SOCI 103M Final Project
Daniel Barquero and Belynda Herrera
24 March, 2023
Section One: Introduction, Research Question, Background, and Hypothesis
Our research focuses on how the presence of hate groups in a state impacts hate crimes in Democratic vs. Republican states. Specifically, we examine whether the political voting habits of a state in presidential elections impacts the number of hate crimes in these states. It is important to note that for our research, we are not selecting any specific type of hate group or hate crime. We are encompassing all hate groups, as defined by the data from the Southern Poverty Law Center, and comparing it to all hate crimes from the data provided by the Department of Justice. We will be utilizing politics in this by establishing the political partisanship for each U.S. state and examining whether it impacts the presence of hate groups and the occurrence of hate crimes in states.
We chose to use political elections to separate the country as a basis for dividing the country into Democratic and Republican states. This division allowed for an analysis of trends in hate group formation and related hate crimes over time, highlighting the divergent patterns observed between regions with varying political affiliations.
This methodology of using political elections to divide the country was important as it allowed for a comparative analysis of patterns in hate group formation and related crimes over time concerning different political affiliations and their impact on the incidence of hate crimes over time. This approach enables researchers to gain a deeper understanding of the dynamics underlying hate crimes in America, enabling them to investigate how political affiliations might impact and shape hate group formation and hate crimes in the country. This type of analysis can inform policymaking and law enforcement strategies aimed at reducing the incidence of hate crimes in America, as it highlights which regions are most susceptible to hate group formation and related crimes. In conclusion, studying the prevalence and patterns of hate crimes across states with varying political affiliations provides important insights into the underlying dynamics that shape these incidents.
The null hypothesis for this research states that there is no correlation between a state’s political stance and the presence of hate groups/ the occurrence of hate crimes in states. We aim to investigate the correlation between political geography and hate crimes, with the objective of verifying the hypothesis that regions considered to be more liberal, such as California, are safer for marginalized communities, in comparison to areas known for being conservative, such as Alabama.
From a conflict theory perspective, this research can be seen as an attempt to understand how power dynamics between different groups of people contribute to the creation of an environment of fear and violence for some members of society. By looking into the relationship between political geography and hate crimes, we can gain a better understanding of how certain political ideologies can lead to certain patterns of hate-motivated violence, and what can be done to reduce those patterns. This research can also help to uncover the underlying mechanisms that lead to the perpetration of such hate-motivated violence, and how this can be addressed in order to create a more equitable and safe society for all.
This research question is important because it brings attention to trends for areas with high crime rates. It can help to raise awareness for hate crimes, as well as provide training for bystanders, police, and policy makers to prevent criminal activity and increase police presence in the states. It also provides the opportunity to study which areas need increased legislation to mitigate the frequency of hate crimes. This study will provide a broad overview of how certain states are affected by hate crime, and it serves to set the foundation for closer studies on a more local level.
This question is also of social, economic, and political importance. The phenomenon of hate-motivated violence is prevalent in many places around the world and has been the subject of extensive research. Other research has suggested conflicting results, as some studies have shown that more liberal areas tend to have lower hate crime rates, while others have found the opposite. Our research could potentially suggest something in opposition to established theories by uncovering the underlying mechanisms that lead to such violence and how this can be addressed in order to create a more equitable and safe society for all.
Hate crimes continue to be a pervasive and unsettling issue throughout the United States, with marginalized communities bearing the brunt of their devastating impact. The presidential election in 2000 saw a greater number of states voting Republican than Democrat, with Democratic-leaning states recording fewer instances of hate groups but higher crime rates - most notably California which reported an alarming 1943 hate crimes. Conversely, Republican-leaning states showed evidence of more active hate groups but lower incidence of hate-related offenses overall. While it may seem from this data that living in a republican state would offer greater safety against these appalling acts, such correlations are far from conclusive proof. Causation is not necessarily linked to political geography alone; comprehensive research must therefore be conducted before making any definitive conclusions about how this factor relates to patterns in incidents involving hatred or discrimination.
In the field of sociology, the causes and effects of violence against marginalized communities have been diligently studied in a plethora of journals, dissertations, and research papers. Our research will look to add on to the work in this field, and provide insightful answers to the questions at hand.
As touched upon earlier, our data for hate group presence in each state comes from The Southern Poverty Law Center. The Southern Poverty Law Center has data as recent as 2021 on how many hate groups are tracked per state, and what these hate groups advocate for. For the purpose of this study, we will focus on all groups included in the data. Our data for hate crime occurrence in each state comes from the Department of Justice’s State by State comparison on reported hate incidents. While the data is no longer available on the Department of Justice’s website, it can be accessed through the Anti-Defamation League (ADL). Lastly, we utilized the resources at the National Archives to find data for the way each state voted in the presidential elections between 2000 and 2014.
Section Two: Literature Review
The article “Making Hate History” explores the work of the New York City Police Department (NYPD) dedicated to eradicating hate crime in New York City. The data collected in this paper shows that an increased passing of legislation, coupled with increased funding to its Hate Crime Task Force, has led to a reduced number of hate crimes in New York City. The methods used in this paper are simple statistics, mainly charts depicting the number of hate crimes per year in a time series. The groups examined are African-Americans, LGBTQ+ identifying individuals, and Jews.
While our research does not employ the split-group theory employed in this paper, we do analyze the impacts of a decrease in violence through our theory. Specifically, democratic states, such as New York are traditionally more open to spending money on large-scale government social programs. Our data will see whether these states do indeed see a drop in the occurrence of hate crimes, or if external factors should also be examined.
Since this data was collected just for New York City, our independent research can contribute to this pre-built research by examining the thesis on a larger scale. Since our data focuses on a state level, applied across the nation, we can see whether the decrease of hate crimes through increased legislation is applicable on a more holistic scale. Our research will serve as a test to see how extendable the conclusions of this prior research are.
The article “White supremacist groups and hate crime” focuses on the increase in hate-based violence in an area when multiple white supremacist organizations are in the area. The null hypothesis for this paper states that white supremacist groups form in response to anti-white based crimes. Additionally, this paper seeks to debunk the theory that hate groups form to create a safe space for individuals to vent racial frustrations and turn away from violent acts. This article uses data from the Southern Poverty Law Center to prove its four hypotheses. The first of these four hypotheses states that white supremacy activity is associated with higher crime rates. The second of these hypotheses states that white supremacy is symptomatic of the overall level of biased-based violence. The third hypothesis states that white supremacist groups form in response to anti-white hate crime. The final hypothesis states that white supremacist groups are correlated with some idiosyncratic factor that influences hate crimes.
While we are not focusing solely on pro-white supremacy groups, we are especially interested in utilizing this paper because it utilizes data from the Southern Poverty Law Center, and FBI’s annual Hate Crime Statistics Report. The Southern Poverty Law Center data is data which we are also utilizing for our independent research. Our research will build upon the theoretical foundations of this paper by examining the hypotheses presented in this paper. Specifically, this paper argues for an increase in hate-based violence in areas where multiple hate groups are present. Since our data groups hate groups by area, we will be looking to see if areas with a plethora of hate crimes experience more hate crimes than areas with fewer or none hate groups.
In terms of expanding upon this research, we will be furthering this research by either agreeing with its hypothesis through our findings, or potentially challenging its results through our own findings. If we come to the conclusion that the number of hate groups in an area does not definitively impact the occurrence of hate crimes in that area, then we would be arguing against the conclusions of this paper. This will be especially interesting to see considering that both our research and this source are using data from the Southern Poverty Law Center.
The article “Community (Dis)Organization and Racially Motivated Crime” examines the relationship between community structural conditions and racially motivated crimes against blacks and whites. Centering on the Chicago, Illinois area, this paper focuses on the social organization of racial hate crimes derived from social disorganization. The methods used in this paper focus on multivariate analysis with controls for spacial autocorrelation. The results of the study show that anti-black hate crimes, in contrast to general forms of crime, are more likely in relatively organized communities with high levels of informal social control. On the contrary, anti-white incidents appear more numerous in traditionally disorganized communities, especially those characterized by residential instability.
Similar to the “Making Hate History” article which focused on hate crime in New York City, this article utilizes the Chicago area as a study for its analysis. As with the “Making Hate History” article, our research will utilize the conclusions seen here on a local scale, and determine whether these conclusions can be applied to a larger national scale. While the “Making Hate History” article focused on legislation towards combating hate crime, this article focuses more on the impacts of social organization and disorganization on hate crime. For our study, considering that Chicago is another primarily liberal area, we will look to see if the crime trends in our democratic states show an increase. Should that be the case, this literature could help explain the trend.
In terms of furthering this research, we can draw parallels between our conclusions and the conclusions of this research. Since this research analyzes the societal characteristics of communities to draw conclusions on the linkage with hate crime, we can also make similar observations in the states we study. Specifically, we can examine commonalities within the trends described here and the wider conclusions created by our study.
The article “Hate Groups and Muslim Population Changes in the Fifty States: Does the Presence of Muslims Encourage Hate Group Formation?” focuses on social dominance theory and group hate theory to determine whether there is a correlation between the size of a Muslim population and the number of hate groups in that area. The article uses a paneled negative binomial regression with a number of relevant control variables to determine that the relationship between these two variables is not statistically significant. The article goes on to unpack the research and discuss the conclusions.
One interesting find in this paper is that the relationship between the size of a Muslim population and the number of hate groups present is not statistically significant. This will be important for our work as we determine whether political voter behavior of a state is statistically significant in comparison to hate groups. If the answer is no, we can look to the conclusions of this paper to help explain why we might be seeing what we are seeing in the data.
In terms of furthering this research, we cannot directly further it because we are not looking at an independent view of each type of hate group and hate crime. What we can do is use the conclusions of this piece as external scholastic information for why we are seeing what we are seeing.
The article “The Determinants of the Number of White Supremacist Groups: A Pooled Time-Series Analysis” examines the political and social mechanisms which explain and promote the growth of white supremacist groups in the United States.This paper uses pooled time-series cross-sectional methods to assess the explanatory power of three racial threat accounts. First, since lynchings indicated intense animosity against blacks that may persist, anti-black hate groups should be especially numerous where lynching rates were substantial in the distant past. Second, where or when white political and cultural dominance is threatened by large or growing black populations, additional white supremacist groups should be present. Third, these anti-black movements often recruit by emphasizing the links between race and violent street crime. This paper further examines stereotypes made by the public. For example, this paper examines the idea that areas where black violence towards whites is high should produce an increase in Anti-Black hate groups. The ultimate goal of this paper is to examine the factors, both political and social, which contribute to the growth of the number of white supremacist groups in an area.
Focusing on our research, we are looking for the linkages between the presence of hate groups in an area, on a national scale, and the corresponding rate of hate crimes in that area. This paper works perfectly as a foundation for that analysis by discussing both the legislative and social reasons for why we might expect to see larger numbers of hate groups in certain areas while seeing smaller numbers in other groups. This understanding will help us unpack our conclusions as to if/how certain legislation can be used to mitigate hate crimes across the nation.
In terms of furthering this research, our research will provide either confirmation or negation of the claims made in this work. While this work serves as a theoretical springboard for our research, we can also use it to draw conclusions about why certain areas have more white supremacist groups than others. Depending on what our research shows, we can use the conclusions to either validate or negate the claims made by this article.
Works Cited Page: In order of appearance above
Levin, Brian, and Sara-Ellen Amster. “Making Hate History: Hate Crime and Policing in America … - Sage Journals.” Sage Journals, 27 July 2016, https://journals.sagepub.com/doi/10.1177/0002764207306062.
Mulholland, Sean E. “White Supremacist Groups and Hate Crime.” Public Choice, vol. 157, no. 1/2, 2013, pp. 91–113. JSTOR, http://www.jstor.org/stable/42003194. Accessed 20 Feb. 2023.
Lyons, Christopher J. “Community (Dis)Organization and Racially Motivated Crime.” American Journal of Sociology, vol. 113, no. 3, 2007, pp. 815–63. JSTOR, https://doi.org/10.1086/521846. Accessed 20 Feb. 2023.
Hummel, Daniel. “Hate Groups and Muslim Population Changes in the Fifty States: Does the Presence of Muslims Encourage Hate Group Formation?” International Journal on Minority and Group Rights, vol. 25, no. 2, 2018, pp. 317–32. JSTOR, https://www.jstor.org/stable/26557897. Accessed 20 Feb. 2023.
Rachel M. Durso, and David Jacobs. “The Determinants of the Number of White Supremacist Groups: A Pooled Time-Series Analysis.” Social Problems, vol. 60, no. 1, 2013, pp. 128–44. JSTOR, https://doi.org/10.1525/sp.2013.60.1.128. Accessed 20 Feb. 2023.
Section Three: Data and Methods
Instead of collecting our own data, we utilized data from the Southern Poverty Law Center’s (SPLC) Hate Map and the state presidential election voter data from the U.S. National Archives for presidential elections from 2000 to 2020. The SPLC dataset speaks to our research by providing figures on the number of hate groups in a state in a given year as well as the number of hate crimes committed in that state. The National Archives data provides state-level voting results in presidential elections from 2000 to 2020. Our plan is to use this data to categorize states as Republican or Democrat. The key variables in our study are: hate group presence in a state, hate crime statistics per state, and state voting results in presidential elections. Our analysis focuses on measuring the occurrence of hate groups in each state per year and filtering the U.S. presidential election data from 2000-2020 to create a binomial variable. States will be given a value of 1 if they voted in accordance with the criteria we present, and 0 otherwise. The first dataset measures hate crimes per state per year, while the second dataset measures each state’s presidential voting pattern. Although our dataset did not contain any missing information, the accuracy of hate group measurements may be lower due to the inability to account for all hate groups.
For our methods, we plan on splitting the data by political party in each election, in a time-series manner (2000-2020). We will then run correlations between hate group presence in each of the states that voted Republican and the number of hate crimes in the states which voted Republican. We will then repeat the process for states that voted Democrat. We have chosen correlations as one of our statistical measures because we are not doing any prediction with our data. We will be doing regression analysis between political voting behavior, hate group presence, and hate crime presence to determine whether there is a causal relationship between these variables. There will be three regressions, one between voting and hate crimes, one between voting and hate groups, and one between hate groups and hate crimes. The goal is to analyze trends in our data and make conclusions.
Section Four: Data Results
Step One: Load the required packages
Step Two: Load in the hatecrimes dataset
## State.Name X2000Crime X2001Crime X2002Crime X2003Crime X2004Crime
## 1 Alabama NA 0 2 1 3
## 2 Alaska 4 20 7 13 9
## 3 Arizona 240 384 238 246 224
## 4 Arkansas 3 3 0 170 93
## 5 California 1943 2246 1648 1472 1393
## 6 Colorado 101 126 96 82 59
## 7 Connecticut 151 169 129 134 116
## 8 Delaware 34 17 13 17 33
## 9 Florida 240 302 257 231 274
## 10 Georgia 35 39 31 23 29
## X2005Crime X2006Crime X2007Crime X2008Crime X2009Crime X2010Crime X2011Crime
## 1 0 1 6 11 9 19 83
## 2 4 6 8 8 9 7 8
## 3 138 149 161 185 219 236 192
## 4 134 113 33 91 74 63 11
## 5 1379 1297 1400 1381 1015 1092 1040
## 6 125 138 156 149 208 154 186
## 7 95 131 127 164 198 147 140
## 8 45 48 49 58 37 20 15
## 9 231 216 166 153 130 136 123
## 10 17 13 13 9 11 17 17
## X2012Crime X2013Crime X2014Crime X2015Crime X2016Crime X2017Crime X2018Crime
## 1 6 6 9 10 14 9 0
## 2 6 8 6 8 11 4 7
## 3 176 155 265 276 213 264 166
## 4 30 27 8 5 12 7 13
## 5 910 843 759 837 931 1094 1063
## 6 189 128 96 107 104 106 121
## 7 149 145 123 93 106 111 81
## 8 10 12 13 11 15 29 16
## 9 144 76 65 72 96 145 141
## 10 21 57 41 44 39 27 35
## X2019Crime X2020Crime X2000Group X2001Group X2002Group X2003Group X2004Group
## 1 0 27 38 35 34 24 27
## 2 11 9 1 4 1 1 2
## 3 209 280 8 9 8 9 8
## 4 9 19 18 13 21 21 23
## 5 1015 1339 29 40 48 45 42
## 6 210 281 7 8 6 7 10
## 7 76 101 3 6 8 8 4
## 8 22 13 2 2 0 2 3
## 9 111 109 38 42 44 38 43
## 10 102 195 30 30 31 54 41
## X2005Group X2006Group X2007Group X2008Group X2009Group X2010Group X2011Group
## 1 22 22 24 36 32 33 32
## 2 2 1 0 0 1 0 1
## 3 12 10 17 19 16 22 17
## 4 21 19 18 20 24 29 26
## 5 49 63 78 84 60 68 84
## 6 10 13 12 15 17 19 15
## 7 4 7 6 5 6 6 5
## 8 3 2 2 4 4 5 5
## 9 49 49 49 56 51 49 55
## 10 41 44 42 40 37 39 65
## X2012Group X2013Group X2014Group X2015Group X2016Group X2017Group X2018Group
## 1 30 22 18 22 27 23 23
## 2 2 1 0 0 0 4 4
## 3 28 20 16 18 18 22 20
## 4 23 24 20 22 16 12 14
## 5 82 77 57 68 79 75 83
## 6 15 17 15 16 16 21 22
## 7 7 5 3 2 5 6 6
## 8 4 4 4 5 4 2 2
## 9 59 58 50 58 63 66 75
## 10 53 50 28 39 32 40 41
## X2019Group X2020Group Rep2000 Dem2000 Rep2004 Dem2004 Rep2008 Dem2008
## 1 16 20 1 0 1 0 1 0
## 2 2 1 1 0 1 0 1 0
## 3 21 26 1 0 1 0 1 0
## 4 15 14 1 0 1 0 1 0
## 5 88 72 0 1 0 1 0 1
## 6 22 17 1 0 1 0 0 1
## 7 8 6 0 1 0 1 0 1
## 8 3 3 0 1 0 1 0 1
## 9 67 68 1 0 1 0 0 1
## 10 38 29 1 0 1 0 1 0
## Rep2012 Dem2012 Rep2016 Dem.2016 Rep2020 Dem2020
## 1 1 0 1 0 1 0
## 2 1 0 1 0 1 0
## 3 1 0 1 0 0 1
## 4 1 0 1 0 1 0
## 5 0 1 0 1 0 1
## 6 0 1 0 1 0 1
## 7 0 1 0 1 0 1
## 8 0 1 0 1 0 1
## 9 0 1 1 0 1 0
## 10 1 0 1 0 0 1Step Three: Linear Regression with state political voting preferences in presidential elections as the predictors and the number of hate crimes across all states as the dependent variables.
##
## Call:
## lm(formula = X2000Crime + X2001Crime + X2002Crime + X2003Crime +
## X2004Crime + X2005Crime + X2006Crime + X2007Crime + X2008Crime +
## X2009Crime + X2010Crime + X2011Crime + X2012Crime + X2013Crime +
## X2014Crime + X2015Crime + X2016Crime + X2017Crime + X2018Crime +
## X2019Crime + X2020Crime ~ Rep2000 + Dem2000 + Rep2004 + Dem2004 +
## Rep2008 + Dem2008 + Rep2012 + Dem2012 + Rep2016 + Dem.2016 +
## Rep2020 + Dem2020, data = hatecrimes)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4456.9 -1994.0 -704.8 466.9 21370.7
##
## Coefficients: (5 not defined because of singularities)
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 11823.0 4316.4 2.739 0.00916 **
## Rep2000 296.6 2818.5 0.105 0.91672
## Dem2000 NA NA NA NA
## Rep2004 -2048.9 2893.2 -0.708 0.48295
## Dem2004 NA NA NA NA
## Rep2008 -520.2 3203.6 -0.162 0.87181
## Dem2008 NA NA NA NA
## Rep2012 -8062.8 5995.6 -1.345 0.18627
## Dem2012 -7096.7 4465.6 -1.589 0.11989
## Rep2016 102.5 2367.3 0.043 0.96567
## Dem.2016 NA NA NA NA
## Rep2020 -139.5 2569.7 -0.054 0.95697
## Dem2020 NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4316 on 40 degrees of freedom
## (2 observations deleted due to missingness)
## Multiple R-squared: 0.1966, Adjusted R-squared: 0.05602
## F-statistic: 1.398 on 7 and 40 DF, p-value: 0.2329Looking at the linear regression model above, we can see that the relationship between voting preferences and the number of hate crimes is not very good. First, the adjusted R-Squared value is only 0.05602. This means that only a meager 5.2% of the variance in the data can be explained by this model. That is not very good. Additionally, there are missing values in the predictor variables (the voting preferences) which negatively impacts the ability of the model to create an accurate causal relationship. The F-Statistic also gives us reasoning to believe that this model is poor. Specifically, the F-Statistic is 1.398 while the p-value is 0.2329. To determine statistical significance, we are shooting for a p-value of 0.05 or lower to determine statistical significance. The goal of the p-value is to reject the null hypothesis. Since the p-value for this model is higher than 0.05 we lack statistical significance and an ability to reject the null hypothesis. Ultimately, the poor R-Squared value paired with the missing values and high p-value show that this model cannot confidently reject the null hypothesis. For our research, this means that we cannot confidently prove a causal relationship between a state’s voting preferences and the number of hate crimes which occur in the state.
Step Four: Linear Regression with state political voting preferences in presidential elections as the predictors and the number of hate groups across all states as the dependent variables.
##
## Call:
## lm(formula = X2000Group + X2001Group + X2002Group + X2003Group +
## X2004Group + X2005Group + X2006Group + X2007Group + X2008Group +
## X2009Group + X2010Group + X2011Group + X2012Group + X2013Group +
## X2014Group + X2015Group + X2016Group + X2017Group + X2018Group +
## X2019Group + X2020Group ~ Rep2000 + Dem2000 + Rep2004 + Dem2004 +
## Rep2008 + Dem2008 + Rep2012 + Dem2012 + Rep2016 + Dem.2016 +
## Rep2020 + Dem2020, data = hatecrimes)
##
## Residuals:
## Min 1Q Median 3Q Max
## -397.50 -209.57 -73.06 132.18 1128.97
##
## Coefficients: (5 not defined because of singularities)
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 695.00 320.72 2.167 0.0360 *
## Rep2000 249.47 209.21 1.192 0.2398
## Dem2000 NA NA NA NA
## Rep2004 -67.85 214.66 -0.316 0.7535
## Dem2004 NA NA NA NA
## Rep2008 -185.24 237.50 -0.780 0.4398
## Dem2008 NA NA NA NA
## Rep2012 -546.76 444.49 -1.230 0.2255
## Dem2012 -452.97 331.09 -1.368 0.1786
## Rep2016 316.34 175.02 1.807 0.0779 .
## Dem.2016 NA NA NA NA
## Rep2020 -127.19 190.48 -0.668 0.5080
## Dem2020 NA NA NA NA
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 320.7 on 42 degrees of freedom
## Multiple R-squared: 0.1397, Adjusted R-squared: -0.003696
## F-statistic: 0.9742 on 7 and 42 DF, p-value: 0.4625Looking at the linear regression model above, we can see that the relationship between voting preferences and the number of hate groups in a state is also not very good. First, the adjusted R-Squared value is negative. Specifically, the value is -0.0037. This means that the model is terrible because it does not explain any of the variance in the data. Additionally, there are missing values in the predictor variables (the voting preferences) which negatively impacts the ability of the model to create an accurate causal relationship. The F-Statistic also gives us reasoning to believe that this model is poor. Specifically, the F-Statistic is 0.9742 while the p-value is 0.4625. To determine statistical significance, we are shooting for a p-value of 0.05 or lower to determine statistical significance. This p-value is even higher than the one of the above model which was already poor. The goal of the p-value is to reject the null hypothesis. Since the p-value for this model is higher than 0.05 we lack statistical significance and an ability to reject the null hypothesis. Ultimately, the terrible R-Squared value paired with the missing values and high p-value show that this model cannot confidently reject the null hypothesis. For our research, this means that we cannot confidently prove a causal relationship between a state’s voting preferences and the number of hate groups which occur in the state.
Step Five: Linear Regression with the number of hate groups across all states as the predictors and the number of hate crimes across all states as the dependent variables.
##
## Call:
## lm(formula = X2000Crime + X2001Crime + X2002Crime + X2003Crime +
## X2004Crime + X2005Crime + X2006Crime + X2007Crime + X2008Crime +
## X2009Crime + X2010Crime + X2011Crime + X2012Crime + X2013Crime +
## X2014Crime + X2015Crime + X2016Crime + X2017Crime + X2018Crime +
## X2019Crime + X2020Crime ~ X2000Group + X2001Group + X2002Group +
## X2003Group + X2004Group + X2005Group + X2006Group + X2007Group +
## X2008Group + X2009Group + X2010Group + X2011Group + X2012Group +
## X2013Group + X2014Group + X2015Group + X2016Group + X2017Group +
## X2018Group + X2019Group + X2020Group, data = hatecrimes)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2780.4 -1044.4 -109.7 901.1 4654.3
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -86.432 584.402 -0.148 0.8836
## X2000Group -376.545 177.365 -2.123 0.0434 *
## X2001Group -119.331 161.217 -0.740 0.4658
## X2002Group 191.348 167.184 1.145 0.2628
## X2003Group 42.056 184.681 0.228 0.8216
## X2004Group 23.414 346.026 0.068 0.9466
## X2005Group -107.481 316.980 -0.339 0.7373
## X2006Group -147.637 265.998 -0.555 0.5836
## X2007Group 275.028 209.497 1.313 0.2007
## X2008Group -130.582 198.067 -0.659 0.5155
## X2009Group -34.196 113.397 -0.302 0.7654
## X2010Group 173.054 135.601 1.276 0.2132
## X2011Group -158.556 109.346 -1.450 0.1590
## X2012Group 217.853 150.617 1.446 0.1600
## X2013Group -3.898 206.568 -0.019 0.9851
## X2014Group 173.168 184.871 0.937 0.3575
## X2015Group -237.130 94.748 -2.503 0.0189 *
## X2016Group 164.464 153.866 1.069 0.2949
## X2017Group -246.795 171.118 -1.442 0.1612
## X2018Group 191.535 202.459 0.946 0.3528
## X2019Group 411.620 200.102 2.057 0.0498 *
## X2020Group -308.696 165.065 -1.870 0.0728 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2108 on 26 degrees of freedom
## (2 observations deleted due to missingness)
## Multiple R-squared: 0.8754, Adjusted R-squared: 0.7748
## F-statistic: 8.7 on 21 and 26 DF, p-value: 4.127e-07The linear regression model between hate groups and hate crimes has its strengths and weaknesses but is ultimately fairly decent. Firstly, the multiple R-Squared value is 0.875 while the adjusted R-Squared is 0.775. This means that once adjusted for multiple independent variables, the model still explains 77.5% of the variance in the data. This is quite good. However, this could also be a result of overfitting the data, which would mean this model is poor. Additionally, the F-statistic of 8.7 has a p-value of 0.0000004127 which is well below the 0.05 threshold. This means that there is great statistical significance between the hate group independent variables and hate crime dependent variables. Additionally, many of the coefficients for the independent variables are at, or above, 0.05. This means that they are confident enough to reject the null hypothesis. Ultimately, while this model is certainly not perfect, it seems to show a strong correlation between the independent and dependent variables. For our research, this means that the number of hate groups does impact the number of hate crimes in states, although political preference of a state does not have an important role.
First Data Visualization: Line Plots
In order to accomplish these visualizations, we first divided up the states. Specifically, we created 12 new dataframes based on how states voted in presidential elections (Republican or Democrat). These results start with the 2000 presidential election and go until the 2020 presidential election.
Lineplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2000.
## [1] "Alabama" "Alaska" "Arizona" "Arkansas"
## [5] "Colorado" "Florida" "Georgia" "Idaho"
## [9] "Indiana" "Kansas" "Kentucky" "Louisiana"
## [13] "Mississippi" "Missouri" "Montana" "Nebraska"
## [17] "Nevada" "New Hampshire" "North Carolina" "North Dakota"
## [21] "Ohio" "Oklahoma" "South Carolina" "South Dakota"
## [25] "Tennessee" "Texas" "Utah" "Virginia"
## [29] "West Virginia" "Wyoming"Lineplot showing the relationship between hate groups and hate crimes for states that voted Democrat in 2000.
## [1] "California" "Connecticut" "Delaware" "Hawaii"
## [5] "Illinois" "Iowa" "Maine" "Maryland"
## [9] "Massachusetts" "Michigan" "Minnesota" "New Jersey"
## [13] "New Mexico" "New York" "Oregon" "Pennsylvania"
## [17] "Rhode Island" "Vermont" "Washington" "Wisconsin"
The presidential election in 2000 saw a greater number of states voting Republican than Democrat, with Democratic-leaning states recording fewer instances of hate groups but higher crime rates - most notably California which reported an alarming 1943 hate crimes. Conversely, Republican-leaning states showed evidence of more active hate groups but a lower incidence of hate-related offenses overall.
Lineplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2004.
## [1] "Alabama" "Alaska" "Arizona" "Arkansas"
## [5] "Colorado" "Florida" "Georgia" "Idaho"
## [9] "Indiana" "Iowa" "Kansas" "Kentucky"
## [13] "Louisiana" "Mississippi" "Missouri" "Montana"
## [17] "Nebraska" "Nevada" "New Mexico" "North Carolina"
## [21] "North Dakota" "Ohio" "Oklahoma" "South Carolina"
## [25] "South Dakota" "Tennessee" "Texas" "Utah"
## [29] "Virginia" "West Virginia" "Wyoming"Lineplot showing the relationship between hate groups and hate crimes for states that voted Democrat in 2004.
## [1] "California" "Connecticut" "Delaware" "Hawaii"
## [5] "Illinois" "Maine" "Maryland" "Massachusetts"
## [9] "Michigan" "Minnesota" "New Hampshire" "New Jersey"
## [13] "New York" "Oregon" "Pennsylvania" "Rhode Island"
## [17] "Vermont" "Washington" "Wisconsin"
The 2004 presidential election was marked by a troubling pattern in which Republican states reported more hate groups but fewer hate crimes, mirroring the same trend observed during the 2000 elections. A closer analysis reveals that states with higher numbers of Republican supporters tended to cluster within a range of 50-100 reported hate groups, while Virginia emerged as the state with the highest number of hate crimes at an alarming rate of 307. On the other hand, Democratic states presented a starkly different picture. Their reportage clustered under the relatively low-range category - around 500 or lower - and California stood out yet again as an outlier for having recorded over thirteen hundred incidents involving hatred based on race or religion.
Lineplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2008.
## [1] "Alabama" "Alaska" "Arizona" "Arkansas"
## [5] "Georgia" "Idaho" "Kansas" "Kentucky"
## [9] "Louisiana" "Mississippi" "Missouri" "Montana"
## [13] "Nebraska" "North Dakota" "Oklahoma" "South Carolina"
## [17] "South Dakota" "Tennessee" "Texas" "Utah"
## [21] "West Virginia" "Wyoming"Lineplot showing the relationship between hate groups and hate crimes for states that voted Democrat in 2008.
## [1] "California" "Colorado" "Connecticut" "Delaware"
## [5] "Florida" "Hawaii" "Illinois" "Indiana"
## [9] "Iowa" "Maine" "Maryland" "Massachusetts"
## [13] "Michigan" "Minnesota" "Nevada" "New Hampshire"
## [17] "New Jersey" "New Mexico" "New York" "North Carolina"
## [21] "Ohio" "Oregon" "Pennsylvania" "Rhode Island"
## [25] "Vermont" "Virginia" "Washington" "Wisconsin"
In the 2008 presidential election, we see a decrease in the crime rate in both democratic and republican states. The trend as prior still follows with republicans having more hate groups with less crime and democratic states having fewer hate groups and more crime.
Lineplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2012.
## [1] "Alabama" "Alaska" "Arizona" "Arkansas"
## [5] "Georgia" "Idaho" "Indiana" "Kansas"
## [9] "Kentucky" "Louisiana" "Mississippi" "Missouri"
## [13] "Montana" "Nebraska" "North Carolina" "North Dakota"
## [17] "Oklahoma" "South Carolina" "South Dakota" "Tennessee"
## [21] "Texas" "Utah" "West Virginia" "Wyoming"Lineplot showing the relationship between hate groups and hate crimes for states that voted Democrat in 2012.
## [1] "California" "Colorado" "Connecticut" "Delaware"
## [5] "Florida" "Hawaii" "Illinois" "Iowa"
## [9] "Maine" "Maryland" "Massachusetts" "Michigan"
## [13] "Minnesota" "Nevada" "New Hampshire" "New Jersey"
## [17] "New Mexico" "Ohio" "Oregon" "Pennsylvania"
## [21] "Rhode Island" "Vermont" "Virginia" "Washington"
## [25] "Wisconsin"
Fast forward to the next election in 2012 when America went to polls once again; data showed that both Democratic and Republican-leaning regions had witnessed a decrease in hate crime rates since their last encounter at voting booths.
Lineplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2016.
## [1] "Alabama" "Alaska" "Arizona" "Arkansas"
## [5] "Florida" "Georgia" "Idaho" "Indiana"
## [9] "Iowa" "Kansas" "Kentucky" "Louisiana"
## [13] "Michigan" "Mississippi" "Missouri" "Montana"
## [17] "Nebraska" "North Carolina" "North Dakota" "Ohio"
## [21] "Oklahoma" "Pennsylvania" "South Carolina" "South Dakota"
## [25] "Tennessee" "Texas" "Utah" "West Virginia"
## [29] "Wisconsin" "Wyoming"Lineplot showing the relationship between hate groups and hate crimes for states that voted Democrat in 2016.
## [1] "California" "Colorado" "Connecticut" "Delaware"
## [5] "Hawaii" "Illinois" "Maine" "Maryland"
## [9] "Massachusetts" "Minnesota" "Nevada" "New Hampshire"
## [13] "New Jersey" "New Mexico" "New York" "Oregon"
## [17] "Rhode Island" "Vermont" "Virginia" "Washington"
During the 2016 presidential election, a concerning trend emerged as reports of hate crimes spiked past 400 incidents in one Republican-led state. Meanwhile, in states led by Democrats, hate crime rates were decreasing along with the number of active hate groups.
Lineplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2020.
## [1] "Alabama" "Alaska" "Arkansas" "Florida"
## [5] "Idaho" "Indiana" "Iowa" "Kansas"
## [9] "Kentucky" "Louisiana" "Mississippi" "Missouri"
## [13] "Montana" "Nebraska" "North Carolina" "North Dakota"
## [17] "Ohio" "Oklahoma" "South Carolina" "South Dakota"
## [21] "Tennessee" "Texas" "Utah" "West Virginia"
## [25] "Wyoming"Lineplot showing the relationship between hate groups and hate crimes for states that voted Democrat in 2020.
## [1] "Arizona" "California" "Colorado" "Connecticut"
## [5] "Delaware" "Georgia" "Hawaii" "Illinois"
## [9] "Maine" "Maryland" "Massachusetts" "Michigan"
## [13] "Minnesota" "Nevada" "New Hampshire" "New Jersey"
## [17] "New Mexico" "New York" "Oregon" "Pennsylvania"
## [21] "Rhode Island" "Vermont" "Virginia" "Washington"
## [25] "Wisconsin"
During the 2020 presidential election, a noteworthy shift in hate crime rates could be observed across different states. Republican states witnessed an unprecedented surge in their rate of hate crimes, as it spiked to over 400 incidents with more hate groups being reported than ever before. This unsettling trend marked a significant continuation from previous years where there was a noticeable increase in hate crimes among these regions. In contrast, in democratic states, the number of hate crimes has been going down since the early 2000s and reached its lowest point in 2020.
Data Visualization Number Two: Scatterplots.
To accomplish this, we used the same parameters as for the lineplots, and simply switched out scatterplots for lineplots.
Scatterplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2000.
Scatterplot showing the relationship between hate groups and
hate crimes for states that voted Democrat in 2000.
Looking at this first scatterplot for the year 2000, we have Republican states on the left and Democratic states on the right. We can see that there is a steeper positive slope in Republican states than in Democrat states. This means that there is a stronger relationship between the number of hate groups and their impact on hate crimes. However, correlation alone is not enough to prove causality. These scatterplots alone are not enough to show a causal relationship between political preference and hate crimes. There is certainly a stronger positive trend in these Republican states when compared to the Democratic states, but that is all it is: a trend. As established earlier in the regression analysis, political voting preference for a state does not strongly impact hate crimes or hate groups, therefore we will need to seek out alternative explanations to this observation.
Scatterplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2004.
Scatterplot showing the relationship between hate groups and
hate crimes for states that voted Democrat in 2004.
Looking at this second scatterplot for the year 2004, we once again have Republican states on the left and Democratic states on the right. We can see that there is a very even division in the steepness of slopes for both sets of states. Both sets of states seem to have a similar positive slope which means that there is no true split. Once again, political preference is not an impactful factor in this analysis, and no real conclusions can be drawn off this scatterplot. This plot fails to give us necessary evidence to disprove the null hypothesis.
Scatterplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2008.
Scatterplot showing the relationship between hate groups and
hate crimes for states that voted Democrat in 2008.
Looking at this third scatterplot for the year 2008, we once again have Republican states on the left and Democratic states on the right. Again, We can see that there is a very even division in the steepness of slopes for both sets of states. Both sets of states seem to have a similar positive slope which means that there is no true split. The Republican states do seem to have a few outliers towards the top of the left plot which may skew the slope a bit. Despite this, there are no blatently obvious trends which prove to be visual evidence that Republican states are more dangerous in terms of hate crimes than Democratic states. Scatterplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2012.
Scatterplot showing the relationship between hate groups and
hate crimes for states that voted Democrat in 2012.
Looking at this fourth scatterplot for the year 2012, we once again have Republican states on the left and Democratic states on the right. This time, we see that the Republican states on the left do not have as strong a correlation as the Democratic states on the right. Specifically, the observations for the Republican states are less centralized and more spread out across the list. While this does not serve as enough information to reject the null hypothesis. One interesting note is that 2012 marks the first year in our data, accounting for election years, that Democratic states seem to have a stronger correlation between hate group numbers and hate crime figures.
Scatterplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2016.
Scatterplot showing the relationship between hate groups and
hate crimes for states that voted Democrat in 2016.
Looking at this fifth scatterplot for the year 2016, we once again have Republican states on the left and Democratic states on the right. We see that the Republican states on the left do not have as strong a correlation as the Democratic states on the right. Specifically, the observations for the Republican states are less centralized and more spread out across the list. There is also a faint trace of heteroscedasticity in both models which could mean that there is some error. The correlations here are not fully robust but they are generally positive for both sets of states. This does not serve as enough information to reject the null hypothesis.
Scatterplot showing the relationship between hate groups and hate crimes for states that voted Republican in 2020.
Scatterplot showing the relationship between hate groups and
hate crimes for states that voted Democrat in 2020.
The sixth and final scatterplot for the year 2020 once again has Republican states on the left and Democratic states on the right. The Democratic states contain a stronger positive correlation than the Republican states. The Democratic states do however have a slight trace of heteroscedasticity which could mean that this plot is corrupted statistically. Ultimately, the correlations here are not fully robust but they are generally positive for both sets of states. These scatterplots do not serve as enough information to reject the null hypothesis.
Section Five: Conclusions
As we wrap up this study on the effect of state partisanship on hate group numbers and hate crime occurrence, we have come to the conclusion that we do not possess enough merit to confidently reject the null hypothesis. Looking first at the regression analysis, we ran three regressions. The first regression looked at the relationship between state partisanship and hate crime occurrence. The model did not sufficiently explain the variance in the data through the adjusted R-Squared, nor did it possess a low enough P-Value to profoundly proclaim statistical significance. The second regression model, which sought a relationship between state partisanship and hate group presence, was plagued by the same issues as the first model. This model possessed a truly horrific adjusted R-Squared, a P-Value even larger than the first model, and several errors in regards to missing data. With these two models, it was evident that the current data is not enough to prove that there is a robust relationship between state political partisanship and hate groups or hate crimes. This alone is enough for us to accept the null hypothesis which states that there is no relationship between these groups.
The data visualizations proved to be a mixed bag. Both the lineplots and scatterplots occassionally favored either Republican or Democratic states as being safer, but no causal relationship can be drawn from these graphs. Additionally, the regression analysis shows that political partisanship does not greatly impact whether a certain set of states will receive higher hate crime rates than others. The trends displayed in these visualizations is a starting point, but each of these states must be examined in a thorough manner to determine the true cause of spikes in hate group presence and hate crime occurrence.
Our data also contained limitations which impacted this study. While the data from the Southern Poverty Law Center and U.S. Department of Justice contained all forms of hate groups and hate crimes, neither source further broke down the data by ethnic group. Therefore, we had to examine the data on a state by state level rather than on a more regional level. The impact of this is that it limited how closely we could study the data. For example, is Los Angeles more prone to hate crimes than San Diego despite them being only 100 miles apart and in the same state? Furthermore, does regional partisanship play more of a role in the occurrence of hate crimes than state-level partisanship? These are the types of questions we will look to examine in further iterations of this study.
Ultimately, we are unable to reject the null hypothesis with our findings in this study. We have greatly enjoyed researching this topic, and are eager to obtain more data and information to reject the null hypothesis in the future.
Section Six: Data Bibliography
- Unknown, Author. “Hate Map.” Southern Poverty Law Center, 15 Mar. 2023, https://www.splcenter.org/hate-map.
- Unknown, Author. “Electoral College Results.” National Archives and Records Administration, National Archives and Records Administration, 16 Apr. 2021, https://www.archives.gov/electoral-college/results.
- Unknown, Author. “Final 2021 State by State Incident Report 2021-2000.” ADL.org, Anti-Defamation League, 2022, https://www.adl.org/sites/default/files/pdfs/2022-12/Final%202021%20State%20by%20State%20Incident%20Report%202021-2000.pdf.