Introduction

Over the past forty-five years the state of New York has implemented many influential criminal justice policies. Legislation including the 1973 Rockefeller Drug Laws, the controversial stop-and-frisk program implemented by Mayor Rudy Giuliani and Police Commissioner William Bratton in the 1990s, and the 2014 decriminalization of 25 grams or less of marijuana in New York City has influenced policy decisions all over the country. I chose to use a dataset of adult arrests by New York State county from 1970 – 2015, available from Data.NY.Gov, in the hopes of gaining a better understanding of the ways that administration and resultant policy decisions affect arrest rates, specifically focusing on the five counties of New York City.

Influential Visualizations

I began my research by exploring different visualizations related to arrest and incarceration statistics. The first visualization that struck my interest was from a book published by the National Research Council entitled “The Growth of Incarceration in the United States: Exploring Causes and Consequences”.

Combined state incarceration rate by crime type, 1980 to 2010. SOURCE: Beck and Blumstein (2012).

 

While this line graph depicts incarceration numbers rather than arrest numbers, its delineation of crime type made it a useful example of the types of visualizations I hoped to create with my dataset. I was particularly interested in the effect of including six different crime categories on one graph. By including multiple types of crimes in one visualization, the viewer is able to truly grasp the incredible spike in drug convictions in the mid-1980s through the early 2000s. I believe that separating these different types of crimes out into six different line graphs would have severely diminished the impact of the data.

The second visualization that informed my own was from a New York Times article on the disproportionately high incarceration rates in a small county in Indiana. This visualization is interactive, allowing users to click on each county in the country to view incarceration rates per 100,000 residents, as well as the imprisonment percentage increase or decrease since 2006 (with the exception of a handful of counties missing data, including the entire states of Vermont, New Hampshire, Connecticut, and Rhode Island).

Due to the time-based nature of my data, a map was not the best choice for visualizing my statistics. However, this visualization still helped me to conceptualize effective ways of highlighting concentrations of incarceration or, in the case of my data, arrests. In the future I would love to create an interactive map depicting arrest statistics over time with some sort of time filter (akin to the Google Gapminder Wealth & Health of Nations visualization).

The third visualization that influenced my own decisions was this filled line graph from the Prison Policy Initiative showing female incarceration rates in the United States from 1910 to 2014 per 100,000 women. While I sometimes find filled line graphs to be somewhat misleading, I was impressed with how visually striking this particular visualization is. The color choice of course plays a large role in the success of this graph. I experimented with filled line graphs, though ultimately found a simple line graph depicting multiple pieces of information on one plane, similar to the National Research Council graph, to be the most effective format for my data.

Materials

The dataset used for this lab, “Adult Arrests by County: Beginning 1970”, came from New York State’s Open Data initiative, Data.NY.Gov. The figures were based on the Division of Criminal Justice Services Computerized Criminal History database for fingerprintable offenses. Two important notes on this dataset: an adult arrest is defined by the state of New York as “an arrest of a person 16 years old or older or a juvenile offender prosecuted in adult court”, and a fingerprintable offense includes “any felony, a misdemeanor defined in the penal law, a misdemeanor defined outside the penal law which would constitute a felony if such a person had a previous judgment of conviction for a crime, or loitering for the purpose of engaging in prostitution as defined in subdivision two of Penal Law §240.37” (New York State Open Data). This dataset includes overall arrests for each county for every year from 1970 to 2015, as well as numbers for total felonies and total misdemeanors. Each of these categories is then broken down to a more granular level, and numbers for drug felonies/misdemeanors, violent felonies/misdemeanors, DWI felonies/misdemeanors, other felonies/misdemeanors, and property misdemeanors are included in the dataset. I exported this dataset to Google Sheets and then, due to the fact that the data was already clean, uploaded it as a .csv file to Tableau Public.

 

Methods

I was particularly interested in exploring long-term trends in New York State arrests as this dataset covers a forty-five year period. I initially attempted to visualize arrest data for all of the counties in New York State. However, given that New York has 62 counties, this was clearly far too much data to include in one visualization. I then selected the ten most populous counties and created individual line graphs for each. This was more effective, though still quite a bit of information to digest in one visualization. Next, I created line graphs for the total number of arrests for each of the five NYC counties, a scale which was far easier to interpret visually. These visualizations were broken up by Tableau into five individual line graphs. I felt that this made it difficult for the viewer to truly compare the arrest statistics for each county, and that including all of the data on the same graph would be more effective. When I added all of the data to the same graph I did in fact feel that it was easier to compare the total number of arrests by NYC county from 1970 to 2015. As I constructed other visualizations using different pieces of the dataset, I made sure to create both individual line graphs for each county, as well as a single graph containing all of the data currently being analyzed. I also experimented with filled line graphs and bar graphs. I felt that neither of these were really able to capture the overall trend of the data as effectively as multiple lines represented on one graph.

 

Results

I ultimately created 13 visualizations and two dashboards for this dataset. I utilized the colorblind palate offered by Tableau to be sure that these visualizations would be accessible to everyone. The two line graphs below depict adult arrests for the five counties of New York City from 1970 to 2015. The first shows the data for each county split into individual graphs, while the second visualization includes the data for all five counties on one graph. This is a good example of the way that including all of the data points in one graph is more effective at displaying trends and differences across counties than splitting the data over multiple graphs.

 

 

 

The below visualizations depict drug felony and drug misdemeanor arrests in NYC between 1970 and 2015. Visualizing this data is particularly useful in helping to identify the long-term trends in felony versus misdemeanor arrests in the state over a forty-five year period. It was necessary to assign the same scale to these two graphs in order to make the comparison accurate.

 

 

 

Many trends became instantly apparent when the above data was visualized. Firstly, the spike in arrests in during the 1980s was clearly seen in the “Total Adult Arrests in NYC” graph, a decade in which New York City was notoriously crippled by drug-related crime. Though the arrest numbers seem to go through an annual valley and peak swing, the overall trend of the data is upwards. All five of the NYC counties achieved somewhat of an arrest plateau between 2007 and 2011, with between ten and thirty-five thousand more people arrested annually than in the late 1970s. Another striking difference that is elucidated by the visualization of this data is the extreme increase in drug misdemeanor arrests, while drug felony arrests are almost identical today as they were in the mid 1970s (though there was a spike in drug felonies between the late 1980s and the early 2000s). This instantly raises questions about the content of these arrests; are they largely the product of stop-and-frisk policing? Were these drug misdemeanors primarily marijuana possession arrests? Perhaps most noteworthy is the clearly defined drop in drug misdemeanors between 2014 and 2015, the year when New York City changed its marijuana laws to merely require a $100 fine for anyone caught carrying 25 grams or less of marijuana, both concealed and out in the open (it is worth noting that this was technically legislated by the 1977 Marijuana Reform Act, pertaining strictly to 25 grams or less of concealed marijuana. The NYPD has long circumvented this law through its stop-and-frisk program which coerced individuals into bringing their marijuana out into the open, thus making it an arrestable offense). Ultimately, visualizing this data allows for simpler analysis on a granular, annual level, as well as on a broader, long-term scale.

 

Future Directions
As previously stated, visualizing this dataset has helped to elucidate a number of trends in New York City arrest statistics over the past forty-five years, triggering questions about the relationship between administration, legislation, and policing. I would love to spend more time with this data in order to determine some explanations for long-term trends, as well as annual increases or decreases in arrests by county. I believe some sort of interactive map would also be highly effective for this dataset, and would allow users to engage more fully with the data, perhaps prompting more questions. I was unable to determine how to add trend lines in Tableau, but I think this could also help to make the overall trend of the data easier to spot, and could hopefully keep viewers from getting distracted by the annual fluctuations. I would also love to augment this data with a more in-depth dataset containing demographic information for arrests in New York State. I believe this could help to shed light on which communities, if any, are disproportionately affected by policies such as stop-and-frisk and accompanying legislation.