For the last one year or so, you probably have seen reports of crime cases on the rise in the news, received security alerts about crime incidents in your community, or had a discussion with others about the worsening crime conditions in the city. So, were there more crime incidents now than a couple years ago? How bad did things get? Why do we feel less safe living in the city now than, say, two years ago? With these questions in mind, I decided to look into the police department's victim based crime data to get a better grasp of the facts.
The data set I downloaded from the Open Baltimore website contains full-year data from 2012 to 2016. There are 241,782 records, each of which is defined by 13 fields describing the day, time, type, weapon involved, neighborhood, and address of the incident. Here is a snapshot of the first few rows in the original date set.
| CrimeDate | CrimeTime | CrimeCode | Location | Description | Inside/Outside | Weapon | Post | District | Neighborhood | Location 1 | Premise | Total Incidents |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 12/31/2016 | 23:51:00 | 4C | 1600 E 25TH ST | AGG. ASSAULT | O | OTHER | 342.00 | EASTERN | Darley Park | (39.3161900000, -76.5938800000) | STREET | 1 |
| 12/31/2016 | 21:27:00 | 4E | 2800 DENHAM CIR N | COMMON ASSAULT | I | HANDS | 922.00 | SOUTHERN | Cherry Hill | (39.2454000000, -76.6287500000) | ROW/TOWNHO | 1 |
| 12/31/2016 | 21:25:00 | 4F | 1300 W LAFAYETTE AVE | ASSAULT BY THREAT | I | NaN | 713.00 | WESTERN | Harlem Park | (39.2989900000, -76.6393400000) | ROW/TOWNHO | 1 |
| 12/31/2016 | 21:10:00 | 3AF | 4500 FAIRFAX RD | ROBBERY - STREET | O | FIREARM | 641.00 | NORTHWESTERN | West Forest Park | (39.3187100000, -76.6910800000) | STREET | 1 |
| 12/31/2016 | 21:05:00 | 4E | 2400 MOSHER ST | COMMON ASSAULT | O | HANDS | 723.00 | WESTERN | Bridgeview/Greenlawn | (39.2995700000, -76.6552900000) | STREET | 1 |
| 12/31/2016 | 20:53:00 | 4E | 3500 ROUND RD | COMMON ASSAULT | I | HANDS | 922.00 | SOUTHERN | Cherry Hill | (39.2424300000, -76.6226600000) | ROW/TOWNHO | 1 |
| 12/31/2016 | 20:45:00 | 4E | 5000 CURTIS AVE | COMMON ASSAULT | O | HANDS | 911.00 | SOUTHERN | Curtis Bay | (39.2208200000, -76.5864600000) | STREET | 1 |
| 12/31/2016 | 20:30:00 | 5A | 3200 E LOMBARD ST | BURGLARY | I | NaN | 222.00 | SOUTHEASTERN | Patterson Park Neighborhood | (39.2913000000, -76.5708500000) | ROW/TOWNHO | 1 |
| 12/31/2016 | 20:30:00 | 5A | NaN | BURGLARY | I | NaN | 513.00 | NORTHERN | NaN | NaN | ROW/TOWNHO | 1 |
| 12/31/2016 | 20:07:00 | 3AF | 4500 PEN LUCY RD | ROBBERY - STREET | O | FIREARM | 822.00 | SOUTHWESTERN | Uplands | (39.2876100000, -76.6922500000) | STREET | 1 |
For this project, I decided to use python and its associated libraries to clean up and analyze the data, use HTML, CSS, and JavaScript to write the webpage report, and, of course, use D3.js for visualization. In this way, the report will have a much smaller file size, which will reduce the time for download and processing in the browser, and ultimately improve the user experience.
Another disclaimer worth mentioning is that this report seeks to identify year-to-year changes. In other words, patterns that were repeated over each of the five years will not be presented here. The list includes variations in the total number of cases with respect to the change in season (warmer months tend to have more incidents), hour of a day (the number of cases follow the pattern of human activity in a 24-hour period), and the temporal and geographical variations of types of crime not mentioned below, etc.
Let's start with getting a big picture of the volume of crime cases. You will be looking at the total number of cases in each of the five years in the following plot. Hover your mouse over (or tap if you are on a mobile device) one of the bars and you will see how the total number of crime cases has changed compared to the 2012 level.
As you can see, the total number has not changed much from 2012 to 2016. If anything, the data show that the number of crime cases has dropped! Clearly, the volume of crime, irrespective of its nature, is not the answer to the question why we feel less safe year after year. In order to find the answer, I am going to dissect the data by its features. I will first check if the nature of crime has changed. Then, I will look at whether the choice of weapons in crime incidents have evolved. In each of these cases, I will look at the geographical distribution of crime across city neighborhoods, with an emphasis on identifying neighborhoods that experienced the largest spikes in crime over the last five years.
Nature of Crime
At first glance, it may appear that the nature of crime in the city has not changed since 2012, with larceny, common assault, burglary, and larceny from auto continue to be the top four most common types of crime in the city. With a bit more scrutiny one would soon realize that, despite their dominance, the annual totals of these four types of crime have in fact decrease sharply (at the level of thousands) over the last couple of years. This decrease, unfortunately, is accompanied by a large increase in the number of violent crimes, such as street robbery, which has had an increase of over 1000 cases since 2012. Other types of crime, while having small base numbers, have also seen steep increase in 2016. For example, in 2016 there were 123% more carjacking, 80% more shooting, 50% more commercial robbery, and 47% more homicide than in 2012.
The following graph depicts the progressions of the annual totals of violent crimes in the last five years. Circles with same color represent the same type of crime, and the size of the circles is proportional to the annual total. Hover or tap any of the circles, the annual totals of that type of crime in the last five years will be shown at the lower right section of the graph.
From this graph, one can see that robbery have overtaken aggravated assault to be the most common violent crime in the city, totaling almost 5,500 cases in 2016. While the numbers of assault by threat and rape have decreased moderately, shooting and homicide numbers are on the rise.
Types of Robbery over The Years
| Crime | 2012 | 2013 | 2014 | 2015 | 2016 |
|---|
This above table shows that while carjacking and commercial robbery saw the steepest percentage increase, street robbery had the largest case number increase (1000+ cases).
Given that robbery incidents are inherently confrontational, and it seems like its increase could be a major factor in contributing to our sense of insecurity. Let's dig deeper to see how things change over time.
To view the time lapse of robbery cases over the years, I will use a calendar to visualize the change in number of robbery cases each day in the past five years. In the following plot, each block represent one year's date, with each cell in the block represents one day in the past five years, and the color of that cell is calibrated by the total number of robberies committed that day. The days are arranged such that the cells in the first row of the block correspond to Sunday, and the the cells in the last row of the block correspond to Saturday. All the days in a given month are grouped together by a thick black border.
Hover your mouse over any cells in the calendar, the full date and the number of robbery cases recorded for that day will be shown in the tooltip.
From the change in hue of the cells, we can see that starting around May 18, 2015, there have been more days with large numbers of robbery cases comparing to the same period of time in previous years. This evolution is both understandable and mind-boggling. It is understandable because we remember the historic civic unrest in the city on April 27, 2015 following Freddie Gray's death. On the other hand, it is mind-boggling because in the three weeks immediately following the day of the unrest, the city, looking through the lens of robbery cases, was relatively calm and experiened robberies at a level that was consistent with those in previous years. However, on the start of the fourth week, things got much worse suddently, and the total number of robbery cases has stayed high ever since. It is worth noting that the cold temperatures in spring tend to have a damping effect on the number of robbery cases, but for 2016 this effect only had a brief presence between the last week of January and March 15.
Robberies by Neighborhood
Another way to understand crime is to look at the spatial distribution of cases. The following set of maps use color to show the yearly aggregates of robbery cases in the city's 278 neighborhoods. The same scale is used to assign color to each city's neighborhood across different years with darker colors correspond to more robbery cases.
Hover your mouse over any point on the map, the name of the neighborhood and the number of robbery cases recorded for that neighborhood for that year will be shown in the tooltip.
The change in the overall color in these maps give us a sense of how robbery cases evolved over the years in the city's neighborhoods. At first glance, you may notice that the colors assigned to the general downtown/mid-town neighborhoods have intensified since 2012. While in 2012 only four neighborhoods (Downtown, Frankford, Belair-Edison, and Cherry Hill) recorded over 70 robberies in a year, in 2016 eight city neighborhoods recorded 70 or more robberies in a year. Downtown, in particular, had 300 robberies in just one year (2016), which almost doubled the number of cases in 2012.
Weapons in Crime
| Weapon | 2012 | 2013 | 2014 | 2015 | 2016 |
|---|
This table tallies the total number of crime committed with a weapon in each of the previous five years. The numbers show that an increase of firearm involved cases (1500+) is accompanied with a drop of cases (2000+) where only hands were used as weapons.
Firearm Involved Crime
Let's see how the number of firearm involved crime changed over time. The following calendar is set up to present data in the same manner as the previous calendar for robbery cases. The only differences are that the color of the cells were calibrated to the firearm data, and the scale is slightly different.
With this calendar, we can see that the large increase in firearm cases preceded the civic unrest, suggesting that the number of firearm cases could potentially be used as a leading indicator for civic unrest in other jurisdictions. Just like the robbery numbers, starting around May 2015 there have been many more days in a year with large numbers of firearm cases.
Firearm Involved Incidents by Neighborhood
In the following set of maps, we are looking at the yearly aggregates of firearm involved incidents in the city's neighborhoods. The same scale is used to assign color to each city's neighborhood across different years with darker colors correspond to more firearm involved cases.
Hover your mouse over any point on the map, the name of the neighborhood and the number of firearm involved cases recorded for that neighborhood for that year will be shown in the tooltip.
Comparing to the set of maps for robbery cases, the change in the overall colors in these maps conveys a change regarding firearm involved cases that was larger in scale and at a much faster pace. Many neighborhoods that were given a light touch of orange in the 2012 map picked up a darker hue in the 2016 map, and some even turned dark red. It draws out attention to neighborhoods such as Downtown, which, you may notice, had the number of cases recorded in 2016 three times that of 2012! Let's find out more specifically how the number of cases changed.
In the following map, color is scaled to the change in number of firearm involved cases between 2012 and 2016. Those neighborhoods with a drop in firearm involved cases are given one of the bluish colors; those neighborhoods with an increase in firearm involved cases are given one of the orange colors. The darker the color the larger the variation in case number. Being painted white means that the neighborhood had the same number of firearm cases in 2016 as in 2012.
Hover your mouse over any point on the map, the tooltip will tell you the name of the neighborhood, the number of firearm cases in 2016, the change in number comparing with the 2012 level, and the change represented in percentages.
One may notice that more neighborhoods are given an orange rather than a bluish color, an indication that firearm has spread to more city neighborhoods. The darkest orange is given to three city neighborhoods: Downtown (95 cases increase), Brooklyn (81 cases increase), and Sandtown-Winchester (52 cases increase). The next color indicator helps identify East Baltimore Midway (48 cases increase), Pigtown/Washington Village (45 cases increase), and Carrollton Ridge (40 cases increase). In all of these neighborhoods, the increase in firearm cases represent a 95% to upward of 200% increase to their 2012 base level. The numbers suggest that these neighborhoods should be the focus of organizations/authorities who want to put a break on the surge of firearm violence in the city.
Notice that geographically close by neighborhoods do not always pick up the same or adjacent colors. For example, despite the fact that they border Downtown, University of Maryland and Downtown West had much fewer firearm cases and saw only a minor increase over the four year period comparing to Downtown's 95-case increase. This suggests that the spread of firearm is not uniform in direction. This suggests that certain forces in these neighborhoods, such as a university's efforts to keep its campus safe, strong presence of police force, or simply a geographic divide, could buck the trend of a general increase of firearm cases in the area.
Intersections of Crime Features
Robbery & Weapon
| Crime | Weapon | 2012 | 2013 | 2014 | 2015 | 2016 |
|---|
The above table breaks down the types of weapon involved in robberies over the years. It shows that all, except residential robbery, have seen a steep increase in the number of cases.
| Crime | Weapon | Change in Number of Crime Cases 2012-2016 | Percent Change in Number of Crime Cases 2012-2016 |
|---|
Between 2012 and 2016, street robbery saw the steepest increase in number, while carjacking saw the greatest percentage increase. Street robbery predominantly involves firearm. This type of robbery has increased by 548 cases or 49% from 2012 to 2016.
Firearm in Shooting & Homicide
Firearms are involved in aggravated assault, homicide, carjacking, commercial robbery, residential robbery, street robbery, and shooting cases. Among these, the percentages of firearm involved shooting and homicide remain relatively steady, as shown in the following table. Given that the number of firearm involved cases is on the rise, it also means that more people are being shot at or killed.
| Year | Number of Firearm Involved Crime | Shooting Cases Percentage | Deadly Cases Percentage |
|---|
Firearm Involved Robbery
The following set of maps zooms in to look at the geographic distribution of robberies that involved firearm. The change in overall tone of the colors in these maps are drastic. Whereas in 2012 only four neighborhoods had 40 or more robbery cases involving a firearm, in 2016 seven neighborhoods are having 40 or more firearm involved robberies in a year. And, many neighborhoods that are geographically adjacent to these epicenters picked up darker colors in the 2016 map as well, an indications that their firearm involved robbery cases were also on the rise.
Let's get more specific about how much was changed from 2012 to 2016. In the following map, color is scaled to the change in number of firearm involved robberies. Those neighborhoods with a drop in firearm involved robberies in 2016 comparing to 2012 are given one of the bluish colors; those neighborhoods with an increase in firearm involved cases are given one of the pinkish-purple colors. Being painted white means that the neighborhood had the same number of firearm involved robberies in 2016 as in 2012.
Hover your mouse over any point on the map, the tooltip will tell you the name of the neighborhood, the number of cases in 2016, the change in number comparing with the 2012 level, and the change represented in percentages.
You may notice that more neighborhoods have picked up a pinkish-purple color than those picking up a bluish color, an indication that firearm had spread to more city neighborhoods. Interestingly, not all neighborhoods, which had large numbers of robbery cases, large numbers of firearm cases, or large increases in either of these categories, had large numbers of firearm involved robberies. For example, Downtown, Brooklyn, and Washington Village/Pigtown are familiar names in this study because they had the largest base numbers and largest increase in robberies and firearm involved cases. On the other hand, data show that Sandtown-Winchester, Frankford, and Carrollton Ridge had only a modest increase of firearm involved robberies over the last five years. This suggests that the police department or neighborhood organizations might want to employ different tactics to tackle crime in different city neighborhoods. Those exhibit a correlation between different features of crime may suggest that the same groups of perpetrators were involved in various crime cases, and those neighborhoods embody a disjoint of different crime features may require authorities to look into different groups of perpetrators for those crimes.
Acknowledgment
The interactive graphics in this project are made with the D3.js library by Mike Bostock. The examples by Mike and many contributors at https://bl.ocks.org and other online discussion forums gave rise to the ideas of the graphs you see here.