Crime distortion within the NYPD: a potential method for estimating crime misclassification within CompStat statistics

ABSTRACT

Since the advent of CompStat in 1994, the NYPD contends that felony crimes have decreased throughout NYC, specifically within the seven-major felony index crime categories. Previous work from Eterno and Silverman has suggested that the scope of the decrease was duplicitous and exaggerated due to the department's widespread practice of crime distortion. However, previous research has never attempted to quantify this phenomenon. This study addresses a critical gap within the existing CompStat literature by attempting to capture the magnitude of crime distortion. Using secondary crime data from the NYPD, the current study will examine burglaries and the suspected downward misclassifications, occurring within all NYC precincts, at the aggregate level, during the period between 2000 and 2013. The goals of this research are to identify and summarize any precinct-level patterns of potential crime distortion using semi-parametric group-based trajectory modeling and multinomial logistic regression techniques.

KEYWORDS:

• CompStat
• NYPD
• crime distortion
• group-based trajectory modeling (GBTM)
• multinomial logistic regression (MNLR)

Acknowledgments

We would like to thank Professor Richard Rosenfeld and Dr. Robert Fornango for providing the precinct-level census data used in the current study. A previous version of this research project was presented at the American Society of Criminology conference in Atlanta, Georgia.

Disclosure statement

The authors declare that they have no competing interests.

Data availability statement

The bulk of the crime data was retrieved from the NYPD’s Historical NYC Crime Data webpage (see https://www1.nyc.gov/site/nypd/stats/crime-statistics/historical.page). The lost property data was obtained through a FOIL request that was filed with the NYPD. Data on the location of public WiFi hotspots and business improvement district (BID) projects were taken from the NYC Open Data webpage (see https://opendata.cityofnewyork.us/). As previously mentioned, the data on the number of persons sentenced to prison from each precinct were provided by the New York State Division of Criminal Justice Services, previously used in Rosenfeld et al.’s (2007) assessment of the impact of order maintenance policing on homicide and robbery in NYC. Finally, the sociodemographic measures used in the analysis for the year 2000 were drawn from the Geolytics Neighborhood Change Database (Tatian, 2003).

Notes

1. Two of the 77 NYPD precincts were removed from the analysis. The first was the 22nd Precinct (Central Park) in Manhattan. It was removed because there are no residential areas within the boundaries Central Park, significantly limiting the number of burglaries occurring within its boundaries. The second removed from the analysis was the 121st Precinct in Staten Island. It was dropped because it was created in July 2013 when the NYPD reconfigured the 120th Precinct and the 122nd Precinct boundaries and therefore only allowed for the review of six months of crime data (NYPD, 2016).

2. Although the original definition of crime distortion highlights the fact that a seven major felony crime can be downcoded to a lower ranking felony crime, we decided to exclude grand larcenies in our analysis. The main reason for exclusion is because based on how the NYPD reports these final numbers we would be unable to tell what type of grand larceny it was recorded as and it could potentially highly skew our results (e.g., according to the NYS Penal Law there are ten different possible qualifications that would allow a crime to be classified as grand larceny in the 4^th degree, which is the lowest of the grand larceny felony crime classifications).

3. Using a GIS program, the proportion of each BID located in each precinct was calculated. Then data on BID spending was weighted using these proportions so that an annual spending total could be calculated. Finally, an average BID spending measure for the years understudy was calculated using the estimated amounts.

4. Results suggested a more complex model fit the data marginally better according to the AIC and BIC scores. However, the four-group trajectory model had one trajectory grouping that consisted of only 4 precincts (or 5.3% of the sample) and included a non-linear trend that was difficult to interpret in relation to the other groups obtained. Therefore, to keep our results as parsimonious and logical as possible, we chose to use the model containing three groups.

5. Group 1 and 2 had three parameters (i.e., intercept, linear and quadratic) and Group 3 had two parameters (i.e., intercept and linear).

6. A second analysis was also run setting Group 2 as the reference group, so that a comparison between Group 2 and Group 3 could be made. These results are discussed in the text.

7. Because no significant effects of commanding officer rank were observed, this measure was omitted from the multivariate analysis to save covariate space.

Additional information

Funding

There was no funding provided for the completion of this research study.

Notes on contributors

Amanda L. Thomas

Amanda L. Thomas is a doctoral student at John Jay College of Criminal Justice, CUNY and a retired sergeant of the NYPD. Her research interests include evidence-based policing, police training, and crime prevention.

Kevin T. Wolff

Kevin T. Wolff is an associate professor at John Jay College of Criminal Justice. He earned his PhD from the College of Criminology and Criminal Justice at Florida State University. His research interests include adverse childhood experiences, juvenile justice, program evaluation, and quantitative methods. His work has appeared in several peer-reviewed journals, including the Justice Quarterly, the Journal of Research in Crime and Delinquency, and the Journal of Youth and Adolescence. He recently received the Feliks Gross Award from the City University of New York.