Munich Personal RePEc Archive

Hate Source: White Supremacist Hate

Groups and Hate Crime

Mulholland, Sean E.

Stonehill College

11 February 2011

Online at https://mpra.ub.uni-muenchen.de/28861/

MPRA Paper No. 28861, posted 21 Feb 2011 00:39 UTC



Hate Source: White Supremacist Hate

Groups and Hate Crime

Sean E. Mulholland
Stonehill College

Abstract

The relationship between hate group activity and hate crime is the-
oretically ambiguous. Hate groups may incite criminal behavior in sup-
port of their beliefs. On the other hand, hate groups may reduce hate
crime by serving as a forum for members to verbally vent their frus-
trations or as protection from future biased violence. I �nd that the
presence of an active white supremacist hate group chapter is associ-
ated with an 18.7 percent higher hate crime rate. White supremacist
groups are not associated with the level of anti-white hate crimes com-
mitted by non-whites, nor do they form in expectation of future hate
crimes by non-whites.
JEL Codes: K14, J15, D71

1 Introduction

The relationship between hate group activity and hate crime is theoretically
ambiguous.1Hate group activities may serve to incite criminal or violent be-
havior in support of their beliefs. On the other hand, if hate groups serve as
support groups for like-minded individuals, members may use meetings and
protests as a way to verbally vent their frustrations. Thus hate group members
may be less likely to carry out these frustrations through criminal or violent
acts. Hate grous may also serve as a means of protection for individuals who
believe they will be victims of biased violence in the future.
From1997 to 2007, the Southern Poverty LawCenter (SPLC), an authority

on hate group activity, declared that the number of active white supremacist
hate group chapters in the United States increased by 63 percent, from 474 in
1997 to 888 in 2007.2 Yet, over the same period, the number of hate crimes

1For a discussion of the historical literature and a fresh look at old analyses of hate
crimes, see Green, Glaser, and Rich [1998].

2The number of groups and the number of a¢liations have changed over time. The
analysis below includes only white supremacist hate groups: the Ku Klux Klan, neo-Nazis,
Racist Skinheads, and Christian Identity Churches from 1997 to 2007. Therefore the terms
hate groups and white supremacist groups will be used interchangeably throughout the
manuscript.

1



recorded in the United States fell by 6.1 percent, from 8443 to 7945.3 Figure
1 depicts the slow and steady increase in the total number of active white
supremacist chapters and the number of hate crimes reported across in the
United States.

<<Figure 1 here>>

Previous studies have focused on economic conditions, demographics, and
historical links between government provided services and the choice to formor
join a hate group [Je¤erson and Pryor, 1999 and Mulholland, 2010]. However,
few have looked at the determinants of hate crimes and all but one fail to in-
clude a measure of hate group activity. McDevitt and Levin [1993] use Boston
data and observe that many hate crimes involved violence directed at dissim-
ilar others moving into a previously segregated area. Medo¤ [1999] performs
a cross-sectional state estimation and �nds that market wages, mean age, and
law enforcement activity predict hateful activity, but urbanization, occupa-
tional status, and social mobility do not. Gale, Heath, and Ressler [2002] use
a panel of 37 states from 1992 through 1995 and �nd that hate crime rates
are positively correlated with unemployment rates, abuse rates, and the parity
of income between blacks and whites. They also �nd hate crime is negatively
correlated with law enforcement expenditures. Ryan and Leeson [2010] use
state level panel data from 2002 through 2008 to �nd that unemployment and
poverty are strongly associated with more hate crime, where as demographics
and, most notably, the number of hate groups are not.
Using U.S. county-level panel data from 1997 through 2007 and control-

ling for unobserved county-level time-invariant heterogeneity, I �nd that hate
crimes are 18.7 percent more likely to occur in counties with active white
supremacist hate group chapters. Excluding hate crimes listed as anti-white,
the presence of a white supremacist chapter is associated with a 19.7 to 24
percent higher rate of hate crimes and between a 22 to 25 percent higher
rate of hate crimes committed by white perpetrators. Anti-white hate crimes
committed by non-white perpetrators are not associated with the presence of
a white supremacist chapter. Rejecting the protection hypothesis, I fail to
�nd white supremacist chapter formation in expectation of future hate crimes
by non-whites. Nor do I �nd chapter formation in response to increases in
anti-white hate crimes perpetrated by non-whites.
Section 2 presents a brief historical look at the evolution of hate crime

legislation in the United States. Section 3 discusses hate groups and their
origins. Section 4 lays out the estimation methodology. Sections 5 and 6
discuss the data and the estimated relationship between white supremacist
chapters and hate crimes. Section 7 investigates further hypotheses on the
relationship between hate groups and hate crimes. Section 8 o¤ers concluding
remarks.

3Multiple hate crimes are sometimes reported for a single hate crime incident. Hate
crime incidents fell by 5.4 percent from 8,049 to 7,624 over the same period.

2



2 Hate Crime

Biased violence directed toward select social characteristics or status, moti-
vated by racism, xenophobia, anti-Semitism, and other biases has been docu-
mented throughout history: from the Old Testament description of the geno-
cides of Amalekites and Midianites to more recent ethnic and religious-based
violence in Bosnia and Sudan. The history of the United States includes bias-
motivated violence against Native Americans, African Americans, and others.
The movement for categorizing certain violent activity directed at individuals
based on their ethnic or social characteristics as �hate crimes" in the United
States, however, is a recent phenomena. According to Jenness and Ryken,
in their book Making Hate a Crime: From Social Movement to Law Enforce-
ment, an"anti-hatecrimemovement"was"energizedbyseveralprevious social
movements" and "emerged in the late 1970s to bring public attention to vio-
lence directed at certain minorities" [Jenness and Ryken, 2001. p.17]. Jenness
and Ryken suggest that the movement to make biased violence a unique type
of crime was born out of the combination of the modern civil rights movement
and the crime victim movement.
The civil rights movement initially concentrated on contesting legal prohi-

bitions based on race. First focusing on voting and then expanding to separate
but equal standards, the civil rights movement sought to obtain similar rights
for those of all races. As groups such as the National Association for the Ad-
vancement of Colored People (NAACP) slowly reduced legal racism, they then
turned to reducing violence directed at those due to ethnic, racial, or religious
status. With these successes, women�s and gay rights groups became more
active in highlighting violence directed at women and gays [Katz, 1976].
In the late seventies, a separate and distinct victim rights movement de-

veloped as a response to the Warren court�s expansion of defendant�s rights
[Weed, 1995 and Maroney, 1998]. The victim rights movement, along with the
women�s rights groups brought attention to the long run trauma of survivors of
violence. These organizations focused their e¤orts on both the direct trauma
resulting from the violence and the abuse that often times took place through
the court system. Lead by Wisconsin in 1981, states began passing victims�
rights legislation. By 1989, forty-two states had passed victims� rights bills
[Weed, 1995].
Growth of the civil rights movement, women�s movement, gay and lesbian

movement, and the victim rights movements set the foundation for discussing
bias-motivated violence toward minorities. Groups that had been a part of
the civil rights movement began focusing their e¤orts on gathering information
about victims of bias-motivated violence. Notable examples include: the Anti-
Defamation League of B�nai B�rith (ADL) which began tracking anti-Semitic
violence in 1979, the Center for Democratic Renewal which began a bimonthly
newsletter in 1979 tracking biased violence, and, the Southern Poverty Law
Center (SPLC) which began tracking Ku Klux Klan activity in 1980 and now
tracks other biased motivated groups.

3



As these organizations continued to track bias-motivated violence, they
slowly began creating "a societal perception that hate crime was a speci�c evil
requiring a speci�c response" Maroney [1998, 579]. The speci�c response often
requested was legal redress. Beginning with Washington and Oregon in 1981,
states began enacting hate crime legislation that criminalized violence direct at
groups due to their racial, ethnic, religious, and various other characteristics.
The activity of organizations interested in victims and civil rights culmi-

nated in the 1990 passage of the Federal Hate Crime Statistics Act (P.L.101-
275). The act de�nes hate crimes as criminal incidents that are at least par-
tially directed against the victims� race, religion, sexual orientation, or eth-
nicity/nationality.4 The act also instructed the FBI to aggregate hate crime
data from agencies actively documenting and willing to voluntarily report hate
crime statistics. In 1992, the FBI began publishing Hate Crime Statistics, a
compilation of hate crime reported by various law enforcement agencies across
the US.5

Di¤erences in how agencies de�ne and report hate crimes are problematic
in any empirical study on crimes in general and hate crimes in particular.
As pointed out by Gale, Heath, and Ressler [2002], while hate crime statis-
tics documentation have improved, there are still serious problems associated
with crime statistics. DiIulio [1996] identi�es two potentially severe measure-
ment errors of crime data: under-reporting by victims and by law enforcement
agencies. Besci [1999] shows that wide variation exists between crime data
sources by comparing the Uniform Crime Report to other sources of crime
data. Furthermore, Grove, Hughes, and Geerken [1985] �nd that the reported
crime rates are closer to the real crime rates for crimes that are less ambigu-
ous between victims and law enforcement, such as homicide, than for more
ambiguous crimes, such as aggravated assault. Grove et al. suggest that this
ambiguity will likely overstate the reported crime rate.
The core concept of a hate crime is ambiguous and therefore can make

classi�cation di¢cult in almost every case [Jacobs and Potter 1998]. However,
Martin [1995, 1996] in two quantitative studies on the Baltimore County and
New York City police departments and Boyd, Berk, and Hamner [1996] in a
study looking at two policing divisions within a large urban police department,
�nd that the process is less problematic than suggested by Jacob and Potter.
"[F]ar from �nding it problematic to interpret and classify speci�c incidents,
police detectives engage in certain routine practices in order to determine the
hate-related status of an incident" [Boyd, Berk, and Hammer 1996, 821]. Thus

4For more information on the Hate Crime Statistics Act, see CRS Report RL33403, Hate
Crime Legislation, by William J. Krouse.

5Especially early on, the concept of a hate crime was nebulous and had yet to be tested
constitutionally. Law enforcement o¢cers may have been reluctant to invest in how or
whether to report a crime as a hate crime. Between 1984 and 1999, the US appellate
court considered the constitutionality of hate crime statutes thirty-eight times, suggesting
to Phillips and Grattet (2000) that the questions of constitutionality and rules had become
settled by the late 1990s.

4



within a division the reporting mechanism for these police departments may be
well de�ned and easier to follow than suggested by Jacobs and Potter [1998].
However, Boyd, Berk, and Hammer show that reporting methods di¤er across
divisions causing comparisons across divisions to be suspect. Therefore, this
manuscript will only look at changes in hate crimes within each county. Even
looking within each county does not eliminate ambiguity and subjectivity that
may occur across each division within each agency. Thus any empirical results
must be interpreted through a critical lens.

3 Hate Groups

The Southern Poverty Law Center de�nes a hate group as an organization
having "beliefs or practices that attack or malign an entire class of people,
typically for their immutable characteristics" [SPLC 2010]. Iannaccone [1992]
and Berman [2000; 2003] demonstrate why rational, utility-maximizing in-
dividuals voluntarily sacri�ce to join religious organizations, fraternities and
sororities, communes, and hate groups. Hate group membership often requires
certain prohibitions and costly signals, such as tattoos, piercings, and acts of
violence. Participation in such prohibitions reveals to other members that
the individual is committed to the organization, and as such, grants an active
member certain privileges and bene�ts. Joining reveals that the bene�ts of
membership are greater than the costs associated with signaling commitment
to the group.
White supremacist hate groups are by no means identical. Each organiza-

tion, as shown in Table 1, has di¤erent levels of prohibitions and goals and thus
solves the problem of collective action in a variety of ways. Racist Skinhead
and neo-Nazi members, unlike Ku Klux Klan and Christian Identity mem-
bers, often display tattoos or wear distinctive clothing to signal membership.
Furthermore, each type and chapter express their biased views in a variety of
manners. Some are physically violent while others are not.

<<Table 1 here>>

Often cited reasons for hate group formation include peer validation, frus-
tration, scapegoating, and boredom. Groups form when a small number of
like-minded individuals believe they have been wronged in a similar manner
and are searching for ways to right perceived wrongs. One hypothesis put
forth by McDevitt and J. Levin [1993] is that the impetus for organizing is
often based on turf issues. Individuals commonly cite the thrill associated with
the victimization [Levin 1993]. By forming a group, these individuals validate
themselves by �committing an act that enhances one�s image internally, with
a peer group, and with society at large� [Levin 1993]. If Levin is correct, then
we should witness more hate crime in counties where hate groups are present.
In a sense, then, many hate groups are much like gangs. In his work on

gangs, Klein [1995] summarizes the psychological factors in his statement that
�the gang is seen as an aggregate of individuals held together more by their

5



own shared incapacities than by mutual goals. Primarily, group identi�cation
is important as it serves individual needs; it leads to delinquent group activity
only secondarily and only in the absence of pro-social alternatives.� Thus, if
Klein is correct, it may be the case that hate groups do not a¤ect the overall
rate of hate crimes if they serve more of a pro-social role.
Furthermore, recent emprical work by Osoba and Sobell [2009] on youth

gangs suggests that gangs form in response to government�s failure to protect
youth against violence.6 Without gangs, the level of violence may actually be
higher. If hate groups form to protect its members from future violence, hate
groups may actually be associated with less crime and even less hate crime.

4 Estimation Method

The goal is to determine whether the hate crime rate in county i in year t is
associated with white supremacist activity in county i for year t. While anec-
dotal evidence suggests that the number of chapters is positively correlated
with the number of members, the SPLC does not report the number of mem-
bers per chapter.7 Therefore, changes in the number of chapters may be a poor
measure of the level of activity. The formation of additional active hate group
chapters from one year to the next may simply be a result of the splintering
of one large group; the reduction in the number of chapters may be the result
of a merger. Following Je¤erson and Pryor [1999] and Mulholland [2010], this
analysis will focus on whether hate crime rates are associated with counties
home to any active white supremacist chapters and not the total number of
chapters.8

To address these issues, I �rst construct a dichotomous variable, activeit:

activeit =

�

1 if number of active white supremacist chapters is >0

0 if no active white supremacist chapters present

�

(1)

where activeit equals one if there are active white supremacist hate group

6Theoretical models of government formation out of anarchy are developed by Nozick
[1974] and Buchanan [1975].

7For two discussions on the correlation between number of active chapters and member-
ship see: Intelligence Report [2000] and The Stephen Roth Institute [2004].

8A county may be home to a hate group even though it is not reported as present for
that calendar year. It is possible that the hate group chapters simply did not draw attention
to themselves for that calendar year. In order to determine the e¤ects of this possibility,
I construct alternative measures that assumed a county was hate group free only if that
county witnessed no hate group activity over multiple years. If an active hate group is
present during any one of these years, I assume that the hate group was simply silent during
the others and continued to be present over the entire time period.
Using this methodology, I construct three alternative measures of white supremacist ac-

tivity: one that assumes a hate group must be silent for two years before I consider it
disbanded; one for three years; and one for four years. Repeating the estimation in Ta-
ble 3 using these three alternative measures reveals qualitatively and quantitatively similar
results. Results are available from the author upon request.

6



chapters in county i at time t and zero if no active white supremacist chapter
is reported.
I then estimate the e¤ects of the presence of one or more active white

supremacist chapters on the overall rate of a hate crime:

hate crimeit = �+��activeit +xit� +�i +�t +"it (2)

where hate crimeit is the hate crime rate in county i in year t;activeit indicates
presence of an active white supremacist chapter, xit is the vector of explana-
tory variables for county i in time period t, and � is the vector of county
parameters to be estimated. The control variables in xit include real median
household income, the unemployment rate, the real unemployment bene�ts
per unemployed citizen, the poverty rate, the percentages of the population
that are black, Hispanic, and white, the population density, the percentage of
males age �fteen through forty-four, and the overall crime rate.9 The Hausman
test rejects the null hypothesis that the coe¢cients estimated by the e¢cient
random e¤ects estimator are the same as the ones estimated by the consistent
�xed e¤ects estimator.10 Therefore I include time-invariant, county-speci�c
e¤ects represented by �i to control for omitted variables that di¤er between
counties but are constant over time. This �xed e¤ects estimation methodology
relies on within county variation. The year dummies, �t, control for omitted
changes over time that a¤ect all counties similarly. The error term, "it; is clus-
tered by county in order to account for non-random errors within each panel
[Rogers, 1993; Williams, 2000; and Wooldridge, 2002].11

There are �aws associated with the collection of hate crime data that may
bias the estimates. First, many states and local governments do not have
formal guidelines for law enforcement o¢cials on how to complete a formal
hate crime report.12 Thus only a fraction all possible hate crimes are reported.
If counties with active hate groups are more likely to report crimes as hate
crimes, then the estimates will be upward biased. If counties with active hate

9I do not include real expenditures on police protection because it is only available for
1997 and 2002 at the county level from the Census Bureau�s Census of Governments. When
including real police expenditures, the presence of an active white supremacist chapter is
associated with a greater increase in all types of hate crimes than the estimates presented
in Table 3.
10The Hausman test value of 73.26 with 20 degrees of freedom result in a p-value of .00,

thus rejecting the null that the more e¢cient random e¤ects estimator returns the same
estimates as the �xed e¤ects estimator.
11Assuming that hate groups did not alter their voting behavior over this time period,

using �xed e¤ects partially addresses any worry that that hate groups "may vote for policies
that reduce the local government provision of services [Glaeser 2005]."
12Hate crime data reported by the FBI are "from all law enforcement agencies that submit-

ted either of the following: 1) at least one National Incident-Based Reporting System Group
A Incident Report, a Group B Arrest Report, or a Zero Report for at least 1 month of the
calendar year; or 2) at least one Hate Crime Incident Report and/or a Quarterly Hate Crime
Report" [http://www2.fbi.gov/ucr/hc2008/data/table_12_dd.html, Viewed 7/11/10].

7



groups are less likely to report crimes as hate crimes, then the estimates will
be downward biased.
Second, in counties where hate crime reporting is conducted by a fraction

of state and local agencies, only a portion of a county�s population may be
covered. If localities within a county where active white supremacist chapters
are present are less likely to be covered by hate crime reporting, this will
downward bias the estimates. Finally, participation of agencies can vary from
year to year. If this is random, it will not bias the estimates. However, if
agencies are less likely to �le a report when chapters are present, the estimates
will be downward biased. If agencies fail to submit hate crime information
when white supremacist chapters are not present, the estimates will be upward
biased.

5 Hate Crime and Hate Group Data

Hate crime data comes from the FBI�s annual report, Hate Crime Statistics.
First published in 1992, Hate Crime Statistics compiles hate crimes reported
by various law enforcement agencies. In 1992, Hate Crime Statistics covered
only 51 percent of the United States population from 42 states. By 1997, 83
percent of the population from 48 states and DC were covered. This is similar
to the coverage in 2007: 85.7 percent from 49 states and DC. Due to the
relative stable percentage of population covered from 1997 to 2007, ranging
from a high of 86.6 percent in 2004 and a low of 80.0 percent in 1998, this
manuscript uses data on hate crimes from 1997 through 2007.
Figure 2 maps the highest hate crime rate for each county over the 1997 to

2007 time period.13 In 1997 approximately 21 percent of counties reported at
least one hate crime incident with 52 percent of incidents resulting in one or
two hate crimes committed. By 2007, 27 percent of counties reported at least
one hate crime incident with just under 50 percent of incidents resulting in one
or two criminal charges. While the overall number of hate crimes fell between
1997 to 2007, the number of counties reporting at least one hate crime rose by
twenty-four percent. The portion of counties reporting at least one hate crime
between 1997 and 2007 is 60.6 percent. The highest hate crime rate over this
period, 37.4 out of 10,000, was reported for Scott County, Arkansas in 2003.
Even though states such as Georgia, Alabama, and Mississippi have no state
datacollectionstatutes, someagenciesvoluntarily submithatecrimereports to
the FBI. The portion of former Confederate States of America (CSA) counties
reporting at least one hate crime is 55.8 percent.

<<Figure 2 here>>

13Hate crimes are reported at the agency level not the county level. Agencies whose
jurisdictions cover multiple counties do not report which county the reported hate crime
took place. Only 5.5 percent of hate crime incidents are reported by agencies covering
multiple counties. For agencies that cover more than one county jurisdiction, I assign the
crimes to the county Federal Information Processing Standards (FIPS) code listed for that
agency in the Law Enforcement Agency Identi�ers Crosswalk [United States], 1996.

8



The Southern Poverty Law Center tracks hate groups by "using hate group
publications and websites, citizen and law enforcement reports, �eld sources
and news reports" [Southern Poverty Law Center]. Initiated in 1981, 1997
marks the �rst year the SPLC attempted to gather data for all known active
hate group chapters by city.14 Although the SPLC now tracks many types of
organizations only the Ku Klux Klan, neo-Nazi, Racist Skinhead, and Chris-
tian Identity groups are available for every year from 1997 to 2007.15 Between
1997 and 2007, 793 US counties, or approximately 25 percent, were home to
at least one active hate group.16 Figure 3 shows the maximum number of Ku
Klux Klan, neo-Nazi, Racist Skinheads, and Christian Identity chapters by
county reported during any calendar year from 1997 to 2007. Harris County,
Texas, in 2006, and Cook County, Illinois, in 1997, 2005, and 2006, were home
to the largest number of active hate group chapters during a single calendar
year. Counties located in states that were part of the CSA are well represented
and make up 43.4 percent of those reporting an active hate group through-
out the period. However, over half of the counties reporting an active hate
group were located outside the former CSA. In fact, of the twelve counties
that report �ve or more active hate group chapters during a calendar year,
only three: Harris, Texas in 2006 and 2007; Tarrant, Texas in 1998, and Spar-
tanburg, South Carolina in 2005 and 2006 are located in states that were part
of the CSA.

<<Figure 3 here>>

Figure 4 depicts the change in the overall presence of active Ku Klux Klan,
neo-Nazi, Racist Skinheads, and Christian Identity chapters between 1997
through 2007. The thirty-two counties in red were always home to at least
one active white supremacist chapter. Seven hundred and sixty one counties,
shown in blue, representing 49 states witnessed at least one year with one
or more active white supremacist chapters and one year without an active
chapter. Of these 761 counties, 215 report an active chapter for only one year.

14Although the SPLC reports hate group location by city or town, the analysis is per-
formed at the county level for theoretical and empirical reasons. First, many hate groups
chapters hold rallies and recruitment meetings outside their home towns in locations nearby
and thus have members from the surrounding towns and townships. Second, because many
of these towns are not in Metropolitan Statistical Areas, county level data represents the
least aggregated measures of crime, unemployment, poverty, and the like that are available.
15In 2000 the Southern Poverty Law Center began monitoring Neo-Confederate organi-

zations. This study does not include those organizations because of their initial omission
by the Southern Poverty Law Center, nor does it include Black Separatists or the "Other"
category.
16Not all active hate groups can be assigned to a single county. For instance the SPLC

reports an active chapter of the Knights of the White Kamellia (KKK) present in NC,
but does not list a city. When no city is reported, the hate group is not included in the
analysis. The portion of active groups not included ranges from 1.2 percent in 1998 to 12.8
percent in 2007. Undercounting active white supremacist chapters likely biases the estimates
downward.

9



All remaining counties, shown in white were never home to an active white
supremacist chapter.

<<Figure 4 here>>

Looking at the individual hate group types, 3.8 percent of the county-year
observations report the presence of at least one active Ku Klux Klan chapter.
Counties reporting an active KKK chapter for at least one year, were home to
an active KKK chapter about 26 percent of the time. Neo-Nazi chapters look
similar toKKKchapterswith4.1percentof county-yearobservations reporting
the presence of a neo-Nazi chapter. Only about 2 percent of counties were
home to a neo-Nazi organization over the ten year period. Racist Skinheads
are much more transient than most other type of hate groups; counties that
were home to a Racist Skinhead chapter reported an active Racist Skinhead
chapter only 15 percent of the time. Christian Identity Churches were present
in 3.5 percent of the counties. Christian Identity Churches were the least likely
to disband. Counties that were home to a Christian Identity Church for at
least one year, were home to an active church 34 percent of time.
Table 2 displays the county-level summary statistics for hate crime rates,

by various types of victims and perpetrators, and white supremacist activity.

<<Table 2 here>>

6 The Relationship Between Hate Groups and Hate

Crime

The results of the �xed e¤ects estimation in equation 2 are reported in column
one of Table 3.

<<Table 3 here>>

Using the overall sample, the presence of an active white supremacist chap-
ter is associated with 286 more hate crimes per 10,000 residents. Given that
the average number of hate crimes per 10,000 residents is 1494, the presence of
an active white supremacist chapter is associated with an 18.7 percent higher
hate crime rate.17 Column 3 displays the same estimation using only those
counties that report one or more hate crimes over the 1997-2007 time frame.
The restricted sample reveals that the presence of a chapter is associated with
335 more hate crime per 10,000 residents or an increase of 14.0 percent.18

17Agencies covering multiple counties do not report the county in which a hate crime
took place. Therefore, I also estimate Equation 2 excluding the 5.5 percent of hate crimes
reported by agencies covering multiple counties. The estimated coe¢cient on active white
supremacist chapters is .029 and signi�cant at the �ve percent level. This corresponds to a
20.2 percent higher hate crime rate.
18The annual average number of hate crimes for counties that submit at least one report

from 1997 through 2007 is 2393 per 10,000 residents.

10



These results suggest the validity of Levin�s (1993) hypothesis that commit-
ting an act of violence may enhance members� image within the group.
Given the annual nature of these data, I perform three tests for autocorre-

lation: the Wooldridge Test, the modi�ed Bhargava et al. Durbin-Watson, and
the Baltagi-Wu LBI test. The Woodlrigde Test barely fails to reject the null of
no autocorrelation.19 The modi�ed Bhargava et al. Durbin-Watson statistic
of 1.78 and the Baltagi-Wu LBI statistic of 1.93 both suggest the presence
of �rst-order autocorrelation.20 Because the dependent variable exhibits �rst-
order autocorrelation in two of the three tests, that is, it is dependent on
its own past realizations, I use the Arellano-Bover (1995) and Blundell-Bond
(1998) System General Method of Moments (GMM) dynamic panel estimator
and include a lagged dependent variable as an explanatory variable.

hate crimeit = �+� �hate crimei;t�1 +��activeit +xit� +�i +�t +"it (3)

In this model speci�cation, the lagged hate crime rate is an endogenous
variable and the overall crime rate is treated as a predetermined variable. All
others variables are treated as exogenous. I address the endogeneity of the
one year lag of hate crimes by using internal instruments, namely lagged levels
and lagged di¤erence of the hate crime rate and the overall crime rate. This
eliminates potential endogeneity and reverse causation. Given the number of
instruments, I conduct a Hansen�s J-statistic overidenti�cation test based on
the weighted matrix.21 With resulting Hansen J values of 13.628 and 14.698,
and the resulting p-values of 0.626 and 0.547, for the full and restricted sample
respectively, I fail to reject the null that these instruments are valid.
Column2 and4 inTable3 reports theestimatedresultsusing the twostage

System-GMMestimation method.22 For the overall sample, the System-GMM
estimator returns a similar signi�cant relationship between white supremacist
hate group presence and a higher hate crime rate: an increase of 18.6 percent.
For the restricted sample, the relationship between hate crimes and white
supremacist hate groups is imprecisely estimated

7 Further Investigation of Alternative Hypotheses

The results above support the theory that white supremacist chapters increase
the hate crime rate. However, the results are also consistent with two compet-

19The Wooldridge test for �rst-level autocorrelation returns an F( 1, 3101) = 4.017 with
a Prob > F = 0.05, just barely failing to reject the null of no autocorrelation.
20For the both the modi�ed Bhargava et al. Durbin-Watson and the Baltagi-Wu LBI

with a large number of cross sections, a test statistic of less than 2 indicates positive serial
correlation.
21One di¢culty with Hansen�s J-statistic test is that its size is distorted as the number of

instruments grows. Therefore, instead of generating an additional instrument for each panel,
lag, and time, I only use the instruments for each panel and lag. Another overidenti�cation
test, the Sargan test is more reliable, but not appropriate if errors are heteroskedastic.
22The Arellano and Bond two stage procedure generates estimates of the standard devia-

tion which can be biased. For this reason the estimated standard errors are reported using
the Windmeijer (2005) correction.

11



ing hypotheses: (1) that the presence of one or more white supremacist groups
is a symptom of the overall level of distrust and violent antagonism amongst
various citizens in a county and (2) that an increase in expected future hate
crime leads to hate group formation for self-defensive purposes. Luckily, Hate
Crime Statistics reports both the type of biased-motivated crime, for exam-
ple anti-black or anti-homosexual, and the race of the perpetrator for each
hate crime. To investigate the �rst competing hypothesis I look at whether
activity by white supremacist chapters is associated with a higher rate of hate
crimes listed as anti-white or hate crimes committed by non-whites. If the
relationship is positive, white supremacist activity may proxy for the overall
biased-motivated antagonism in a county and not be directly related to white
supremacist group activity. Alternatively if white supremacist activity is as-
sociated with more hate crimes excluding those listed as anti-white or more
hate crimes committed by white perpetrators, the initial hypothesis that hate
crimes are caused by white supremacist groups is more plausible.

7.1 Non-anti-white hate crimes and white perpetrators

To disentangle these hypotheses, I �rst repeat the estimations in equations 2
and 3 but replace the dependent variable with the hate crimes, excluding those
listed as anti-white.23 As reported in column 1 and 2 of Table 4, I �nd that
the presence of a white supremacist chapter is associated with between 245
to 300 more non-anti-white hate crimes per 10,000 residents or an increase of
between 19.7 to 24.1 percent. This is a slightly larger result than those found
in Table 3 when looking at all types of hate crime.
However, a portion of these hate crimes are committed by people who

are unlikely to be members of white supremacists groups, namely non-whites.
To account for this, I estimate the relationship between white supremacist
groups and hate crimes committed by whites with non-white victims.24 The
estimates are reported in columns 3 and 4 of Table 4. The presence of a
white supremacist chapter is associated with between 136 and 155 more white
on non-white hate crimes. Recognizing the smaller number of white-on-non-
white hate crimes, this is an increase of between 22.0 and 25.1 percent. In
both cases I �nd that the presence of an active white supremacist group is
associated with a greater likelihood of non-anti-white hate crimes and hate
crimes with non-white victims committed by whites. The percentage increase
for these types of hate crimes is greater than the increase estimated for all
types of hate crimes shown in Table 3. This suggests that white supremacist
chapters play a direct role in hate crime rates.

<<Table 4 here>>

23
hate crime(excluding anti-white)

i;t
is constructed using the entire county population

in the denominator.
24The variable hate crime(white-on-other)

i;t
uses the total population as its denominator

because biased motivation can include racial as well as non-racial characteristics such as
gender, religion, and sexual identity.

12



7.2 Anti-white hate crimes and non-white perpetrators

White supremacist group members or those sympathetic to white supremacist
organizations are not likely to perpetrate hate crimes classi�ed as anti-white.
They are also unlikely to be non-white. By looking at hate crimes by non-
whites and those by non-white listed as anti-white, I am able to di¤erentiate
whether or not white supremacist groups are symptomatic of the overall level
of antagonism. A positive relationship between these types of hate crimes and
white supremacists groups would reject the hypothesis that white suprema-
cist groups cause hate crimes. A negative or non-relationship would further
support the hypothesis that active white supremacist groups directly increase
hate crime rates. I �rst estimate the relationship between white supremacist
chapters and hate crimes committed by non-whites by replacing the dependent
variable in equation 2 with the rate of hate crimes by non-whites.
The results are reported in columns 1 and 2 in Table 5. The �xed e¤ect

estimation shows no statistical relationship between white supremacist groups
and hate crimes with non-white perpetrators. However, the System-GMM
reports a positive relationship, suggesting that presence of a hate group may
be symptomatic of overall antagonism. This results may come about if white
supremacist groups in�ame anti-gay or anti-black, or other biases.
Hate crimes listed as anti-white, however, are unlikely to be perpetrated

by white supremacists Therefore, I replace the dependent variable with hate
crimes listed as anti-white committed by non-whites. The estimated the re-
lationship is reported in columns 3 and 4 in Table 5. Both the �xed e¤ects
and the System-GMM reject the hypothesis that white supremacist groups are
simply a proxy for violent antagonism.

<<Table 5 here>>

7.3 Protection from future hate crimes

The results in Tables 4 and 5, save for column 2 in Table 5, reject the hypoth-
esis that white supremacist groups are symptomatic of overall antagonism in
a county. Instead the results suggest that white supremacists are the source
of greater hate crime rates. White supremacist chapters may, however, be a
response to expected future increases in hate crime activity. Thus the above
results may not reveal a causal e¤ect of white supremacist chapters on hate
crimes. Rather the estimates may reveal the formation of chapters in antici-
pation of future increases in hate crime by non-whites or hate crimes with an
anti-white bias.25

To determine whether current changes in hate crimes perpetrated by non-
whites is related to past changes in white supremacist activity or lagged
changes in hate crimes by non-whites, I estimate Equation 4. Equation 4

25Heckman [2000] mentions that future yt often determines current xt in a dynamic eco-
nomic model.

13



estimates whether a change hate crimes committed by non-whites is associ-
ated with past changes in the presence of active white supremacist chapters.
The regression controls for the lagged changes in hate crimes by non-whites as
well.

�hate crime(by non-white)
it
=�+

P

2

r=1

r ��hate crime(by non-white)i;t�r

+
P

2

r=1
�r ��activei;t�r

+xit� +�t +"it
(4)

The results in column 1 of Table 6 show that the coe¢cient estimates of
lagged chapter activity are not signi�cant. This reinforces the notion that
non-whites do not commit more hate crime when a county witnesses a change
in the activity of white supremacist chapters. To test whether chapters form
in anticipation of future changes in hate crimes by non-whites, I �rst construct
the trichotomous variable �activeit where:

�activeit =

8

<

:

�1 if a county is no longer home to an active white supremacist chapter
0 if a county realizes no change in the presence or lack of presence of a chapter
1 if a county is no longer free of white supremacist activity

9

=

;

(5)
I then replace the dependent variable in Equation 4 with, �activeit, the change
in the probability an active white supremacist chapter is present, and estimate
the ordered logistic using Equation 6.

�activeit =�+
P

2

r=1

r ��hate crime(by non-white)i;t�r

+
P

2

r=1
�r ��activei;t�r +xit� +�t +"it

(6)

Column 2 in Table 6 shows the results. The coe¢cient on hate crimes by
non-whites is again not signi�cant, suggesting that current white supremacist
activity is not related to past hate crimes committed by non-whites. White
supremacist groups do not appear to form for protection from hate crimes
committed by non-whites.
Lookingmore deeply, however, potential white supremacymembers maybe

less concerned with hate crimes committed by non-whites and more concerned
with hate crimes by non-whites that are listed as anti-white. Table 7 repeats
the exercise in Table 6 but replaces the dependent variable with anti-white
hate crimes committed by non-whites and estimates each of the following two
equations:

�hate crime(other-on-white)
i;t
=�+

P

2

r=1

r ��hate crime(other-on-white)i;t�r

+
P

2

r=1
�r ��activei;t�r

+xit� +�i +"it
(7)

�activei;t =�+
P

2

r=1

r ��hate crime(other-on-white)i;t�r

+
P

2

r=1
�r ��activei;t�r +xit� +�i +"it

(8)

14



I again �nd no relationship. The change in anti-white hate crimes commit-
tedbynon-whitesdoesnotappear related topast changes inwhite supremacist
activity. Neither do white supremacist chapters appear to formto protect their
members from anti-white hate crimes.

8 Conclusion

Counties with active white supremacist chapters experience greater rates of
hate crime. In particular, hate crimes not listed as anti-white and hate crimes
committed by white perpetrators increase by a greater percentage than all
types of hate crimes. White supremacists groups do not appear to increase the
rate of anti-white hate crime perpetrated by non-whites. These results rejects
the possibility that white supremacist groups are symptomatic of changes in
the overall level of bias-motivated antagonism. I also reject the protection
hypothesis. White supremacistgroupsdonotappear to formtoprotectagainst
expected future increases in hate crimes by non-whites. Nor do they form to
protect against future anti-white crimes perpetrated by non-whites.
Previous research by Ryan and Leeson [2010] �nds no relationship between

hate groups and hate crimes. There are several possible reasons why these
results di¤er. First, Ryan and Leeson perform their estimation at the state
level whereas this analysis is at the county level. If hate groups are a local
phenomena, thenwhite supremacist groupsaremuchmore likely todirect their
frustration at local residents. If a hate group member commits a hate crime, it
is likely that theywill do soagainst thosewhomthey feelhavedirectlywronged
them such as others in the community. Due to local di¤erences, aggregation
may result in lost variation and reduce the estimated e¤ect. Moreover, because
the SPLCdoes not report chapter size, the numberof chapters manybe a noisy
measure of the number of white supremacists.
Second, even if a hate group is not actively seeking to commit crimes, its

presence may signal to local community members that a particular bias is
acceptable. Non-members may act out against those viewed as inferior by the
white supremacists because such views are present in the community. This
may partially explain why I �nd a positive relationship between current white
supremacist activity and hate crimes by non-white perpetrators, though I do
not �nd a relationship when looking at the lagged changes. Spill over e¤ects,
if they exist at all, are much more likely to be local in nature and thus lost
when aggregating to the state level.
Third, these results include only the KKK, neo-Nazi, Racist Skinhead

Chapters, and Christian Identity Churches that are listed with a city or county
location. Ryan and Leeson include all types of hate groups or all types
sans Patriot Groups listed in a state. Their estimates include three addi-
tional hate groups: neo-Confederates, Black Separatists, and Other. This
"Other" category includes organizations that may be biased in their views
of those covered or protected by hate crime legislation, but do not advo-
cate violent acts against those whom they are biased. This category in-

15



cludes scienti�c research institutes, recording studios, religious organizations,
rights organizations, and retail outlets . Included in the "Other" category
is the Charles Darwin Research Institute (CDRI), "a scienti�c and educa-
tional foundation established to honor and extend the scienti�c revolution
inaugurated by one of the greatest �gures in the history of human thought"
[http://www.charlesdarwinresearch.org, Viewed 11/10/10] and the Family Re-
search Institute whose goal is "to generate empirical research on issues that
threatenthe traditional family, particularlyhomosexuality, AIDS, sexual social
policy, anddrugabuse" [http://www.familyresearchinst.org, Viewed11/10/10]
Although both organizations may hold biased views that may "malign an en-
tire class of people," they do not advocate violence. This "Other" category
may produce a spill over e¤ect, and increase hate crime, but it is unlikely to
occur through direct action of the organization�s members.
White supremacist chapters do not appear to form in response to hate

crimesperpetratedbynon-whitesorhatecrimes listedasanti-white. In this re-
spect white supremacists groups are unlike gangs, where "violent crime causes
an increase in gang membership" [Osoba and Sobel, 2010]. This di¤erence
may be a function of how law enforcement agencies address hate crimes and
gang crimes. When it comes to hate crimes, agencies may more closely follow
Osoba and Sobel�s suggestion that "governments should try harder to protect
the rights of individuals who are the victims of violence." This may in part be
due to how hate crime legislation attempts to protect the rights of individuals
who are victims of bias-based violence.
Further research is needed to determine the source of these con�icting re-

sults. In addition, further queries are required to determine if black separatist
organizations, such as the new Black Panthers, and Patriot organizations are
positively related to hate crimes in the United States.

Acknowledgement 1 Bene�cial suggestions were received from seminar par-

ticipants at the College of the Holy Cross, Lebanon Valley College, Macon State

College, Mercer University, Stonehill College and the Southern Economic As-

sociation, and the Association for Private Enterprise Education. I wish to

extend my gratitude to Angela K. Dills, Rey Hernandez-Julian, Peter Leeson,

Matt Ryan, and Robert Tollison, for their valuable comments and suggestions.

Errors or de�ciencies that have to this point survived this counsel are most

assuredly mine alone.

References

[1] Anti Defamation League Website. http://www.adl.org/. (Accessed Feb-
ruary 4, 2005).

[2] Berman, Eli. 2000. "Sect, Subsidy, and Sacri�ce: An Economist�s View of
Ultra-Orthodox Jews," Quarterly Journal of Economics. 115(3), 905-953.

[3] ____________. "Hamas, Taliban, and the Jewish Underground:
An Economist�s View of Radical Religious Militias." National Bureau of
Economic Research Working Paper. No. 10004.

16



[4] Besci, Z. 1999. "Economics andCrime in the State."Federal Reserve Bank
of Atlanta Economic Review, 38-56.

[5] Boyd, Elizabeth, Richard Berk, and Karl Hamner.1996. "Motivated by
Hatred or Prejudice : Categorization of Hate-Motivated Crimes in Two
Police Divisions." Law & Society Review, 30(4): 819-850.

[6] Buchanan, James M. 1975. The Limits of Liberty: Between Anarchy and
Leviathan.University of Chicago Press: Chicago, IL.

[7] DiIulio, J. J., Jr. 1996. "Help Wanted: Economists, Crime and Public
Policy." The Journal of Economic Perspectivies. 3-24.

[8] Gale, Lewis R., Will C. Heath, and Rand W. Ressler. 2002. "An Economic
Analysis of Hate Crime." Eastern Economic Journal 28: 203-15.

[9] Glaeser, Edward L. 2005. "The Political Economy of Hatred," Quarterly
Journal of Economics. 120(1), 45-86.

[10] Green, Donald P., Jack Glaser, and Andrew Rich. 1998. "From Lynching
to Gay-Bashing: The Elusive Connection Between Economic Conditions
andHateCrime."Journal of Personality and Social Psychology, 75, 82-92.

[11] Grove, W.R., M. Hughes, and M. Geerken. 1985. "Are Uniform Crime
Reports a Valid Indicator of Index Crimes? An A¢rmative Answer with
Minor Quali�cations." Criminology, 451-501.

[12] Iannaccone, Laurence R. 1992. "Sacri�ce and Stigma: Reducing Free-
riding in Cults, Communes, and Other Collectives," Journal of Political
Economy, 100(2), 271-291.

[13] Jacobs, James and Kimberly Potter, 1998. Hate Crimes: Criminal Law
and Identity Politics, New York: Oxford University Press.

[14] Je¤erson, Philip N, and Frederic L. Pryor, 1999. "On the Geography of
Hate," Economics Letters, 65(3), 389-395.

[15] Jenness, Valerie and Ryken Grattet. 2001. Making Hate a Crime: From
Social Movement to Law Enforcement. New York: Russell Sage Founda-
tion.

[16] Katz, Jack, 1976. Gay American History: Lesbians and Gay Men in the
U.S.A. New York: Thomas Y. Crowell.

[17] Kleg, Milton. 1993. Hate Prejudice and Racism. (The State of New York:
Albany Press).

[18] Klein, M. W. 1995. The American street gang: Its nature, prevalence and
control. New York: Oxford Univerity Press.

[19] Krouse, William J. 2007. �CRS Report for Congress: Hate Crime Legis-
lation.�

[20] Levin, Brian. 1993. "ADreamDeferred: the Social and Legal Implications
of Hate Crimes in the 1990s". The Journal of Intergroup Relations, vol.
xx, no. 3, Fall.

[21] Maroney, Terry. A., 1998. "The Struggle Against Hate Crime: Movement
at a Crossroads." New York University Law Review, 73(2): 564-620.

[22] Martin, Susan. 1995. "A Cross-Burning is Not Just an Arson: Police
Social Construct of Hate in Baltimore County." Criminology. 33(3): 303-
326.

17



[23] Martin, Susan. 1996. "Investigating Hate Crimes: Case Characteristics
and Law Enforcement Responses." Justice Quarterly 13(3): 455-480.

[24] McDevitt, Jack. and Jack Levin. 1993. "Hate Crimes: The Rising Tide of
Bigotry and Bloodshed". 9-19.

[25] Medo¤, Marshall. 1999. "Allocation of Time and Hateful Behavior: A
Theoretical Analysis of Hate and Hate Crimes." American Journal of
Economics and Sociology 58: 959-973.

[26] Nozick, Robert. 1974. Anarchy, State, and Utopia. Basic Books: New
York, NY.

[27] Phillips, Scott and Ryken Grattet. 2000. "Judicial Rhetoric, Meaning-
Making, and the Institutionalization of Hate Crime Laws." Law & Society
Review 34(3): 567-602.

[28] Rogers, W. H. 1993. Regression standard errors in clustered samples.
Stata Technical Bulletin 13: 19�23. Reprinted in Stata Technical Bulletin
Reprints, vol. 3, 88�94.

[29] Ryan, Matt E., and Peter T. Leeson. 2010. "Hate Groups and Hate
Crimes". Working Paper.

[30] Sobel, Russell S. and Brian J. Osoba. 2009. "Youth Gangs as Pseudo-
Governments: Implications for Violent Crime," Southern Economic Jour-
nal 75:4, pp. 996-1018.

[31] Southern Poverty Law Center Website: http://www.splcenter.org/. (Ac-
cessed 07/18/08).

[32] The Southern Poverty Law Center. 1998. "474 Hate Groups Blanket
America." Intelligence Report. Issue 89.

[33] ____________. 1999. "Hate Groups Top 500." Intelligence Report.
Issue 93.

[34] ____________.2000."TheDecade inReview." Intelligence Report.
Issue 97.

[35] ____________. 2001. "Blood on the Border." Intelligence Report.
Issue 101.

[36] ____________. 2002. "The Year in Hate." Intelligence Report. Is-
sue 105.

[37] ____________.2003."HateTakesaHit." Intelligence Report. Issue
109.

[38] ____________.2004."AgeofRage." Intelligence Report. Issue114.
[39] ____________. 2005. "Holy War". Intelligence Report. Issue 117.
[40] ____________. 2006. "The Year in Hate". Intelligence Report. Is-

sue 121.
[41] ____________. 2007. "The Year in Hate". Intelligence Report. Is-

sue 125.
[42] ____________. 2008. "The Year in Hate". Intelligence Report. Is-

sue 129.
[43] The Stephen Roth Institute. 2004. For The Study of Contemporary Anti-

semitism and Racism Annual Report United States 2004.

[44] U.S. Dept. of Justice, Bureau of Justice Statistics. LAW ENFORCE-

18



MENT AGENCY IDENTIFIERS CROSSWALK [UNITED STATES],
1996 [Computer �le]. ICPSR ed. Ann Arbor, MI: Inter-university Consor-
tium for Political and Social Research [producer and distributor], 2000.
doi:10.3886/ICPSR02876

[45] United States Census. 1992 Census of Government. Volume 4. Govern-
ment Finances. Number 5, Compendium of Government Finances

[46] ____________. 1997 Census of Government. Volume 4. Govern-
ment Finances. Compendium of Government Finances.

[47] Weed, Frank, 1995. Certainty of Justice: Refonn in the Crime victim
Movement. New York,NY: Aldine De Gruyter.

[48] Williams, R. L. 2000. A note on robust variance estimation for cluster-
correlated data. Biometrics 56: 645�646.

[49] Windmeijer, Frank, 2005. "A �nite sample correction for the variance
of linear e¢cient two-step GMM estimators," Journal of Econometrics,
126(1), 25-51.

[50] Wooldridge, J. M. 2002. Econometric Analysis of Cross Section and Panel
Data. Cambridge, MA: MIT Press.

A Data Appendix

crime data comes from the U.S. Dept. of Justice, Federal Bureau of Investiga-
tion. Uniform Crime Reporting Program Data: County-Level Detailed Arrest
And O¤ense Data, 1997 -2007. Ann Arbor, MI: Inter-university Consortium
for Political and Social Research (Retrieved 5/20/2006 & 10/20/10).
1997-2007. (Retrieved June 20, 2006), from the University of Virginia,

GeospatialandStatisticalDataCenter: http://�sher.lib.virginia.edu/collections/stats/crime/.
income is the median household income in thousands ($1000s) of 2006

dollars by county as reported by the U.S. Census Bureau, "State and County
Income and Poverty Estimates (SAIPE) from 1997 through 2007. http://
www.census.gov/hhes/www/saipe/county.html. Viewed (10/20/10).
unemployment rate is the annual average percent of county residents

who are unemployed and looking for a job as reported by the United States De-
partment of Labor: Bureau of Labor Statistics. Labor Force Data by County,
Annual Averages (Viewed (07/10/10):
ftp://ftp.bls.gov/pub/special.requests/la/laucnty97.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty98.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty99.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty00.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty01.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty02.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty03.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty04.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty05.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty06.txt
ftp://ftp.bls.gov/pub/special.requests/la/laucnty07.txt

19



poverty is the portion of a county�s residents that are below a household
income threshold as reported by the U.S. Census Bureau, "State and County
Income and Poverty Estimates (SAIPE) from 1997 through 2007. http://
www.census.gov/hhes/www/saipe/county.html. Viewed (07/10/08).
adult male is the percentage of the county population that is male and

between the ages of 15 and 44. U.S. Census Bureau, Population Division:
County estimates by demographic characteristics - age, sex, race, and Hispanic
Origin. http://www.census.gov/popest/datasets.html (Viewed 7/15/08) U.S.
Census Bureau, Population Division:
adult female is the percentage of the county population that is female

and between the ages of 15 and 44. U.S. Census Bureau, Population Division:
County estimates by demographic characteristics - age, sex, race, and Hispanic
Origin. http://www.census.gov/popest/datasets.html (Viewed 7/15/08)
male youth is thepercentageof thecountypopulationthat ismaleandbe-

tween the ages of 5 to 14 (1997 - 1999) or 5 to 13 (2000 - 2005) U.S. Census Bu-
reau, Population Division: County estimates by demographic characteristics -
age, sex, race, andHispanicOrigin. http://www.census.gov/popest/datasets.html
(Viewed 7/15/08)
female youth is thepercentageof thecountypopulationthat is femaleand

between the ages of 5 to 14 (1997 - 1999) or 5 to 13 (2000 - 2005). U.S. Census
Bureau, Population Division: County estimates by demographic characteris-
tics -age, sex, race, andHispanicOrigin. http://www.census.gov/popest/datasets.html
(Viewed 7/15/08)
black is thepercentageof countypopulationthat isblack. U.S.CensusBu-

reau, Population Division: County estimates by demographic characteristics -
age, sex, race, andHispanicOrigin. http://www.census.gov/popest/datasets.html
(Viewed 7/15/08)
hispanic is the percentage of the county population that is Hispanic. U.S.

Census Bureau, Population Division: County estimates by demographic char-
acteristics -age, sex, race, andHispanicOrigin. http://www.census.gov/popest/datasets.html
(Viewed 7/15/08)
msa is the Metropolitan Statistical Area reported by the Population Divi-

sion, U.S.CensusBureau in1999andupdated in2003: http://www.census.gov/population/estimate
city/99m�ps.txt (Viewed 1/29/2007).
population density is calculated from the county population in U.S. Cen-

sus Bureau, Population Division: County estimates by demographic charac-
teristics - age, sex, race, and Hispanic Origin, (Viewed 7/15/08) divided by the
land area in square miles from the U.S. Census Bureau (Viewed 1/29/2007).
hate group total is the total number of hate group chapters active in

a county. This variable is the sum of all Ku Klux Klan, Neo-Nazi, Racist
Skinheads, and Christian Identity Chapters active at the county level. These
data are reported annually at the city level by the Southern Poverty Law
Center in their quarterly publication the Intelligence Report. Each annual
report reports activity for the previous year.
- 1998. "474 Hate Groups Blanket America." The Southern Poverty Law

20



Center. Issue 89.
- 1999. "Hate Groups Top 500." The Southern Poverty Law Center. Issue

93.
- 2000. "The Decade in Review." The Southern Poverty Law Center. Issue

97.
- 2001. "Blood on the Border." The Southern Poverty Law Center. Issue

101
- 2002. "The Year in Hate." The Southern Poverty Law Center. Issue 105.
- 2003. "Hate Takes a Hit." The Southern Poverty Law Center. Issue 109.
- 2004. "Age of Rage." The Southern Poverty Law Center. Issue 114.
- 2005. "Holy War". The Southern Poverty Law Center. Issue 117
- 2006. "The Year in Hate". The Southern Poverty Law Center. Issue 121.
- 2007. "The Year in Hate". The Southern Poverty Law Center. Issue 125.
- 2008. "The Year in Hate". The Southern Poverty Law Center. Issue 129.
kkk is the total number of Ku Klux Klan Chapters active in a county.

These data are reported annually at the city level by the Southern Poverty
Law Center in their quarterly publication the Intelligence Report, Issues 85,
89, 93, 97, 101, 105, 109, 114, 117, 121, 125, 129.
nazi is the number of Neo-Nazi chapters active in county. These data are

reported annually at the city level by the Southern Poverty Law Center in
their quarterly publication the Intelligence Report, Issues 85, 89, 93, 97, 101,
105, 109, 114, 117, 121, 125, 129.
skinheads is the number of racists skinhead chapters active in a county.

These data are reported annually at the city level by the Southern Poverty
Law Center in their quarterly publication the Intelligence Report, Issues 85,
89, 93, 97, 101, 105, 109, 114, 117, 121, 125, 129.
identity is the number of Christian Identity Churches active in a county

.These data are reported annually at the city level by the Southern Poverty
Law Center in their quarterly publication the Intelligence Report, Issues 85,
89, 93, 97, 101, 105, 109, 114, 117, 121, 125, 129.
consumer price index is from the United States Department of Labor:

Bureau of Labor Statistics: Consumer Price Index. Washington, D.C. ftp://
ftp.bls.gov/pub/special.requests/cpi/cpiai.txt (viewed 7/10/08)
hate crime data come from the United States Department of Justice.

Federal Bureau of Investigation. Uniform Crime Reporting Program Data
[UnitedStates]: HateCrimeData, 1997-2007. AnnArbor, MI: Inter-university
Consortium for Political and Social Research.

21



Figure 1: Figure 1: Number of Hate Crimes and Number of Ku Klux Klan,
Neo-Nazi, Skinhead, and Christian Identity Chapters in the US: 1997 - 2007

22



Figure 2: Maximum Hate Crime Rate in a Given Year: 1997 - 2007

2
3



Figure 3: Maximum Annual Number of Active Hate Group Chapters Present In Between 1997 and 2007

2
4



Figure 4: Change in Hate Group Presence From 1997 Through 2007

2
5



Table 1: Hate Group Characteristics

Movement Ku Klux Klan neo-Nazi Racist Skinheads Christian Identity
Year Established 1866 1913 1970s 1940s

Estimated Membership 6,000 6,000 20,000-50,000

Prohibitions/Sacri�ce

Clothing
Scar
Race
Ethnic
Religious
Sexual

No
No
Yes
Yes
Yes
Yes

Yes
Yes
Yes
Yes
Yes
Yes

Yes
Yes
Yes
Yes
Yes
Yes

No
No
Yes
Yes
Yes
Yes

Membership Bene�ts Mutual Insurance
Mutual Insurance
Educational Services

Safety

Mutual Insurance
Safety

Mutual Insurance
Educational Services

Safety
Sources: Anti-Defamation Leage, Southern Poverty Law Center, Kleg (1993)

2
6



Table 2: Summary Statistics

Mean Standard Deviation Maximum Minimum

Number of Hate Crimes 2.877 18.261 925 0

Hate Crime Rate (10000) 0.16 0.58 37.443 0

Hate Crime Rate Excluding Anti-White (10000) 0.132 0.434 19.231 0

Hate Crime Rate Comitted by Whites Excluding Anti-White (10000) 0.066 0.241 8.251 0

Hate Crime Rate Comitted by Non-Whites (10000) 0.085 0.39 19.231 0

Anti-White Hate Crime Rate by Non-Whites 0.018 0.173 11.872 0

Hate Group Present (Yes=1) 0.091 0.288 1 0

Real Median Household Income ($2006) 41379.68 10442.02 106782.4 17808.63

Unemployment Rate 5.226 2.145 33.2 0.7

Bene�ts per Unemployed 4.36 2.174 32.402 0

Percent in Poverty 14.188 5.752 55.9 1.7

Percent Black 9.438 14.711 86.127 0

Percent Hispanic 6.869 12.611 104.657 0

Percent White 86.445 15.908 157.395 4.786

Population Density 268.643 1792.164 70786.2 0.043

Percentage Male Age 15-44 20.765 3.359 55.507 10.66

Crime Rate (10000) 272.326 171.97 3888.889 0.405

N 30621

2
7



Table 3: Hate Crime Rate 1997 - 2007: All Counties and Only Those Reporting At Least One Hate Crime

All Counties Counties Reporting At Least One Hate Crime

Fixed E¤ects Dynamic Panel System GMM Fixed E¤ects Dynamic Panel System GMM

Hate Group Present(Y e � =1) 0.029 * * 0.028 * 0.034 * * 0.008

[0.013] [0.014] [0.016] [0.016]

Hate Crime Ratet�1 0.435 * * * 0.418 * * *

[0.124] [0.113]

Income ($1000) 0.000 0.000 * * * 0.000 0.000

[0.000] [0.000] [0.000] [0.000]

Unemployment Rate 0.001 -0.002 0.001 -0.008 * *

[0.003] [0.002] [0.005] [0.004]

Bene�ts per Unemployed -0.001 0.004 -0.002 0.005

[0.004] [0.003] [0.007] [0.004]

Percent in Poverty -0.002 0.001 -0.004 0.003

[0.002] [0.002] [0.005] [0.003]

Percent Black 0.001 -0.000 0.003 -0.002 *

[0.004] [0.001] [0.007] [0.001]

Percent Hispanic 0.005 * * -0.000 0.007 * -0.001

[0.002] [0.001] [0.003] [0.001]

Population Density -0.000 0.000 * * * -0.000 0.000 * * *

[0.000] [0.000] [0.000] [0.000]

Crime Rate 0.00 * * * -0.000 0.001 * * * 0.000

[0.000] [0.000] [0.000] [0.00]

Adj. R-squared 0.003 0.006

R-squared Within 0.004 0.007

Correlation between u i and X b -0.776 -0.824

Rho 0.458 0.471

Hansen J statistic 13.552 14.302

Hansen p-value 0.632 0.576

AR(2) test statistic 1.628 1.617

AR(2) p-value 0.104 0.106

Di¤erence-in-Hansen GMM 0.741 0.826

No. of Instruments 38 38

No. of groups 3116 3115 1931 1931

No. of Observations 33948 30872 21200 19271

Standard errors in brackets. All Estimations Include Year Dummies.

Columns 1 and 3 are �xed e¤ects regressions with standard errors clustered by county.

Columns 2 and 4 are two-step system GMM with Windmeijer (2005) corrected standard errors.

The collapsed instruments used in di¤erenced equations of the GMM estimation are: Hate Crime Ratet�r ; Crime Ratet� r for r=3,4, ...8

The collapsed instruments used in level equations of the GMM estimation are: � Hate Crime Ratet�2 ; � Crime Ratet�2
Signi�cant at * 10 % , * * 5 % , * * * 1%

1



Table 4: Hate Crime Rate Excluding Anti-White and Hate Crime Rate with White Perpetrator with Non-White Victim

Hate Crime Excluding Anti-White White Perpetrator with Non-White Victim

Fixed E¤ects Dynamic Panel System GMM Fixed E¤ects Dynamic Panel System GMM

Hate Group Present (Yes=1) 0.025 � � 0.030� � � 0.016 � 0.014 �

[0.011] [0.010] [0.009] [0.008]

Excluding Anti-White Ratet�1 0.196

[0.140]

White Perpetrator with -0.182

Non-White Victim Ratet�1 [0.234]

Income ($2006) 0.000 � 0.000� � � 0.000 0.000 � � �
[0.000] [0.000] [0.000] [0.000]

Unemployment Rate 0.001 -0.001 -0.000 -0.000

[0.0020] [0.002] [0.001] [0.001]

Bene�ts per Unemployed ($2006) -0.001 0.005� � -0.001 0.004 � � �

[0.003] [0.002] [0.002] [0.002]

Percent in Poverty -0.002 0.003� -0.001 0.001

[0.002] [0.002] [0.001] [0.001]

Percent Black 0.002 -0.001 0.001 -0.001 � � �

[0.002] [0.000] [0.001] [0.00]

Percent Hispanic 0.004 � � -0.001 � 0.001 -0.000

[0.002] [0.000] [0.001] [0.000]

Population Density -0.000 0.000� � � -0.000 � 0.000 �

[0.000] [0.000] [0.000] [0.000]

Crime Rate (10000) 0.000 � � � 0.000 0.000 � � � 0.000

[0.000] [0.000] [0.000] [0.000]

Adj. R-squared 0.005 0.003

R-squared Within 0.005 0.003

Correlation between ui and � b -0.839 -0.710

Rho 0.557 0.359

Hansen J statistic 16.499 20.251

Hansen p-value 0.419 0.209

AR(2) test statistic 0.964 -0.552

AR(2) p-value 0.335 0.581

Di¤erence-in-Hansen GMM 0.787 0.509

No. of Instruments 38 38

No. of groups 3116 3115 3116 3115

No. of Observations 33948 30872 33948 30872

Standard errors in brackets. All Estimations Include Year Dummies.

Columns 1 and 3 are �xed e¤ects regressions with standard errors clustered by county.

Columns 2 and 4 are two-step system GMM with Windmeijer (2005) corrected standard errors.

The collapsed instruments used in di¤erenced equations of the GMM estimation are: Hate Crime Rate Typet�4,

Hate Crime Rate Typet�5 , ...Hate Crime Rate Type1997;Crime Ratet�1,Crime Ratet�2,...Crime Rate1997.

The collapsed instruments used in level equations of the GMM estimation are: � Hate Crime Rate Typet � � Crime Ratet�4
Signi�cant at � 10 � , � � 5 � , � � � 1�

2



Table 5: Hate Crimes Committed by Non-Whites and Hate Crimes Commotted by Non-Whites Listed as Anti-White

Hate Crimes with Non-White Perpetrators Anti-White Hate Crimes with Non-White Perpetrators

Fixed E¤ects Dynamic Panel System GMM Fixed E¤ects Dynamic Panel System GMM

Hate Group Present (Yes=1) 0.011 0.017 � 0.002 0.000

[0.007] [0.009] [0.004] [0.004]

Hate Crime Rate with 0.184

Non-White Perpetratort�1 [0.217]

Anti-White Hate Crime Rate 0.237

with Non-White Perpetratort�1 [0.333]

Income ($2006) 0.000 0.000� � � -0.000 0.000�

[0.000] [0.000] [0.000] [0.000]

Unemployment Rate 0.001 -0.002 -0.001 -0.001

[0.002] [0.001] [0.001] [0.000]

Bene�ts per Unemployed ($2006) 0.000 0.004 � -0.000 0.001

[0.003] [0.0020] [0.001] [0.000]

Percent in Poverty -0.000 0.003 � � 0.001 0.000

[0.002] [0.001] [0.001] [0.000]

Percent Black 0.002 0.000 0.000 -0.000

[0.003] [0.000] [0.001] [0.000]

Percent Hispanic 0.003 � � -0.000 � 0.000 -0.000 � � �

[0.001] [0.000] [0.001] [0.000]

Percent White -0.001 -0.000 -0.000 -0.000

[0.002] [0.000] [0.001] [0.000]

Population Density -0.000 0.000 � � � 0.000 0.000

[0.000] [0.000] [0.000] [0.000]

Percent Male Age 15-44 0.005 0.002 � � 0.002 0.000

[0.004] [0.001] [0.002] [0.000]

Crime Rate (10000) 0.000 � � � 0.000 0.000� 0.000

[0.000] [0.000] [0.000] [0.000]

Adj. R-squared 0.002 0.000

R-squared Within 0.003 0.001

Correlation between ui and � b -0.769 -0.207

Rho 0.444 0.195

Hansen J statistic 20.190 18.170

Hansen p-value 0.265 0.378

AR(2) test statistic 0.044 0.082

AR(2) p-value 0.965 0.935

Di¤erence-in-Hansen GMM 0.157 0.776

No. of Instruments 39 39

No. of groups 3116 3115 3116 3115

No. of Observations 33948 30872 33948 30872

Standard errors in brackets. All Estimations Include Year Dummies.

Columns 1 and 3 are �xed e¤ects regressions with standard errors clustered by county.

Columns 2 and 4 are two-step system GMM with Windmeijer (2005) corrected standard errors.

The collapsed instruments used in di¤erenced equations of the GMM estimation are: Hate Crime Rate Typet�3,

Hate Crime Rate Typet�4 , ...Hate Crime Rate Type1997;Crime Ratet�1,Crime Ratet�2,...Crime Rate1997.

The collapsed instruments used in level equations of the GMM estimation are: 	 Hate Crime Rate Typet 
 	 Crime Ratet�4
Signi�cant at � 10 � , � � 5 � , � � � 1 �

3



Table 6: Expected Future Hate Crimes Committed by Non-Whites and Hate Group Activity

� Hate Crime Rate by � Hate Group Presencet
Non-White Perpetratort

� Hate Crime Rate by -0.576 
 
 
 0.003

Non-White Perpetratort�1 [0.071] [0.052]

� Hate Crime Rate by -0.317 
 
 
 0.037

Non-White Perpetratort�2 [0.067] [0.050]

� Hate Group Presencet�1 0.018 -3.456 
 
 

[0.014] [0.075]

� Hate Group Presencet�2 -0.003 -2.455 
 
 

[0.010] [0.109]

Adj. R-squared 0.284

Psuedo-Rsquared 0.156

No. of Observations 24662 24662

All Estimations include, � incomet, � unemployment ratet, � unemployment rate bene�tst, � poverty ratet, � percentage blackt,

� percentage Hispanict, � percentage whitet, � population densityt, � percentage male 15-44t, � crime ratet, and year dummies.

Column 1 is ordinary least squares. Column 2 is an ordered logit.

Robust standard errors in brackets.

Signi�cant at 
 10 � , 
 
 5 � , 
 
 
 1 � .

4



Table 7: Expected Future Hate Crimes Committed by Non-Whites Listed as Anti-White and Hate Group Activity

� Anti-White Hate Crime Rate � Hate Group Presencet
by Non-White Perpetratort

� Anti-White Hate Crime Rate -0.594 � � � 0.083

by Non-White Perpetratort�1 [0.118] [0.126]

� Anti-White Hate Crime Rate -0.331 � � � -0.030

by Non-White Perpetratort�2 [0.101] [0.096]

� Hate Group Presencet�1 0.010 -3.455 � � �

[0.009] [0.075]

� Hate Group Presencet�2 -0.001 -2.456 � � �

[0.006] [0.109]

Adj. R-squared 0.296

Psuedo-Rsquared 0.156

No. of Observations 24662 24662

All Estimations include, � incomet, � unemployment ratet, � unemployment rate bene�tst, � poverty ratet, � percentage blackt,

� percentage Hispanict, � percentage whitet, � population densityt, � percentage male 15-44t, � crime ratet, and year dummies.

Column 1 is ordinary least squares. Column 2 is an ordered logit.

Robust standard errors in brackets.

Signi�cant at � 10 � , � � 5 � , � � � 1 � .

5