The Complexity of Hate Crime and Bias Activity: Variation across Contexts and Types of Bias

¹ The terms “hate crime” and “bias crime” are used interchangeably. Additionally, for the sake of brevity and to avoid unnecessary repetition, any reference to “hate crime” throughout this manuscript includes racially- or ethnically-motivated crime unless otherwise noted.

² In this manuscript, we provide only cursory attention to anti-Semitic and anti-sexual orientation events due to empirical concerns. Population estimates of certain minority groups (e.g. Jews, homosexuals) are inadequately detailed to allow modeling of crime rates. As Berk (Citation1990) states, “For instance, who qualifies as being at risk for a hate-motivated assault against gay people? And how many such people are there? This is important because for crime rates to interpretable, the standardizing base must be sensible.” Census estimates indicate that less than 3% of the population in Pennsylvania is non-heterosexual, but this figure cannot be disaggregated by jurisdiction. Modeling the crime rate as a function of a global constant ignores potentially substantial variation in the population at risk across places. Scholars have for decades cautioned against using inappropriate denominators in the study of crime rates (e.g. Steffensmeier & Harer, Citation1987) and hate crime rates (Berk, Citation1990); doing so can yield distorted or invalid inferences. To err on the side of caution, we restrict our discussion of incident rates to three racial and ethnic groups for which we have reasonably accurate population counts.

³ As an example, 5 of 67 counties (7.5%) reported no hate- or bias-motivated activities from 2000 to 2011 compared to 1,898 of 2,563 municipalities (74.1%). Relying on counties as the unit of analysis may obscure important relationships.

⁴ The heterogeneity index is included in all analyses because social disorganization theory predicts that heterogeneity should predict all manner of criminal activity within a neighborhood. Accordingly, racial/ethnic heterogeneity should also be an important indicator of other hate crime, including those targeting victims for sexual orientation and other characteristics.

⁵ Specifically, the 2000 Census measures tenure as residing in the same household since 1995 or before, while the ACS estimates tenure as having lived in the same household for one year or more. Because these measures are clearly not comparable, we opted to avoid averaging the two.

⁶ We considered the use of zero-inflated models due to the large number of municipalities where no hate activity was reported. However, we opted against this approach. Substantively, researchers must hypothesize why some municipalities are “always zeroes” and why others are “potential zeroes.” There is no theoretical reason to expect why a municipality must have no hate crime. Even municipalities with no minority population could use bias activities to promote in-group solidarity and deter potential immigrants from moving to the area. Methodologically, a test of the zero-inflated model yielded poor model fit. Based on these two points, we decided to use the negative binomial models presented here.