Have crime rates changed under Mayor DeBlasio? I examine this question by looking at NYPD court summons using an NYC OpenData Extract of NYPD Shooting Incident reports. It covers 2006 to 2019. Students who are interested in exploring this data can download a CSV version of the set from NYC OpenData, and can review my R Markdown file here.

Loading the Set

The data set is distributed as a comma-separated values file, and so we call it up using read.csv():

data <- read.csv("NYPD_Shooting_Incident_Data__Historic_.csv")

# Trimming data set, to ease display in this post: 
data <- data[c(1,2,4,5,8)]

Appraising the Data for Cleaning Strategy

After loading the data, my next step is to examine the data and develop a strategy for cleaning it into a tidy, compact table:

head(data)

##   ï..INCIDENT_KEY OCCUR_DATE          BORO PRECINCT STATISTICAL_MURDER_FLAG
## 1       201575314 08/23/2019        QUEENS      103                   false
## 2       205748546 11/27/2019         BRONX       40                   false
## 3       193118596 02/02/2019     MANHATTAN       23                   false
## 4       204192600 10/24/2019 STATEN ISLAND      121                    true
## 5       201483468 08/22/2019         BRONX       46                   false
## 6       198255460 06/07/2019      BROOKLYN       73                   false

Each row is a separate shooting, so I can get a count of shootings by counting rows. Between January 1, 2006 and December 31, 2019, how many shootings were there in NYC? I count using nrow()

nrow(data)

## [1] 21626

The answer is 21,626. Of these, how many were flagged as murders? We use table() to get counts of a dichotomous variable:

table(data$STATISTICAL_MURDER_FLAG)

## 
## false  true 
## 17499  4127

R 4,127

My strategy is to collapse these tables into annual counts. To do that, I am going to extract the year from the date field, and then use frequency tables of that new variable to get row counts:

#Create new variable with year extracted from OCCUR_DATE
#As dates are all same number of characters with year in same
#position, you can extract using **substring()**

data$YEAR <- as.numeric(substring(data$OCCUR_DATE,7,10))

Have Shootings Become More or Less Common?

A count of rows will give us a count of how many reported shootings are in the database by year:

table(data$Year)

## < table of extent 0 >

Visualized, using ggplot2:

data.1 <- data.frame(table(data$YEAR))
library(ggplot2)
ggplot(data.1, aes(x = Var1, y = Freq, group = 1)) + geom_point() + geom_line()

Shootings seem to be way down. It may be that shootings went up during COVID-19, but it seems more likely that COVID-19 itself would precipitate a rise in shootings, than it is a policy change indepdently initiated by DeBlasio .

How about Murders?

We can perform the same operations, only looking at shootings flagged as murders:

data.2 <- subset(data, STATISTICAL_MURDER_FLAG == "true")
data.3 <- data.frame(table(data.2$YEAR))
data.3

##    Var1 Freq
## 1  2006  445
## 2  2007  373
## 3  2008  362
## 4  2009  348
## 5  2010  403
## 6  2011  373
## 7  2012  287
## 8  2013  223
## 9  2014  248
## 10 2015  283
## 11 2016  223
## 12 2017  174
## 13 2018  202
## 14 2019  183

ggplot(data.3, aes(x = Var1, y = Freq, group = 1)) + geom_point() + geom_line()

Murders also seem down

Verdict: Improvements until 2019

Overall, our analysis concludes that shootings have not gotten worse under DeBlasio. This may change with 2020 data, although one might question whether such changes are attributable to DeBlasio, the COVID-19 crisis, political unrest, or some other factor.

Photo Credit. By Gage Skidmore, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=81590858