Exploring violent crimes in Chicago during the COVID-19 pandemic: do location, crime type, and social distancing type matter?

ABSTRACT

The current study estimates the impact of the SAH order on violent crimes across public and residential locations: assault, battery, homicide, robbery, and sexual assault. Using interrupted time series analyses, it analyzes weekly crime data in Chicago, Illinois, from 2017 to 2020. The SAH order caused significant decreases in battery and sexual assault across public and residential locations. It also decreased assault in public locations only. Such decreases in assault, battery, and sexual assault were greater under the SAH order when social distancing was strictly enforced, as opposed to during the relaxation of social distancing. On the other hand, there were significant increases in homicide across public and residential locations. Robbery increased in public locations only. There were greater increases in homicide and robbery during the relaxation of social distancing, as opposed to under the SAH order. The study ultimately indicates that the impacts of the containment measures are conditional upon the offense location, type of crime, and level of social distancing being mandated. From a public policy perspective, it is important to allocate staffing and resources for law enforcement accordingly during the enduring pandemic.

KEYWORDS:

• COVID-19
• relaxation of social distancing
• stay-at-home orders
• violence

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. One reviewer recommended including the descriptions of crime trends prior to the study period. In the three years immediately preceding this study period, Chicago had experienced increases in violent crime, especially in 2016. Examining Uniform Crime Report (UCR) data, the rate of violent crime increased from 884.29 per 100,000 population in 2014 to 903.84 in 2015 to 1105.48 in 2016. While virtually all types of violent crime increased in 2016, the rise was especially pronounced for murder, the rate of which increased from 17.52 per 100,000 in 2015 to 28.07 in 2016.

2. One reviewer had concerns over the stability of the high-order ARMA models for public battery (27,0,0), and logged residential sexual assault (17,0,0). As seen in (Tables 3–4), the AR terms are significant and less than unity in absolute value. The authors also estimated the inverse roots of the AR characteristic polynomial. Given that all inverted roots are inside the unit circle, the ARMA models are stationary and stable. In addition, the reviewer suggested that the authors conduct negative binomial analyses as a complement to the ARMA models for public battery and residential sexual assault. Overall, the effects of the SAH and relaxed SAH interventions in the negative binomial models remained similar to those in the ARMA models (see Tables 3–4) in terms of the direction and significance. An examination of the residuals indicates no autocorrelation for both models. It is noted that a lagged dependent variable was included in the models to address the problem of autocorrelation in the residuals. The residuals in the model for public battery approximate a normal distribution, but those for residential sexual assault are not normally distributed. Finally, the correlograms of standardized residuals squared for both models present no heteroscedasticity in the residuals.

3. Given the presence of ARCH effects in the residuals for robbery in public areas, this study estimated an ARCH (1) model to take account of heteroscedasticity. The ARCH term of e^2(− 1) in the variance equation was positive and significant (b = .39, p < .01), meeting the condition for stability. The relaxed SAH variable in the ARIMA-ARCH model caused a larger impact on robbery in public areas (b = 31.58, p < .01), as opposed to that in the ARIMA model (b = 28.70, p < .05). In addition, there are no significant effects of other variables (SAH, BLM, and Q 2-3). There are no problems with serial correlation, non-normality, or heteroscedasticity in the ARIMA-ARCH model.

4. Given the presence of zero values in the data, this study adds a constant value of one to each value before taking the natural log of the time series.

Additional information

Notes on contributors

Dae-Young Kim

Dae-Young Kim is an Associate Professor in the Criminal Justice Department at SUNY Buffalo State. His research interests include prisoner reentry, problem-based learning, criminal justice policy and program evaluation, and political economy of crime and punishment. His work has appeared in journals such as Criminal Justice and Behavior, Journal of Crime and Justice, Journal of Criminal Justice, Journal of Research in Crime and Delinquency, and The Prison Journal.

William P. McCarty

William P. McCarty, Ph.D. in Criminal Justice, is an Associate Professor in the Department of Criminology, Law and Justice at the University of Illinois at Chicago (UIC). He is also the director of the Center for Research in Law and Justice at UIC. His research interests include criminal justice organizations, police supervision, technology in law enforcement, and officer stress and health. His recent academic publications have appeared in Crime and Delinquency, Police Quarterly, and Journal of Crime and Justice.