Crime Science(2021) 10:10Stacy et al. Crime Scihttps://doi.org/10.1186/s40163-021-00146-9Open AccessRESEARCHThe impact of gunshots on place‑levelbusiness activityChristina Stacy1, Yasemin Irvin‑Erickson2* and Emily Tiry3AbstractObjectives: Gun violence can negatively affect business activity at the place-level through a variety of mechanisms.However, estimating this effect is difficult since reported crime data are biased by factors that are also associated withbusiness health. Despite some of its limitations, data from gunshot detection technology has been shown as a newvaluable source of data on gun violence (Irvin-Erickson et al. in Appl Geogr 86: 262–273, 2017a). In this study, we usegunshot detection data to explore the spatial relationship between gunshots and business activity at the neighbor‑hood level in Washington, DC between 2010 and 2012.Methods: In this exploratory study, we create spatial buffers of 500 and 1000 feet around each block and sum up thetotal number of gunshots and business births, deaths, sales, and number of employees within these buffers each yearand estimate a spatial fixed effects panel model.Results: Gunshots within 1000 feet of a block increase the number of business deaths by 4.3% within that buffer onaverage, and gunshots within 500 feet of a block decrease the total number of service and retail businesses, the num‑ber of employees employed by businesses within that buffer, and total sales for those businesses (although not at astatistically significant rate). Gunshots on blocks with the lowest initial levels of gunshots increase business turnoverand reduce the total number of businesses present by 0.5%, and gunshots on blocks with the highest initial levels ofgunshots cause an increase in the number of business deaths by 7.5%.Conclusion: Results suggest that efforts to improve distressed neighborhoods should target both areas with lowerand higher pre-existing levels of gunshots.Keywords: Gunshots, Business, ShotSpotter, Spatial econometricsIntroductionCrime, particularly violent crime, has been shown toimpose a variety of economic costs on individuals, communities, and society as a whole. These costs includeincreased health care costs (Howell et al., 2014; Milleret al., 1993), costs associated with lost productivity (Cooket al., 1999), costs associated with police, courts, and correctional institutions (Cook & Ludwig, 2000; Shapiro& Hassett, 2012), reduced property values (Hipp et al.,*Correspondence: [email protected] of Criminology, Law and Society, George Mason University,4400 University Dr, Fairfax, VA 22030, USAFull list of author information is available at the end of the article2009; Irvin-Erickson, Lynch, et al., 2017; Kirk & Laub,2010; Lynch & Rasmussen, 2001; Shapiro & Hassett,2012; Tita et al., 2006), lost time from work (Cook & Ludwig, 2000; Perkins et al., 1996), and costs associated withvictims’ efforts to avoid revictimization such as relocation of victims (Dugan, 1999).Noticeably absent from the literature on the economicimpacts of crime is the impact of gun violence on localbusinesses. This is partially due to the lack of availablemicro-geographical level data sources on business activity until very recently. The scarcity of research on thistopic is also in line with the omission of businesses fromstudies on the impact of crime on neighborhoods, despitethe importance of local business activity as an indicator The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) andthe source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party materialin this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If materialis not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds thepermitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Stacy et al. Crime Sci(2021) 10:10of the local economy and of the quality of life for residents, non-residents, and investors (Fisher, 1991; Skogan,1986). This omission is surprising considering that thelack of legitimate local jobs for youth, and especiallyminority youth, has been shown to increase the likelihood that these youth engage in criminal activity (Ihlanfeldt, 2002). Furthermore, economic development effortswithin Business Improvement Districts have been shownto be related to reductions in community-level incidencesof interpersonal violence, which is largely experienced byyouth and young adults (MacDonald et al., 2009).Business revenues can be affected by gun violencethrough a variety of mechanisms. People have beenshown to be afraid of places where they know violentcrimes happen or where they perceive that they have ahigh likelihood of victimization from violent crime (seeFisher, 1991 and Skogan, 2012 for a detailed discussion)and this can affect business revenues. Although theresearch on the impacts of fear of crime on individuals’behavior and routine activities is somewhat inconsistent, it nonetheless suggests that some individuals mayalter their routine activities and constrain their outdooractivity in response to increased perceived risk of crime(see Foster et al., 2014; Foster & Giles-Corti, 2008; IrvinErickson, Lynch, et al., 2017b; Liska et al., 1988; Lorencet al., 2012; Markowitz et al., 2001; Mesch, 2000; Oh &Kim, 2009; Otis, 2007; Ross, 1993; Skogan & Maxfield,1981; Stafford et al., 2007). However, it is importantto note that individuals’ perceptions of social disordercan moderate the relationship between perceived riskand routine activities (Rengifo & Bolton, 2012). Rengifoand Bolton (2012) found that individuals who perceivea higher level risk of victimization and a lower level ofdisorder engage in significantly more voluntary activities in comparison to individuals who perceive a higherlevel risk of victimization and a higher level of disorder. Wesley Skogan (1986; 2012) also provides contextinto the relationship between social disorder on fear ofcrime. According to Skogan (1986; 2012), disorder canindependently, but in parallel with crimes in communities, increase fear of crime and discourage investmentsin neighborhoods. On the topic of business patronage,Skogan (1986; 2012) and other authors (Bowes, 2007;Fisher, 1991) suggest that crime and fear of crime canreduce business revenues due to residents in high-crimeneighborhoods limiting their activities and not patronizing businesses. Unsurprisingly, when business revenuesare reduced, there are fewer jobs at businesses for localresidents (Hamermesh, 1999; Levi, 2001).Business owners can also change business operationsand business decisions in response to crime (Bowes,2007; Fisher, 1991; Hamermesh, 1989, 1999; Levi, 2001;Skogan, 1986; 2012). Businesses have been shown to bePage 2 of 9negatively impacted by reduced business hours, difficultyhiring employees or having employees work at undesirable evening and night business hours, and increasedinsurance costs due to crime. For instance, a study byFisher (1991), based on interviews with business ownersin the Hilltop Community in Columbus, Ohio, demonstrated business owners’ difficulty in hiring or retainingemployees who are worried about working in an environment where they perceive that they are likely to be victimized. The same study showed further harmful impactsof crime on business operations such as reduced businesshours and increased business insurance costs (Fisher,1991). Another study by Hamermesh (1989), linking Current Population Survey (CPS) data to FBI crime reports,studied time use as a nonmonetary cost of time andfound that higher homicide rates in large metropolitanareas are related to a lower propensity of workers to workevenings and nights. Other studies further showcase thatcrime and the fear of crime can be related to decreasedbusiness investment (such as the opening of new businesses or the expansion of existing businesses) in areaswith a reputation as high-crime areas (Bowes, 2007;Fisher, 1991; Skogan, 1986; 2012).The aforementioned studies on the economic impactsof crime, along with the wider literature on the impact ofcrime and fear of crime on routine activities, suggest thatlocal businesses may have difficulty attracting customers, attracting and retaining employees, or maintainingregular hours in response to heightened gun violence.To the best of our knowledge, only three studies haveestimated the impact of violent crime on local businesses using business data. Rosenthal and Ross (2010)estimated the impact of violent crime on the location ofbusinesses in Atlanta, Chicago, Houston, Indianapolis,and Seattle at the Census tract level via a cross-sectionalstudy. The authors used two datasets for their analysis:reported crime data from local police agencies and business activity data from Dunn and Bradstreet, a for-profitfirm. According to this study, an increase in violent crimeduring prime dinner hours (5 pm to 9 pm) reduced thepresence of high-end restaurants by roughly 40 percentage points when considering the spread of minimum andmaximum number of violent crimes in Census tractsobserved in the study period. In this study, restaurantsare defined as high-end “if they have 1–24 employeesand sales are greater than 0.5 million, 25–49 employeesand sales are greater than 1.0 million, or 50–99 employees and sales are greater than 2.5 million” (Rosenthal &Ross, 2010, p. 142).In addition to Rosenthal and Ross (2010) cross-sectional study, only two longitudinal studies have beenconducted on the effects of violent crime on local business. Greenbaum and Tita (2004) used longitudinal
Stacy et al. Crime Sci(2021) 10:10business and homicide data at the ZIP code level toexplore the impact of homicide surges on the creation,closing, and growth of businesses in Chicago, Houston,Miami, Pittsburgh, and St. Louis between 1987 and1994. The authors found that local increases in lethalviolence caused existing businesses to downsize andled to fewer new businesses forming. These effects wereconcentrated in ZIP codes where homicides were lessfrequent, suggesting that surges in violence in neighborhoods that already have high levels of violence maynot increase the perceived risk of violence to the pointof affecting business activity. The study also found thatestablished businesses were less affected, as surges inviolence had no significant impact on prompting business closures. Finally, the impact of homicide crimewas greatest among personal service and retail businesses, indicating that jobs relying on face-to-faceinteraction between employees and customers maybe most susceptible to the effects of increased violentcrime (Greenbaum & Tita, 2004).Irvin-Erickson, Lynch, et al. (2017) estimated theimpact of a sudden increase in gun homicides and gunshots on local business growth, home values, homeownership rates, and credit scores in five US cities at theCensus tract level. The authors found that gun homicidesurges in Census tracts reduced the growth rate of newretail and service establishments by 4% in Minneapolis,Oakland, San Francisco, and Washington, DC. The samestudy also found that gun homicide surges in Censustracts slowed home value appreciation by 3.9% in BatonRouge, Minneapolis, Oakland, San Francisco, and Washington, DC. Similarly, the authors found that gunshotsurges in Census tracts slowed home value appreciationby 3.6% in Oakland, Rochester, San Francisco, and Washington, DC.We expand upon these previous studies by estimatingthe relationship between detected gunshots and business births, deaths, sales, and number of employees atthe Census block level using data from the NationalEstablishment Time Series Database and ShotSpotter, agunshot detection system. While we acknowledge thatgunshot detection technology has its own limitations indetecting gunshots at certain times of the day and theyear and at a farther distance from acoustic sensors, datafrom gunshot detection technology has been shown as avaluable new data source on gun violence in the recentliterature (Irvin-Erickson, La Vigne, et al., 2017a). Inour study, the availability of data directly from gunshotdetection technology allows us to measure the impact ofactual gunshots on businesses, rather than the impact ofreported gunshots on businesses, in which overreportingor underreporting are likely endogenous to neighborhood characteristics and business activity.Page 3 of 9MethodologyData sources and measuresThe data for this study were collected from two mainsources. The first is a gunshot detection system thatuses a network of acoustic sensors to identify the uniqueaudio signature of a gunshot pinpoint the location ofa gunshot (Bieler & La Vigne, 2014). The second set ofdata comes from the National Establishment Time SeriesDatabase, which provides point-level data on establishments including industry, location, sales, and numberof employees. These data allow us to accurately measuregunshots without concern about over or underreportingin certain neighborhoods, and to link it at the point levelto business location and behavior. Block-level descriptivestatistics can be found in Table 1.Gunshot dataThe gunshot data used in our analysis are based on information collected by gunshot detection technology (GDT)in Washington, DC. GDT uses a network of acousticsensors to identify the sound of a gunshot and triangulate its position. The time and location of the gunshotare then recorded and sent to law enforcement personnel (Eng, 2004; Showen, 1997; Siuru, 2007). The newestversions of this technology have been found to accuratelyrecord gunshots under most conditions (Goode, 2012).Data used in this study come from ShotSpotter, a GDTvendor for Washington, DC. The GDT sensors covered17.3 square miles of the city (about 25% of the total areaof Washington, DC) at the time of the study (see Fig. 1).The data were made publicly available online by the Metropolitan Police Department in response to a Freedom ofInformation Act request. We remove incidents detectedon January 1 and July 4 to minimize false positive detections from firework detonations, and we aggregate thenumber of gunshots detected within 500 and 1000 feetfrom the block edges (see Fig. 2).Establishment dataThe business location, revenue, and employment datain our analysis come from the National EstablishmentTime Series Database, which is an annual snapshot ofDun and Bradstreet’s data on establishments, including industry, location, sales, and number of employees,among other indicators (Walls & Associates, 2012). Tolimit our study to retail and service establishments withwhich customers are likely to interact at the establishment’s physical location, we use establishments withinthe retail trade; accommodation and food services;personal and household goods repair and maintenanceservices; and personal services industries, based on theestablishment’s North American Industry ClassificationSystem (NAICS) code (we used establishments whose
Stacy et al. Crime Sci(2021) 10:10Page 4 of 9Table 1 Block level descriptive statisticsVariableObsMeanStd. DevMinMaxTotal establishments8370.871.57012Total births8370.190.4804Total deaths8370.010.1201Total employees8374.2015.090228Total sales837 320,708 1,413,300 0 20,300,000ShotSpotter Incidents 500 ft83719.6617.450122ShotSpotter Incidents 500–1000 ft83730.4420.151168Total establishments8370.871.57013Total births8370.010.1101Total deaths8370.090.3202Total employees8374.1714.870228Total sales837 302,056 1,334,378 0 19,600,000ShotSpotter Incidents 500 ft83722.6721.950142ShotSpotter Incidents 500–1000 ft83736.4929.462209Total establishments8370.931.70014Total births8370.160.4704Total deaths*n/aTotal employees8374.8817.750233Total sales837 340,812 1,501,971 0 22,200,000ShotSpotter incidents 500 ft83714.6313.64090ShotSpotter incidents 500–1000 ft83724.1018.450127201020112012*Total deaths not calculable for 2012 since we lack data from 2013NAICS codes began with 44, 45, 72, 811, or 812). In2012, approximately 25% of these types of establishments in Washington, DC fell within the GDT coveragearea. Additionally, we limit the establishments to thosethat existed at least 1,000 feet within the GDT coverage area to ensure consistency in measuring gunshotsacross establishments. We geocode each establishmentin these data and calculate the total sales, employees,establishment births, and establishment deaths for2010 through 2012 at both the establishment and blocklevel. Births and deaths were defined based on establishments entering or leaving Census blocks in a givenyear, rather than their existence in the dataset. A birthestablishment is an establishment that did not exist atits same location in the year prior to the year of consideration. A death establishment, on the other hand, isan establishment that did not exist at the same locationwhere it had existed a year prior. Establishments thatremained in existence but moved to a different Censusblock are therefore counted as birth and death establishments. We convert sales for each year into 2010 dollars using Consumer Price Index conversion factors.ModelingTo estimate the relationship between gunshots and business activity, we estimate both a cross sectional Poissonand fixed effects Poisson panel model as follows:Cross sectional OLS:Yi δ0 β1 gunshotsi β2 gunshots500to1000fti β2 Xi uiFixed effects panel:Yit δ0 β1 gunshots500ftit β2 gunshots500to1000ftit β3 Xit γi t uitwhere each equation is estimated at the block level, andYit is a vector of outcome measures, including total salesand total number of employees for the establishmentson block i in year t, gunshots500ftit is the total numberof gunshots within 500 feet of block i or on block i inyear t, gunshots500to100ftit is the total number of gunshots between 500 and 1,000 feet of block i (or businessi) in year t, Xit is a vector of control variables, and γi andt are block and year fixed effects, respectively. We first
Stacy et al. Crime Sci(2021) 10:10Page 5 of 9and where xi β xi β is specified the same as the abovemodels.By including fixed effects in the panel models, weremove any unobserved characteristics that are timeinvariant, like the geography of the area, the type of business that we are examining (for those that do not change),and any other unobserved characteristics of the businessor neighborhood that do not change over time.Fig. 1 Shotspotter coverage area in Washington, DCResultsResults indicate that across blocks, those with a highernumber of gunshots also have a greater number of businesses.1 However, over time as gunshots increase, thenumber of service and retail businesses decline (althoughthis effect is not statistically significant for the sample asa whole).Table 2 displays these results. Cross sectionally, a blockwith 1 unit higher level of gunshots within 500 feet isassociated with a 0.6% greater number of establishments(Column 1). However, when examined over time (andwithin block), these results switch sign, indicating thatthe effects seen in the cross-sectional model are due toendogeneity between gunshots and establishments ratherthan a causal effect: it is not the gunshots that are causingthe a greater number of establishments, but rather thegeneral correlation between commercial areas and gunshots that is causing this relationship. In fact, gunshotswithin 500 feet of a block decrease the total number ofservice and retail businesses, the number of employeesemployed by businesses within that buffer, and total salesfor those businesses, although not at a statistically significant rate. The 95% confidence interval in the fixed effectsmodels is 0.3% to 0.1% for establishments, 2.1% to1.1% for births, 2.8% to 2.2% for deaths, 2,297 to 555 for sales, and 0.022 to 0.006 for employees.Spatial displacement and spilloverFig. 2 Example block and buffer, 500 and 1000 feetmeasure the model without the spatial lag, and then addit in to examine spillover and displacement effects. Wealso examine the effect of gunshots on the total numberof employees using the above model, without the naturallog.We also examine whether gunshots have any impact onthe number of establishments, births, and deaths on eachblock using a cross sectional and fixed effects Poissonmodel, where yi given xi takes on a Poisson distributionWe might be concerned that our results are attenuated because gunshots are reducing business sales in theimmediate area surrounding the store and displacing itinto nearby areas. To test for this, we run a spatial lagmodel where we include gunshots within 500 feet of theblock as well as gunshots between 500 and 1000 feet ofthe block.Results show that gunshots within both ranges areassociated with reductions in the number of establishments, although not statistically significantly (Table 3,Column 1). Gunshots between 500 and 1000 feet ofthe block are, however, significantly associated with anincrease in business deaths: for every additional gunshot1Results are robust to a number of alternative specifications, includingremoving blocks with no establishments.
Stacy et al. Crime Sci(2021) 10:10Page 6 of 9Table 2 Effect of gunshots on the number of service and retail establishments, births, deaths, and employeesNumber of ShotSpotterincidents within 500 irthsDeathsTotal SalesEmployeesPoissonPoisson FEPoisson FEPoisson FEFEFE0.006***(0.002) 0.001 0.005(0.001) 0.003(0.008) 870.90(0.013)(727.40) 0.008(0.007)Block fixed effectsNoYesYesYesYesYesYear fixed 25111119Number of blocks83737320879837373Robust standard errors in parentheses. Fixed effects estimates are clustered at the block level***p 0.01, **p 0.05, *p 0.1Table 3 Displacement and spillover effects of gunshots on business l SalesEmployeesPoisson FEPoisson FEPoisson FEFEFENumber of ShotSpotter incidentswithin 500 ft(0.001) 0.001Number of ShotSpotter incidentsbetween 500 and 1,000 ft(0.0005) 0.0001 0.006 0.013(0.008)(0.014) 0.002(0.007) 912.50(817.00)0.043**105.1(0.018)(466.8) 0.007(0.007) 0.002(0.006)Block fixed effectsYesYesYesYesYesYear fixed 11Number of establishments37320879837837Robust standard errors in parentheses. Fixed effects estimates are clustered at the block level***p 0.01, **p 0.05, *p 0.1between 500 and 1000 feet of a block, that block sees anincrease of 4.3% of business deaths. This may be relatedto the fact that 1000 feet is the greatest distance withinwhich a gunshot can be heard indoors (Bieler & La Vigne,2014), and that the power of the regression is strongerwhen larger areas are included in the gunshot count.However, these results hold whether we create the buffersas 500 and 1000 feet, or 1000 and 2000 feet, suggestingthat the impacts on neighborhoods may be larger thanthe audible distance of a gunshot.Heterogeneous effects by initial level of gunshotsEven small increases in violence in areas with lower initiallevels of gunshots may have a large impact on businessoutcomes. Therefore, gunshots may have heterogeneouseffects on business revenues based on the initial level ofgunshots in the area of the business; an additional gunshot on a block with a high starting level of gunshotsmight not have as much of an effect as an additional gunshot on a block that usually has no gunshots at all.To examine this hypothesis, we estimate the relationship between gunshots and commercial revenues usingsubsets of the data based on initial gunshot levels in 2010.Table 4 below show the descriptive statistics for thesequartiles for both the block and establishment level.These results show that impacts are, in fact, concentrated in areas with lower initial levels of gunshots. Inthe first quartile of blocks (i.e., blocks with the lowestnumber of initial gunshots), one additional gunshot isassociated with a reduction of 0.5% of establishments.Gunshots on these blocks also reduce the number ofbusiness births: each additional gunshot is associatedwith a reduction in business births of 9.4%. Surprisingly,deaths also go down on these blocks by 13.5% for eachgunshot, but the overall effect on total number of establishments is still negative.
Stacy et al. Crime Sci(2021) 10:10Page 7 of 9Table 4 Descriptive statistics for quartiles of initial levels of shotspotter incidents within 500 ftVariableQuartileObsMeanStd. DevMinMaxShotSpotter incidents 500 ft11553.711.7506ShotSpotter incidents 500 ft221710.002.01713ShotSpotter incidents 500 ft323017.462.791423ShotSpotter incidents 500 ft423541.2518.6324122Table 5 Effect of gunshots on service and retail establishments by beginning level of gunshotsQuartileNumber of ShotSpotter incidents within 500 EmployeesPoisson FEFEPoisson FEFEFE 0.005* 0.094*(0.003)Number of ShotSpotter incidents within 500 ft2(0.052)0.004 0.001(0.003)Number of ShotSpotter incidents within 500 ftNumber of ShotSpotter incidents within 500 ft34(0.034) 0.135**(0.066) 0010.0010.0050.001(0.001)(0.031)(0.026)(0.001) 0.001 0.005(0.001)(0.009) 0.213(0.179) )Robust standard errors in parentheses. Fixed effects estimates are clustered at the block levelNumber of observations and R2 vary by subgroup, so suppressed for brevity***p 0.01, **p 0.05, *p 0.1We also see, however, a statistically significant increasein business deaths on blocks that begin with the highest initial levels of gunshots (Table 5). For each gunshoton these blocks, business deaths increase by 7.5%. Thisimplies that even areas with higher initial levels of gunshots are still affected by subsequent gunshots at thebusiness level.Discussion and conclusionsThe dearth of studies exploring the relationship betweengun violence and business activity represents a critical gap, since the level of economic activity in a neighborhood is an important indicator of the ability of itsresidents to build wealth. Furthermore, as discussed atthe beginning of our paper, lack of local employmentopportunities have important consequences, especiallyfor minority youth (Ihlanfeldt, 2002; MacDonald et al.,2009). Absent an understanding of the specific impacts ofgun violence on business activity and the causal mechanisms underlying these impacts, it is difficult to makeinformed policy recommendations to boost businessactivity in areas beset by gun violence.Our research addresses this gap by leveraging the precision of NETS and GDT data to explore how changesin gunshots affect business revenue in areas withboth higher and lower pre-existing levels of gunshots.GDT data allow us to measure actual gunshots ratherthan reported gunshots, avoiding a fundamental challenge in crime literature relating to the endogeneityof crime reporting with outcomes of interest. Specifically, reported crime data may be endogenous to thecharacteristics of neighborhoods since neighborhoodswith lower income, younger, and male victims are morelikely to have underreporting of crimes while neighborhoods with a large number of homeowners are lesslikely to have underreporting (Skogan, 1999). Additionally, crime reporting may be higher in neighborhoodswith more eyes on the street—or people to report thosecrimes.Results show that, cross sectionally, blocks with ahigher number of gunshots also have a greater numberof businesses. However, over time as gunshots increase,the total number of service and retail businesses decline(although not at a statistically significant rate). Additionally, gunshots on nearby blocks increase the number of business deaths by 4.3%, and gunshots on blockswith the lowest initial levels of gunshots (i.e., blocks inthe first quartile) increase business turnover and reducethe total number of businesses present by 0.5%. Perhapssurprisingly, even on blocks with the highest initiallevels of gunshots (i.e., blocks in the fourth quartile),increased gunshots cause an increase in the number
Stacy et al. Crime Sci(2021) 10:10of business deaths by 7.5%. These results suggest thatgunshots affect businesses not only in areas with lowerinitial levels of gunshots, but also in areas with higherinitial levels of gunshots.These results add valuable insights into the relationship between gun violence and business activity andreinforce Rosenthal and Ross’ finding that efforts tomake distressed portions of cities more vibrant mustgive consideration to the need to ensure that suchareas are safe (Rosenthal & Ross, 2010). As discussedearlier, findings from Greenbaum and Tita (2004) andIrvin-Erickson, Lynch, et al. (2017) show that a suddenincrease in gun violence reduces the growth of localbusinesses, especially in areas with higher pre-existinglevels of gun violence. Our findings support and expandupon these studies, showing that gunshots affect thehealth of local businesses in both areas with the lowest and areas with the highest pre-existing levels ofgunshots. Accordingly, efforts to improve distressedneighborhoods should target both areas with lower andhigher pre-existing levels of gunshots. In distressedneighborhoods, business owners have a strong incentive to collaborate with local officials to reduce gun violence. Gun violence r
ties, increase fear of crime and discourage investments in neighborhoods. On the topic of business patronage, Skogan (1986; 2012) and other authors (Bowes, 2007; Fisher, 1991) suggest that crime and fear of crime can reduce business revenues due to residents in high-crime neighborhoods l