1   Hate Groups and Hate Crime Matt E. Ryan† Peter T. Leeson‡ Abstract This paper is the first to empirically investigate the relationship between hate groups and hate crime. We do so using panel data for the U.S. states between 2002 and 2008. Contrary to conventional wisdom, we find that hate groups have no influence on hate crime in the United States. Instead we find that economic hardship does. Unemployment and, to a lesser extent, poverty, are strongly associated with more hate crime, particularly crimes that are sexually, racially, and religiously motivated. Demographic variables aren’t significant determinants of hate crime in the United States. Taken together our results support the “frustration-aggregation thesis.” They suggest that economics, not hate groups or demographics, drive hate crime in America.                                                              † Email: RyanM5@DUQ.edu. Address: Duquesne University, Department of Economics and Quantitative Sciences, 802 Rockwell Hall, Pittsburgh, PA 15282. ‡ Email: PLeeson@GMU.edu. Address: George Mason University, Department of Economics, MS 3G4, Fairfax, VA 22030. 2   1 Introduction Hate groups are organizations of individuals whose “beliefs or practices . . . attack or malign an entire class of people, typically for their immutable characteristics,” such as race or sexual orientation, but sometimes for their mutable ones, such as religious beliefs (Southern Poverty Law Center 2010). They include organizations such as the Ku Klux Klan, neo-Nazi groups, white nationalist groups, neo-Confederate groups, and black separatist groups. Hate crimes are “crimes that manifest evidence of prejudice based on race, religion, disability, sexual orientation, or ethnicity” (DOJ and FBI 2004). The potential connection between hate groups and hate crime is obvious. Yet no one has explored it empirically.1 Indeed, very little work in economics has explored hate crime at all. Dharmapala and Garoupa (2004) examine various aspects of hate crime policy. But their analysis is theoretical and focuses on how the law influences hate crime. Medoff (1999) considers hate crime’s determinants empirically.2 But his study is based on a single cross-section meant to append a theory of hateful behavior and doesn’t consider hate groups. Gale, Heath and Ressler’s (2002) important paper, which is closest to ours, takes a more serious empirical approach to investigating hate crime’s determinants. But it, too, neglects the potentially important influence of hate groups. Our paper is a first step toward filling this lacuna. It empirically investigates the relationship between hate groups and hate crime using panel data on the U.S. states between 2002 and 2008. American hate groups grew significantly over the past decade. Between 2002 and 2008 the number of hate groups per capita increased 25 percent. But American hate crime didn’t. Between                                                              1 Two papers consider the determinants of hate groups (Jefferson and Pryor 1999 and Mulholland 2010). 2 Glaeser (2005) presents a more general theory of the political economy of hatred. 3   2002 and 2008 the number of hate crimes per capita actually decreased 1.3 percent. This stylized fact foreshadows a key result of our econometric analysis: contrary to conventional wisdom, the rise of American hate groups over the past decade hasn’t contributed to a rise in hate crime. In fact, in terms of hate crime, hate groups haven’t mattered at all. There are lots of reasons to loath hate groups. But their influence on hate crime isn’t one of them. This study considers two other types of potential determinants of hate crime in America: economic determinants, such as the extent of unemployment and poverty, and demographic ones, such as race and urbanism. In examining economic and demographic determinants of hate crime, this paper follows the small, existing literature that addresses hate crime empirically. That literature’s focus on economic variables derives from Dollard et al.’s (1939) “frustration- aggregation thesis.” According to that thesis, when people endure economic hardship they get frustrated. They take their frustration out on vulnerable social groups, such as ethnic, sexual, and religious minorities. Existing empirical support for the frustration-aggregation thesis is mixed. In their paper that examined the American South in the late 19th and early 20th centuries, Hovland and Sears (1940) found a strong relationship between lynchings of blacks and poor economic conditions. However, subsequent work showed their results to be very fragile (see Mintz 1946; Hepworth and West 1988; Beck and Tolnay 1990; Olzak 1990; Tolnay and Beck 1995; and Green, Glaser and Rich 1998). Krueger and Pischke (1997) find no relationship between economic conditions and racially motivated crime against foreigners in early 1990s Germany. However, Green and Rich (1998) find weak links between unemployment and assorted hate crimes in North Carolina between 1987 and 1993. Green, Glaser and Rich (1998) find no consistent relationship between the 4   unemployment rate and a range of racially motivated crimes in New York City between 1987 and 1995. But Gale, Heath and Ressler (2002) find evidence of a relationship between economic factors and hate crime in the United States. Our analysis finds that unemployment and, to a lesser extent, poverty, is strongly associated with more hate crime in America, particularly crimes that are sexually, racially, and religiously motivated. Unlike previous studies, ours disaggregates hate crime data. In doing so we find evidence for the frustration-aggregation thesis that previous work overlooks. The existing literature’s focus on potential demographic determinants of hate crime derives from the observation that crime tends to be higher in urban areas and that potential conflicts leading to, and opportunities for, hate crime are greater in areas that have a higher concentration of socially vulnerable groups, such as racial, sexual, and religious minorities. Gale, Heath and Ressler (2002) find that demographic determinants aren’t strong predictors of hate crime in America. Our disaggregated study confirms the unimportance of demographic variables. Take together our results support the frustration-aggregation thesis. They suggest that economics, not hate groups or demographics, drive hate crime in America. 2 Data We collect data on hate crime from the Federal Bureau of Investigation’s (FBI) Uniform Crime Reporting (UCR) Program’s Hate Crime Statistics (HCS) Program. The Hate Crime Statistics Act of 1990 brought these data into existence.3 That Act requires the Attorney General to collect annual data on “crimes that manifest evidence of prejudice based on race, religion, disability, sexual orientation, or ethnicity, including where appropriate the crimes of murder, non-negligent                                                              3 Hate Crime Statistics Act, 28 USC § 534, 1990. 5   manslaughter; forcible rape; aggravated assault, simple assault, intimidation; arson; and destruction, damage or vandalism of property” (DOJ and FBI 2004). The Department of Justice made these data available at the state level beginning in 1995. State hate-crime reporting to the HCS is voluntary. Within reporting states, local jurisdictional reporting is incomplete and varies by year. Because of this, hate-crime reporting is imperfect. Still, it’s very good. In our sample, which covers forty-nine U.S. states (Hawaii didn’t report hate crime activity) and the District of Columba from 2002 and 2008, the percentage of the American population covered by hate-crime reporting to the HCS ranges from 82.9 percent in the lowest-covered year to 88.6 percent in the highest-covered year.4 We return to the issue of imperfect hate-crime reporting and how we address this issue in our empirical model below. We collect data on the number of hate groups in each state between 2002 and 2008 from the Southern Poverty Law Center (SPLC). The SPLC provides these data in an annual report called “The Year in Hate.” The SPLC separates hate groups into the following categories: Ku Klux Klan, Neo-Nazi, White Nationalist, Racist Skinhead, Christian Identity, Neo-Confederate, Black Separatist, and General Hate. They also report the number of patriot groups in each state. Our empirical model uses three sets of key regressors. The first is our measure of hate crime, discussed above. This variable is unique to our analysis. It allows us to investigate the relationship between organizations such as the KKK, neo-Nazis, and so on, and hate crime. The other two sets of regressors focus on the hate crime determinants that previous studies of those determinants have examined. One of these is a set of economic variables. These variables allow us to investigate the frustration-aggregation thesis. They include the state unemployment rate for persons 16 years old and older (Unemployment), the percentage of the state’s population that                                                              4 Data are missing for five state-years in our sample: Alabama 2005; Arkansas 2002; Mississippi 2005, 2006, 2007. 6   lives in poverty as defined by the U.S. Census Bureau (Poverty), and gross state product per capita (GSPpc). The other set of regressors we consider includes demographic variables. These are the percentage of the state’s population living in a metropolitan area (Metro); the percentage of the state’s population that’s African-American (Black); the percentage of the state’s population that’s Jewish (Jewish); and percentage of children under 18 years old in the state who were victims of abuse or neglect (Abuse). Finally, though it’s not a demographic variable, we control for the percentage of state and local government spending devoted to police protection (Police). These regressors are same ones Gale, Heath, and Ressler (2002) use in their study of hate crime’s determinants.5 We collect data for these variables for each state and year in our sample. We get data for each of them from the Statistical Abstract of the United States. Table 1 provides summary statistics for all our variables. To account for the measurement problem that imperfectly reported hate crimes create, our empirical model also includes a variable that controls for the percentage of the population covered by crime-reporting statistics in each state in each year (PopShare). We get these data from the HCS reports. As noted above, hate-crime reporting covers a larger percentage of the population in some states and years than others. PopShare accounts for this.                                                              5 We omit one of their controls—a constructed ratio between white and black income—because it’s unavailable as a state-level annual metric. 7   3 Empirical Approach To identify the potential relationship between hate groups, economic determinants, demographic determinants, and hate crime, we estimate the following two-way fixed effects model with standard errors clustered by state:6 HateCrimei,t = α + β1HateGroupsi,t + β2Economici,t + β3Demographici,t + β4PopSharei,t + ϕi + ζt + εi,t (1) HateCrimei,t is the number of hate crimes per capita in state i in year t. HateGroupsi,t is the number of hate groups per capita in state i in year t.7 Economici,t is a matrix of economic variables discussed above. Demographici,t is a matrix of demographic variables discussed above. PopSharei,t is the percentage of the population covered by HCS hate-crime reporting in state i in year t. ϕi is a vector of comprehensive state-specific fixed effects. It controls for any time-invariant differences across states that might affect hate crime, such as a long-standing culture of certain kinds of prejudices or tolerance. ζt is a vector of comprehensive year-specific fixed effects. It controls for any features that are common across states but vary across time that might affect hate crime, such as the national population’s changing attitudes toward various racial groups, religious groups, and changing views about homosexuality. εi,t is a random error term. To see if different kinds of hate groups and our economic and demographic determinants are associated differently with different kinds of hate crimes, in addition to considering an aggregate measure of total hate crime, we disaggregate our HateGroups and HateCrime variables. We                                                              6 The Hausman test confirms the superiority of a fixed-effects model over a random-effects one. 7 Following Gale, Heath and Ressler (2002), the population figure we use to generate per capita measures is the population covered by hate-crime reporting. 8   consider a version of our HateGroups variable that measures all hate groups including patriot groups, a version that measures all hate groups excluding patriot groups, and a version that measures just the number Ku Klux Klan, Neo-Nazi, White Nationalist and Racist Skinhead groups. We consider a version of our HateCrime variable that measures the total number of hate crimes per capita, as well as versions that separately measure hate crimes based only on race, those based only religion, and so on for sexual orientation, ethnicity, and disability. One potential concern with our empirical model is that hate crime may be endogenous. Hate groups may influence hate crime’s prevalence. But hate crime’s prevalence may also influence the formation of hate groups. For example, hate groups may find it easier to form where hate crime is more common and thus there is a larger base of potential members to recruit. The potential for endogeneity is important in principle. But it turns out to be unimportant in practice. As we discuss below, hate groups and hate crime are uncorrelated: there’s no statistically relevant relationship between them at all. The absence of such a relationship precludes the possibility of endogeneity bias. 4 Results Tables 2, 3 and 4 present our estimates when HateGroups measures all hate groups including patriot groups, all hate groups excluding patriot groups, and the total number of Ku Klux Klan, Neo-Nazi, White Nationalist and Racist Skinhead groups respectively. 9   4.1 Hate Groups Column 1 presents hate groups’ relationship to total, aggregate hate crime. HateGroups has the expected sign in all three tables. But none of these regressions show a statistically (or economically) meaningful relationship to hate crime. The other columns in Tables 2-4 break aggregate hate crime activity into component crimes. Hate groups still don’t matter. This is true for every individual kind of hate crime. This (non-)result contradicts conventional wisdom, according to which hate groups matter for hate crime. For example, Levin (2007) argues that hate groups provide the “situational facilitator” needed to translate a latent, widespread dislike of minority groups into violence. Jenness et al. (2000) postulate that white supremacist movements encourage people to commit hate crimes. Levin and McDevitt (1993: 103) claim that hate groups actively foster aggressive behavior towards innocents and that, even when their members don’t themselves commit hate crimes, they persuade others to, who “are . . . inspired by the presence of such groups.” Berrill (1992: 31) sees hate groups as playing a similar role, providing potential offenders with “encourage[ment] by . . . rhetoric.” These scholars may be right. Hate groups may be interested in committing hate crimes and encouraging others to commit them. But, if they are, they’re exceptionally bad at doing so. Alternatively, hate groups, though populated by hateful people, may be a lot of hateful bluster. Hate group members may say they’re interested in harming the minorities. But, in practice, they may not actually commit hate crimes or convince others to. Either way our (non-)result is a happy one from the perspective of those who are concerned about the presence and growth of hate groups in the U.S. over the last decade. There are many reasons to be concerned about hate groups’ growth apart from their effect on hate crime. But the 10   most immediate and direct danger hate groups might pose—increasing hate crime’s prevalence—doesn’t seem to be a danger at all. 4.2 Economic Determinants Our model finds that hate groups don’t matter for hate crime. But it’s clearly capable of delivering strong, positive results. Consider the set of economic determinants we estimate in Tables 2-4: Unemployment and Poverty. Column 1 provides the closest match to the specification previous research uses to identify hate crime’s determinants. This column uses total hate crime as the regressand. Unemployment is an important determinant of hate crime. Higher unemployment is positively and significantly associated with more hate crime in all three tables. According to our estimates, a 1 percentage-point increase in a state’s unemployment rate increases the number of hate crimes in that state by between 3.66 to 3.86 crimes per million residents, depending on the specification one considers. This effect is large. In 2008 a one standard deviation increase in unemployment at the mean population value is associated with nearly 28 additional hate crimes over the course of the year. That represents an 18 percent increase in the number of hate crimes from the mean. Unemployment’s relationship to specific kinds of hate crime displays a pattern. Examining columns 4 and 5 in each table, it’s clear that unemployment’s positive association with hate crime overall stems from its positive association with hate crime relating to sexual orientation and ethnicity. When unemployment is higher, hate crimes against homosexuals and ethnic minorities are higher too and vice versa. 11   Our other measure of economic hardship—poverty—displays a positive relationship with hate crime as well. But one must disaggregate hate crime to find it. The hate crimes poverty is associated with are specific: those relating to religion. Our results indicate that where more of the population is impoverished, religiously motivated hate crime is more common and vice versa. Religiously motivated hate crimes aren’t a large enough fraction of total hate crime to make poverty a significant determinant of hate crime overall. However, at the disaggregated level, it’s clear that more poverty has an important and positive association with more religiously motivated hate crime. Average income is positively associated with religiously (see column 3 in Tables 2-4) and sexually (see column 4 in Tables 2-4) oriented hate crime. However, this isn’t a strike against the frustration-aggregation thesis. Lower income is associated with economic hardship. That suggests we should find a negative relationship between incomes and hate crime. But this overlooks the fact that many hate crimes are “crimes-for-profit” that happen to be directed at the members of a socially vulnerable group because of the perpetrator’s hatred. If this is true, we should expect to find more hate crime in richer states, not poorer ones. This may help explain the positive relationship we find between GSP per capita and hate crime. 4.3 Demographic Determinants Finally, consider our results for potential demographic determinants of hate crime, measured by the Metro, Black, Jewish, and Abuse variables in Tables 2-4. Consistent with existing research, we find that, in general, these variables don’t matter. None of them has a significant relationship to the overall level of hate crime, regardless of the specification we consider. Only two consistently produce significant estimates when we disaggregate hate crime. The percentage of 12   the population living in a metropolitan area is negatively related to religiously motivated hate crime: more rural areas have less religious hate crime. And the percentage of the population that’s black is negatively related to sexually motivated hate crime: states with higher concentrations of blacks have fewer hate crimes against homosexuals. Our results that consider the relationship between state and local government police funding and hate crime suggest that more police funding isn’t associated with less hate crime overall. With one exception it isn’t associated with less hate crime of any specific kind either. The single exception is for hate crime relating to sexual orientation. However, the coefficient on our police funding variable is positive rather than negative. Where police funding is higher, sexually motivated hate crime is too. We don’t interpret this as evidence that police funding leads to sexually motivated hate crime. Rather, it may reflect that police activity is especially concerned with sexually motivated hate crime. Thus, where such crime is higher, so is police funding. 5 Concluding Remarks This paper is first to explore the relationship between hate groups and hate crime empirically. Contrary to conventional wisdom, our analysis finds that hate groups have no influence on hate crime in the United States. Instead we find that economic hardship does. Our results suggest that unemployment and, to a lesser extent, poverty, are strongly associated with more hate crime, particularly crimes that are sexually, racially, and religiously motivated. Consistent with existing work, our findings suggest that demographic variables aren’t important predictors of hate crime in the United States. Taken together our results support the frustration-aggregation thesis. They suggest that economics, not hate groups or demographic factors, drive hate crime in America. 13   References Beck, E. M. and Stewart E. Tolnay. 1990. “The Killing Fields of the Deep South: The Market for Cotton and the Lynching of Blacks, 1882-1930.” American Sociological Review 55: 526-39. Berrill, Kevin T. 1992. “Anti-Gay Violence and Victimization in the United States: An Overview.” In Gregory M. Herek and Kevin T. Berrill (eds.), Hate Crimes: Confronting Violence against Lesbians and Gay Men. Newbury Park, CA: Sage. Department of Justice and Federal Bureau of Investigation. 2004. “Crime in the United States 2004.” Available at:  http://www.fbi.gov/ucr/cius_04/offenses_reported/hate_crime/index.html. Last accessed: 27 July 2010. Dharmapala, Dhammika and Nuno Garoupa. 2004. “Penalty Enhancement for Hate Crimes: An Economic Analysis.” American Law and Economics Review 6: 185-207. Dollard, J., L. Doob, N. Miller, O.H. Mowrer, and R.R. Sears. 1939. Frustration and Aggregation. New Haven: Yale Press. Gale, Lewis R., Will C. Heath, and Rand W. Ressler. 2002. “An Economic Analysis of Hate Crime.” Eastern Economic Journal 28: 203-15. Glaeser, Edward L. 2005. “The Political Economy of Hatred.” Quarterly Journal of Economics 120: 45-86. Green, Donald P., Jack Glaser, and Andrew Rich. 1998. “From Lynching to Gay Bashing: The Elusive Connection between Economic conditions and Hate Crime.” Journal of Personality and Social Psychology 75: 82-92. Green, Donald P. and Andrew Rich. 1998. “White Supremacist Activity and Crossburnings in North Carolina.” Journal of Quantitative Criminology 14: 263-82. Hepworth, Joseph T. and Stephen G. West. 1988. “Lynchings and the Economy: A Time-Series Reanalysis of Hovland and Sears (1940).” Journal of Personality and Social Psychology 21: 296-301. Hovland, C. I. and R. R. Sears. 1940. “Minor Studies of Aggression: VI. Correlation of Lynchings with Economic Indices.” Journal of Psychology: 301-310. Jefferson, Philip N. and Federic L. Pryor. 1999. “On the Geography of Hate.” Economic Letters. 65: 389-395. Jenness, Valerie, Abby L. Ferber, Ryken Grattet, and James F. Short, Jr. 2000. Hate crime in America: What do we know? Washington, D.C.: American Sociological Association. 14   Krueger, Alan B. and Jorn-Steffen Pischke. 1997. “A Statistical Analysis of Crime against Foreigners in Unified Germany.” Journal of Human Resources 32: 182-209. Levin, Jack and Jack McDevitt. 1993. Hate Crimes: The Rising Tide of Bigotry and Bloodshed. New York: Plenum Press. Levin, Jack. 2007. The Violence of Hate. Boston: Allyn and Bacon. Medoff, 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. Mintz, Alexander. 1946. “A Re-examination of Correlations between Lynchings and Economic Indices.” Journal of Abnormal and Social Psychology 41: 154-160. Mulholland, Sean E. 2010. “Hate Fuel: On the Relationship between Local Government Policy and Hate Group Activity.” Mimeo. Olzak, Susan. 1990. “The Political Context of Competition: Lynching and Urban Racial Violence, 1882-1914.” Social Forces 69: 395-421. Southern Poverty Law Center. 2010. “Hate Map.” Available at: http://www.splcenter.org/get- informed/hate-map. Last Accessed: 27 July 2010. Tolnay, Stewart E. and E. M. Beck. 1995. A Festival of Violence: An Analysis of Southern Lynchings, 1882-1930. Urbana: University of Illinois Press. 15     Variable Definition Mean S.D. HateCrime Total crimes attributed to all bias motivations per 1,000,000  covered population, by state 30.82 21.73 Total crimes attributed to race bias per 1,000,000 covered  population, by state 16.36 13.03 Total crimes attributed to religion bias per 1,000,000 covered  population, by state 4.74 5.17 Total crimes attributed to sexual orientation bias per  1,000,000 covered population, by state 5.86 7.69 Total crimes attributed to ethnicity bias per 1,000,000 covered  population, by state 3.55 3.00 Total crimes attributed to disability bias per 1,000,000  covered population, by state 0.32 0.90 HateGroups Total number of hate groups, including patriot groups, per  1,000,000 residents, by state 3.89 2.76 Total number of hate groups, excluding patriot groups, per  1,000,000 residents, by state 3.14 2.59 Total number of Ku Klux Klan, Neo‐Nazi, White Nationalist  and Racist Skinhead groups per 1,000,000 residents, by state 1.80 1.42 GSPpc Gross State Product per capita, in thousands of 2000 dollars 36.80 13.18 Unemployment State unemployment rate for individuals 16 years and older 5.04 1.16 Poverty Percentage of state population in poverty 12.58 3.24 Police Percentage of state and local government expenditures on  police protection 2.60 0.61 Popshare Percentage of state population covered by HCS reporting 82.38 26.35 Metro Percentage of state population residing within a  metropolitan statistical area (MSA) 85.51 12.84 Abuse Percentage of children under 18 years of age who were  victims of abuse or neglect 1.17 0.67 Black Black population as a percentage of state population 11.29 11.46 Jewish Jewish population as a percentage of state population 1.37 1.70 Data sources:  Hate Crime Statistics, Southern Poverty Law Center, Statistical Abstract of the United States. Table 1. Summary Statistics 16   All Hates Crimes Race Religion Sex. Orient. Ethnicity Disability 1 2 3 4 5 6 HateGroups 1.108 0.665 0.118 0.277 0.047 0.001 (0.870) (0.611) (0.146) (0.245) (0.135) (0.052) GSPpc 0.935 ‐0.304 0.445*** 0.581** 0.142 0.070 (0.877) (0.615) (0.147) (0.247) (0.136) (0.053) Unemployment 3.662** 1.300 0.305 1.539*** 0.481* 0.037 (1.771) (1.242) (0.297) (0.499) (0.274) (0.106) Poverty 1.491 0.603 0.410** 0.462 0.033 ‐0.017 (1.131) (0.794) (0.190) (0.319) (0.175) (0.068) Police 3.623 1.073 ‐0.228 2.491* 0.477 ‐0.188 (4.647) (3.260) (0.778) (1.310) (0.719) (0.279) Popshare ‐0.638*** ‐0.398*** ‐0.016 ‐0.157*** ‐0.058*** ‐0.009* (0.088) (0.062) (0.015) (0.025) (0.014) (0.005) Metro 0.136 0.195 ‐0.092** 0.004 0.034 ‐0.005 (0.230) (0.161) (0.039) (0.065) (0.036) (0.014) Abuse ‐1.039 ‐1.622 ‐0.033 0.848 ‐0.016 ‐0.216 (3.324) (2.332) (0.557) (0.937) (0.515) (0.199) Black ‐2.870 0.749 ‐0.410 ‐4.171*** 0.757 0.203 (3.116) (2.186) (0.522) (0.879) (0.482) (0.187) Jewish 3.225 ‐0.370 0.942 1.679 0.754 0.221 (8.042) (5.642) (1.347) (2.268) (1.245) (0.482) N 292 292 292 292 292 292 R 2 0.24 0.18 0.11 0.32 0.11 0.03 Table 2. Hate Groups' Relationship to Hate Crime: All Hate Groups, Including Patriot Groups Notes: All regressions include year‐ and state‐fixed effects (not reported) and robust standard errors  clustered by state (in parentheses). *** = 1%; ** = 5%; * = 10%.     17   All Hates Crimes Race Religion Sex. Orient. Ethnicity Disability 1 2 3 4 5 6 HateGroups 0.404 0.086 0.160 0.207 ‐0.053 0.004 (0.971) (0.681) (0.162) (0.273) (0.150) (0.058) GSPpc 1.158 ‐0.132 0.441*** 0.612** 0.168 0.069 (0.874) (0.613) (0.146) (0.246) (0.135) (0.052) Unemployment 3.858** 1.434 0.314 1.577*** 0.497* 0.036 (1.769) (1.241) (0.295) (0.498) (0.273) (0.106) Poverty 1.453 0.577 0.408** 0.454 0.030 ‐0.017 (1.134) (0.795) (0.189) (0.319) (0.175) (0.068) Police 3.719 1.134 ‐0.221 2.512* 0.482 ‐0.188 (4.661) (3.268) (0.778) (1.312) (0.719) (0.278) Popshare ‐0.639*** ‐0.400*** ‐0.016 ‐0.156*** ‐0.059*** ‐0.906* (0.088) (0.062) (0.015) (0.025) (0.014) (0.528) Metro 0.142 0.197 ‐0.090** 0.007 0.033 ‐0.005 (0.231) (0.162) (0.039) (0.065) (0.036) (0.014) Abuse ‐0.771 ‐1.408 ‐0.044 0.879 0.019 ‐0.217 (3.335) (2.338) (0.557) (0.939) (0.515) (0.199) Black ‐2.987 0.637 ‐0.391 ‐4.171*** 0.733 0.204 (3.131) ‐2.195 (0.523) (0.882) (0.483) (0.187) Jewish 3.408 ‐0.216 0.928 1.695 0.781 0.220 (52.456) (5.657) (1.346) (2.272) (1.245) (0.482) N 292 292 292 292 292 292 R 2 0.24 0.18 0.11 0.32 0.11 0.03 Table 3. Hate Groups' Relationship to Hate Crime: All Hate Groups, Excluding Patriot Groups Notes: All regressions include year‐ and state‐fixed effects (not reported) and robust standard errors  clustered by state (in parentheses). *** = 1%; ** = 5%; * = 10%.    18   All Hates Crimes Race Religion Sex. Orient. Ethnicity Disability 1 2 3 4 5 6 HateGroups ‐0.179 ‐0.069 0.077 ‐0.203 ‐0.007 0.022 (1.107) (0.776) (0.185) (0.312) (0.171) (0.066) GSPpc 1.252 ‐0.113 0.480*** 0.659*** 0.156 0.070 (0.843) (0.591) (0.141) (0.237) (0.130) (0.050) Unemployment 3.857** 1.427 0.349 1.549*** 0.49* 0.042 (1.788) (1.253) (0.299) (0.503) (0.276) (0.107) Poverty 1.438 0.572 0.409** 0.441 0.031 ‐0.016 (1.136) (0.796) (0.190) (0.320) (0.175) (0.068) Police 3.794 1.161 ‐0.245 2.592* 0.484 ‐0.196 (4.680) (3.280) (0.782) (1.318) (0.722) (0.279) Popshare ‐0.643*** ‐0.401*** ‐0.016 ‐0.159*** ‐0.058*** ‐0.009* (0.088) (0.062) (0.015) (0.025) (0.014) (0.005) Metro 0.137 0.196 ‐0.092** 0.005 0.034 ‐0.005 (0.231) (0.162) (0.39) (0.065) (0.036) (0.014) Abuse ‐0.575 ‐1.356 ‐0.016 1.017 0.004 ‐0.223 (3.341) (2.342) (0.558) (0.941) (0.516) (0.200) Black ‐3.030 0.639 ‐0.462 ‐4.153*** 0.750 0.195 (3.147) (2.206) (0.526) (0.886) (0.486) (0.188) Jewish 3.531 ‐0.188 0.970 1.763 0.766 0.220 (8.066) (5.654) (1.348) (2.271) (1.245) (0.482) N 292 292 292 292 292 292 R 2 0.24 0.18 0.11 0.31 0.11 0.03 Table 4. Hate Groups' Relationship to Hate Crime: KKK, Neo‐Nazi, White Nationalist and Racist Skinhead Notes: All regressions include year‐ and state‐fixed effects (not reported) and robust standard errors  clustered by state (in parentheses). *** = 1%; ** = 5%; * = 10%.