Predictive Modeling Of Categorical Data And Modeling Dependence For Hate Crime Occurrences In New York City PDF Download

The fledgling field of hate crime has turned out mostly descriptive research since the Hate Crime Statistics Act was enacted in 1990. Hate crime causation, an exceedingly important area of inquiry, has been severely neglected. In a novel approach, literature on the causes of historical lynchings and contemporary intergroup conflict was reviewed in addition to hate crime research. This dissertation also applied Robert Agnew's macro-level version of his general strain theory (MST) to empirically examine hate crime causality. The basic premise of MST is that an increase in community crime rates is a common consequence of the strain produced by those communities. The primary dependent variable of the study was the rate of hate crime with the other dependent variables representing the offense types and bias motivations of hate crime. The types of offenses chosen for analysis were crimes against persons and crimes against property.^Anti-racial, anti-religion, anti-ethnicity/national origin, anti-sexual orientation, and anti-disability comprised the general bias motivations while the categories of specific bias motivations included anti-white, anti-black, and anti-Hispanic. The independent variables were used as measures of MST and were meant to embody the categories of economic deprivation, social cleavages, and family disruption. Economic deprivation was operationalized using variables consisting of poverty, unemployment, and educational attainment. Indicators of social cleavages included variables associated with heterogeneity, density/overcrowding, and population mobility. Family disruption measures related to divorce or separation and single parent families. Control variables were also incorporated because they have been found to have significant effects on crime. These control variables corresponded to race, region, age, and gender.^Data for the hate crime measures came from the Uniform Crime Report while the indicators of strain and control variables were created from Census data. The two data sets for the year 2000 were linked using FIPS county codes, which established a unique, never-before-used data set. Informed by the literature reviewed and the theoretical framework employed, hypotheses were generated and analyses were performed. Results for the correlation and unstandardized coefficients for the bivariate and multivariate analyses, respectively, yielded mixed results and only partial support for Agnew's macro-level general strain theory.