Recently, Forbes posted America’s Most Dangerous Cities of 2015 which attempted to rank the most violent cities with populations over 200,000 persons. Despite urgings by the Federal Bureau of Investigation to refrain from making violent crime rankings, several groups such as the former Morgan Quitno and, presently, Forbes continue to put forth rankings. In this article, we take a closer look at those numbers.
Most Violent Cities with Population 200,000+
Before we begin, let’s put together the basic list and information typically used in crime statistics.
- Detroit, Michigan
- Ranking: First
- Population: 684,694
- Violent Crimes: 13,616
- Violent Crime Rate: 1988.63 per 100,000 persons
- Memphis, Tennessee
- Ranking: Second
- Population: 654,922
- Violent Crimes: 11,399
- Violent Crime Rate: 1740.51 per 100,000 persons
- Oakland, California
- Ranking: Third
- Population: 409,994
- Violent Crimes: 6,910
- Violent Crime Rate: 1685.39 per 100,000 persons
- St. Louis, Missouri
- Ranking: Fourth
- Population: 318,574
- Violent Crimes: 5,348
- Violent Crime Rate: 1678.73 per 100,000 persons
- Birmingham, Alabama
- Ranking: Fifth
- Population: 212,115
- Violent Crimes: 3,369
- Violent Crime Rate: 1588.29 per 100,000 persons
- Milwaukee, Wisconsin
- Ranking: Sixth
- Population: 600,374
- Violent Crimes: 8,864
- Violent Crime Rate: 1476.41 per 100,000 persons
- Baltimore, Maryland
- Ranking: Seventh
- Population: 623,513
- Violent Crimes: 8,246
- Violent Crime Rate: 1338.54 per 100,000 persons
- Cleveland, Ohio
- Ranking: Eighth
- Population: 388,655
- Violent Crimes: 5,186
- Violent Crime Rate: 1334.35 per 100,000 persons
- Stockton, California
- Ranking: Ninth
- Population: 299,519
- Violent Crimes: 3,988
- Violent Crime Rate: 1331.47 per 100,000 persons
- Indianapolis, Indiana
- Ranking: Tenth
- Population: 858,238
- Violent Crimes: 10,768
- Violent Crime Rate: 1254.66 per 100,000 persons
If you are familiar with this list, then you know that this list is identical to the Forbes list. So let’s take a look at how Forbes created their list.
What Goes Into The Simple Model?
First we identify the definition of a violent crime, as defined by the Unified Crime Report (UCR) by the FBI as murders and non-negligent manslaughters, rapes, robberies, and aggravated assaults. Violent crimes are defined as those offenses that involve force or threat of force.
The data that is presented is reported by the individual policing districts to the best of their abilities. That means, the crimes are listed as offenses known to law enforcement. The data that is presented by the FBI is then reported using the hierarchy rule. This rules indicates that if multiple crimes occur simultaneously, such as a robbery and murder, then the more serious crime is reported. The order of seriousness is given in the respective order above.
Rapes also have two separate definitions under the FBI UCR: legacy and updated definitions. The legacy definition is defined as the carnal knowledge of a female forcibly and against her will. In 2013, the definition was updated by dropping the word forcibly. In either definition attempts and assaults in attempt to commit rape are included; however statutory rape and incest are not included. To date some municipalities use the new definition, some use the legacy definition, and some report using a non-allowable definition. For example, cities in Illinois do the latter, which exempted Illinois cities from rankings such as the Morgan Quitno.
The simple model tallies the number of violent crimes and then calculates the rate per 100,000 persons. For example in the city of Simi Valley, California, a city of 126,604 persons, there were 140 total violent crimes reported in the FBI UCR. This total results in 110.58 violent crimes per 100,000 persons.
This is the same method as the famous Morgan Quitno.
In the Morgan Quitno, the number of cities were cut down to populations of 75,000+. This resulted in 444 total cities. The rankings is then built off the highest violent crime rates within this subgroup of 444 cities. Since the Morgan Quitno violent crime report has not been published for several years, we can reconstruct their old listings:
- Detroit, Michigan (1988.63 per 100,000; pop. 684,694)
- Memphis, Tennessee (1740.51 per 100,000; pop. 654,922)
- Flint, Michigan (1708.25 per 100,000; pop. 99,166)
- Oakland, California (1685.39 per 100,000; pop. 409,994)
- St. Louis, Missouri (1678.73 per 100,000; pop. 318,574)
- Birmingham, Alabama (1588.29 per 100,000; pop. 212,115)
- Milwaukee, Wisconsin (1476.41 per 100,000; pop. 600,374)
- Little Rock, Arkansas (1391.91 per 100,000; pop. 198,217)
- Baltimore, Maryland (1338.54 per 100,000; pop. 623,513)
- Cleveland, Ohio (1334.35 per 100,000; pop. 388,655)
Before we continue we note that Camden, New Jersey, is not on the “top 10” list. This is because Camden, New Jersey is not reported in the FBI UCR. In fact, the FBI will not report a city in the Uniform Crime Report if the city is unable to supply statistics to their standards for certain amount of time during the year. Due to this, we also have no results for Honolulu, Hawai’i.
Let’s do this for a larger grouping of cities.
Now if we restrict to populations of 1,000+ persons, we have a slightly different most violent cities:
- Oceana, West Virginia (4175.99 per 100,000; pop. 1,341)
- Wellston, Missouri (3936.67 per 100,000; pop. 2,337)
- Darby, Pennsylvania (3855.15 per 100,000; pop. 10,687)
- East St. Louis, Illinois (3645.89 per 100,000; pop. 26,523)
- Cairo, Illinois (3527.13 per 100,000; pop. 2,580)
- Mangonia Park, Florida (3331.62 per 100,000; pop. 1,951)
- Lithonia, Georgia (3154.73 per 100,000; pop. 1,997)
- Florida City, Florida (2771.84 per 100,000; pop. 12,158)
- West Wendover, Nevada (2677.58 per 100,000; pop. 4,519)
- Iowa, Louisiana (2669.21 per 100,000; pop. 3,147)
We see immediately that Detroit is no longer a most violent city. In fact, in the 1,000+ population groups, Detroit is listed at 22nd.
So Oceana, West Virginia is More Violent than Detroit, Michigan…
By using the traditional rankings, this is indeed the case. Oceana is known for another dubious distinction: Oxycotin abuse. Breaking down the statistics between the two cities we see that Oceana and Detroit are entirely different cities:
Crime Oceana, WV Detroit, MI
Population 1,341 684,694
Murder 0 298
Rapes 0 557
Robberies 1 3,570
Assaults 55 9,191
Thus, despite Oceana being more violent, almost every crime in Oceana was an assault. Rate-wise, Detroit is infinitely deadlier than Oceana. This is a primary problem with rates being used for rankings. In regions of smaller populations, a single assault can skyrocket a violent crime rate. This is a large reason as to why lists select larger populations, such as 75,000+.
However, this shows boundary problems just from a numbers perspective: Cities that may include or exclude specific neighborhoods from reporting (Las Vegas Metro covers 1,000,000+ persons for reporting) are adding and subtracting just these smaller populations. Thus we need to find a stronger method to compare distributions of crimes and probability of crimes occurring at random.
Let’s Compare Houston, TX and Philadelphia, PA
We will take a look at larger cities and compare them directly using basic statistics. Houston, Texas and Philadelphia, Pennsylvania have fairly similar statistics across the board. The rankings for the city, using the Forbes rankings, Philadelphia is listed at 19th with Houston at 22nd. Only Minneapolis, Minnesota and San Bernadino, California are between these two cities.
Crime Philadelphia Houston
Population 1,559,062 2,219,933
Ranking 19 22
Murder 248 242
Rapes 1,207 812
Robberies 6,970 10,186
Assaults 7,500 10,768
Philadelphia has 134.1 square miles of land coverage while Houston has 634 square miles. That is, Houston is approximately 450% the size of Philadelphia. Using this knowledge, we find that the population density of Houston (3,501.47 persons per square mile) is much less than Philadelphia (11626.11 persons per square mile). So despite the crimes occurring at roughly the same frequencies, the spatial distributions are much different.
By considering all murders as random events, we maximize the probability of an average person of becoming a victim of murder. In this case, since people are typically 3-4 feet wide, we need to take into account spatial distributions; IE: population densities. Using the common rule of thumb that there are typically 17 city blocks per mile, or 289 square city blocks in 1 square mile, we see that Houston has approximately 12 people per average city block. For Philadelphia this is approximately 40 people per average city block. If we erroneously assume uniform distribution of murders across the entire city (since it’s simple), we see that Houston experiences 0.00132077 murders per city block. This leads to a person expecting a 0.0109% chance of becoming a victim of murder at random. In Philadelphia, this number is 0.015907% chance. This suggests that Houston’s murder rate is truly 31% less than that of Philadelphia, when uniform spatial distribution is considered.
Comparing this to the numbers reported in basic rankings models, Philadelphia has a rate of 15.907 per 100,000 versus Houston’s rate of 10.901 per 100,000; which is 31% less than that of Philadelphia. This is identical to the spatial distribution! This is expected as we assumed a uniform distribution of crimes. From our previous posts on Oakland, Milwaukee, and Washington D.C., this is not necessarily the case.
So How Do We Rank?
We then rank by taking into account a method for estimating distributions of violent crimes. In the case of Washington, D.C., most crimes occur in three specific regions of the city, less than 30% of the actual city. In fact, downtown D.C. experiences very little crime when compared to most major cities. If we take an expected value based on the distribution of violent crimes, we will see that tighter concentrated cities will have a better “crime ranking.” This suggests that cities have contained crime.
However, we haven’t touched on many factors that aren’t available by the FBI UCR. For instance, economic factors lead to crime; drug availability affects crime; social factors affect crime. More importantly, targeting dramatically increases or decreases probabilities for the average person to become a victim of a crime.
Since these methods are difficult to quantify, we merely attempt to collect probabilities of such events occurring and measuring their correlative effect on the crime rates in the specific areas.
For our black box purposes, we will only make a reference towards such research and treat our results as spatial analysis devoid of unquantified social factors.
OK, so what about Oceana, WV?
While we focused on using spatial distributions having an impact on rankings, we still note that Oceana is considered “more violent” than Detroit. Recalling that this was due to smaller sample sizes and almost all violent crimes being assaults, we note that there is a need to weight the type of violent crime.
Basic Rankings Adjustment With Weights.
The FBI UCR inherently does this through hierarchy: Murder, Rape, Robbery, Assault. Therefore, we should weight the crimes accordingly. One method to do this is weight according to minimum first time offender sentences: Murder 20 years, Rape 5 years, Robbery 4 years, Assault 1 year.
Using the weight redistribution with a uniform distribution, the most violent cities in the U.S. for populations over 75,000+ would then be given as follows:
- City, State (weighted score)
- Oakland, CA (160.24)
- Detroit, MI (156.84)
- St. Louis, MO (148.28)
- Cleveland, OH (148.21)
- Memphis, TN (131.91)
- Baltimore, MD (130.26)
- Birmingham, AL (129.46)
- Newark, NJ (128.23)
- Milwaukee, WI (126.11)
- Flint, MI (118.22)
- Buffalo, NY (114.04)
- Trenton, NJ (112.27)
- Atlanta, GA (109.57)
- Jackson, MS (108.74)
- Washington, D.C. (108.15)
- Little Rock, AR (106.20)
- Indianapolis, IN (105.13)
- Gary, IN (103.62)
- New Orleans, LA (103.38)
- Philadelphia, PA (99.15)
- Cincinnati, OH (98.71)
- Kansas City, MO (98.05)
- Minneapolis, MN (97.67)
- Stockton, CA (97.37)
- New Haven, CT (94.26)
So we see the adjustments for severity of crime, but we take into account the fact that our rankings still use a uniform distribution. As a note, Houston drops down to 29th in the weighted rankings.
What do we learn?
The thought study of crime rankings is to show that we cannot simply use rates to justify the rankings for a city’s crime. Instead, we need to look at the spatial tendencies, the population distribution, and other various factors that go into determining a probability of a crime occurring. In doing this, we can then start to answer the more difficult, but more effective question: can we predict crimes based on social and spatial criteria.
The answer is yes, but to an extent. We can use these probabilities to rank crimes, but also give weight to where crimes are more likely to happen. This would in turn help determine required resources and hopefully improve the quality of policing specific regions of cities. In the end, the information is to provide quality assessment of crime occurrence; as opposed to xenophobia over personally, potentially becoming a victim; which has already been shown to be a low probability event to begin with… even in one of our “most violent” cities.