A Localized Culture of Violence

H.L. Mencken once said “For every complex problem there is an answer that is clear, simple, and wrong.” This quote seems applicable to the activists who blame a “culture of violence” for the high rate of firearms-related homicides in the United States and propose restrictions on or elimination of 2nd Amendment rights as the solution. A previous article “Guns Don’t Kill People, Democrats Kill People” examined three diverse states and demonstrated that the majority of firearm homicides in 2013 occurred in the small fraction of the voter precincts that overwhelmingly voted for Democrat candidates.This dataset of homicides will be mapped against socioeconomic Census data to quantify the characteristics of the neighborhoods where firearm homicide is prevalent. The intent is not to generate counterintuitive or surprising results, but to provide data that supports the premise that the United States does not have a systemic firearm violence problem.  Then, by identifying the characteristics of these localized areas, suggest that more effective, albeit more difficult, approach to reducing firearm homicides would be to address the root causes of the “culture of violence” that appears to exist in these neighborhoods.

The previous article describes how a dataset containing a year of firearm homicides was compiled for Louisiana (a solidly red state with the highest firearm homicide rate in the country), Virginia (an election battleground state with a homicide rate that is close to the average firearm homicide rate for the country), and Minnesota (a solidly blue state with a low firearm homicide rate).  The location of each firearm homicide was mapped to a voting precinct.  The results indicated that over half of the firearm homicides occurred in the small percentage of voting precincts that gave President Obama more than 70% of the vote in the 2012 election.

This same dataset is usedherein to map the location of each firearm homicide to a census tract. The Census Bureau defines a census tract as a “small, relatively permanent statistical subdivision of a county that contains an average of 4,000 people.”  The Census database contains census tract socioeconomicfactors (e.g., population, area, per capita income, ethnic composition) to provide a characterization of the neighborhood where the firearm homicides took place and to identify similarly composed census tracts.  This dataset does not presume to describe the characteristics of any individuals committing the firearm homicides or their victims; the publically available databases found to date have this information obscured to preclude the ability to make any such assessment.

The first breakdown was based on the percentage of the census tract population that was “Black or African American” (Census ID DP05).  The firearm homicide rate (per 100,000 people) was plotted as a function of the percentage of the census tract population that was African American, as shown in Figure 1.  The firearm homicide rate was determined by considering all census tracts with similar population compositions.  The results for the three states are given as blue, purple, and red, for Minnesota, Virginia, and Louisiana, respectively.  The firearm homicide rates are below 1 per 100,000 for census tracts with few African American residents and rises almost monotonically to rates above 20 for increasing African American population composition. Note that none of the census tracts in Minnesota had an African American population that exceeded 90%, hence the firearm homicide rate is displayed as zero. A firearm homicide rate of less than 1 per 100,000 is similar to that of European countries with severe restrictions of firearm ownership rights and rates approaching 20 are similar to that of the most violent Third World countries.The majority of the firearm homicides and the highest firearm homicide rates occur in the small subset of the census tracts with an African American racial composition of greater than 30%.


The influence of census tract per capita income (Census ID DB03) and population density was examined in the plots shown in Figures 2 to 4 for Virginia, Minnesota, and Louisiana, respectively. The census tracts with racial composition greater than 30% African American were shown Figures 2a, 3a, and 4a, while those in Figures 2b, 3b, and 4b have a racial composition less than 30% African American. The black and white symbols indicate census tracts with firearm homicides.  The dashed black lines highlight the census tracts with a per capita income less than $25,000 (lowest quadrant) and a population density greater than 640 people per square mile (1 per acre).  The census tracts within the dashed black lines consist of the lower income, urban neighborhoods.  Two observations are immediately apparent from the data in the three figures:

  1. The firearm homicides that occurred in census tracts with a mostly white racial composition (less than 30% African American) were randomly distributed across the per capita income and population density spectrum.
  2. The firearm homicides that occurred in census tracts with an African American composition of greater than 30% were highly concentrated in lower income, urban neighborhoods.

The localization of firearm homicides in census tracts characterized by lower incomes (LI), high density (HD), and predominantly African American populations is highlighted in Figure 5.  These census tracts, with about 15% of the population and about 1% of the land area, account for up to 67% of the firearm homicides.  Two additional observations can be made from the data in this figure:

  1. The firearm homicide rates are significantly lower for census tracts with predominantly African American populations that do not have the LI/HD socioeconomic characteristics than for the predominately African American census tracts with LI/HD socioeconomic characteristics.
  2. The census tract with predominantly White populations that have LI/HD social-economic characteristics do not have the dramatic increase in firearm homicide rates that were observed in the LI/HD African American census tracts.

The data for Minnesota, Virginia, and Louisiana indicate that 67% of the firearm homicides (a rate of more than 14 per 100,000 people) occurred in neighborhoods with a racial composition of more than 30% African American, a per capita income of less than $25,000, and a population density greater than 640 people per square mile. The firearm homicide rate in the remaining 99% of the land area was 1.3 per 100,000 people. Predominantly African American communities that are not LI/HD and predominantly white communities that are LI/HD do not exhibit the Third World levels of firearm violence seen in predominantly African American LI/HD communities.  Thus, the United States may have a culture of violence, but one that is isolated in easily identifiable communities.

The results from Virginia, Minnesota, and Louisiana bring into question the veracity, pragmatism, or true objectives of activists who propose broad restrictions of Constitutional rights when high firearm homicide rates are almost exclusive to less than 1% of the land area and a small fraction of the population. A more logical and effective approach would be to address the root causes of the culture of violence in these communities, but this is more difficult and the required introspective self-analysis appears to be outside of the comfort zone of most community activists.

David Waciski is a “Big Data” engineer and writer. He can be contacted on Twitter @DWaciski.