Health and Society, Life in Cities
giStrat | Jul 2016

Murder in America

Updated: Jul 22
  • UPDATED: Jul 22

    What Game Theory Tells Us About Reducing Violence Between Police and Communities in America

  • PUBLISHED: Jul 22

    Murder in America

     

    The United States has some of the most violent cities in the world.

    St. Louis (15), Baltimore (19), Detroit (28), and New Orleans (32), all rank within the global top fifty. This is a stark contrast to places like Europe, Asia, Australia, and Canada, all of which do not have a single city in the top fifty. In fact, the United States, South Africa, and Jamaica are the only countries outside of Central and South America to make the list.

    Compared to the rest of the developed world, American cities are unusually violent. An enormous amount of academic, anecdotal, and speculative analysis has been devoted to trying to explain this problem. We decided to conduct our own analysis, reviewing the academic literature and running statistical models to create a risk index that forecasts the cities that will be the most violent in the near future–and, more importantly, to identify the underlying causes of these crimes.

    This is the first step in a larger analysis we are undertaking with the goals of identifying cities at high risk, understanding the factors driving murders in those cities, and then discovering the most successful efforts to combat those factors and mapping a path forward for solutions.

    Here is what we have discovered so far.

    Download the Full the Report Here

What Game Theory Tells Us About Reducing Violence Between Police and Communities in America

giStrat applied game theory to model the ongoing cycle of violence between the police and the citizenry, particularly African-American communities, as part of our series on violence in America. We identified key actions needed to reduce the lethal use of force by the police and end retaliatory actions by civilians. Our analysis suggests the following:

  • Transparent in-group accountability mechanisms can effectively reduce the lethal use of force by the police and end retaliatory killings of police officers.
  • Police in-group accountability may include greater oversight on the excessive use of force by fellow police through measures that include: 1) body cameras worn by police; 2) reformed sentencing and conviction laws for shooting unarmed citizens, particularly for cases in which the officer is involved in multiple shootings of unarmed suspects; 3) a federal database of officers involved in multiple shootings of unarmed suspects; and 4) public condemnation of the excessive use of force by fellow police officers.
  • Citizen in-group accountability, especially for African-American communities, will require public condemnation of retaliatory actions against the police, and greater community representation within law enforcement. A police force should demographically reflect the community that it seeks to protect.

Excessive Use of Force by the Police and Retaliatory Actions by Civilians

As mobile phone cameras have become more prevalent, police departments across the country are under greater scrutiny for the excessive use of lethal force, especially against African-Americans. The use of social media enables people to easily share footage of these killings recorded by witnesses. This magnifies the grievances of African-American communities nationwide by bringing examples of police violence to the attention of the nation. Viral recordings of these killings have inspired lone individuals across the country to retaliate by shooting police officers.

The deaths of Eric Garner in New York and Michael Brown in Missouri sparked mass protests against the disproportionate use of lethal force against black men and women by police officers. The fatal shootings of Philando Castile and Alton Sterling in Minnesota and Louisiana reignited calls for public demonstrations against police brutality. During one of these marches in Dallas, violence escalated when an African-American man murdered five police officers in retaliation against police violence. Ten days later, another African-American man with similar motives shot and killed three police officers and wounded three more in Baton Rouge, Louisiana.

One day later an unarmed African-American man was shot while lying on his back and holding his hands up in the air. Fortunately, he was not killed.

Intergroup Violence Spirals

Game theory defines these ongoing retaliatory responses as inter-group violence spirals, characterized as events of short duration in which two or more groups commit violent acts in response to ongoing grievances. Repeated acts of police brutality and retaliation by the citizenry lead to social fractionalization, which exacerbates violence between police and the aggrieved communities at the national level.

A growing number of Americans distrust the legitimacy of our legal system because law enforcement officials often fail to hold officers accountable for their actions. According to the Washington Post, despite “thousands of fatal shootings at the hands of police since 2005, only fifty-four officers have been charged.” Most of these officers were cleared or acquitted. giStrat used a game theory model for violence spirals to show that greater police accountability can successfully disrupt the cycle of violence.

There are two possible outcomes when a dispute or single act of violence arises between groups: 1) violence breaks out en masse because one group retaliates against the transgression of a person from another group by holding responsible all members of that group; 2) violence does not escalate because an in-group accountability mechanism mediates the conflict. If a member of law enforcement shoots and kills an unarmed citizen under questionable circumstances, law enforcement and courts of law keep the peace by identifying and punishing the perpetrator. In the absence of this mechanism, violence escalates. Various studies have shown that current legal mechanisms fail to respond justly to fatal police shootings. Under these circumstances, we can expect escalating violence in periodic tit-for-tat violence spirals.

How Violence Spirals 
  • A community member retaliates indiscriminately against law enforcement personnel for the transgression of one particular law enforcement official.
  • Mass or indiscriminate retaliation may ensue by law enforcement officials, resulting in a complete breakdown of inter-group relations.
  • Violence escalates, resulting in a violence spiral.

Obstacles to In-Group Accountability: A Case Study

The current in-group accountability mechanisms within the law enforcement community have resulted in few prosecutions and even fewer convictions. Officers tend to protect each other in matters of misconduct, a practice so rampant that terms such as “testilying,” “the blue code,” and “the blue wall of silence” are used to describe the police culture around accountability. In California, where about one-tenth of all officers in the country work and reside, state laws have barred prosecutors from accessing police personnel files. Under this system, dishonest officers are allowed to testify and remain on the job, and defendants are denied impeachment evidence. Too often officers who choose to become whistleblowers are taunted and refused backup.

The public is increasingly aware of cases in which an officer’s initial reports are later found to be false after video evidence is discovered. For example, in the recent shooting death of Laquan McDonald, officers initially stated that McDonald was slashing tires with a knife and, when confronted, lunged at an officer with said knife. The officer claimed to shoot McDonald in self-defense. An organization later found dashboard video evidence that proved McDonald was shot while walking diagonally away from officers, holding the knife in the hand furthest from them. McDonald was shot sixteen times, contradicting officer claims of self-defense. Civilian witnesses were shooed away from the scene by law enforcement. One police officer deleted footage from a security camera at a Burger King in the vicinity of the incident. Many officers remained silent or engaged actively in the cover up. This system needs to be replaced with one in which officers hold each other accountable and discipline wrongdoers for misconduct. 

Ending the Violence Spiral: Reducing the Excessive Use of Force and Deadly Retaliation

The violence spiral can be broken by creating and enforcing in-group accountability measures, an approach proposed by political scientists James Fearon and David Laitin. This approach suggests that when an unarmed and innocent member of a community is killed by a police officer, the law enforcement agency itself must condemn the attack, sanction the offending officer, and support all efforts to serve justice. This will reduce violent retaliation by the community.

Escalating violence between police and civilians is not inevitable. Our game-theory analysis is consistent with the academic literature, showing that an in-group accountability mechanism can effectively reduce the lethal use of force by the police and end deadly retaliatory actions by the citizenry.

In-Group Accountability Mechanisms for Police
  • Body cameras worn by officers on patrol.
  • Reformed, transparent conviction laws for shooting unarmed citizens.
  • Federal database of officers involved in multiple shootings of unarmed suspects.
  • Public condemnation of the excessive use of force by the police officer.

To date, not a single police officer since 2014 has been found guilty for the 391 incidents in which an unarmed citizen was killed. In these incidents, 37% of the suspects shot were African-American. For communities disproportionately affected by police violence, in-group accountability will require greater community representation within police forces. A police force should demographically reflect the community that it seeks to protect.

Numerous law enforcement officials have called for more diverse representation in police forces. At a press conference following the retaliatory murder of five officers, Dallas Police Chief David Brown stated, “We’re hiring. Get off that protest line and put an application in… We’ll put you in your neighborhood. And we’ll help you resolve some of the problems you’re protesting about.”

In-Group Accountability Mechanism for Communities
  • Greater proportion of police officers originating from communities they serve.
  • Public condemnation of violence or retaliatory actions against the police.

gametheorytree4

Our game theory model shows an obvious but important reality: the absence of fair, transparent, and honest policing obstructs the path to building more stable, safer communities while also endangering the lives of police officers. Lack of justice and accountability pushes the communities to resort to retaliation.

The best way to disrupt this cycle of violence and reach the optimal outcomes for both parties is to create in-group accountability mechanisms within law enforcement and the communities they police.

Share this update

An Econometric Analysis of Violence in America

 

The United States has some of the most violent cities in the world.

St. Louis (15), Baltimore (19), Detroit (28), and New Orleans (32), all rank within the global top fifty. This is a stark contrast to places like Europe, Asia, Australia, and Canada, all of which do not have a single city in the top fifty. In fact, the United States, South Africa, and Jamaica are the only countries outside of Central and South America to make the list.

Compared to the rest of the developed world, American cities are unusually violent. An enormous amount of academic, anecdotal, and speculative analysis has been devoted to trying to explain this problem. We decided to conduct our own analysis, reviewing the academic literature and running statistical models to create a risk index that forecasts the cities that will be the most violent in the near future–and, more importantly, to identify the underlying causes of these crimes.

This is the first step in a larger analysis we are undertaking with the goals of identifying cities at high risk, understanding the factors driving murders in those cities, and then discovering the most successful efforts to combat those factors and mapping a path forward for solutions.

Here is what we have discovered so far.

Download the Full the Report Here

Analysis

Many American cities with high murder rates have the following characteristics:

  • Disrupted families (high divorce rates)
  • High poverty rates
  • Low high school graduation rates
  • Large proportion of young adults (Age 20-34)

Looking into the factors that often accompany those listed above, we start to get a more detailed picture of the landscape in which high murder rates often exist.

Broken Windows and More

How smaller crimes play a role in greater crimes

Many people are familiar with the broken windows theory of crime. It was popularized in Malcolm Gladwell’s bestseller The Tipping Point but had been around for decades before the book’s publication. The theory has its detractors. However, our analysis shows that the general idea of a connection between less serious crimes and murders does exist.

Our findings indicate murder rates rise through an urban decay process that begins with less violent crime, eventually elevating to more violent ones. In evaluating 150 American cities we find a statistically significant linkage between property crimes leading to increases in robberies, which in turn lead to more murders. This is in line with the classic Broken Windows theory of crime, whereby less violent crimes such as burglary, larceny, and vandalism set an urban atmosphere where disregard for the law becomes a norm.

Through detailed econometric analysis we established the following link with a high level of statistical significance.

progressionchart

This chart shows the major factors our research revealed as being associated with each type of crime, including the less violent crimes as a major factor in more violent crimes.

Property Crime

Our analysis shows the main statistical drivers for property crime in a city are:

  • A large youth population
  • High divorce rates
  • High levels of poverty
  • Low high school graduation rates

When all these factors are in place we see high levels of property crime. We can then see that property crime becomes one of the main factors present in more serious crimes such as robbery.

Young Adults

Our findings indicate cities with a higher proportion of young adults (age 20-34) experience increased rates of property crime but are not directly linked to murder rates. In accordance with the academic literature, urban areas with larger youth cohorts tend to experience higher crime rates partially due to crowding of the labor market.

Robbery

Robbery is primarily associated with:

  • High levels of property crime
  • Low high school graduation rates
  • High divorce rates

Murder

Finally we found the following factors associated with high murder rates:

  • High rates of robbery
  • High divorce rates
  • Large overall population in the city
  • Low high school graduation rates

Divorce and Families

When considering the process of urban decay, as measured by property crimes, robberies, and murders, the most significant societal factors across America are the divorce rate and poverty. Divorce rates are typically an indicator for disruption in the family unit. Coupled with deprivation of resources, as measured by the poverty rate, we found that African American communities in particular are at the highest risk of falling prey to cycles of crime and violence.

Low High School Graduation Rates

Low high school graduation rates are also a significant indicator for areas with high robbery and murder rates. This contributes to higher unemployment and lower life expectancy, and is highly correlated with divorce and poverty rates.

Race

Areas with large African American populations often coincided with high levels of crime in our analysis. However, it is important to remember that these populations experience higher levels of poverty, lower graduation rates, and higher divorce rates, all of which are themselves factors listed above. In addition, historical factors such as legal and housing discrimination, many of which still occur or have results that can be seen today, caused us to exclude race from our analysis.

Cities at Risk

To better predict which cities are at the greatest risk, we created a risk index. This index is composed of two main components:

  • Event Risk – based on murder rates in these cities in the past
  • Structural Risk –  projections based on the factors we identified above

These two rankings were combined to produce overall risk as shown in the table below. Factors were normalized to a 0-100 score, with 100 being the highest level of risk.

MurderChart

Here is the list viewed on a map:

map_linecentered

Based on the rankings above we looked more closely at a few cities facing the greatest risk.

Detroit

Detroit ranks at the top of our murder risk in America when accounting for both socio-economic structural factors and the actual murder rate. 

The most significant drivers of Detroit’s murder rate are poverty coupled with low high school graduation rates. It is by far the poorest city in America with nearly 35% of all families living under the poverty line, and it has the highest unemployment rate at 14.4%. Out of the top five highest risk cities, Detroit has the highest high school dropout rate (22%) coupled with only 13% of young adults receiving a college degree or higher. 

St. Louis

St. Louis is ranked second in murder risk. In 2015 it had the highest murder rate at approximately 38 murders for every 100,000 people, making it the bloodiest year for the city in over twenty years. Of the top five highest risk cities, St. Louis has the largest youth cohort with more than 27% of the population comprised of young adults aged 20-34. This creates a crowding effect and makes competition for jobs even more difficult in a city with high unemployment and more than 22% of families living below the poverty line.

Jackson

Out of the top five highest murder risk cities, Jackson has the highest divorce rate (13% of females and 11% of males age 15 or higher). This is an indicator for a disruption in the family unit and an absence of father figures for young adult males. Although Jackson has a relatively low unemployment rate of 7.8% compared to America’s most violent cities, nearly 25% of households live below the poverty line with African Americans comprising of 80% of the population. A highly homogenous minority population coupled with high levels of poverty tends to be a significant driver of homicide in America. Most homicides in America are intra-racial, with 84% of white victims killed by whites, and 93% of black victims killed by blacks.

Baltimore

Baltimore has characteristics similar to the other top five murder risk cities ranging from high poverty rates to low high school graduation rates. However, what sets Baltimore apart in comparison to these other cities is its population density. It has a density of 7,672 people per square mile, more than 2,500 people more than the next densest city (St. Louis). Population density is often viewed as a significant factor in murder rates and general crime by making it more difficult for police to catch and identify perpetrators.

New Orleans

New Orleans has the second highest divorce rate among the top five highest risk cities (just below 13% of female and 11% of males age 15 or higher), a significant indicator for disruption in the family unit. It has most of the same characteristic socio-economic drivers of murder in comparison to the other top cities.

Conclusion

The unusually high murder rate in American cities is a complex problem caused by a number of intertwined social factors. Each city must address the problem with its own targeted solutions, and we hope to elaborate on this with a detailed example in part two of this report. For now we can look at the major factors that need to be addressed nationally.

When analyzing data across 150 major cities, we discovered that poverty and the breakdown of the family consistently emerged as systemically linked to the crime and murder rate in America. Although there are numerous drivers of violence, efforts focused on reducing poverty and divorce rates in America will produce the greatest results on reducing murders, robberies, and property crimes.

Methodology

We applied four steps to analyze this problem.

First, we aggregated rates on murders, robberies, and property crime across 150 American cities. Then we identified general factors that have a medium to high degree of consensus among scholars as drivers of homicide in the United States.  Next, we tested these factors across the most recent data to determine the extent of their impact on the murder and forecasted rate of murders in 2016. Finally, we built an empirical index that ranked the murder risk across the 150 cities.  Once we completed this analysis, we extrapolated what factors must be present for a high murder rate and what adjustments can be made to bring these rates down.

We applied OLS (Ordinary Least Squares) regressions to test the relationship between murder rate and several other socio-economic factors. The variable we tested consisted of population, population density, percentage of people aged 20–34 (youth cohorts), education levels (percentage who did not complete high school and percentage who have a bachelor degree or higher), divorce rates (both male and female older than 15), unemployment rates, and poverty levels.  We also evaluated percentage of African American or Hispanic populations given that most of the expert literature places these minority populations at highest risk. We controlled for whether the city is in the south to determine if there is any geographical bias. Last, we also tested for different types of crime that may contribute to the murder rates such as robberies and property crimes.

In order to evaluate the future risk of murders across the most violent American cities, we considered both the number of murders that occurred in the cities along with statistically significant socio-economic factors (structural risk). For structural factors, we used the regression coefficient of each factor multiplied by the base value of the corresponding factor for each of the 150 cities. The weighted factor was added to derive a total score for each city.  We then normalized the total score from 0 to 100. We also normalized the murder rate per 1,000,000 from 0 to 100. We then combined the murder and structural risk scores to derive overall risk.

Contributors

Amir Bagherpour, PhD, Chief Political Scientist
Jeff Li, Research Associate

Share this project
Top