COVID-19 and crime: Analysis of crime dynamics amidst social distancing protocols.

In response to the pandemic in early 2020, cities implemented states of emergency and stay at home orders to reduce virus spread. Changes in social dynamics due to local restrictions impacted human behavior and led to a shift in crime dynamics. We analyze shifts in crime types by comparing crimes before the implementation of stay at home orders and the time period shortly after these orders were put in place across three cities. We find consistent changes across Chicago, Baltimore, and Baton Rouge with significant declines in total crimes during the time period immediately following stay at home orders. The starkest differences occurred in Chicago, but in all three cities the crime types contributing to these declines were related to property crime and statutory crime rather than interpersonal crimes.

Introduction

Crime is a major public concern in the United States. Gun violence leads to the deaths of 36,000 individuals and the non-fatal injuries of 85,000 others [1]. Other crime types can produce property damage, trauma for victims, and fractures in community trust [2]. Across the five major crime categories (personal, property, inchoate, statutory and financial), some rely on regular interpersonal interactions and social dynamics [3].

In early 2020, COVID-19 emerged in the United States and case numbers quickly grew across the country. In response to the pandemic, many cities implemented a variety of mitigation policies to minimize spread. A major strategy for preventing viral spread focuses on reducing contacts between individuals using stay at home orders, which closed non-essential businesses, mandated wearing masks, and encouraged citizens to minimize all non-essential visits [4]. Adherence to these policies was initially strong, but faded over time, as evidenced by self-reporting and cell phone movement data (see S1-S3 Figs in S1 File) [5]. Implementing these types of policies resulted in drastic changes in the behaviors of citizens. Because crime is associated with human behavior, many have questioned how reductions in human contact have changed the dynamics of crimes across the country [6-15].

In a pandemic which strains health care resources, collateral mortality may increase, due to the inability of the health care system to respond as effectively to crime-derived injuries [16]. Evaluating the impact of the pandemic on crime in a rigorous manner can assess the potential for significant changes in crime-derived mortality.

We test whether there are significant differences in the crime dynamics in Chicago, IL when comparing the time periods pre- and post-establishment of the stay at home orders, and then compare these results to Baton Rouge, LA and Baltimore, MD. We choose to analyze the first two weeks following the stay ay home order implementation because adherence to the mandates was strongest immediately following implementation (see S1-S3 Figs in S1 File) [5]. Our main focus is Chicago, due to its high rates of both COVID-19 and crime. Comparison of the crime dynamics in the three cities determines whether similar patterns exist (Fig 1). We hypothesize that changes in crime dynamics in all three cities following the implementation of stay at home orders are not uniform across all crime types and that there are differences between crimes associated with property, those which are statutory, and those involving interpersonal interactions. We define property crimes (denoted with P for this study) as those which involve interference with the property of another [3]. We define statutory crimes (denoted with S for this study) as those which are proscribed by statute [3]. We define interpersonal crimes (denoted with I for this study) as those that result in physical or mental harm to another person [3]. We find that total crimes declined significantly in the two weeks following implementation of stay-at-home orders across all three cities. In Chicago total crimes declined 31.5%. The property crimes with significant declines were burglary (22.9%), criminal trespass (50.1%), robbery (25.8%), and theft (41.0%). The statutory crimes with significant declines were interference with public officers (93.1%), narcotics (86.1%), and other offenses (41.2%). The interpersonal crimes with significant declines were assault (19.4%), and criminal sexual assault (56.0%). Baltimore also experienced significant declines in total crimes (25.9%). The crime types showing significant declines were auto theft (29.9%), burglary (29.2%), and larceny (35.0%), which are all property crimes. Baton Rouge also experienced significant declines in total crimes (22.4%). The crime types that showed significant declines were narcotics (52.3%) and other crimes (41.5%), which are both statutory crimes.

Fig 1

The total crime trends seen in Chicago, Baltimore, and Baton Rouge from the beginning of the year through the end of our study period.

Each point is the total number of crimes observed on that day. The vertical line represents the day when the stay at home order was implemented for each city. The dark line represents a moving average of the data with k = 5 to observe shifts in dynamics over a five day temporal window [23]. More information regarding these methods is available in the S1 File.

Materials and methods

Data

Chicago, Baltimore, and Baton Rouge each have publicly available crime datasets through citywide data portals [17-19]. Due to lack of consensus on collecting, defining, and reporting crime, care should be taken when comparing one dataset to another. Descriptions of each crime type included are available in the (see S2 Table in S1 File). These three cities were chosen because they had openly available crime data and they differ in demographics (see S1 Table in S1 File) [20]. Chicago has a 4.5 times larger population than Baltimore and 12 times larger than Baton Rouge [20]. All three cities have fairly high poverty rates and regularly receive media attention for their crime dynamics. Crime in all three cities remained fairly consistent from 2017-2019 (see S4-S6 Figs in S1 File). They also exist in three different regions of the country, making them interesting comparisons for this study [20].

All data and code files used in this analysis are available through GitHub at https://github.com/shelbymscott/COVIDandCrime.

Chicago data

The Chicago City Data Portal provides crime reports from 2001-present [19]. Data are extracted from the CLEAR (Citizen Law Enforcement Analysis and Reporting) system. The information from this dataset used for this analysis includes the date, primary type, and description, but a number of other items are available. We isolated the crimes which occurred from January 1, 2020 through April 4, 2020, which gave data from before the onset of the pandemic (01/01/2020—03/08/2020), during the state of emergency (03/09/2020—03/20/2020), and after the stay at home order was put in place (03/21/2020—04/04/2020) for temporal comparisons. There are 32 different crime types available, but for this analysis we choose 19 of the crime types with high occurrence and denote them as property (P), statutory (S), or interpersonal (I). They include:

Total Crimes

Gun Crimes (pulled from crime descriptions that included firearm use)

Arson (P)

Burglary (P)

Criminal Damage (P)

Criminal Trespass (P)

Robbery (P)

Theft (P)

Weapons Violation (P)

Interference with Public Officer (S)

Narcotics (S)

Other Offense (S)

Public Peace Violation (S)

Assault (I)

Battery (I)

Criminal Sexual Assault (I)

Homicide (I)

Sex Offense (I)

These crime types were chosen because they composed a high proportion of total crimes or were present in the Baltimore or Baton Rouge datasets for potential comparison. Definitions of these crime types are available in the (see S2 Table in S1 File). Those crime types excluded due to low numbers were: concealed carry license violations, deceptive practices, gambling, human trafficking, intimidation, kidnapping, liquor law violations, obscenity, offenses involving children, other narcotics violations, prostitution, public indecency, and stalking. We determined whether the crime types should be included based on if they comprised greater than or equal to 0.2% of total crimes. This number was found by determining the lowest percentage from the Baltimore and Baton Rouge datasets and only including those crime types which exceeded this amount.

Baltimore data

Baltimore, Maryland has Victim Based Crime data available for public download [18]. The data are preliminary and therefore may be subject to change [18]. The data provided includes a number of different options from which to choose. The relevant information we use includes the date of the crime and the crime type. We isolate those crime types which occurred from the beginning of the year (1/1/20) to when the stay at home order was put in place (3/29/20) and the two weeks following the implementation (3/30/20—4/13/20). The 11 crime types available for analysis are denoted as property (P) or interpersonal (I) as this dataset did not account for statutory crimes since it only provides crimes where an individual was victimized. Definitions of these crime types are available in the (see S2 Table in S1 File). The dataset includes:

Total Crimes

Gun Crimes (determined by finding those crimes which involved a firearm as a weapon)

Arson (P)

Auto Theft (P)

Burglary (P)

Larceny (P)

Robbery (P)

Assault (I)

Homicide (I)

Rape (I)

Shooting (I)

Baton Rouge data

Crimes reported in Baton Rouge are handled by the Baton Rouge Police Department and are publicly available [17]. The data are pulled from police reports using automatic statistical reporting, which can lead to some errors [17]. We obtained the dates and crime types from the dataset and parse for the time period from the beginning of the year to when the stay at home order was put in place (1/1/20—3/21/20) and the two weeks following the stay at home order (3/22/20—4/5/20). The 15 crime types analyzed are denoted as property (P), statutory (S), and interpersonal (I). Definitions of these crime types are available in the (see S2 Table in S1 File). The dataset includes:

Total Crimes

Burglary (P)

Criminal Damage (P)

Robbery (P)

Theft (P)

Juvenile (S)

Narcotics (S)

Nuisance (S)

Other (S)

Vice (S)

Assault (I)

Battery (I)

Firearm (I)

Homicide (I)

Sexual Assault (I).

t-test analysis

In order to determine whether the differences seen in crime dynamics are significant or not, we performed numerous t-tests that compared whether different years and time periods differed significantly [21]. T-tests determine whether the means of two sets of data are significantly different from each other [21]. Because we were performing a number of statistical tests simultaneously, we used the Bonferroni correction to be sure that spurious positives were not included by random chance [22]. The Bonferroni correction sets the significance value, α, for the entire set of n t-tests run equal to α* by taking: Formally, given n t-tests T for hypotheses H (i between 1 and n) under the assumption H0 that all hypotheses H are false, and if the individual test critical values are less than or equal to α/n, then the critical value is less than or equal to α. In equation form, if: for 1 ≤ i ≤ n, then which follows from the Bonferroni inequalities [21]. The p-value result from a t-test represents the probability of obtaining test results at least as extreme as the results actually observed, under the assumption that the null hypothesis is correct [21].

There were t-tests performed on the datasets in order to determine whether what has been observed in 2020 is significantly different behavior from other years. These included:

Year to year tests to determine whether the 2020 observed behavior in Chicago differs from what we would have expected from previous years (Table 1, S4-S6 Figs in S1 File).

Table 1

Comparisons of each Chicago crime type in the first three months 2019, 2018, and 2017 compared to crime types in the same time period of 2020.

The degrees of freedom for these analyses are 179 and α = .05 is adjusted after Bonferroni correction (with n = 18) to α = 0.0027. The values for mean μ, standard deviation σ, and percent change are also provided. Bolded crime types show significant differences between years and crime categories are denoted by (P) for property, (S) for statutory, and (I) for interpersonal crimes.

	Early 2019 Dataset (μ, σ)	Early 2020 Dataset (μ, σ)	p-value	Percent Change
	Total Crimes (650, 84.3)	Total Crimes (575, 90.6)	7.45 × 10−7	-11.5%
	Gun Crimes (33.6, 9.37)	Gun Crimes (36.4, 9.38)	0.038	8.33%
P	Arson (0.872, 0.845)	Arson (0.884, 1.07)	0.933	-1.38%
P	Burglary (23.7, 7.37)	Burglary (20.6, 5.16)	8.53 × 10−4	-13.1%
P	Criminal Damage (61.7, 13.8)	Criminal Damage (59.1, 12.4)	0.167	-4.40%
P	Criminal Trespass (17.5, 5.02)	Criminal Trespass (16.1, 5.70)	0.076	-8.00%
P	Robbery (18.4, 5.59)	Robbery (20.4, 6.11)	0.019	10.9%
P	Theft (149, 23.7)	Theft (130, 28.9)	1.09 × 10−6	-12.8%
P	Weapons Violation (14.5, 5.93)	Weapons Violation (15.6, 5.80)	0.186	7.59%
S	Interference with Public Officer (3.40, 1.65)	Interference with Public Officer (3.32, 2.19)	0.755	-2.35%
S	Narcotics (44.1, 11.0)	Narcotics (31.0, 13.6)	8.41 × 10−12	-29.7%
S	Other Offense (47.8, 9.80)	Other Offense (39.4, 9.98)	2.89 × 10−8	-17.6%
S	Public Peace Violation (3.81, 2.18)	Public Peace Violation (2.87, 2.06)	0.003	-24.7%
I	Assault (50.8, 9.89)	Assault (47.4, 8.02)	0.010	-6.69%
I	Battery (122, 22.5)	Battery (115, 21.7)	0.033	-5.74%
I	Criminal Sexual Assault (3.81, 2.44)	Criminal Sexual Assault (3.73, 2.47)	0.818	-2.10%
I	Homicide (0.989, 1.05)	Homicide (1.08, 1.03)	0.534	9.2%
I	Sex Offense (3.28, 2.37)	Sex Offense (2.78, 1.65)	0.095	-15.2%
	Early 2018 Dataset	Early 2020 Dataset
	Total Crimes (655, 65.2)	Total Crimes (575, 90.6)	7.63 × 10−11	-12.2%
	Gun Crimes (35.1, 7.99)	Gun Crimes (36.4, 9.38)	0.301	3.70%
P	Arson (0.787, 0.914)	Arson (0.884, 1.07)	0.504	12.3%
P	Burglary (27.6, 6.31)	Burglary (20.6, 5.16)	1.14 × 10−14	-25.4%
P	Criminal Damage (66.2, 14.8)	Criminal Damage (59.1, 12.4)	4.00 × 10−4	-10.7%
P	Criminal Trespass (19.0, 3.78)	Criminal Trespass (16.1, 5.70)	5.32 × 10−5	-15.3%
P	Robbery (25.1, 6.38)	Robbery (20.4, 6.11)	7.45 × 10−7	-18.7%
P	Theft (150, 21.1)	Theft (130, 28.9)	1.19 × 10−7	-13.3%
P	Weapons Violation (12.1, 4.37)	Weapons Violation (15.6, 5.80)	7.84 × 10−6	28.9%
S	Interference with Public Officer (3.30, 1.75)	Interference with Public Officer (3.32, 2.19)	0.951	0.606%
S	Narcotics (37.3, 9.65)	Narcotics (31.0, 13.6)	3.66 × 10−4	-16.9%
S	Other Offense (45.4, 7.94)	Other Offense (39.4, 9.98)	1.11 × 10−5	-13.2%
S	Public Peace Violation (3.22, 1.79)	Public Peace Violation (2.87, 2.06)	0.214	-10.9%
I	Assault (48.7, 9.40)	Assault (47.4, 8.02)	0.295	-2.67%
I	Battery (119, 19.9)	Battery (115, 21.7)	0.137	-3.36%
I	Criminal Sexual Assault (3.74, 2.67)	Criminal Sexual Assault (3.73, 2.47)	0.961	-0.267%
I	Homicide (1.30, 1.18)	Homicide (1.08, 1.05)	0.188	-16.9%
I	Sex Offense (2.61, 3.23)	Sex Offense (2.78, 1.65)	0.644	6.51%
	Early 2017 Dataset	Early 2020 Dataset
	Total Crimes (686, 74.7)	Total Crimes (575, 90.6)	7.48 × 10−17	-16.2%
	Gun Crimes (37.7, 8.86)	Gun Crimes (36.4, 9.38)	0.336	3.45
P	Arson (1.21, 1.15)	Arson (0.884, 1.07)	0.044	-26.9%
P	Burglary (35.5, 10.6)	Burglary (20.6, 5.16)	6.24 × 10−26	-42.0%
P	Criminal Damage (74.8, 14.1)	Criminal Damage (59.1, 12.4)	6.27 × 10−14	-21.0%
P	Criminal Trespass (18.5, 4.54)	Criminal Trespass (16.1, 5.70)	0.002	-13.0%
P	Robbery (29.5, 7.27)	Robbery (20.4, 6.11)	4.86 × 10−17	-30.8%
P	Theft (154, 21.4)	Theft (130, 28.9)	5.99 × 10−10	-15.6%
P	Weapons Violation (10.8, 4.23)	Weapons Violation (15.6, 5.80)	7.26 × 10−10	44.4%
S	Interference with Public Officer (2.77, 1.47)	Interference with Public Officer (3.32, 2.19)	0.044	19.9%
S	Narcotics (33.9, 8.77)	Narcotics (31.0, 13.6)	0.091	-8.55%
S	Other Offense (49.9, 8.96)	Other Offense (39.4, 9.98)	1.41 × 10−12	-21.0%
S	Public Peace Violation (3.87, 2.12)	Public Peace Violation (2.87, 2.06)	0.001	-25.8%
I	Assault (46.4, 8.24)	Assault (47.4, 8.02)	0.386	2.16
I	Battery (121, 19.6)	Battery (115, 21.7)	0.037	-4.96%
I	Criminal Sexual Assault (4.07, 3.56)	Criminal Sexual Assault (3.73, 2.47)	0.436	-7.62%
I	Homicide (1.65, 1.66)	Homicide (1.08, 1.04)	0.006	-34.5%
I	Sex Offense (2.67, 3.78)	Sex Offense (2.78, 1.65)	0.798	4.12%

Within year analysis to determine whether the behavior before pandemic response (1/1/20—3/20/20) and after the stay at home orders were put in place in Chicago (3/21/20—4/3/20) differ from one another (Table 2, S7-S10 Figs in S1 File).

Table 2

Comparisons of the time period before the stay at home order was put in place and the two weeks after it was put in place in Chicago, Baltimore, and Baton Rouge.

The observed time period for the stay at home order spans from 03/21/2020—04/04/2020 in Chicago. The degrees of freedom for these analyses are 89 and α = .05 is adjusted after Bonferroni correction (with n = 18) to α = 0.0027. The two weeks after Baltimore’s stay at home period span from 03/30/20—04/13/20. The degrees of freedom for these analyses are 101 and α = .05 is adjusted after Bonferroni correction (with n = 11) to α = 0.0045. The two weeks after Baton Rouge’s stay at home order span from 03/22/20—04/05/20. The degrees of freedom for these analyses are 93 and α = .05 is adjusted after Bonferroni correction (with n = 15) to α = 0.0033. The values for mean μ, standard deviation σ, and percent change are also provided. Bolded crime types show significant differences between years and crime categories are denoted by (P) for property, (S) for statutory, and (I) for interpersonal crimes.

	Crime Types Pre-Stay at Home Orders (μ, σ)	Crime Types Post Stay at Home Orders (μ, σ)	p-value	Percent Change
	Chicago
	Total Crimes (606, 59.4)	Total Crimes (415, 47.8)	4.61 × 10−20	-31.5%
	Gun Crimes (37.1, 9.47)	Gun Crimes (33.1, 8.42)	0.139	-10.8%
P	Arson (0.863, 1.09)	Arson (1.00, 1.00)	0.651	15.9%
P	Burglary (21.4, 5.08)	Burglary (16.5, 3.52)	6.89 × 10−4	-22.9%
P	Criminal Damage (59.6, 12.5)	Criminal Damage (56.5, 11.5)	0.375	-5.20%
P	Criminal Trespass (17.5, 5.08)	Criminal Trespass (8.73, 2.05)	3.38 × 10−9	-50.1%
P	Robbery (21.3, 5.88)	Robbery (15.8, 5.28)	0.001	-25.8%
P	Theft (139, 21.3)	Theft (82.0, 11.0)	1.83 × 10−16	-41.0%
P	Weapons Violation (15.8, 6.02)	Weapons Violation (14.3, 4.42)	0.339	-9.49%
S	Interference with Public Officer (3.89, 1.89)	Interference with PublicOfficer (0.267, 0.458)	7.46 × 10−11	-93.1%
S	Narcotics (35.9, 8.07)	Narcotics (5.00, 2.56)	7.09 × 10−26	-86.1%
S	Other Offense (42.2, 8.15)	Other Offense (24.8, 4.54)	3.27 × 10−12	-41.2%
S	Public Peace Violation (3.14, 2.11)	Public Peace Violation (1.47, 0.916)	0.003	-53.2%
I	Assault (48.9, 7.28)	Assault (39.4, 7.14)	1.07 × 10−5	-19.4%
I	Battery (117, 21.7)	Battery (100, 15.15)	0.004	-14.5%
I	Criminal Sexual Assault (4.09, 2.50)	Criminal Sexual Assault (1.80, 1.08)	7.76 × 10−4	-56.0%
I	Homicide (1.15, 1.08)	Homicide (0.733, 0.704)	0.155	-36.3%
I	Sex Offense (2.91, 1.65)	Sex Offense (2.07, 1.49)	0.067	-28.9%
	Baltimore
	Total Crimes (103, 15.4)	Total Crimes (76.3, 12.4)	6.39 × 10−9	-25.9%
	Gun Crimes (12.8, 6.70)	Gun Crimes (11.1, 6.21)	0.359	-13.3%
P	Arson (0.216, 0.441)	Arson (0.133, 0.352)	0.493	-38.4
P	Auto Theft (8.85, 2.68)	Auto Theft (6.20, 2.70)	6.01 × 10−4	-29.9%
P	Burglary (11.4, 4.07)	Burglary (8.07, 2.52)	0.003	-29.2%
P	Larceny (34.3, 8.83)	Larceny (22.3, 4.68)	1.24 × 10−6	-35.0%
P	Robbery (12.1, 4.30)	Robbery (8.73, 4.64)	0.007	-27.9%
P	Shooting (1.49, 1.68)	Shooting (1.33, 1.35)	0.735	-10.7%
I	Assault (33.2, 7.84)	Assault (28.3, 6.74)	0.024	-14.8%
I	Homicide (0.761, 1.13)	Homicide (0.933, 0.799)	0.575	22.6%
I	Rape (0.511, 0.773)	Rape (0.333, 0.488)	0.391	-34.8%
	Baton Rouge
	Total Crimes (113, 20.0)	Total Crimes (87.7, 15.8)	1.42 × 10−5	-22.4%
P	Burglary (13.7, 5.75)	Burglary (12.1, 4.40)	0.296	-11.7%
P	Criminal Damage (9.03, 3.59)	Criminal Damage (10.0, 3.82)	0.342	10.7%
P	Firearm (5.21, 3.82)	Firearm (4.53, 2.59)	0.511	-13.1%
P	Robbery (1.31, 1.23)	Robbery (1.27, 1.44)	0.898	-3.05%
P	Theft (29.3, 7.39)	Theft (24.7, 7.47)	0.028	-15.7%
S	Juvenile (1.13, 1.27)	Juvenile (0.867, 0.834)	0.450	-23.3%
S	Narcotics (9.23, 5.97)	Narcotics (4.40, 2.72)	0.003	-52.3%
S	Nuisance (1.78, 1.26)	Nuisance (1.33, 1.72)	0.245	-25.3%
S	Other (23.4, 8.94)	Other (13.7, 6.01)	1.06 × 10−4	-41.5%
S	Vice (0.575, 0.708)	Vice (0.600, 0.910)	0.905	4.35%
I	Assault (6.45, 3.03)	Assault (5.40, 1.84)	0.198	-16.3%
I	Battery (9.84, 4.38)	Battery (7.00, 1.96)	0.016	-28.9%
I	Homicide (1.20, 1.24)	Homicide (1.53, 0.916)	0.324	27.5%
I	Sexual Assault (0.512, 0.675)	Sexual Assault (0.333, 0.488)	0.330	-35.0%

Time period comparisons over Baltimore and Baton Rouge to determine whether the patterns seen in response to COVID in Chicago are consistent across numerous cities (Table 2, S11 and S12 Figs in S1 File).

Comparison tests with past years to determine whether the observed behavior is due to seasonality or differing temporal dynamics during the pandemic (S3-S19 Tables in S1 File, S4-S6 Figs in S1 File).

Results

Comparing January through early April Chicago crime data (early 2020) to the same time period in each of the three previous years determines which crime types in 2020 are significantly different [19]. The results (Table 1, S3-S5 Tables and S9-S17 Tables in S1 File) show that between the early months of 2019 and 2020, there were decreases in total crimes, burglaries, narcotics, other offenses, and thefts. Between 2018 and 2020, total crimes, burglaries, criminal damages, criminal trespasses, narcotics, other offenses, robberies, and thefts all decreased, while weapons violations increased. Between 2017 and 2020, total crimes, burglaries, criminal damages, criminal trespasses, other offenses, public peace violations, robberies, and thefts all decreased, while weapons violations increased. These comparisons include the time period from January 1 to April 4.

To determine whether there were significant differences in crime between time periods, we performed a series of t-tests [21, 22]. While Chicago had three distinct time periods: pre-COVID, a state of emergency, and a stay at home order, the results from the state of emergency were not as stark as comparisons between the time period before the stay at home order was implemented and the two weeks after implementation (see supplementary text, S3-S17 Tables in S1 File, and S4 Fig in S1 File). Total crimes, assaults, burglaries, criminal sexual assaults, criminal trespasses, interference with public officers, narcotics, other offenses, robberies, and thefts each had significant decreases from the time period before the stay at home order was put in place to the two weeks following implementation (Table 2, Fig 2a).

Fig 2

All crime types across Chicago, Baltimore, and Baton Rouge split into pre- and post-stay-at-home order time periods.

We also split the crime types into the three crime categories: interpersonal, statutory, and property. The crime types that show significant differences are denoted with an asterisk.

In order to determine whether similar patterns have been observed in other cities with different demographics (S1 Table in S1 File), we test victim-based crime data from Baltimore, before the stay at home order was put in place and two weeks after implementation (Table 2) [18]. We find that total crimes, auto thefts, burglaries, and larceny all showed significant declines between the two time periods (Fig 2b).

As a further comparison to a smaller population city with different demographics (S1 Table in S1 File), crime data from Baton Rouge were analyzed (Table 2) [17]. We find that total crimes, narcotics, and other crimes all decreased significantly after the stay at home order was put in place (Fig 2c).

Discussion and conclusions

The analysis of crime data from three cities indicates significant impacts on certain crime types arising from changes in social dynamics due to regulations implemented in response to the COVID-19 pandemic. The results of pair-wise t-tests, appropriately corrected, show that the crime dynamics experienced in 2020 significantly differ from previous years and that the implementation of strict stay at home orders in all three cities initially impacted crime.

Chicago’s total crimes during the first three months of the year (including the time period following the stay at home order) declined in 2020 compared to 2017, 2018, and 2019 (see S3-S17 Tables in S1 File). Before the time period during which COVID protocols were put in place, there were no significant changes in total crimes compared to 2018 and 2019. There were some changes in particular crime types that may have been due to social factors, policing protocols, or other mechanisms occurring in past years. There were significant changes in total crimes between 2017 and 2020, with all significant crime types but weapons violations showing declines (S11 Table in S1 File). Once the stay at home order was put in place, there were more stark significant differences between past years and 2020 (see S12-S14 Tables in S1 File). Total crimes during this time period declined compared to the past three years. There are a number of different crime types contributing to this decline, and few of the contributing crime types are interpersonal (see S7-S10 and S14 Figs in S1 File). The exceptions are significant declines in assaults and batteries between 2017 and 2019 compared to 2020 (S12 and S14 Tables in S1 File). These results show that the crime dynamics of 2020 are significantly different from those in past years and that the changes would not have been expected based on time-based observations prior to the pandemic outbreak.

We also compared the crime numbers in Chicago between different time periods. Comparing crime types before the stay at home order was put in place to the two weeks after implementation indicates that most of the crime types tested showed significant declines. They include total crimes, assaults, burglaries, criminal sexual assaults, criminal trespasses, interference with public officers, narcotics, other offenses, robberies, and thefts (Table 2, Fig 2, S14 Fig in S1 File). This shows that the immediate time period after stay at home orders were announced does correlate with decreased total crimes, but that the crimes contributing to this decline are related more to property crimes and statutory crimes than to interpersonal crimes (S7-S10 Figs in S1 File).

To determine whether this pattern holds in other cities, we also carried out time period comparisons for Baltimore, MD and Baton Rouge, LA (Table 2). Both cities showed similar results. In Baltimore, comparing the time period before the stay at home order was implemented and the two weeks after shows that total crimes, auto thefts, burglaries, and larceny showed significant differences (Fig 2, S14 Fig in S1 File). For Baton Rouge, total crimes, narcotics, and other crimes showed significant differences when comparing the two weeks after the stay at home order was put in place and the time period before the order was implemented (Fig 2, S14 Fig in S1 File).

In all of our tests across different years and different time periods, we find that the implementation of social distancing and quarantine protocols led to significant decreases in crime in the first two weeks. There were declines during the state-of-emergency time period (March 9—March 21), but they were not statistically significant (S7 Table in S1 File). Total crimes declined in Chicago after the stay at home order was put in place, but the crime types contributing to this decline are mostly property-based and statutory rather than interpersonal (Fig 2, S7-S10 and S14 Figs in S1 File). Similar patterns hold for both Baltimore, Maryland and Baton Rouge, Louisiana. Both cities showed declines in crimes after the introduction of stay at home orders. As in Chicago, the crime types contributing to this decline are more often property-based or statutory rather than individuals (S11 and S12 Figs in S1 File). These patterns may have only been present in the initial time period following implementation and may not persist over longer time periods.

The observed patterns could be the result of several different mechanisms. First, there may be differences in policing and reporting under stay at home orders. For interpersonal crimes, victims who are quarantined with their abusers may be less likely to speak up. Social distancing may prevent law enforcement officers from responding to and reporting certain crime types. This pattern may also arise from individuals spending less time in public spaces and therefore participating in social interactions. The exact reasons contributing to this change in crime dynamics are difficult to determine, but it is clear that in the immediate time period following implementation of stay at home orders, there was a significant change in crime in Chicago, Baltimore, and Baton Rouge.

This study has a number of limitations. There is differential data availability across the three cities. Chicago has far more crime types publicly accessible for analysis, and there is no consensus on how the data are collected and reported nationally. We have also only analyzed the data for three cities. This pattern therefore may not exist across all regions, especially if adherence to stay at home orders differs between cities. The use of t-tests also limits the information we can obtain from these data. Finally, we have only observed the two weeks following the implementation of stay at home orders. The declines in crime may be limited to this time period and not subsist. With this analysis, we cannot pinpoint exactly when the decline began and when crime returned to previously expected levels. Recent reports show that violent crime has increased as cities are reopening [6].

These limitations leave open questions for future research. First, there is a need to explore the data presented here more in depth using various statistical tests, expanding the temporal window, or exploring confounding variables in all three cities. Additionally, analysis of patterns of crime data from additional cities is necessary to verify that the observed changes are national in scope. There is also a need to determine some of the mechanisms which have produced these declines and how adherence to stay at home orders impacts the crimes which occur. Overall, stay at home orders produced declines in crime over the initial time period and the crimes contributing to this decline were mainly property-based. The COVID-19 pandemic responses present a forced social experiment impacting many behavioral components and providing opportunities for novel explorations of the connection between behavioral constraints and crime. Given our results, the vigorous public policy debates regarding the impacts of potential interventions on violent crime, particularly gun crime, could benefit from further detailed analysis of imposed regulations arising from the pandemic.

Complete supplementary text, figures, and tables.

(PDF)

Click here for additional data file.

11 Nov 2020

PONE-D-20-31968

COVID-19 and crime: Analysis of crime dynamics amidst social distancing protocols

PLOS ONE

Dear Dr. Scott,

Thank you for submitting your manuscript to PLOS ONE. Please see below for comments from the referees. They have some concerns about methodology, clear communication of results, and a few other issues. The referees have thus suggested various helpful methodological and stylistic improvements.

You'll see that the referees have both recommended "major revision" for your manuscript, and I agree with this assessment. I feel moved to say that I think the type of work you have done is extremely valuable and so I sincerely hope you will choose to undertake the revision.

Please submit your revised manuscript by Dec 26 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

We look forward to receiving your revised manuscript.

Cheers and best,

Chad

Chad M. Topaz

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: I Don't Know

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors aim to analyze the effect of COVID-19 on crime by studying a data set from Chicago, and comparing to similar data sets in Baltimore and Baton Rouge. They used t-tests and the Bonferroni correction to determine which crimes were significantly affected by stay-at-home orders due to COVID-19. The authors found that the greatest change in behavior of crime occurred in Chicago, where they compared crime types in early 2020 to those same crimes in earlier years (2019, 2018 and 2017). Then, they compared changes in crime between three time periods in 2020 across the three cities. The authors found that many crimes did indeed have a significant decrease after the stay-at-home orders were imposed in each city and that most crimes that were found to have significantly decreased were property crimes, not interpersonal crimes.

I have some questions and concerns with the organization of the paper:

- I don’t understand the choice to only analyze two weeks following the stay-at-home orders. I think the paper would be significantly strengthened by including more data after the stay-at-home orders. We are now well past the end date of those orders, so it makes sense to check and see if crime indeed went back up (or didn’t).

- I feel like the authors were very thorough in comparing changes in the 2020 data to previous years to make sure the effects were not due to seasonality for Chicago, but why not for the other two cities? Is it fair to only use the 2020 data for those two cities after performing such an in-depth analysis of Chicago to claim that the changes in their crimes were in fact due the stay-at-home order?

- Why scatter plots? They are super hard to read. For the t-test, aren’t you just taking total number before and after? Why not show just that? Or a moving average if you want to show time effects?

- What were the stay-at-home orders in the different cities? Did they all have similar stay-at-home orders Similar punishments for breaking the order? Any way to gauge how well people actually listened?

- The introduction seems to concentrate on the idea of interpersonal vs. property crimes, but there doesn’t seem to be much discussion or analysis of this in the text. Perhaps a visualization here would be nice where you color-code crimes that are considered personal and property and show in which city they increased, decreased or stayed the same.

-All of the tables in the main text and the SI are too hard to digest. I think summarizing the results in figures in the main paper and referencing those tables might be better.

In general I feel that this paper would be much easier to understand if the authors put some thought into creating figures that summarize the data and support their conclusions. Here are some suggestions:

- For Figure 1, plot the total crime in first three months of 2020 compared to 2019 and 2018 and compute the percent change.

- For Figure 2, color code the crimes as property or interpersonal and plot the change in number of crimes for each city that were statistically significant. I’d like to know from this graph, which crimes had similar behavior across the three cities and by how much did they change

- In the SI text, a comparison of the total number (or average number) of crimes between same time periods of 2019 compared to 2020 (summarizing Tables S1 - S7) would be good. You can even include this in the first figure summarizing Table 1 by breaking down the first three months to the three time periods that are analyzed in the SI text.

Again, I wouldn’t use scatter plots to visualize the data since the actual statistical analysis that’s being performed is simply on the total number of crimes before and after a date.

Smaller concerns:

- The introduction is lacking a bit and I would appreciate a bit more motivation for choosing these three particular crime sets, what they have in common, and what you found (be specific, by how much did crime overall seem to change during the stay-at-home order?).

- Write out less than or equal to rather than using the math symbol within paragraphs (e.g., lines 68, 123, 124)

- Additional and in line 133

- Say what sigma and mu are in Table 1 and 2 caption.

- A but more explanation on the Bonferonni correction and what you are using in the t-tests might be nice for readers that don't have a background in statistics

Reviewer #2: Please see the document attached above for my commentary on this paper. I forgot to mention some typos spread throughout the paper and things like "and and" that the authors should check on their manuscript.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: PONE-D-20-31968.pdf

Click here for additional data file.

24 Dec 2020

Dear Dr. Topaz,

First, we would like to thank you and the two anonymous reviewers for the comments and for the

helpful feedback provided, which has enhanced the manuscript in many ways. We greatly appreciate the

effort that the reviewers made to constructively provide advice about the manuscript and believe the

resulting aper is much stronger due to their reviews. Below is our response to the comments provided on

our manuscript entitled, “COVID-19 and crime: Analysis of crime dynamics amidst social distancing

protocols.” Our responses follow each comment.

Reviewer #1 Comments

1) I don’t understand the choice to only analyze two weeks following the stay-at-home orders. I

think the paper would be significantly strengthened by including more data after the stay-athome

orders. We are now well past the end date of those orders, so it makes sense to check

and see if crime indeed went back up (or didn’t).

We chose to analyze the two weeks following the implementation of stay-at-home orders due

to the likelihood that stronger adherence to stay at home orders would occur in the initial

time period following the announcement (see figures S1-S3). From the data, all three metrics

showed the best adherence to social distancing protocols immediately following their

implementation. The crime data from 2020 (see figures S4-S6) also show that April was the

month with the lowest number of crimes. Therefore, determining whether there is a

correlation between social distancing protocols and crime will be most clear when tested in

this time period. We are also aware that the full time period should be tested now that more

data are available but wanted to highlight the initial impacts of these protocols. We have

added more explanation within the manuscript as to our decision to test these two weeks,

which is further supported by new figures within the supplemental information.

2) I feel like the authors were very thorough in comparing changes in the 2020 data to previous

years to make sure the effects were not due to seasonality for Chicago, but why not for the

other two cities? Is it fair to only use the 2020 data for those two cities after performing such

an in-depth analysis of Chicago to claim that the changes in their crimes were in fact due the

stay-at-home order?

Chicago is often mentioned as a major hub for gun crime in the United States, making it an

important study system for crime in general. Also, it was one of the major cities in the United

States which experienced an early COVID-19 outbreak. The initial purpose of this study was

to observe changes that occurred in Chicago following the implementation of stay-at-home

orders, but we then determined that it would be helpful to determine whether these patterns

held across other cities. To address this in the manuscript, we have altered our paper

objectives to state that Chicago is our main study area and that the comparisons with

Baltimore and Baton Rouge were to support our conclusions about Chicago. We have also

added line plots highlighting the temporal changes in Baltimore and Baton Rouge over the

last three years to allay concerns that differences in the stay-at-home time period were due to

seasonality. In the supplementary material, the line plots added also include monthly total

crime data for 2020 through October as a basis for potential future discussions of longer term

impacts of behavior on total crime.

3) Why scatter plots? They are super hard to read. For the t-test, aren’t you just taking total

number before and after? Why not show just that? Or a moving average if you want to show

time effects?

We chose to use scatter plots in order to highlight the variation present in the datasets and

to highlight the differences in pre- vs. post-stay-at-home order crime. We have improved the

clarity of the scatter plots and added in a 5-day moving average trend line and discussed the

calculation of this in the supplementary material. This moving average analysis did not add

any new conclusions but may improve clarity of the underlying trends for readers. In terms of

the t-tests, we have further clarified the objectives of the t-tests to compare the average

number of crimes per day pre- and post-stay-at-home orders (as well as for comparisons of

time periods from previous years over the same seasons).

4) What were the stay-at-home orders in the different cities? Did they all have similar stay-athome

orders Similar punishments for breaking the order? Any way to gauge how well people

actually listened?

The stay-at-home orders were generally the same across the three cities. They closed all

non-essential businesses, mandated wearing masks, and encouraged citizens to minimize

non-essential visits outside their homes. The punishments for stay-at-home orders included

fines or potential jail time in Maryland and Illinois, but Louisiana did not enforce fines.

Information regarding the mandates and adherence to stay-at-home orders have been added

to the manuscript. We note that in the three metrics for adherence (change in distance

traveled, non-essential visits, and encounter density) across all three cities, the lowest values

over the year (through November) occurred during the initial time period following

implementation of stay-at-home-orders (see figures S1-S3). We have added figures

highlighting adherence to the supplemental information.

5) The introduction seems to concentrate on the idea of interpersonal vs. property crimes, but

there doesn’t seem to be much discussion or analysis of this in the text. Perhaps a

visualization here would be nice where you color-code crimes that are considered personal

and property and show in which city they increased, decreased or stayed the same.

This is a great suggestion. The crime types for each city are now broken down into

interpersonal vs. property vs. statutory crimes using (1, 2, and 3) to improve understanding

within the manuscript. The specific crime types are defined in the supplementary material

(table S2). We created a bar chart of the different crime types pre- and post-stay-at-home

orders that is now in the supplementary material (see figures S7-S12) to show how crime

categories shifted over these time periods. We have also added a statement regarding the

lack of uniformity of crime data availability across the U.S., which constrains the ability to

match exact crime types between cities.

6) All of the tables in the main text and the SI are too hard to digest. I think summarizing the

results in figures in the main paper and referencing those tables might be better.

We have constructed bar charts showing the differences in the three crime categories

between time periods (see figures S7-S12). We have also added the percent and direction

change to the tables in order to clarify the importance of each crime type and category’s

contributions to the overall decline in total crimes across the three cities.

7) In general, I feel that this paper would be much easier to understand if the authors put some

thought into creating figures that summarize the data and support their conclusions. Here are

some suggestions:

a. For Figure 1, plot the total crime in first three months of 2020 compared to 2019 and

2018 and compute the percent change.

This is a great suggestion. The percent change has been included for each table (see

tables 1 and 2, S3-S19).

b. For Figure 2, color code the crimes as property or interpersonal and plot the change in

number of crimes for each city that were statistically significant. I’d like to know

from this graph, which crimes had similar behavior across the three cities and by how

much did they change.

We have created bar charts for changes pre- and post-stay-at-home-order for the three

crime categories in each of the three cities to allow comparisons across cities (see

figures S7-S12). Due to the lack of uniformity in publicly available data, comparisons

of exact crime types between cities are difficult (e.g., theft data in Chicago differs from

auto theft data in Baltimore), but we hope the inclusion of crime categories and

percent change within the table clarifies the patterns observed in the data.

8) In the SI text, a comparison of the total number (or average number) of crimes between same

time periods of 2019 compared to 2020 (summarizing Tables S1 - S7) would be good. You

can even include this in the first figure summarizing Table 1 by breaking down the first three

months to the three time periods that are analyzed in the SI text. Again, I wouldn’t use scatter

plots to visualize the data since the actual statistical analysis that’s being performed is simply

on the total number of crimes before and after a date.

The purpose of a t-test is to compare the means across the two time periods. The data show

the day-to-day variability in occurrence of each crime. Scatter plots allow this variability to

be demonstrated for readers to encourage appreciation of this variability in cases for which

there are significant differences seen across the two time periods. We have added line graphs

showing the monthly total crime dynamics across the three cities in each year 2017-2019 and

January-October of 2020 (see figures S4-S6). We have also shown in tables S3-S19 the

percent change to further clarify the dynamics between the time periods of interest.

Smaller concerns:

9) The introduction is lacking a bit and I would appreciate a bit more motivation for choosing

these three particular crimes sets, what they have in common, and what you found (be

specific, by how much did crime overall seem to change during the stay-at-home order?).

We have added more information in the introduction to clarify the motivation and

conclusions of this study for both reviewers.

10) Write out less than or equal to rather than using the math symbol within paragraphs (e.g.,

lines 68, 123, 124).

This has been updated.

11) Additional and in line 133

This change has been made.

12) Say what sigma and mu are in Table 1 and 2 captions.

Mu and sigma have been defined in the table captions.

13) A bit more explanation on the Bonferroni correction and what you are using in the t-tests

might be nice for readers that don't have a background in statistics.

Further explanation of t-tests and the Bonferroni correction have been added to the

manuscript.

Reviewer #2 Comments

1) For example, for Chicago they use 20 out of 32 types of crime for their statistics, with a

cutoff of each crime being at least 0.2 percent of total tally. For the other cities, no criteria

are presented. The crime lists don’t include anything else than the names of the crimes (we

don’t get to learn about numbers, locations, definitions, which ones are considered felonies

etc). In some case the nomenclature is bizarre - is “juvenile” a crime? How is “vice” defined?

How are “gun crimes” different than “shooting”? What is the crime of “criminal damage”?

For Baton Rouge and Baltimore, all crime types available have been included. Meanwhile,

the Chicago dataset has far more crime types available. Some of them only have one or two

occurrences throughout the entire year and no occurrences during the time periods we were

observing in this study. To determine the appropriate percentage of total crime cutoff, we

found the crime type in Baltimore or Baton Rouge with the lowest percentage of total crimes

and used this as the baseline to determine our cutoff for Chicago. We have highlighted this

explanation in the manuscript for clarity. Each dataset includes different information since

there is no consensus on how to collect, define, and present publicly available crime data.

We have added a statement about data quality to the manuscript and included definitions for

each crime type in the supplemental information (see table S2).

2) Similarly, the t-Test section is poorly presented. We are not told what alpha is, what is n,

how the Bonferroni (or is it Bonferonni? It is spelt in two different ways) correction is

supposed to affect the t-Test. They speak of an experiment within the context of the t-Test,

but it is not clear what experiment they are referring to. What is a “year in year” test? Later

they call it a “pair-wise t-test appropriately corrected for replication” without any context of

what the replication is.

The t-test is a comparison of means from two different populations (in this case, different

time periods), in order to determine whether the data come from the same distribution. Alpha

is the level of significance desired, n is the total size of our sample, and the Bonferroni

correction keeps random results from showing significance due to the number of t-tests we

are using. When a researcher completes a large volume of any sort of statistical test, there is

a random chance some of them will come up as significant due to repetition. The Bonferroni

correction helps to tighten the restrictions on what is significantly different in order to

prevent these errors due to replication. We have added further descriptions and definitions to

the manuscript in order to clarify the statistical methods used and the comparisons

undertaken. We have also corrected the spelling of Bonferroni within the manuscript.

3) Finally, in the results section they speak of “victim-based crime data from Baltimore”, what

exactly does this mean? Is it specific to Baltimore? Is crime from Chicago not victim-based?

The crime data available from Baltimore is narrowed to only crimes in which there are

victims. From the dataset, victim-based crime is defined as crimes in which someone is

victimized (either personally or in terms of property). Meanwhile, Chicago and Baton Rouge

have crime datasets that are comprised of more crime types. We have added clarifications

into the materials and methods about what each of these crime datasets include. We have

also divided the crime types into three different crime categories (property, statutory, and

interpersonal). These crime categories are highlighted in the tables (1 and 2), as well as

being used to create bar graphs that show shifts between pre- and post-stay-at-home order

time periods (see figures S7-S12). The supplemental information now also includes

descriptions of each crime type used from the datasets (see table S2).

4) They also discuss Baton Rouge as a “city with different demographics” what does this mean?

Is it in terms of race? age? wealth? Up until this point (page 5) there has been no discussion

on demographics of the different cities. Indeed, we are not even told what the population

numbers are.

Baton Rouge, Chicago, and Baltimore are all cities regularly studied for their crime

dynamics. Chicago is the largest city, followed by Baltimore and then Baton Rouge. Each of

them has differing poverty levels, but all fall above the national average (10.5%). Chicago is

the densest, followed by Baltimore and then Baton Rouge. Therefore, these cities have

similarities, but also differences that make them interesting comparisons. We have added

more information about the cities themselves into the descriptions of the materials and

methods to address this. We have also added a table of demographic information to the

supplemental material for further clarification (see table S1).

5) In Table 1 \\sigma and \\mu are not defined. Nor is the p-value.

All of these terms have been defined in the manuscript.

6) In the Conclusions it is not clear what their message is. Consider these two statements made

by the authors:

1) Chicago's total crimes during the first three months of the year (including the time period

following the stay-at-home order) declined in 2020 compared to 2017, 2018, and 2019.

2) Before COVID protocols were put in place, there were no significant changes in total

crimes compared to 2018 and 2019.

Since COVID-19 became a problem in March 2020, and all the protocols came after March

2020, what are we to believe? Did crime within January 2020-March 2020 decline (as the

first sentence would imply!) or not (as the second sentence would imply)?

Then they add nuances that confuse the reader even more:

3) There were some changes in particular crime types that may have been due to social

factors, policing protocols, or other mechanisms. There were also significant changes in total

crimes between 2017 and 2020, with all but weapons violations showing declines.

So, did all these crimes except for weapons violations decrease? Also, what does it mean that

there were changes in “crime types?” Did the definition change? This is unclear. They also

discuss interpersonal crimes, but do not define an interpersonal crime. Is robbery an

interpersonal crime? Unclear.

The major conclusion of this paper is that there were significant changes that occurred in the

two weeks after the stay-at-home orders were put in place across all three cities and that

more work needs to be done to analyze the impacts that changes in social dynamics have on

crime. This conclusion was supported by first making sure that 2020 was not showing

unexpected crime dynamics in Chicago before the pandemic began. We then checked to be

sure that there were not seasonal dynamics in place by comparing time periods of past years

to the crime dynamics in 2020 (see figures S4, S13). Once we were certain that, generally,

there are no existing temporal patterns at play, we then test to see whether the stay-at-home

orders had a significant impact on the crime dynamics in Chicago and then, for comparison,

two other cities. We have clarified the conclusions of the manuscript and added more figures

to the supplementary material for ease of reader understanding. Specifically, we have added

descriptions of crime types and data quality, divided the crime types into crime categories,

and carried these definitions throughout the manuscript.

Sincerely,

Shelby M. Scott

National Defense Science and Engineering Graduate Fellow

Department of Ecology and Evolutionary Biology

Louis J. Gross

Chancellor’s Professor and Alvin and Sally Beaman Distinguished Professor of Ecology and Evolutionary

Biology and Mathematics

Director, National Institute for Mathematical and Biological Synthesis

Director, The Institute for Environmental Modeling, University of Tennessee

Past-President, The Society for Mathematical Biology

Submitted filename: Response to Reviewers.pdf

Click here for additional data file.

29 Jan 2021

PONE-D-20-31968R1

COVID-19 and crime: Analysis of crime dynamics amidst social distancing protocols

PLOS ONE

Dear Dr. Scott,

Thank you for submitting your manuscript to PLOS ONE. We appreciate the substantial improvements to the manuscript and one reviewer has signed off on it. The second reviewer has some additional constructive suggestions that we ask you to address. These suggestions are largely centered around bolstering the clarity of the argument you are making.

Please submit your revised manuscript by Mar 15 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Chad M. Topaz

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The manuscript is much improved. The main takeaway of the article is much clearer. However, it’s still not clear to me that the figures and tables are the best way to support the hypothesis of the paper. It seems that the main point of the manuscript is that not all crimes changed in the same way after the stay-at-home (SAH) order; specifically, it seems that property and statutory crimes experienced significant decreases in number during the two weeks post SAH order, but interpersonal crimes did not.

Major points:

- The introduction should clearly state the main conclusion of the paper. It currently states “changes in crime dynamics in all three cities .. are not uniform .. there are differences” What are those differences? Being clear about which city had significant changes in which type of crime and by how much up front will make the rest of the paper easier to digest.

Some ways to better support the hypothesis in the text:

- Label the crime types using P, S, and I rather than (1), (2), and (3) so it is easier for the reader to understand

- Rearrange Table 1 and Table 2 to group crimes of a similar type rather than alphabetically. This way, the reader can easily see that the greatest change in crime was in P and S types, not I. Similarly for the list of crimes in each city in the Data section

- Re-think Figure 2. The purpose of this figure is unclear to me. It is referenced in the text (e.g., in line 249) as supporting the hypothesis that P and S type crimes significantly decreased during SAH orders. This figure, however, doesn’t clearly show that. Instead, it seems to show only two crimes that significantly decreased during this time period for each city. It doesn’t even state which category those crimes are in. I wonder if it would be more useful to show ALL crimes that had significant decreased in each city and again order them by crime type (or color code by crime type, not city) to see that they are mostly type S or P.

Figure 2a,b is not referenced in the text.

- Along the same lines as the point above, Fig S10-S12 does attempt to demonstrate this point (and is referenced in line 249), but is lacking. I would include all crime types for each city (also label them by type instead of number, or remind the reader in the caption what the number represents) and clearly label which exhibit significant decreases (this is typically done with an asterisk above the two bars). It might also be clearer to color code by crime type, not city, since the main point is that all three cities demonstrated the greatest change in similar crime types.

Other points:

- The t-test indicates that there was a significant change in the number of crimes before and after the SAH order was implemented, but the scatter plots show a decrease in crime count prior to this date. This point should be addressed in limitations. The type of analysis that was performed cannot determine the date at which the crime changed. For example, if you chose to perform the same test before and after 3/1 instead, it looks from the scatter data that there would still be a significant decrease in total crimes in Chicago (Fig 1a).

- It’s unclear what is considered as the time period before SAH (e.g., in caption of Table 2 there are dates for the SAH and the two weeks after. What are the dates for before? Jan 1 - 3/21?). This should be clearly stated.

- The SI text needs a bit more information to be readable. It should include more text in the caption of the tables and figures, or should just be put into the main manuscript. For example, Fig S4 — cited in line 190. It would be useful to include a sentence after the title of the figure to describe what the reader should take away from that figure.

Minor errors:

- There are inconsistencies in how the manuscript refers to the SI figs (e.g., line 233 vs. line 236) and Figure vs. Fig (for example, see Fig in line 249 and Figure in line 244).

- The font size in every table is way too small

- Extra 5 in line 27 and 14.

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

9 Mar 2021

To whom it may concern,

First, we would like to thank the anonymous reviewer for their service and for the helpful feedback provided on our revision, which has enhanced the manuscript in many ways. Below is our response to the comments provided on our revision of the manuscript entitled, “COVID-19 and crime: Analysis of crime dynamics amidst social distancing protocols.” Our responses follow each comment in italics. Actions that have been taken or sections that have been added are colored in green.

Major points

1. The introduction should clearly state the main conclusion of the paper. It currently states “changes in crime dynamics in all three cities .. are not uniform .. there are differences” What are those differences? Being clear about which city had significant changes in which type of crime and by how much up front will make the rest of the paper easier to digest.

This is helpful feedback. We have explicitly listed the percent and direction change for each city’s overall crime, each of the crime types that significantly changed, and their percent and direction of change. The rest of the paper then outlines the support for these claims.

2. Some ways to better support the hypothesis in the text:

a. Label the crime types using P, S, and I rather than (1), (2), and (3) so it is easier for the reader to understand

We have added letter rather than number designations for crime categories throughout the manuscript.

b. Rearrange Table 1 and Table 2 to group crimes of a similar type rather than alphabetically. This way, the reader can easily see that the greatest change in crime was in P and S types, not I. Similarly, for the list of crimes in each city in the Data section.

We have rearranged the tables and lists within the manuscript to reflect this change.

c. Re-think Figure 2. The purpose of this figure is unclear to me. It is referenced in the text (e.g., in line 249) as supporting the hypothesis that P and S type crimes significantly decreased during SAH orders. This figure, however, doesn’t clearly show that. Instead, it seems to show only two crimes that significantly decreased during this time period for each city. It doesn’t even state which category those crimes are in. I wonder if it would be more useful to show ALL crimes that had significant decreased in each city and again order them by crime type (or color code by crime type, not city) to see that they are mostly type S or P.

Figure 2a, b is not referenced in the text.

Thank you for this incredibly helpful feedback. We have created a new Figure 2 that shows all of the crime types for the three cities, grouped by crime category, comparing average daily crime pre- and post-stay-at-home order implementation. The significant crime types are denoted with an asterisk next to the label. The past figure has been moved into the supplemental information (Fig S14) and now shows the time series with moving average trendline for all of the significant crime types in the three cities.

d. Along the same lines as the point above, Fig S10-S12 does attempt to demonstrate this point (and is referenced in line 249) but is lacking. I would include all crime types for each city (also label them by type instead of number or remind the reader in the caption what the number represents) and clearly label which exhibit significant decreases (this is typically done with an asterisk above the two bars). It might also be clearer to color code by crime type, not city, since the main point is that all three cities demonstrated the greatest change in similar crime types.

We have added a new Figure 2 that includes all the available crime types, grouped by crime category and have indicated which show significant different using an asterisk. We have maintained the color schemes based on cities for continuity between the main text and supplemental information but have split the crime types into crime categories throughout the main text for ease of understanding.

Other points

1. The t-test indicates that there was a significant change in the number of crimes before and after the SAH order was implemented, but the scatter plots show a decrease in crime count prior to this date. This point should be addressed in limitations. The type of analysis that was performed cannot determine the date at which the crime changed. For example, if you chose to perform the same test before and after 3/1 instead, it looks from the scatter data that there would still be a significant decrease in total crimes in Chicago (Fig 1a).

This is a great point, and we agree that there was a visual decline prior to implementation of the stay-at-home order. In the supplementary information (Table S7), we show that total crimes did decline in the state-of-emergency time period between the pre-COVID and stay-at-home order windows, but that this decline was not significant. We have added this information to the manuscript and also added to our limitations that we cannot know the exact date of the change in dynamics.

2. It’s unclear what is considered as the time period before SAH (e.g., in caption of Table 2 there are dates for the SAH and the two weeks after. What are the dates for before? Jan 1 - 3/21?). This should be clearly stated.

The time period before spans from the beginning of the year 1/1/20 until before the stay-at-home order was implemented on 3/21/20. We have clarified this in the methods section of the manuscript.

3. The SI text needs a bit more information to be readable. It should include more text in the caption of the tables and figures or should just be put into the main manuscript. For example, Fig S4 — cited in line 190. It would be useful to include a sentence after the title of the figure to describe what the reader should take away from that figure.

We have added more text to the supplementary information and also have updated the captions to make the figures capable of being understood without having to refer to the supplement text. Because of journal limitations, we unfortunately cannot add more figures or tables into the main text of the manuscript.

Minor errors

1. There are inconsistencies in how the manuscript refers to the SI figs (e.g., line 233 vs. line 236) and Figure vs. Fig (for example, see Fig in line 249 and Figure in line 244).

Thank you for catching these discrepancies. They have been updated.

2. The font size in every table is way too small

We have increased the font size for the tables in the main manuscript before forcing a split onto multiple pages.

3. Extra 5 in line 27 and 14.

This has been rectified.

Submitted filename: Reponse to Reviewers.pdf

Click here for additional data file.

18 Mar 2021

COVID-19 and crime: Analysis of crime dynamics amidst social distancing protocols

PONE-D-20-31968R2

Dear Dr. Scott,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Also, you'll see a few typo corrections suggested by one reviewer.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Chad M. Topaz

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: (No Response)

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: (No Response)

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: (No Response)

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: (No Response)

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I’m happy with the edits the authors have made. Here are just a few small editorial comments:

Line 27: should read “main focus is Chicago”

Line 186 should read S3-S5 Figs

Line 199: should say S4 Fig

Line 216: $t$-tests

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

22 Mar 2021

PONE-D-20-31968R2

COVID-19 and crime: Analysis of crime dynamics amidst social distancing protocols

Dear Dr. Scott:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Chad M. Topaz

Academic Editor

PLOS ONE