This is not merely a disciplinary bias, for some of the first sociologists who studied lynching
emphasized the importance of inhibitory legal institutions on its level. (Brundage 1993, 9)

8.In fact, they often simply assume that law enforcement officials were virtually unanimous in
being willing to allow lynching. (207)

9.They remark (193-94, 197) that they tested other models, perhaps along this line, but they do
not describe them and do not give any results. A proper test of such models would require data
on which party controlled the sheriff or other law enforcement officer in each area, which may
not be adequately proxied by the returns from gubernatorial or presidential elections. For
instance, Populists and Republicans often fused on local tickets, with representatives of each
party taking some places on a common electoral ticket. In a few cases, notably in Mississippi
from 1875 to 1890, Republicans and Democrats fused on some local tickets.

10.They do briefly discuss the data, which they believe, no doubt correctly, is of “questionable
accuracy.” (203) But is it worse than assuming that the probability of an unsuccessful attempt at
lynching was the same everywhere?

11.Wright (1990, 1-10) emphasizes political reasons for lynching in Kentucky, where the rate of
killings of blacks by whites from 1865 to 1880 exceeded that in the last two decades of the
century, as it probably did throughout the South.

12.The widespread publicity and outrage in the North over such urban “race riots” as those in
New Orleans and Memphis in 1866, which were reported almost like Civil War battles, indicates
the importance of reporting in determining the attention given by the public, as well as by later
historians, to violent acts.

13
13.The lynching trends in the 1920s and 30s, economically depressed in the rural South, severely
undermine Williamson’s psychosexual thesis. Why would southern white males be so much
more secure, if lynching was a sign of insecurity, in the dusty, boll-weeviled 20s and 30s than in
the hungry 90s? And if the beginning of white women’s rebellion sparked lynchings in the 90s,
how could lynching have declined in the eras of enfranchised flappers and female cabinet
officers?

14