The COVID-19 Pandemic,Stay-at-Home Orders,and Gun Violence: A Story of Two Cities

Abstract

This study examines the impact of the pandemic on gun violence in Philadelphia and Washington DC. Interrupted time-series analysis is used to examine weekly data from January 2017 to March 2021. Robust diagnostic checks confirm the validity of the fitted models. There were significant increases in gun violence during the pandemic, especially in the staged relaxation of social distancing. The timing of the increases in gun violence varies by location and fatality. Criminal justice agencies should place more attention and reallocate resources on gun violence in a timely manner in the volatile state of the nation. Finally, this study concludes with a discussion of the findings, limitations, and implications for future research.

Keywords

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

Introduction

Since the first cases of COVID-19 emerged in December 2019 in China, it has exponentially spread to other countries. The United States is not an exception. As of March 2021, there have been almost 30 million confirmed cases, and more than 540,000 people have died of COVID-19 (Centers for Disease Control and Prevention, 2020). Besides the vaccination and medical approach, U.S. governments have imposed various containment measures to decrease the risk of COVID-19 transmission, ultimately flattening the curve. The implementation of the stay-at-home (SAH) order and social distancing policies resulted in the closing of nonessential workplaces, restaurants, schools, and borders. The negative consequences of the pandemic and containment measures inflicted upon individuals are well documented in the media, scientific journals, and government reports (Pew Research Center, 2020; World Health Organization [WHO], 2020). For example, there were significant increases in unemployment (Falk et al., 2020), alcohol consumption (Pollard et al., 2020), and social isolation (Clair et al., 2021), and many individuals experienced increased stress and struggled with new or preexisting mental and behavioral problems (Clair et al., 2021).

There have been significant increases in gun violence across the United States during the pandemic (Donaghue, 2020; Everytown Research & Policy, 2021). Policymakers have been struggling with COVID-19 and gun violence. Despite the seriousness of these public health crises, little research has been available on the impact of the pandemic on gun violence, and its findings offered no simple, consistent conclusion (Abrams, 2020; Campedelli, Aziani, et al., 2020; Kim, 2022a; Kim & Phillips, 2021; Rosenfeld & Lopez, 2020), warranting further research endeavors. Focusing on crime data from two cities, Philadelphia and Washington DC, this study explores whether gun violence increased during the pandemic and when and how much such changes occurred between the SAH order and relaxation of social distancing. Both cities experienced significant increases in gun violence in the pandemic era, especially during the relaxation of social distancing. The timing of such increases differs by location and fatality. The comparison of both cities allows us to examine whether the effects of the pandemic were limited to a particular city or not. Finally, this study concludes with a discussion of implications for further research and policy.

Literature Review

Theoretical Backgrounds

Strain theory is useful to discuss an increase in gun violence during the pandemic (Agnew, 2006; Merton, 1938). It expounds on how crime is associated with three types of strain: the failure of accomplishing monetary success, presentation of negatively valued stimuli, and removal of positively valued stimuli. Individuals commit crimes as a result of experiencing these types of strain. During the pandemic, individuals were exposed to various negative life events. Unemployment has been one of the biggest negative stimuli to which most individuals and families were exposed. In Philadelphia, there was a significant increase in the unemployment rate from 6.7% in March to 18.4% in June 2020, and in DC, from 5.2% to 8.9% (U.S. Bureau of Labor Statistics, 2021). Due to the mandatory closure of nonessential workplaces, many individuals were laid off and had difficulty in making ends meet. Individuals may turn to gun violence out of strain and/or anger when there are few or no legitimate opportunities to achieve culturally valued goals, including job security and wealth (Merton, 1938; Messner & Rosenfeld, 2006). In addition, the pandemic caused a significant increase in alcohol use (The Nielsen Company, 2020; Pollard et al., 2020). To deal with the stress of financial hardship and social isolation, individuals were drinking more than usual during the pandemic. Excessive drinking led to increases in (mental) health risks and deviant behaviors (Greenfeld, 1998; Pollard et al., 2020).

In these highly volatile and stressful circumstances, there was a significant increase in gun sales (Federal Bureau of Investigation, 2021). According to Pennsylvania State Police (2020), 1,445,910 background checks were processed in 2020, which represents a 47.2% increase over the previous year. A similar trend was found in Washington DC; there was a 25% increase in gun registrations from 2019 to 2020 (Mullins, 2020). The fear of the unknown drove millions of people across the country to purchase guns for the first time or additional ones for personal protection (Everytown Research & Policy, 2021; Kraviz-Wirtz et al., 2020). There is a positive relationship between gun availability and gun violence (Siegel et al., 2013). In sum, individuals are more likely to engage in gun violence in the context of high strain, unemployment, alcohol use, and gun ownership during the prolonged pandemic.

While strain theory offers an explanation for the increase in gun violence during the pandemic, routine activity and situational crime prevention theories provide insights for understanding the impact of the SAH order and social distancing policies. Routine activity theory associates the pattern of crime with changes in the daily routines of individuals. Crimes occur when three factors converge in time and place: motivated offenders, suitable targets, and a lack of capable guardians (Cohen & Felson, 1979). In addition, situational crime prevention theory (Clarke, 1980) helps gain an accurate understanding of the mechanisms underlying the effect of the SAH order on crime. It posits that crime is a combination of choices made by individuals and situational criminal opportunities.

To place the theories into context, there would be a decrease or at least no significant increase in gun violence under the SAH order. Given that mandatory physical isolation restricts individuals’ mobility, there would be fewer opportunities for victims and offenders to come into contact in public areas, eventually decreasing situational opportunities for gun violence. On the contrary, the relaxation of social distancing can lead to an increase in gun violence due to increased contact among people. Individuals began spending more time on the streets during the relaxation of social distancing, which brings victims and offenders into contact in the context of high strain, unemployment, alcohol use, and gun ownership during the prolonged pandemic. A combination of those factors might increase situational opportunities for gun violence.

Empirical Evidence

Prior research examined whether there were significant changes in crime during the pandemic (Andresen & Hodgkinson, 2020; Ashby, 2020; Balmori de la Miyar et al., 2020; Payne et al., 2020; Piquero et al., 2020). Most studies analyzed daily or weekly counts of gun violence at the city level using time-series analyses. The literature review focuses on changes in U.S. gun violence during the pandemic.

The early literature showed that the pandemic and social distancing restrictions did not result in significant changes in homicide and gun violence across U.S. cities (Abrams, 2020; Campedelli, Aziani, et al., 2020). Campedelli, Aziani, et al. (2020) conducted a Bayesian structural time-series examination of daily data from January 2017 to March 2020. The results showed no evidence of significant changes in homicides and assaults with a deadly weapon in Los Angeles. In addition, a difference-in-difference examination of week-to-week data from February 2015 to May 2020 presented no significant changes in homicides and shootings across 25 U.S. cities after the SAH orders were in effect (Abrams, 2020). Using crime data in Los Angeles from January 2016 to September 2020, Brantingham et al. (2021) examined temporal trends in gang- and nongang-related violent crimes using the Autoregressive Integrated Moving Average (ARIMA) models. They compared the observed values for each crime type with what were predicted after March 16, 2020, with 95% and 99% confidence intervals. Overall, there were no significant impacts of the SAH order on both gang- and nongang-related violent crimes. However, the SAH order caused several discrete spikes in gang-related all violent crime, aggravated assaults, and gun violent crime. Similar spikes were also detected for nongang-related aggravated assault and gun violent crime.

On the contrary, Kim and Phillips (2021) examined gun violence in one city, Buffalo, NY, focusing on fatal versus nonfatal shootings and gang versus nongang-related shootings. They used ARIMA and poisson models to examine weekly data from January 2017 to October 2020, indicating significant gradually accruing increases in both nonfatal and gang related shootings. In addition, Kim (2022a) examined the varying effects of the pandemic on gun violence by containment measure type, fatality, and location. Box-Cox transformation and GARCH models were used to analyze weekly data in New York City from January 2016 to December 2020. He found significant increases in fatal and nonfatal shootings during the pandemic, especially in urban boroughs. Such increases mostly occurred during the staged relaxation of social distancing, not under the SAH order, when individuals were allowed to interact with others. In addition, there was some evidence of high volatility in gun violence during the pandemic, which has led to greater perceived uncertainty, insecurity, and fear of gun violence among people. Finally, Rosenfeld and Lopez (2020) analyzed weekly crime data in 28 U.S. cities from January 2017 to October 2020. While adjusting for seasonality, they found that there were significant changes in violent crime rates during the pandemic. Specifically, homicide, aggravated assault, and gun assault in 2020 increased by 29%, 9.8%, and 10.1%, respectively, over the same period from January to October in 2019.

Summary of Prior Research

In the United States, few studies were conducted to examine changes in gun violence during the pandemic, relative to other crime types. The empirical findings remain inconclusive and vary across temporal and geographic contexts. Early studies found no significant impact of the SAH orders on homicides and shootings across U.S. cities. However, these results should be interpreted with caution, given the use of short-term postintervention data. By design, they overlooked that the impacts of the SAH orders may not be immediate and instead realized through a gradual process. Using the relatively long-term postintervention data, other studies found evidence of significant increases in gun violence across cities during the pandemic, especially in the staged relaxation of social distancing. Given the lack of research and its conflicting evidence, much work is warranted to have a complete understanding of the pandemic and gun violence. Thus, this study attempts to examine long-term postintervention changes in gun violence in Philadelphia and Washington DC.

The Present Study

The research is based on two city-specific case analyses. According to the U.S. Census Bureau (2021), Philadelphia is one of the most populous cities in the United States with a population of 1,603,797 residents in 2020. The population per square mile in 2010 for Philadelphia was 11,379.5. Relatively, Washington DC. has a smaller population size (689,545), and is also less densely populated (9,856.5). Furthermore, Philadelphia (19.4) has a higher poverty rate, as opposed to Washington DC (15). 34.3% of the population in Philadelphia was White alone, not Hispanic or Latino, while in DC that percentage was 37.5%. According to the 2018 crime in the United States data (FBI, 2018), DC (20.78 per 100,000) had a higher murder rate than Philadelphia (17.38).

Based on the abovementioned theoretical and empirical frameworks, this study explores two research questions. First, it examines whether there were increases in gun violence during the pandemic, while seasonality and other covariates are held constant. Second, it assesses varying impacts of the pandemic across levels of social distancing, as in Andresen and Hodgkinson’s (2020) study.

Method

Data

The data were retrieved from the open data portals of Philadelphia and Washington DC. Open data portals refer to any online platforms that government agencies have established to share their data with the public. Their open data sources facilitate public monitoring of government functions, which ultimately contributes to greater transparency and accountability in government.

This study aggregated individual cases to weekly segments from January 1, 2017, to March 14, 2021. Interrupted time-series analysis was used to examine the impact of the SAH order and relaxation of social distancing on gun violence. This study analyzed shooting incidents only known to police, which possibly underestimated the true prevalence of gun violence during the pandemic. However, this type of bias may not be large in this study given that serious crimes, including gun violence, are more likely to be reported to the police.

There are three dependent variables: fatal, nonfatal, and the total combined shootings. It is important to collapse data into subcategories. Aggregate data do not provide a clear picture of crime because it is skewed by the crime with the highest number of occurrences such as nonfatal shootings in this study. The large volume of nonfatal shootings overshadowed more serious but less frequently committed fatal shootings, which can lead to an underappreciation of changes in fatal shootings and ultimately skew our understanding of the pandemic–gun violence association.

The independent variables are the SAH order and relaxation of social distancing. The execution of quarantine and social distancing policies differed across jurisdictions in onset and length. Philadelphia and Washington DC issued a SAH order on March 22 and April 1, respectively, which closed all nonessential businesses and prohibited social gatherings of any size. The orders expired on June 4 in Philadelphia and June 8 in Washington DC. The SAH order is captured in a binary variable. All observations under the SAH order period are coded as one, and otherwise coded as zero. Another dummy variable measures the staged relaxation of social distancing. It takes the value of one after the expiration of the SAH order for each city and coded all preintervention data as zero.

Several control variables are included in the study: seasonality, Black Lives Matter (BLM) protests, and/or the riot at the Capitol. First, three quarterly dummy variables are constructed to control for seasonality over time. The literature has revealed seasonal patterns in violent crime (McDowall & Curtis, 2015; McDowall et al., 2012). Each variable for the second, third, and fourth quarter is coded as one and otherwise coded as zero. The respective variable is compared with the first quarter for interpretation.

Second, the impact of the BLM protests is dichotomously measured (1 = the weeks of 5/24 to 6/21; 0 = otherwise). Following the death of George Floyd on May 25, there were a series of protests and riots in June, which might influence the volume of gun violence. The motivations behind gun violence during the BLM protests are ambiguous. BLM protesters can be victims as well as offenders. In addition, given that protesters showed strong opposition toward the police, officers had difficulties in enforcing laws and could fall back from their work in public safety (Cassell, 2020; Nix et al., 2018; Phillips, 2020; Rushin & Edwards, 2017), in turn causing an increase in gun violence in the absence of law enforcement.

Third, the Capitol riot occurred in Washington DC on January 6, 2021, leading to a significant number of injuries and deaths during the riot. The week of January 3 to 9 is coded as one to control for the potential impact of the Capitol riot on gun violence, and all other observations are coded as zero. Finally, this study could not include other control variables due to the unavailability of weekly information at the city level. Instead, it includes more than 3 years of preintervention data to reduce possible threats to internal validity and control over secular trends.

Statistical Analysis

There are several procedures for statistical analysis. First, this study offers descriptive statistics and corresponding t tests to see whether there were significant differences between the preintervention and postintervention means of gun violence. Second, it conducts unit root and normality tests to understand the basic features of the time series. Third, it uses interrupted time-series analyses to estimate the impact of the pandemic on gun violence while controlling for control variables and secular patterns. This study considers three impact patterns to model the impact of the interventions: an abrupt-temporary change, an abrupt-permanent change, and gradual-permanent change (McDowall et al., 1980). The Akaike information criterion and Schwarz information criterion are used as guides for selecting the appropriate model. Finally, diagnostic checks are performed to evaluate the validity of fitted models in terms of autocorrelation, normal distribution, and heteroscedasticity.

Results

Descriptive Statistics and T-Tests

Figures 1 and 2 demonstrate substantial increases in gun violence during the pandemic. This study compares the preintervention and postintervention means using the SAH order as the starting point of the pandemic. As seen in Table 1, the t tests indicate significant increases in all types of gun violence during the pandemic. In the following sections, this study examines the varying effects of the pandemic across the cities by the types of social distancing and gun violence, while the covariates remained constant. The issues in question are when such increases in fatal and nonfatal shootings occurred between the SAH order and the relaxation of social distancing.

Figure 1.

Gun violence in Philadelphia.

Figure 2.

Gun violence in Washington DC.

Table 1.

Preintervention and Postintervention Means of the Time Series.

Variable	Preintervention, M (SD)	Postintervention, M (SD)	Change in mean	t-test
Philadelphia
Total	26.23 (7.35)	43.48 (13.97)	+17.25	−11.65**
Fatal	5.55 (2.32)	7.23 (4.08)	+1.68	−3.73**
Nonfatal	20.68 (6.62)	36.25 (11.53)	+15.57	−12.21**
Washington DC
Total	30.40 (7.29)	40.88 (11.57)	+10.48	−7.75**
Fatal	2.19 (1.63)	3.27 (2.24)	+1.08	−3.80**
Nonfatal	28.21 (6.95)	37.61 (11.00)	+9.40	−7.29**

Significant at α = .01.

Stationarity and Normality Tests

Table 2 reports the findings of unit root and normality tests. For all the time series, the null hypothesis that the series has a unit root is rejected. Thus, they do not possess a unit root and are stationary. In addition, while most time series approximate a normal distribution, fatal shootings in Philadelphia and Washington DC are not normally distributed.

Table 2.

Unit Root and Normality Tests of the Preintervention Time Series.

Variable	Aug. DF	PP	HEGY	J-B
Philadelphia
Total	−10.39**	−11.03**	6.97*	4.04
Fatal	−10.92**	−10.82**	11.72*	7.55*
Nonfatal	−7.10**	−12.45**	14.96*	4.00
Washington DC
Total	−10.03**	−10.59**	33.54**	1.05
Fatal	−11.69**	−11.69**	53.79**	16.87**
Nonfatal	−10.08**	−10.53**	30.80**	.48

Note. Aug. DF-GLS = augmented Dickey Fuller; PP = Phillips-Perron; HEGY = Hylleberg, Engle, Granger, and Yoo; J-B = Jarque-Bera.

Significant at α = .05. **Significant at α = .01.

Gun Violence in Philadelphia

Table 3 shows the results of Autoregressive Moving Average (ARMA) models for Philadelphia. Diagnostic checks confirm no serial correlation nor heteroscedasticity in the residuals. Given that the residuals in the models for the total combined and fatal shootings are not normally distributed, this study conducts additional analyses with the Box-Cox transformed data.

Table 3.

ARMA Models by Fatality, Philadelphia.

Variable/model	Total		Fatal		Nonfatal
Variable/model	Original	Box-Cox ^{λ = .343}	Original	Box-Cox ^{λ = .384}	Original	Box-Cox ^{λ = .465}
Constant	18.02 (1.87)**	5.39 (.12)**	5.38 (.48)**	2.62 (.13)**	16.07 (.96)**	5.61 (.17)**
SAH	9.37 (2.88)**	1.06 (.30)**	−.40 (1.08)	−.19 (.29)	10.76 (2.44)**	1.86 (.44)**
Relaxed SAH	14.62 (2.07)**	1.69 (.16)**	1.71 (.53)**	.45 (.14)**	15.83 (1.32)**	2.62 (.24)**
BLM	7.20 (4.17)^†	.61 (.44)	2.69 (.97)*	.80 (.35)*	5.72 (3.59)	.75 (.65)
Q2	4.52 (1.74)**	.69 (.18)**	−.30 (.63)	−.07 (.18)	5.61 (1.45)**	1.18 (.26)**
Q3	8.19 (1.71)**	1.05 (.17)**	−.63 (60)	.20 (.17)	8.86 (1.37)**	1.66 (.25)**
Q4	4.49 (1.66)*	.64 (.17)**	−.49 (.57)	.12 (.16)	5.03 (1.37)**	1.00 (.25)**
δY_t-1	.16 (.07)*	–	–	–	–	–
AR(7)	–	–	.21 (.06)**	.17 (.08)*	–	–
AR(8)	–	–	–	−.14 (.07)^†	–	–
Adj. R²	.49	.45	.11	.09	.51	.48
SIC	7.27	3.02	5.04	2.65	6.95	3.54
Ljung-Box Q₃₆	45.42	42.69	45.67	42.07	37.03	32.16
Jarque-Bera	17.60**	10.14**	72.20**	3.37	1.96	2.07
ARCH Obs × R^2	.03	.84	.71	1.28	.14	.76

Note. ARMA = Autoregressive Moving Average; SAH = stay at home; BLM = Black Lives Matter; SIC = Schwarz information criterion.

†

Significant at α = .10. * Significant at α = .05. ** Significant at α = .01.

The intervention coefficient indicates the change in the number of shootings following the intervention. Table 3 indicates a significant increase in the total combined shootings on the original matrix, independent of the levels of social distancing. There were increases of 9.37 total combined shootings per week under the SAH order and 14.62 during the phased relaxation of social distancing. When the data are disaggregated by fatality, there is a difference in the estimates of the SAH intervention in terms of significance and direction. There were significant increases in nonfatal shootings during both the SAH order and the relaxation of social distancing. Fatal shootings significantly increased during the relaxation of social distancing only, not under the SAH order. There was a decrease in fatal shootings under the SAH order, but its effect was not significant.

A lagged dependent variable is included in the models for the total combined shootings to estimate the gradual, permanent effects of the SAH and relaxation of social distancing. It is not included for fatal and nonfatal shootings due to its insignificance and to find the best-fitting model. The gradual, permanent impact of the SAH order on the total combined shootings is 11.15, or 9.37/(1–.16). The cumulative effect of the intervention for each passing week can be also computed using the equation: Z_t = ∑ δ₁^j-1ω₀ (McDowal et al., 2019). The cumulative impact of the SAH order on the total combined shootings is 9.37 in the first week of the SAH order, 10.87 in the second week, 11.11 in the third week, and almost 11.15 in the fourth week. Given the parameter of Y_t–1 is relatively small, the impact of the intervention arose more rapidly within four weeks. In addition, the gradually accruing impact of the relaxed SAH variable is 17.40, or 14.62/(1–.16). Its cumulative impact is 14.62 in the first week of the intervention, 16.96 in the second week, 17.33 in the third week, 17.39 in the fourth week, and so on.

The BLM protests are significant for fatal shootings only. They were associated with an increase of 2.69 fatal shootings. In addition, seasonal dummy variables were significant predictors of the total combined and nonfatal shootings. The total combined and nonfatal shootings are likely to increase in the second, third, and fourth quarters of the year as opposed to the first quarter. On the contrary, there were no seasonal patterns in fatal shootings.

Given that the residuals in the models for the total combined and fatal shootings are not normally distributed, this study conducts the Box-Cox transformation to reduce the problem of nonnormality (Box & Cox, 1964). A range of log-likelihood values are considered to identify the best power value to which each dependent variable should be raised for optimal transformation (see Figure 3). The Box-Cox transformation equation is written as: y^(λ) = (y^λ−1)/λ. Then, it builds the models with the transformed data. Table 3 reports the best lambda values: .343 for the total combined shootings, .384 for fatal shootings, and .465 for nonfatal shootings. The residuals of the model for fatal and nonfatal shootings are normally distributed, whereas nonnormality still exists in the residuals of the model for the total combined shootings.

Overall, there were no notable differences between the original and transformed models in the coefficient estimates of all variables in terms of their significance and direction. The Box-Cox transformation is effective in reducing nonnormality in the residuals. However, it introduces complications for interpreting the models due to the transformed data. For example, there was an increase in the total combined shootings, on average, by 1.69 units of the transformed series ^{λ = .343} when the SAH expired and social distancing restrictions were loosened. It is also implausible to interpret them in a consistent way due to the use of different power values for all the variables.

Violence in Washington DC

Table 4 presents the results of ARMA models by fatality. The residual diagnostics indicate no serial correlation or heteroscedasticity in the models. According to the Jarque-Bera test, there is evidence of nonnormality in the residuals of all models. This study conducts additional analyses with the Box-Cox transformed data. Table 4 reports the best λ values: .505 for the total combined shootings, .343 for fatal shootings, and .505 for nonfatal shootings. The residuals in all models approximate a normal distribution on the transformed matrix.

Table 4.

ARMA Models by Fatality, Washington DC.

Variable/model	Total		Fatal		Nonfatal
Variable/model	Original	Box-Cox ^{λ = .505}	Original	Box-Cox ^{λ = .343}	Original	Box-Cox ^{λ = .505}
Constant	21.00 (2.08)**	7.33 (.37)**	1.79 (.23)**	1.11 (.11)**	20.07 (1.96)**	7.07 (.36)**
SAH	−.35 (2.59)	−.11 (.46)	.19 (.63)	.15 (.28)	−.48 (2.50)	−.15 (.46)
Relaxed SAH	10.02 (1.62)**	1.62 (.28)**	1.16 (.33)**	.42 (.15)**	9.32 (1.52)**	1.57 (.28)**
BLM	1.28 (3.62)	35 (.64)	.85 (.87)	.46 (.39)	.61 (3.50)	.24 (.64)
CAPITOL	−.36 (7.51)	.18 (1.33)	.05 (1.81)	.24 (.82)	−.56 (7.25)	.13 (1.33)
Q2	3.02 (1.48)*	.60 (.26)*	.39 (.36)	.16 (.16)	2.68 (1.43)^†	.57 (.26)*
Q3	4.57 (1.42)**	.90 (.25)**	.62 (.33)^†	.30 (.15)*	4.15 (1.36)**	.84 (.25)**
Q4	6.71 (1.48)**	1.21 (.26)**	.68 (.33)*	.31 (.15)*	6.29 (1.42)**	1.18 (.26)**
δY_t−1	.20 (.07)**	.04 (.01)**	–	–	.18 (.07)*	.03 (.01)**
Adj. R²	.40	.39	.05	.06	.37	.36
SIC	7.01	3.54	2.55	2.73	6.94	3.54
Ljung-Box Q₃₆	25.95	28.53	22.95	23.34	22.78	32.76
Jarque-Bera	24.95**	2.08	27.11**	1.38	37.08**	2.77
ARCH Obs × R^2	.61	1.06	.47	.06	.57	1.81

Note. ARMA = Autoregressive Moving Average; SAH = stay at home; BLM = Black Lives Matter; SIC = Schwarz information criterion.

†

Significant at α = .10. * Significant at α = .05. ** Significant at α = .01.

Based on the original data, all models show significant increases in gun violence in the staged relaxation of social distancing only. For example, there were increases of 10.02 total combined, 1.16 fatal, and 9.32 nonfatal shootings in Washington DC during the relaxation of social distancing. However, no significant changes were found under the SAH order. As the lagged rate parameter is significant at the .01 level, it is useful to compute the gradually accruing impact of the relaxation of social distancing on the total combined shootings (12.53 or 10.02/[1–.20]) and nonfatal shootings (11.37, or 9.32/[1−.18]). In addition, seasonality is an important predictor of all types of shootings. Finally, both the BLM and Capitol variables were not significant. The Box-Cox transformed data also display similar results (see Table 4).

Figure 3.

The Box-Cox transformation of the time series.

Given the availability of disaggregated data in DC, this study performs additional analyses of nonfatal gun violence associated with assault and robbery (see Appendix). Diagnostic checks confirm no serial correlation and heteroscedasticity in the residuals. Based on the original data, nonfatal assault and robbery shootings significantly increased by 3.46 and 5.99 incidents, respectively, during the staged relaxation of social distancing. The gradual, permanent impact of the relaxation of social distancing on robbery shootings is 7.88, or 5.99/(1–.24). On the contrary, there were no significant changes in either nonfatal assault or robbery shootings under the SAH order. In addition, significant seasonal cycles appear in both nonfatal assault and robbery shootings. No BLM protest and Capitol riot variables are significant in explaining variations in nonfatal assault and robbery shootings. Given that the residuals in the models for nonfatal robbery shootings are not normally distributed at the .01 level, this study conducts the Box-Cox transformation to reduce the problem of nonnormality. The Appendix reports the best lambda values: .505 for nonfatal assault shootings and .424 for nonfatal robbery shootings. There are similarities in the estimated coefficients between the original and transformed data. There is no serial correlation, nonnormality, nor heteroscedasticity in the residuals at the .05 signficance level.

Discussions and Conclusions

To reiterate, this study examined increases in gun violence during the pandemic and whether such increases appeared under the SAH order and/or relaxation of social distancing. Several important issues in the research are worthy of attention. First, gun violence significantly increased during the pandemic in both Philadelphia and Washington DC. The current finding is in line with strain theory. During the pandemic, individuals encountered various negative life events, including unemployment, isolation, and mental distress. For example, the unemployment rate dramatically increased in Philadelphia from 6.7% in March to 10.2% in December 2020 with a peak in July at 19.5%, and in DC from 5.2% to 8.8% with a peak in April at 11.1% (U.S. Bureau of Labor Statistics, 2021). In these highly stressful situations, there were increases in alcohol consumption (The Nielsen Company, 2020; Pollard et al., 2020) and gun sales (Federal Bureau of Investigation, 2021). All of which might have increased gun violence in both cities.

The next issue in question is when such increases in gun violence occurred between the SAH order and relaxation of social distancing. Overall, the impact of the SAH order differs by social distancing type, location, and fatality. Philadelphia experienced significant increases in both fatal and nonfatal shootings during the relaxation of social distancing. Under the SAH order, nonfatal shootings significantly increased in Philadelphia. In Washington DC, there were significant increases in both fatal and nonfatal shootings only during the relaxation of social distancing. Additional analyses of disaggregated data in DC reveal that nonfatal gun violence associated with assault and robbery significantly increased only after the expiration of the SAH order.

More specifically, despite the increased strain and mental health problems during the pandemic, most time series did not significantly decrease or increase under the SAH orders. Andresen and Hodgkinson (2020) found significant decreases in various crimes across Queensland’s districts in Australia during the lockdown period. In this study, the timing of the increases in gun violence varies by location and fatality. In Washington DC, gun violence did not increase under the SAH order and began significantly increasing during the relaxation of social distancing. In Philadelphia, however, there was a significant increase in nonfatal shootings under the SAH order, as well as during the relaxation of social distancing. This finding might result from the significant spike in the unemployment rate from 6.7% in March to 18.4% in June 2020 under the SAH order, which is much higher than the increase from 5.2% to 8.9% in Washington DC during the same period (U.S. Bureau of Labor Statistics, 2021). The current findings are somewhat consistent with Kim and McCarty’s (2021) study. In Chicago, homicides increased significantly during both the SAH order and relaxation of social distancing when the unemployment rate exploded from 6.1% in March to 18.3% in April to 13.2% in June (U.S. Bureau of Labor Statistics, 2021). In sum, Philadelphia was more susceptible to the economic fallouts of the pandemic, ultimately leading to a substantial increase in nonfatal shootings immediately following the announcement of the SAH order (see Figure 1). During the pandemic, there were greater increases in gun sales in Philadelphia, as opposed to Washington DC. Philadelphia also had higher levels of poverty, minority populations, and population density. Future research should expore what socio-economic factors were related to the increases in gun violence.

There is strong evidence that the relaxation of social distancing led to significant increases in gun violence regardless of city or fatality. These findings are in line with Andresen and Hodgkinson’s (2020) and Kim’s (2022a) studies showing that violence significantly increased when the social distancing policies were relaxed upon the expiration of the SAH order. The relaxation of social distancing increased physical contact among individuals, potentially leading to significant increases in both fatal and nonfatal shootings due to the prolonged stress linked to the pandemic and social distancing policies. On the other hand, given that most time series did not significantly increase under the SAH orders, the imposition of strict restrictions on individuals’ routine activities might decrease or at least not increase the situational opportunities for gun violence. These findings reflect routine activity theory, as well as strain theory.

There were nationwide BLM protests during the pandemic, some of which resulted in violent riots. Overall, BLM protests did not have significant effects on gun violence in either Philadelphia or Washington DC. One exception was a significant increase in fatal shootings in Philadelphia during the period of BLM protests. These findings are somewhat contrary to what was found in Kim and Phillips’s (2021) and Kim’s (2022a) studies. In their studies, the effects of BLM protests were often significant in the models for nonfatal shootings, whereas they were not significant for fatal shootings. These results do not necessarily mean that BLM protesters are responsible for the increases in gun violence. They can be victims of gun violence, while they can take part in gun violence. In addition, the BLM protests might simply provide opportunities for those who were already motivated to gun violencein the absence of law enforcement. Further research is warranted to reveal the motivations of crime occurring during the BLM protests. In addition, there is wide variation in measuring the effect of the BLM protests and de-policing phenomenon, and their influence depends on how the variable is operationalized. This issue is also in need of research attention.

In this study, the impact of seasonality varied by fatality. Nonfatal shootings show significant seasonal patterns, whereas the impact of seasonality on fatal shootings is often insignificant and much weaker. These findings are in line with those of prior research that seasonality is more apparent in nonfatal violence as opposed to fatal violence (McDowall & Curtis, 2015). In addition, there were local variations. Seasonality is important to explain the mechanisms underlying fluctuations in both fatal and nonfatal shootings in Washington DC, whereas it was a significant predictor of nonfatal shootings only in Philadelphia.

In the United States, guns have been an important part of society. The Second Amendment and the Supreme Court (District of Columbia v. Heller, 2008) have conferred an individual right to bear and keep guns. There are three reasons for gun acquisition: recreation, self-defense, and Second Amendment mobilization (Boine et al., 2020; Yamane, 2017). In recent years, the most popular element is self-defense, subsequently followed by Second Amendment activism, and recreation. There were substantial increases in gun violence when COVID-19 arrived and clashed with the U.S. gun culture (Everytown Research & Policy, 2021). Many people bought new guns to deal with feelings of isolation, uncertainty, insecurity, and fear resulting from the public health and economic emergency (Kraviz-Wirtz et al., 2020). During the relaxation of social distancing, gun violence might have increased due to a confluence of motivated offenders with financial distress, suitable targets, gun availability, and decreased criminal justice interventions in the uncertain and volatile state of the nation.

There are several study limitations. First, given that the pandemic is a universal phenomenon across the country, it is difficult to identify appropriate comparison groups that were not exposed to the intervention. Instead, this study conducted two city-specific case analyses to examine which city more amenable to the impact of the pandemic. It also used long-term data to reduce internal validity threats and control over possible rival hypotheses. Second, police departments have collected and made crime data available to the public using open web portals. Advances in data collection and availability have promoted more research with various designs, ultimately resulting in a better understanding of the pandemic-gun violence association. Despite the value of open portals, they are not without their limitations: under-reporting and under-recording. Police statistics include crimes known to police only. Police also do not always make official records of crime incidents reported by victims or witnesses. It is noted that due to its visibility, gun violence is more likely to be reported to and recorded by the police. Third, this study analyzed gun violence at the city level. However, it overlooked the heterogeneity of policy impacts across different communities in each city. The poor in inner-city neighborhoods have been hit hard by the pandemic. Due to the high density of population and poverty, they were disproportionately infected with and died from COVID-19 (Adhikari et al., 2020; Berkovitz, 2020). According to Falk et al. (2020), the pandemic also had significantly decreased job availability in service sectors for leisure and hospitality, which exacerbated the economic difficulty of undereducated and underemployed individuals in inner-city neighborhoods. Future work should examine whether the pandemic had varying effects on gun violence across small geographic units within a given city. Only a few studies explored the impact of containment measures on change in various types of crime across communities or districts (Campedelli, Favarin, et al., 2020; Kim, 2022b; Sun et al., 2021).

Despite the importance of understanding the association between the pandemic and gun violence for both theory and policy, only a handful of studies have empirically examined this potential relationship. It is plausible that the impacts of the pandemic differ by types of gun violence, such as domestic violence and suicide. Whenever the data are allowed, it is very important to use disaggregated data for a nuanced understanding of the pandemic-gun violence association. For example, National Incident-Based Reporting System (NIBRS) data can be used to examine the impact of the pandemic on crime. The NIBRS data is incident-driven and has detailed information about the situations and contexts surrounding criminal incidents, such as location, the weapon used, personal characteristics of the offender and victim, and offender–victim relationships. It will allow researchers to examine how personal and situational characteristics influence gun violence during the pandemic. Using disaggregated data will help explain a complex picture of the pandemic’s impact patterns across different types of gun violence, warranting more work with disaggregated data.

Policy Implications

This study has several policy implications to reduce gun violence. There were significant increases in gun shootings in the context of high unemployment, poverty, and economic inequality during the pandemic. Many individuals were deprived of legitimate means for acquiring jobs, housing, and other necessities, which might facilitate their involvement in gun violence from financial strain. It is imperative to implement economic stimulus policies for disadvantaged individuals who experience financial distress and to buffer them from the harshest effects of the pandemic. In addition, the pandemic and social distancing have increased people’s social isolation and mental disorders, such as stress, anxiety, depression, and suicidal thoughts (Clair et al., 2021). Given the impulsive nature of violent acts, it is imperative to identify high-risk individuals for gun violence and provide immediate mental health interventions at the community level (Stone et al., 2017). In this study, gun violence tends to increase during the relaxation of social distancing policies when individuals have more interactions with others. The relaxation of social distancing might allow for more social gatherings and interactions, resulting in increased intentional or accidental shootings. Thus, governments should plan optimal levels of staffing and resources for law enforcement and mental health before they relax social distancing restrictions.

During the pandemic, many people bought new guns and ammunition (Federal Bureau of Investigation, 2021; Kraviz-Wirtz et al., 2020), and/or consumed more alcoholic beverages (Pollard et al., 2020), to deal with feelings of uncertainty and insecurity about the public health and economy. It is reasonable to assume that the confluence of high strain, alcohol consumption, and gun availability might have resulted in more gun violence incidents during the pandemic. This study brings us to an unsettled and grave policy question of how to restrict the possession and use of guns. Although this issue is still debatable and beyond the scope of this study, it is important to briefly discuss policy recommendations for safe gun storage. During the pandemic, many gun owners tended to keep loaded and/or unlocked guns accessible, within easy reach (Kraviz-Wirtz et al., 2020). People should be warned that loosened gun storage methods can increase the risk of intentional and accidental shootings, and it is important to store the guns unloaded and locked up in a secure place.

Footnotes

Appendix

Table A1.

ARMA Models for Nonfatal Gun Violence associated With Assault and Robbery, Washington DC.

Variable/model	Assault		Robbery
Variable/model	Original	Box-Cox ^{λ = .505}	Original	Box-Cox ^{λ = 424}
Constant	11.53 (.54)**	4.74 (.15)**	9.58 (1.14)**	3.83 (.23)**
SAH	.78 (1.43)	.21 (.40)	−1.04 (1.91)	−.34 (.38)
Relaxed SAH	3.46 (.75)**	.91 (.21)**	5.99 (1.13)**	1.03 (.22)**
BLM	1.32 (2.00)	.32 (.55)	−.24 (2.68)	.05 (.53)
CAPITOL	3.01 (4.15)	.89 (1.15)	−3.97 (5.54)	−.48 (1.09)
Q2	2.326 (.82)**	.69 (.23)**	.67 (1.09)	.19 (.21)
Q3	2.52 (.77)**	.72 (.21)**	2.04 (1.03)*	.50 (.20)*
Q4	1.39 (.77)^†	.39 (.21)^†	4.90 (1.09)**	1.00 (.22)**
δY_t-1	–	–	.24 (.07)**	.05 (.01)**
Adj. R²	.14	.13	.38	.36
SIC	5.831	3.23	6.40	3.15
Ljung-Box Q₃₆	43.01	40.60	30.09	34.69
Jarque-Bera	5.67^†	1.12	85.83**	5.20^†
ARCH Obs × R^2	.64	.44	.78	3.06^†

Note. ARMA = Autoregressive Moving Average; SAH = stay at home; BLM = Black Lives Matter; SIC = Schwarz information criterion.

†

Significant at α = .10. * Significant at α = .05. ** Significant at α = .01.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Dae-Young Kim

Author Biography

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, Homicide Studies, Journal of Criminal Justice, Journal of Crime and Justice, Journal of Research in Crime and Delinquency, and The Prison Journal.

References

Abrams

D. S.

(2020). COVID and crime: An early empirical look. Faculty Scholarship at Penn Law. https://scholarship.law.upenn.edu/faculty_scholarship/2204

Adhikari

Pantaleo

N. P.

Feldman

J. M.

Ogedegbe

Thorpe

Troxel

A. B.

(2020). Assessment of community-level disparities in coronavirus disease 2019 (COVID-19) infections and deaths in large US metropolitan areas. JAMA Network Open, 3(7), 1–4.

Agnew

(2006). Pressured into crime: An overview of general strain theory. Roxbury Publishing.

Andresen

M. A.

Hodgkinson

(2020). Somehow I always end up alone: COVID-19, social isolation and crime in Queensland, Australia. Crime Science, 9(1), 1–20.

Ashby

(2020). Initial evidence on the relationship between the coronavirus pandemic and crime in the United States. Crime Science, 9(6), 1–16. https://crimesciencejournal.biomedcentral.com/track/pdf/10.1186/s40163-020-00117-6.pdf

Balmori de la Miyar

J. R.

Hoehn-Velasco

Silverio-Murillo

(2020). Druglords don’t stay at home: Covid-19 pandemic and crime patterns in Mexico City. Journal of Criminal Justice, 72, 101745. https://doi.org/10.1016/j.jcrimjus.2020.101745

Berkovitz

(2020, May 19). Environmental racism has left black communities especially vulnerable to COVID-19. The Century Foundation. https://tcf.org/content/commentary/environmental-racism-left-black-communities-especially-vulnerable-covid-19/?agreed=1&agreed=1

Boine

Siegel

Ross

Fleegler

E. W.

Alcorn

(2020). What is gun culture? Cultural variations and trends across the United States. Humanities & Social Sciences Communications, 7, Article 21. https://doi.org/10.1057/s41599-020-0520-6

Box

G. E.

Cox

D. R.

(1964). An analysis of transformation. Journal of the Royal Statistical Society. Series B (Methodological), 26(2), 211–252.

10.

Brantingham

P. J.

Tita

G. E.

Mohler

(2021). Gang-related crime in Los Angeles remained stable following COVID-19 social distancing orders. Criminology & Public Policy, 20, 423–436. https://doi.org/10.1111/1745-9133.12541

11.

Campedelli

G. M.

Aziani

Favarin

(2020). Exploring the immediate effects of COVID-19 containment policies on crime: An empirical analysis of the short-term aftermath in Los Angeles. American Journal of Criminal Justice, 46, 704–727. https://doi.org/10.1007/s12103-020-09578-6

12.

Campedelli

G. M.

Favarin

Aziani

Piquero

A. R.

(2020). Disentangling community-level changes in crime trends during the COVID-19 pandemic in Chicago. Crime Science, 9(21), 1–18.

13.

Cassell

(2020). Explaining the recent homicide spikes in U.S. cities: The “Minneapolis Effect” and the decline in proactive policing. Federal Sentencing Reporter, 33(1–2), 83–127. https://doi.org/10.1525/fsr.2020.33.1-2.83

14.

Centers for Disease Control and Prevention. (2020). United States COVID-19 cases and deaths by state. https://covid.cdc.gov/covid-data-tracker/#cases_totalcases

15.

Clair

Gordon

Kroon

Reilly

(2021). The effects of social isolation on well-being and life satisfaction during pandemic. Humanities & Social Sciences Communications, 8(28), 1–6. https://doi.org/10.1057/s41599-021-00710-3

16.

Clarke

R. V. G.

(1980). Situational crime prevention: Theory and practice. The British Journal of Criminology, 20(2), 136–147.

17.

Cohen

L. E.

Felson

(1979). Social change and crime rate trends: A routine activity approach. American Sociological Review, 44, 588–608.

18.

Donaghue

(2020). Coronavirus and gun violence: Mayors fight a double public health crisis. https://www.cbsnews.com/news/coronavirus-gun-violence-double-public-health-crisis/

19.

Everytown Research & Policy. (2021). Gun violence and COVID-19 in 2020: A year of colliding crises. https://everytownresearch.org/report/gun-violence-and-covid-19-colliding-public-health-crises/

20.

Falk

Carter

J. A.

Nicchitta

I. A.

Nyhof

E. C.

Romero

P. D.

(2020). Unemployment rates during the COVID-19 pandemic: In brief. Congressional Research Service.

21.

Federal Bureau of Investigation. (2018). Crime in the United States, 2018. Washington, DC: U.S. Department of Justice.

22.

Federal Bureau of Investigation. (2021). NICS firearm checks: Month/year. https://www.fbi.gov/file-repository/nics_firearm_checks_-_month_year.pdf/view

23.

Greenfeld

L. A.

(1998). Alcohol and crime: An analysis of national data on the prevalence of alcohol involvement in crime. Bureau of Statistics, U.S. Department of Justice.

24.

Kim

D.-Y.

(2022a). COVID-19 and gun violence: Keeping unknown shocks and volatility in perspective. Criminal Justice Review. Advance online publication. https://doi.org/10.1177/07340168221088571

25.

Kim

D.-Y.

(2022b). The impact of COVID-19 on gun violence across census tracts in NYC. Homicide Studies. Advance online publication. https://doi.org/10.1177/10887679221077036

26.

Kim

D.-Y.

McCarty

W. P.

(2021). Exploring violent crimes in Chicago during the COVID-19 pandemic: Do location, crime type, and social distancing type matter? Journal of Crime and Justice. Advance online publication. https://doi.org/10.1080/0735648X.2021.2014934

27.

Kim

D.-Y.

Phillips

S. W.

(2021). When COVID-19 and guns meet: A rise in shootings. Journal of Criminal Justice. https://doi.org/10.1016/j.jcrimjus.2021.101783

28.

Kraviz-Wirtz

Aubel

Schleimer

Pallin

Wintemute

(2020). Violence, firearms, and the coronavirus pandemic: Findings from the 2020 California Safety and Wellbeing Survey. medRxiv. https://doi.org/10.1101/2020.10.03.20206367

29.

McDowall

Curtis

(2015). Seasonal variation in homicide and assault across large U.S. cities. Homicide Studies, 19, 303–325.

30.

McDowall

Loftin

Pate

(2012). Seasonal cycles in crime, and their variability. Journal of Quantitative Criminology, 28(3), 389–410. https://doi.org/10.1007/s10940-011-9145-7

31.

McDowall

McCleary

Bartos

B. J.

(2019). Interrupted time series analysis. Oxford University Press.

32.

McDowall

McCleary

Meidinger

E. E.

Hay

(1980). Interrupted time series analysis. SAGE.

33.

Merton

(1938). Social structure and anomie. American Sociological Review, 3, 672–682.

34.

Messner

S. F.

Rosenfeld

(2006). Crime and the American dream. Wadsworth.

35.

Mullins

(2020). Gun ownership seems to be rising in the DC area: At a time of unease, some locals are buying firearms. https://www.washingtonian.com/2020/11/06/gun-ownership-seems-to-be-rising-in-the-dc-area/

36.

The Nielsen Company. (2020). Rebalancing the “COVID-19 effect” on alcohol sales. https://nielseniq.com/global/en/insights/analysis/2020/rebalancing-the-covid-19-effect-on-alcohol-sales/

37.

Nix

Wolfe

S. E.

Campbell

B. A.

(2018). Command-level police officers’ perceptions of the “war on cops” and de-policing. Justice Quarterly, 35(1), 33–54.

38.

Payne

J. L.

Morgan

Piquero

A. P.

(2020). COVID-19 and social distancing measures in Queensland, Australia, are associated with short-term decreases in recorded violent crime. Journal of Experimental Criminology, 18, 89–113. https://doi.org/10.1007/s11292-020-09441-y

39.

Pennsylvania State Police. (2020). Firearms annual report 2020. https://www.psp.pa.gov/firearms-information/Firearms%20Annual%20report/Pennsylvania_State_Police_2020_Firearms_Annual_Report.pdf

40.

Pew Research Center. (2020). In their own words, Americans describe the struggles and silver linings of the COVID-19 pandemic. https://www.pewresearch.org/2021/03/05/in-their-own-words-americans-describe-the-struggles-and-silver-linings-of-the-covid-19-pandemic/

41.

Phillips

S. W.

(2020). Exploring law enforcement’s views of de-policing: A vignette research design. Criminal Justice Studies, 33(2), 178–191.

42.

Piquero

A. R.

Riddell

J. R.

Bishopp

S. A.

Narvey

Reid

J. A.

Piquero

N. L.

(2020). Staying home, staying safe? A short-term analysis of COVID19 on Dallas domestic violence. American Journal of Criminal Justice, 45, 601–635.

43.

Pollard

M. S.

Tucker

J. S.

Green

H. D.

(2020). Changes in adult alcohol use and consequences during the COVID-19 pandemic in the US. JAMA Network Open, 3(9), Article e2022942. https://doi.org/10.1001/jamanetworkopen.2020.22942

44.

Rosenfeld

Lopez

(2020). Pandemic, social unrest, and crime in U.S. cities: November 2020 update. Council on Criminal Justice.

45.

Rushin

Edwards

(2017). De-policing. Cornell Law Review, 102(3), 721–782.

46.

Siegel

Ross

C. S.

King

(2013). The relationship between gun ownership and firearm homicide rates in the United States, 1981–2010. American Journal of Public Health, 103(11), 2098–2105.

47.

Stone

D. M.

Holland

K. M.

Bartholow

Crosby

A. E.

Davis

Wilkins

(2017). Preventing suicide: A technical package of policies, programs, and practices. National Center for Injury Prevention and Control, Centers for Disease Control and Prevention.

48.

Sun

Huang

Yuan

Chan

T. O.

Wang

(2021). Spatial patterns of COVID-19 incidence in relation to crime rate across London. International Journal of Geo-Information, 10(53), 1–13.

49.

U.S. Bureau of Labor Statistics. (2021). Economy at a glance. https://www.bls.gov/eag/

50.

U.S. Census Bureau. (2021). QuickFacts. https://www.census.gov/programs-surveys/sis/resources/data-tools/quickfacts.html

51.

World Health Organization. (2020). Impact of COVID-19 on people’s livelihoods, their health and our food systems. https://www.who.int/news/item/13-10-2020-impact-of-covid-19-on-people%27s-livelihoods-their-health-and-our-food-systems

52.

Yamane

(2017). The sociology of US gun culture. Sociology Compass, 11(7), Article e12497. https://doi.org/10.1111/soc4.12497

53.

District of Columbia v. Heller, 554 U.S. 570 (2008).