Since 1990, the U.S. Federal Bureau of Investigation (FBI) has collected data on criminal offences designated as “hate crimes,” defined as offences that “manifest evidence of prejudice based on race, religion, sexual orientation, or ethnicity.” (28 U.S.C. § 534) Local law enforcement agencies collect the information on hate crimes, and report them to the federal datbase through the Uniform Crime Reporting Program, and the National Incident-Based Reporting System. The FBI then makes these data available in a series of summary data files, published on the Hate Crime Statistics web page.

Because the collection of hate crimes data relies on the cooperation of local law enforcement agencies, our ability to draw inferences about national patterns will depend on the distribution and abundance of agencies that report those data. Reporting hate crime data to the federal databases may require additional time and effort, and some agencies, especially those in more rural or resource-limited areas, may choose not to participate in the program. This would then limit our knowledge of the frequency and types of hate crimes occurring in those areas. The FBI reports that in 2013, over 15,000 law enforcement agencies participated in reporting hate crime data (1). These agencies served over 93% of the total U.S. population (2). But the level of reporting, and the amount of the population covered, may vary by state. For example, Hawai’i does not participate in the hate crimes reporting program (3). So the first question I was interested in examining is the degree to which reporting varies by state; specifically, how does the percentage of the population covered by agencies that report hate crimes vary from state to state?

The FBI also reports that participation varies by the type of jusrisdiction served by the law enforcement agency (1). In general, fewer agencies in rural areas report hate crimes data, so a smaller proportion of the population in those areas is covered by hate crimes reporting. So a follow-up question on the distribution of reporting was: do states with a higher percentage of rural population have a lower percentage of their population covered by agencies that report hate crimes?

In addition to these questions about reporting, I was interested in the relationship between racial and ethnic diversity, on one hand, and the prevalence of hate crimes on the other. Obviously, in an area completely lacking in diversity, there are no targets for hate crimes. But what happens as diversity increases? One possibility is that increasing levels of diversity beget more hate crimes, as contact between people of different backgrounds come together more often. An opposing prediction is that increasing diversity leads to fewer hate crimes, because increased contact among people of different backgrounds leads to greater understanding and decreased animosity. So I wished to examine the question: do states with higher levels of racial (or ethnic) diversity have higher, or lower rates of hate crimes based on race (or ethnicity)?

All of these questions required combining state-level data from the FBI Hate Crimes Statistics page with data from the U.S. Census Bureau’s “Fact Finder” data interface.

How does hate crimes reporting vary from state to state?

Methodology

Data on the number of agencies reporting hate crimes to the FBI database, and the population covered by those agencies, in each state, were obtained from Table 12 of the 2013 hate crimes data:

##        State NoParticipating PopCovered AgenciesReporting No.Incidents
## 1    Alabama              48    1147612                 4            6
## 2     Alaska              33     730950                 1            8
## 3    Arizona              81    6358545                21          155
## 4   Arkansas             265    2809536                20           27
## 5 California             732   38324460               229          843
## 6   Colorado             226    5163276                48          128

Data on the urban and rural populations in each state, as well as the total population, were obtained from the the U.S. Census bureau’s “Fact Finder” database:

##            Id Id2            State  TotalPop     Urban UrbanAreas
## 1   0100000US   0    United States 308746065 249253271  219922123
## 2  0200000US1   1 Northeast Region  55317240  47016291   44082415
## 3  0200000US2   2   Midwest Region  66927001  50771646   42026807
## 4  0200000US3   3     South Region 114555744  86876545   75800759
## 5  0200000US4   4      West Region  71945553  64588789   58012142
## 6 0400000US01   1          Alabama   4779753   2821804    2325304
##   UrbanClusters    Rural
## 1      29331148 59492267
## 2       2933876  8300949
## 3       8744839 16155355
## 4      11075786 27679199
## 5       6576647  7356764
## 6        496500  1957932

In addition to the state-level data, the census dataset has information on other geographic regions, which I didn’t need for the analysis. I combined the two data sets, removing the rows in the census dataset that did not match rows in the FBI dataset. Then I calculated the proportion of each state’s population that is covered by agencies that report hate crimes, and the proportion of the population that is in areas designated as “rural” by the census bureau. To examine the distribution of coverage by state, I created a cloropleth map, and to examine the relationship between coverage and rural population, I used a scatterplot.

Results

The percentage of the population in a state covered by agencies reporting hate crimes ranged from a low of 24% to a high of 107%. Values above 100 percent result from differences in the way the census bureau and the FBI estimated 2013 population numbers. Both agencies based their estimates on the 2010 decennial census data, and project population forward to 2013 using specific assumptions. Slight differences in those methods meant that the populations covered by agencies reporting hate crimes (the FBI data) could be higher then the total state population estimated by the census bureau. The distribution of coverage values was highly skewed, with more than half of the states having 100% coverage, and a small number of states having lower coverage:

The interactive map below shows how states differ in the percentage of their popualtion covered by agencies that report hate crime data. Hovering over a state will show the state name and the percentage of the population covered. Alabama, New Mexico, and Mississipi have the lowest levels of coverage, whith only half or less of their population covered.

The fact that most states have all of their population covered by agencies reporting hate crimes means that the relationship between coverage and rural population is unlikely to be significant. The scatterplot below confirms that this is the case.

While there is a slight decline in coverage rural population increases, some states with high proportions of rural residents still have 100% of their population covered by agencies that report hate crimes. The states with the lowest levels of coverage fall in the middle of the distribution for rural population.

Conclusions and limitations

It appears from this analysis that overall, the vast majority of the population in the U.S. is covered by law enforcement agencies that report hate crimes. That bodes well for the validity of statistical conclusions based on those data. On the other hand, some states lag far behind in hate crimes reporting, with half or less of their populations covered by reporting agencies. Those states are not the most rural in the country, although the states with the lowest coverage are clustered in the South and Southwest regions. The proportion of a state’s population in urban or rural areas does not seem to affect its level of hate crimes reporting coverage.

The main limitation in this analysis is that it is based on data from only one year. It is unlikely that the number of agencies participating in the hate crimes statistics program would change dramatically from one year to the next, but a sample over a five-year span, say, would provide more confidence in the generality of the conclusions.