My doctoral dissertation was a study of homicides in Baltimore between 2005 and 2017 in order to better understand what is going on. Right off the start, I used data to show that, indeed, there is an epidemic of homicides in Baltimore that started right around April of 2015, and that intensified since the unrest following the homicide of Freddie Gray Jr. From there, the dissertation was presented in three parts.
The first part of the dissertation looked at the 3,366 homicide victims reported between January 1, 2005, and December 31, 2017. The age, gender and other individual characteristics of those victims were analyzed. Here are the major findings:
- The majority of homicide victims were African American males between the ages of 15 and 34. Even when adjusting for population differences between age groups, this group constituted the majority of victims, and most of them were killed by firearm.
- Female victims were less likely than their male counterparts to be killed by firearm, and they were more likely to be victims of intimate partner violence. They were also more likely to have been killed at home.
- Most homicide victims were also unemployed or under employed, and the proportion of them who did not finish high school was significant.
- Adjusting (accounting) for differences in the population by age group, there were eight (8) African American homicide victims for every two (2) Hispanic victims and for every one (1) White victim.
The second part looked at the neighborhood characteristics where the homicides happened. For that, I took the addresses of the homicide locations and geocoded them, overlaying them on a map. That allowed me to determine in which Community Statistical Area (CSA) the homicide happened. I then took the characteristics of the CSAs and looked for any trends and associations between those characteristics and the homicide rates. Here are the major findings:
- Adjusting (accounting) for other characteristics, poverty is associated with an increase in the homicide rate at the CSA level. For example, a CSA with 20% of households under the poverty level had a homicide rate about 21% higher than a CSA with 10% of households under the poverty level.
- Adjusting (accounting) for other characteristics, physical disorder (e.g. broken street lights or trash on the street) is associated with an increase in the homicide rate at the CSA level. For example, a CSA with a physical disorder index of 2 will have a homicide rate about 47% higher than a CSA with a physical disorder index of 1.
- The 18 poorest neighborhoods had 50% of the homicides during the study period (2005 to 2017). The 18 wealthiest neighborhoods had 10% of the homicides during the study period. Both groups (the 18 poorest and the 18 wealthiest) each have about 32% of the population, so it’s not a matter of differences in the number of people in those neighborhoods.
- Geographic hot spots of homicides varied by time. That is, where a hot spot is located in Baltimore City depended on when the homicides happened, with some hot spots disappearing and others appearing. Also, when looking only at homicides of African American men between 15 and 34 years of age, the hot spots were different than when looking at all homicides.
- When looking at these data, it is important to take person, place and time into account. You have to look at data in four dimensions, not one or two or even three.
The third and final part looked at the available violence interventions in Baltimore City and how they address the individual and neighborhood characteristics associated with homicides described in the first two parts. Unfortunately, there is no one master list of interventions in Baltimore City. There are many organizations, directed by the government and civil society (or a combination thereof), and they are sometimes duplicating their work. Here are some of the key findings:
- Intervention programs are not being periodically evaluated for effectiveness, for the most part. And, if they are, those results are not readily available to the public. For interventions to succeed and be effective and efficient, they need to be evaluated, and the results from those evaluations need to be used to modify the intervention.
- Many interventions are not based on objective data, or there is no objective measure done to know if those interventions work. There may be some anecdotal evidence, but it doesn’t take into account modifying factors or confounders that could be obscuring the true association between the intervention and the outcome.
- I also proposed a rating system for interventions. It’s only a first step in how to rate them, something I hope will get the conversation started.
If you’re interested in watching me present all of this, please check out the video below.