Eric Jefferis Kent State University. Stephen M. Haas West Virginia Statistical Analysis Center. Erica Turley West Virginia Statistical Analysis Center

Transcription

1 Gun Availability and Crime 139 * Gun Availability and Crime in West Virginia: An Examination of NIBRS Data Stephen M. Haas West Virginia Statistical Analysis Center John P. Jarvis Behavioral Science Unit Federal Bureau of Investigation Eric Jefferis Kent State University Erica Turley West Virginia Statistical Analysis Center * Abstract The National Incident-Based Reporting System (NIBRS) data afford unique opportunities to inform crime analytic efforts. Recently reported relationships between the availability of firearms and violent crime constitute one such effort. While available criminological literature concerning the relationships between guns and crime are methodologically varied and equivocal, few studies have examined the issue of the prevalence of illegal versus legal guns and their impacts on reported violent crime. Inspired by an earlier effort to examine this question (Stolzenberg & D Alessio, 2000), this study attempts to replicate these findings through county-level analyses of West Virginia National Incident-Based Reporting System (NIBRS) data. Moreover, additional proxy indicators are used to identify significant clusters of illegal guns and examine the relationship between illegal guns and violent crime. Particular attention is paid to the spatial distribution of the variables and the potentially confounding effect of spatial correlation of crime rates across contiguous counties. The results indicate that counties with high concentrations of both legal and illegal guns are associated with violent crime, gun crime, and knife crime. These findings partially substantiate results from previous studies. This work was presented at the 2005 American Society of Criminology meeting. Initial funds supporting the work resulting in this manuscript were provided to the West Virginia Statistical Analysis Center by the Bureau of Justice Statistics (Grant# 2004-BJ-CX-K004). The viewpoints expressed herein are solely those of the authors and do not reflect the opinions of the U.S. Department of Justice, the Office of Justice Programs, the Bureau of Justice Statistics, the West Virginia Division of Criminal Justice Services, or Kent State University. JUSTICE RESEARCH AND POLICY, Vol. 9, No. 2, Justice Research and Statistics Association

2 140 JUSTICE RESEARCH AND POLICY Much empirical attention has focused on the steady and substantial decline in crime rates during the 1990s. Results from the National Crime Victimization Survey (NCVS) indicate a 50% decrease in violent crime from 1993 to 2001, which includes a 10% reduction in the most recent period, (Rennison, 2002). Overall, this represents a reduction in risk of violent victimization for those aged 12 and older from 50 to 25 victimizations per 1,000 individuals. Although violent crime victimization rates are at the lowest levels since the NCVS began collecting such statistics in the early 1970s, teens and blacks continue to be at the greatest risk for such victimization (Rennison, 2002). Furthermore, the decline in violent crime rates in the 1990s was preceded by a decade of extraordinary growth in risk of youth to violent crime (Blumstein, 2002). Consequently, although violent crime rates have substantially declined since 1993, firearm death rates for children and adolescents are still higher than historical rates in the United States and higher than those of similarly developed countries (Fingerhut & Christoffel, 2002). Blumstein (2002) and others (e.g., Cork, 1999; Lattimore, Trudeau, Riley, Leiter, & Edwards, 1996) have attributed much of the rise and fall of homicide rates over the past two decades to the emergence and subsequent stabilization of the crack cocaine market in the 1980s and 1990s, as well as a related increase in gun carrying by youth who were recruited into those crack markets. While the crack cocaine market may have stabilized, the increased lethality of firearms violence continues to plague youth in the United States (Fingerhut & Christoffel, 2002). As reported in recent Uniform Crime Reports (UCR), firearms were used in 68% of murders committed in 2006 (Federal Bureau of Investigation, 2007). NCVS statistics estimated that 29% of all reported violent victimizations in 1993 involved an offender armed with a firearm (Zawitz, 1995). In sum, while rates of victimizations have improved over the past decade, violent victimizations by firearms continue to be a serious problem, particularly among the youthful population. In response to this persistent problem of violent crime in the United States, the federal government has recently undertaken a major national initiative, Project Safe Neighborhoods, to reduce gun violence across the country. Critical to the current federal effort is its emphasis on data-driven decision-making and measurable accountability. In short, this initiative brings together multiagency collaborators in each of the 94 federal judicial districts to identify specific gun crime problems in their communities, locate where the problems are concentrated, implement intervention strategies to eradicate those gun crime problems, and evaluate the success of the efforts (U.S. Department of Justice, 2001). The background information above provides some insight into the scope of the gun violence problem in the United States and the current national strategy that has been developed to deal with the problem. Clearly, the success of this effort, like many other criminal justice programs and policies, largely relies on the use of empirical data to shape, direct, and measure the effort s impact. This

3 Gun Availability and Crime 141 study responds to this specific need for empirical analysis that may inform such programs or policies by exploring the connection, if any, between the availability of guns and crime in West Virginia. * Relevant Literature Guns pose unique challenges in that the use of such lethal weapons makes it possible for perpetrators to attack another individual with minimal effort and threat to the attacker (Hoskin, 2001; Wells & Horney, 2002). Varying aspects regarding the use of firearms in violent behavior have been the topic of several recent studies. In fact, much research has been dedicated to understanding the relationship between gun availability and violent crime (e.g., Cook & Ludwig, 2004; Hepburn & Hemenway, 2004; Hoskin, 2001; Moody & Marvell, 2005; Stolzenberg & D Alessio, 2000). For example, Hoskin (2001) contends that, violent crime statistics in the United States are very similar to those of other industrialized nations; however, the level of lethal violence is much higher in the United States (p. 570). In his comparative analysis of violent crime rates, Hoskin (2001) proposed two reasons to explain why the U.S. lethal crime rate is so high as it relates to gun availability: 1) the prevalence of handguns is much higher than in other nations, according to available data; and 2) the difference between gun homicide rates versus non-gun homicide rates is greater in the United States than in other countries. Yet, Hoskins suggests that the relationship between firearms and violence can be viewed in three different ways. First, it is possible there is no relationship between guns and violent crime. Second, gun availability contributes to elevations in violent crime. And third, the presence of guns actually reduces the crime rate. Others have explored the possibility that even if guns were not available, violent individuals would likely select another type of lethal weapon to conduct their criminal behavior (Stolzenberg & D Alessio, 2000; Wells & Horney, 2002; Griffiths & Chavez, 2004; Moody & Marvell, 2005). Still others have focused on the outcomes of criminal behavior when guns are involved. Not surprisingly, Wells and Horney (2002) found that injuries from a gun are more likely to be lethal than injuries sustained by other weapons. In a different vein, a study by Stolzenberg and D Alessio (2000) explored the firearm/violent crime relationship with the integration of two seemingly contradictory hypotheses: the objective dangerousness hypothesis and the self-defense hypothesis. The objective dangerousness hypothesis posits that the availability of illegal guns has a positive relationship with violent crime. The self-defense hypothesis argues that the availability of legal guns serves as a deterrent to criminals and thereby has a negative relationship with violent crime. In testing these contentions, Stolzenberg and D Alessio (2000) found that while the prevalence of legal gun ownership showed no consistent relationship to violent

4 142 JUSTICE RESEARCH AND POLICY crime, there were consistent findings to support the objective dangerousness hypothesis namely, that availability of illegal guns has a significant positive relationship to violent crime. Considering this objective dangerousness notion, Griffiths and Chavez (2004) found evidence to support the relationship between firearms and violent crime in their analysis of homicide trajectories. Their study found that of the total 12,541 homicides in the city of Chicago from , a total of 6,291 (50.2%) were committed with the use of street guns. Additionally, street guns were a significant factor in Chicago s violent crime spike of the early 1990s. Interestingly, these authors noted a weapon substitution effect in the crime statistics of Chicago over those years. That is, they found that while some sections of the city experienced rising street gun homicides, those same areas of the city experienced a decline in homicides committed by the use of other weapons (Griffiths & Chavez, 2004). In contrast, Cook and Ludwig (2004) found that on a national level, the violent crime rate among adolescent age groups experienced a spike in the mid-1980s to early 1990s. As a result, they concluded that the prevalence of guns carried and used were to blame for the dramatic increase in homicide rates. However, the authors suggested that one of the primary reasons that adolescent males carried guns was for self-defense. Complicating this debate, however, Hoskin (2001) did not find evidence that increased homicide rates encouraged citizens to purchase firearms for self-defense. Clearly not all research supports the theory that gun prevalence is directly related to increases in violent crime. Moody and Marvell (2005) conducted a study to test the assumption that guns are related to increased levels of violent activity. They included data on gun ownership (specifically handguns) and conducted analyses that would link gun ownership to violent crime. They found that even though guns may be more prevalent in some crime categories (e.g., rape), overall guns had a minimal net effect on crime. In fact, their findings showed that if the reported gun ownership were to double from 26% to 52%, major crime would increase by only 1% (Moody & Marvell, 2005). The results of Moody and Marvell s research raise some important questions regarding the relationship between guns and violent crime. Does this study truly suggest that guns do not contribute to crime, or can their findings be explained by the fact that legal gun owners are not significant contributors to violent crime? Or, is it possible, as they suggest, that if guns are readily available, they use guns, if not, they use other weapons, the rate of crime being unaffected by the presence or absence of guns (Moody & Marvell, 2005, p. 735). It is clear that research findings on the relationship between gun prevalence and violent crime are equivocal. However, the complexities of the guns and crime issues do not end here. Wells and Horney (2002) took a slightly different approach in examining this relationship by seeking to determine whether violence could be reduced through restricting the availability of guns and whether

5 Gun Availability and Crime 143 the individual s intent to cause injury was somehow related to the use of firearms. Their rationale was that an aggressor with some motivation to attack, but without the intentions to do serious harm would probably not make the effort required to produce serious injuries (Wells & Horney, 2002, p. 271). Although Wells and Horney found no relationship between weapons and an individual s intentions to inflict injury, they did find a significant relationship between weapons and violent crime. Specifically, they found that the likelihood of an attack increased fivefold when a weapon was used (compared to no weapon use). Interestingly, while non-firearm weapons increased the likelihood of an attack by fivefold, guns only increased the probability of an attack by about a factor of two, leading to the conclusion that in relation to other weapon attacks, gun attacks significantly reduce the chances of an opponent being injured (Wells & Horney, 2002, p. 285). Yet they also find that while guns may not increase the likelihood of an attack compared with other weaponry per se, when a gun is used, the probability of inflicting serious injury is increased by a factor of 60, and a fourteenfold increase can be observed when comparing firearms to other weapon attacks (Wells & Horney, 2002). They explain this by noting: Guns are believed to produce more serious injuries and a greater likelihood of death than are other weapons because of their technical properties and because doing serious harm with a gun requires less strength and sustained effort on the part of an assailant (p. 291). Thus, many questions remain unanswered regarding gun prevalence and violent crime, and the extant research remains mixed. In particular, the role of illegal and legal guns continues to be debated. The National Research Council (2004) recently concluded that ecological studies do not credibly demonstrate a causal relationship between the ownership of firearms and criminal violence because of the limitations of existing data and methods. Others argue that more attention should be given not only to those who own guns, but also to those who are involved in activities or situations that require guns for self-defense. While this notion suggests a mitigating impact of guns on crime, some suggest that differential effects of available legal and illegal guns specify the impact of illegal rather than legal guns on crime rates. To differentiate their impacts on crime trends, however, measurements of both illegal and legal guns are essential. Research by Stolzenberg and D Alessio (2000) and others has extended this interesting debate amongst social scientists examining the gun availability-crime link (for example, see Blumstein, 1995; Lott, 1998). In particular, Stolzenberg and D Alessio (2000) demonstrated that counties with high concentrations of illegal guns are associated with violent crime, gun crime, and juvenile gun crime. Yet their research showed no such relationship exists among legal gun availability and measures of violent crime. In addition, Stolzenberg and D Alessio s (2000) analysis of four years of South Carolina NIBRS data showed that illegal gun availability had no impact on crimes committed with a knife, indicating no displacement of violent gun crime to the alternative form of knife violence.

6 144 JUSTICE RESEARCH AND POLICY While these findings are consistent with strategies of removing illegal guns from the community and provide valuable analytic insights into how to direct available resources to those communities with the greatest need, the question remains as to whether these findings are applicable to other localities where the dynamics of gun ownership (whether legal or illegal) are likely to be different. The present study is devoted to this particular question. * Data and Methods As a result of expanded analytical and mapping possibilities, incident-based police data are perhaps ideal for identifying the details of gun-related crime. Police incident data often include such variables as whether a weapon was used, type of weapon, relationship between the offender and victim, and where the incident occurred. The inherent detail of incident-based data also offers an opportunity to examine changes in gun-related crime over a period of time. The National Incident-Based Reporting System (NIBRS) data collected by the Federal Bureau of Investigation contains such details and is utilized in the analysis here. The use of NIBRS data for these types of strategic crime analysis is not new (e.g., Chilton & Jarvis, 1999; Faggiani & McLaughlin, 1999; Turley & Haas, 2006). However, NIBRS data are particularly appropriate for cross-jurisdictional efforts because of the uniform manner in which crime incidents are defined and data collected (FBI, 1992). Indeed, with the simple addition of a geographic referent, NIBRS data facilitate many kinds of strategic crime analyses as described here, although comprehensive and timely reporting of NIBRS data is not yet a reality in many states. Though limitations exist, 1 NIBRS data are uniquely capable of informing crime control strategies in at least one important way. NIBRS analyses can illuminate efforts to target high-risk communities and remove illegally obtained firearms from adults and youth who may use or have already used them in the commission of a violent or property crime. In support of this contention, we use NIBRS data to identify significant concentrations of and measure the effects of illegal gun availability across NIBRS reporting jurisdictions in West Virginia. In particular, we utilize county-level incident data from the 55 counties in West Virginia over a three-year period ( ) to derive not only measures for 1 The most common limitation to the use of NIBRS data is that many agencies across the nation are not yet reporting NIBRS-type data to the FBI. West Virginia was the 16 th state to receive certification from the FBI and began accepting data in the West Virginia Incident-Based Record System (WVIBRS) format in January WVIBRS data are estimated to cover 100% of the population. While individual agency participation levels do vary across the state and over time, we believe this source of error is negligible in the analyses presented here.

7 Gun Availability and Crime 145 various crime rates but also a measure of illegal gun availability in these counties in support of three research objectives: 1) to examine the spatial dynamics of guns and crime in West Virginia; 2) to replicate previous research investigating the relationship between guns and crime (Stolzenberg & D Allesio, 2000); and 3) to examine the utility of NIBRS data for examining criminal justice policy and practice. Data for the proposed analyses are derived from several sources including, but not limited to: (1) NIBRS reporting agencies in West Virginia, (2) U.S. Bureau of the Census (Census 2000) data downloadable from the Inter-university Consortium for Political and Social Research (ICPSR) data archive, and (3) countylevel demographic data from various governmental agencies in West Virginia that track such information as aid to dependent children, divorce rates, and school dropout rate. A detailed description of the data sources and variables included in the current analysis is presented below. Dependent Variables Consistent with Stolzenberg and D Alessio (2000), we derived three dependent variables from the NIBRS data. First, violent crime rate is defined as the number of violent crimes reported during the study period, divided by the county population for the year Second, gun crime rate is similarly operationalized as the number of violent crimes in which a gun was used by at least one offender in the reported incident during the study period. Finally, we examine violent crimes committed with a knife, knife crime rate, to determine if illegal gun availability is associated with a change in other forms of violent crime. This measure is similarly operationalized as the number of violent crimes in which a cutting instrument was used by at least one offender in the reported incident during the study period. Independent Variables Measuring gun availability, whether legal or illegal, is particularly difficult and methodologically challenging. By reviewing over two dozen gun measures, Kleck (2004) highlights this point and challenges much of the guncrime research by concluding that most measures have significant limitations that diminish the usefulness of extant research. We do not refute Kleck s validity analyses relating to gun availability measures. However, we do adopt the measures used previously by Stolzenberg and D Alessio (2000) for two reasons. First, one of the principal points of our study is not to debate the adequacy 2 Violent crime is operationalized as incidents in which at least one of the following offenses was reported: murder, non-negligent manslaughter, forcible rape, aggravated assault, robbery, kidnapping, forcible sodomy, sexual assault with an object, forcible fondling, or extortion/blackmail.

8 146 JUSTICE RESEARCH AND POLICY of proxy gun measures but to attempt to replicate the findings and analyses conducted in South Carolina using the same data collection instruments and strategies employed in the earlier work. Second, while there are important issues in the measurement of gun availability, few widely accepted measures are currently found in the existing literature. Therefore, measurement issues notwithstanding, in order to examine this dimension of gun prevalence further and to retain the particular strength of the earlier work in this replication effort, we follow Stolzenberg and D Alessio (2000) by measuring illegal gun availability as the number of guns reported stolen in each jurisdiction. In contrast, legal gun availability, which is not available in NIBRS, was measured as the number of concealed weapon permits issued in each county per 100,000 population lagged one year. Concealed weapon permits are issued by county sheriffs and compiled statewide by the West Virginia State Police. Similar to Kleck (2004), we acknowledge that each of these gun measures is inherently imperfect and not an ideal metric for the dynamics of gun availability. 3 However, our contention is that these proxy measures, like similar limitations inherent in reported crime data, are adequate for examining, at least in part, the dynamics of the gun-crime debate as analyzed here. Control Variables Eight variables are also included in this study to control for community structural factors and to most closely reflect the earlier analyses conducted by D Allessio and Stolzenberg (2000). These include: 1) population density, county-level population density rates calculated from the 2000 census; 2) divorce, obtained from the Health Statistics Center, Department of Health and Human Resources (DHHR); 3) TANF, county-level Temporary Assistance for Needy Families (TANF) numbers extracted from the DHHR Web site; 4) dropout rate, the number of dropouts and total enrollment obtained from the U. S. Department of Education (county rates calculated by dividing the number of dropouts by the total enrollment); 5) unemployment, the number of persons in the annual average civilian labor force who are unemployed by county obtained from the Bureau of Employment Programs Web site; 6) prison admissions, the number of inmates admitted to the Division of Corrections annually by county. These data were obtained from National Corrections Reporting Program files provided by the Division of Corrections; 7) juvenile commitments, the number 3 For example, these measures may not fully account for stolen guns not reported to police, illegal gun sales, gun in-migration or out-migration, gun trades for other in-kind valuables, or other secondary market sales. Nonetheless, we contend that these measures are conservative measures of the availability of guns, which should bias the analyses toward null findings. We also acknowledge that these measurement limitations may be important for considering and interpreting the utility of the findings that result.

9 Gun Availability and Crime 147 of juvenile offenders committed to juvenile detention and correctional facilities. Data were obtained from a database maintained by the West Virginia Division of Juvenile Services and are based on county intakes and represent regular commitments and others under supervision; and 8) arrest rate, the total number of UCR arrests as reported in annual FBI Uniform Crime Reports. These data were obtained from UCR summary data files and extracted from the Criminal Justice Archive at the Inter-university Consortium for Political and Social Research at the University of Michigan. Analytical Strategy We begin the analysis by first conducting univariate analyses of the variables of interest. Second, county-level analyses of the association between illegal gun availability and the three measures of violent crime are conducted using both geographic maps and statistical modeling to inform cross-jurisdictional collaborative efforts and to illustrate the practical utility of geographically referencing NIBRS data. These analyses may not only serve to inform and direct the West Virginia crime control initiatives but they also provide a partial replication of the earlier work as previously noted. Lastly, ordinary least squares regression (OLS) analyses are conducted to further demonstrates the utility of NIBRS data for use by criminal justice policymakers and practitioners. In all, these mapping efforts and statistical models provide an opportunity to examine the counties and jurisdictions with significant concentrations of gun violence and illegal gun availability, as well as potentially determine the effect of illegal guns on four measures of violence (net of the community structure control variables). * Results Univariate Analyses Univariate descriptive statistics show that in West Virginia the mean violent crime rate is per 100,000 resident population. This compares to a mean gun crime rate that is about 20% of that figure at and a knife crime rate that measures just about 10% of all violent crime at In terms of gun availability, the summary measures reported in Table 1 show more prevalence of legal guns than illegal guns, with respective mean rates of and However, recall that the illegal gun measure is only a proxy or reflection of those stolen guns reported to police. While elusive, more comprehensive measures of true illegal gun activity are likely to be substantially higher than those reflected in these statistical results.

10 148 JUSTICE RESEARCH AND POLICY * Table 1 Descriptives for All Variables of Interest Standard Variable N Minimum Maximum Mean Deviation Violent Crime Rate Gun Crime Rate Knife Crime Rate Gun Theft Rate Suicide Rate Total Population 165 5, , , , Population Density Gun Carry Permits (lagged) Percent TANF Divorce Rate , Juvenile Commitment Rate Percent Unemployed Prison Admission Rate Percent HS Dropouts Arrest Rate, , , Valid N (listwise) 165 Note. Crime rates calculated per 100,000 population, while population was calculated per square mile. These measures of central tendency, however, do not reveal particular concentrations of these variables across the various counties. To examine this level of specificity, the geographical distribution of all county-level variables was examined by creating choropleth maps using ArcView 3.3. For brevity, only the crime and gun variables are depicted here (see Figures 1-5). Figure 1 shows the highest concentrations of violent crime rates in the southern part of the state (especially Charleston and Bluefield), with some high concentrations in other, more distant cities (i.e., Morgantown and Martinsburg). Also of importance is the relatively low concentration of violent crime in certain areas of the state, as represented by the white portions of the graphic. In all, this geographic analysis highlights the fact that the concentration of violent crime is not uniformly distributed throughout the state. A similar pattern emerges for gun crime rates presented in Figure 2. While gun crimes are less prevalent and also less concentrated than violent crimes, the rates for gun crimes generally seem to overlay in areas where high concentrations of violent crime were present in Figure 1. There is some variability (e.g., Morgantown and Parkersburg), but overall, gun crime incidents that come to the attention of the police appear in the same geographical areas as violent crime. As shown in Figure 3, the distribution of knife crimes is also comparable

11 Gun Availability and Crime 149 * Figure 1 Violent Crime Rate in West Virginia per 100,000 Population Source: WV Incident-Based Reporting System data. * Figure 2 Gun Crime Rate in West Virginia per 100,000 Population Source: WV Incident-Based Reporting System data.

12 150 JUSTICE RESEARCH AND POLICY * Figure 3 Knife Crime Rate in West Virginia per 100,000 Population Source: WV Incident-Based Reporting System data. * Figure 4 Concealed Weapon Permit Rate in West Virginia per 100,000 Population Source: WV Incident-Based Reporting System data.

13 Gun Availability and Crime 151 * Figure 5 Gun Theft Rate in West Virginia per 100,000 Population Source: WV Incident-Based Reporting System data. to other types of crime. This map displays the same concentrations in Charleston and Bluefield as violent and gun crime analyses showed. In the case of knife crimes, however, the prevalence is much less than for violent and gun crimes. Moreover, the distribution across the state is also much thinner as evidenced by the numerous white portions in Figure 3. Conversely, different patterns emerge when examining how gun availability is distributed at the county level. In terms of legal gun availability, Figure 4 shows legal gun (concealed weapon permit) rates as concentrated in areas other than where the violent, gun, and knife crime rates were found. Rather than the persistent concentrations of crime in Bluefield and Charleston, legal guns appear to have the greatest concentrations in adjacent (or not too distant) geographical areas to these cities. In fact, Figure 4 shows a concentration of a sort that could be loosely described as an arc from Parkersburg to Beckley, with some other higher concentrations south and west of that arc. This interpretation notwithstanding, the figure suggests that there is a relative prevalence of guns throughout the state, with the highest concentrations in the more rural localities rather than the more urban centers. Lastly, Figure 5 depicts the distribution of gun theft rates in West Virginia. This map shows just two areas with high concentrations of stolen guns. The high concentration in and around Bluefield as well as in the vicinity of Charleston suggests a modest relationship between crime rates and illegal guns that is consistent with the objective dangerousness hypothesis.

14 152 JUSTICE RESEARCH AND POLICY In summary, these analyses have identified significant hot spots and cold spots of illegal gun availability across West Virginia through the use of spatial analysis tools. These graphs also depict co-occurring and notable pockets of both criminal incidents and presence of guns (both legal and illegal) rather than a more uniform distribution of crime and guns across the state. However, these descriptive analyses do not account for the possibility that other community-level variables may impact these apparent relationships. To account for this possibility, we used statistical tests available in GEODA (e.g., Moran s I and Anselin s Local Indicators of Spatial Autocorrelation) to identify potential problems of spatial autocorrelation across the counties that may be present in subsequent multivariate modeling efforts, which are presented below. 5 In order to address the potential problem of spatial autocorrelation across the counties, the GEODA software package was used to construct a spatial weights matrix using county centroids, and a spatial regression model was also calculated. Although Moran s I was significant in the multivariate model, the independent variables were not substantially impacted. Therefore, OLS results are presented here for ease of interpretation. Finally, one poor reporting county (Ohio) was removed from the analysis, which also made no substantive difference to the final models. Multivariate Analyses With the above results serving as a foundation, three multivariate models were examined to determine the impact of guns on crime. In particular, ordinary least squares regression models for violent crime, gun crime, and knife crime are derived and reported in Table Spatial autocorrelation can occur when the enumeration units being analyzed share boundaries that result in a structure to the data. Units that share physical boundaries or are closer to each other are likely more similar than those not sharing physical boundaries or those that are farther away from each other. Thus, geographic data may often fail to meet the assumption of normal-theory statistical methods that the observations be independent of one another. Some suggest that spatial autocorrelation can be interpreted as a nuisance or a substantively more interesting phenomenon (e.g., Morenoff, Sampson, & Raudenbush, 2001). However, the problem is that the errors will not be independent a necessary condition for valid hypothesis tests. The consequences of spatial autocorrelation are the same as in temporal autocorrelation; that is, in the linear regression model the standard errors will be biased, and t-statistics used to test the null hypothesis may seriously overstate the statistical significance of an effect. This can lead to the mistaken conclusion that variables are related when they are not (Odland, 1988). 6 A series of diagnostic tests were performed to assess the appropriateness of the final models. Multicollinearity tests (Variance Inflation Factor [VIF], and Tolerances) were performed to identify problems with collinear variables. These tests revealed no significant multicollinearity problems with the variables presented in the final models.

15 Gun Availability and Crime 153 * Table 2 Ordinary Least Squares Multivariate Regression Models Model 1 Model 2 Model 3 Violent Crime Rate Gun Crime Rate Knife Crime Rate Independent Variables b SE Beta b SE Beta b SE Beta Gun Theft Rate (t-1) * * * Concealed Carry Rate (t-1) * * * Year (2001) * Population Density Pct TANF (t-1) * * Divorce Rate (t-1) Juvenile Commit Rate (t-1) * * Pct Unemployed Prison Admit Rate (2001) * Pct High School Dropouts (t-1) Arrest Rate * * * Constant F-Value * * * Adjusted R * Indicates statistical significance at the.05 or greater level.

16 154 JUSTICE RESEARCH AND POLICY The violent crime model (column 1) shows a significant impact of both legal and illegal guns on violent crime. The gun theft rate (b =.713) has the largest effect in this model compared to its effects in the other models and exhibits a statistically significant positive effect on the violent crime rate. In comparison, the concealed carry rate (b =.195) also shows a significant positive effect on violent crime. However, this impact is about a quarter that of illegal guns. Stated differently, the standardized regression coefficients show that a unit change in the illegal guns measure has about a.33 impact on a unit change in violent crime compared to just a.22 change in violent crime for legal guns. 7 In contrast, the gun crime model (column 2) shows, as expected, different coefficients but similar results. That is, consistent with the previous findings in Model 1, these results reveal significant positive effects of both illegal (b =.269) and legal gun (b =.064) measures on gun crime rates. Again, the larger effects for illegal guns and the attendant beta coefficients suggest about a 70% greater impact of illegal guns than legal guns on gun crime rates. Finally, turning to Model 3 we again see a similar pattern with significant and positive impacts of gun availability on knife crimes. While the effects of parameters in this model are much reduced compared to the other models in Table 2, the pattern is consistent with larger impacts of the illegal gun measure on knife crimes. When beta coefficients are examined, however, the relative impact of illegal guns (beta =.299) is nearly commensurate with that of legal guns (beta =.292). Or, as expressed above, these beta coefficients suggest just a 2% greater impact of legal guns on knife crime than illegal guns. Nonetheless, both of these indicators imply a significant impact of the availability of guns (whether legal or illegal) on the rate of knife crimes, even after controlling for the other community context variables included in the model. Yet, these results are of a smaller magnitude than the findings of the previous models of gun and violent crime. With that said, however, these results are consistent with those of Models 1 and 2, yielding an overall pattern that suggests not only significant positive effects of guns on crime, but also differential effects of illegal versus legal guns on crime rates. * Discussion and Conclusions This study has been devoted to utilizing both NIBRS data and other county-level data to examine the connections, if any, between guns and crime in West Virginia. To date, the empirical literature has offered several differing contentions about the relationship between guns and crime. These include arguments that 7 One alternative interpretation of these standardized regression results would suggest about a 50% larger impact of illegal guns compared to legal guns (.33 versus.22) on unit changes in violent crime.

17 Gun Availability and Crime 155 imply little or no relationship may exist between gun availability and violent crime to the contention that the presence of guns can actually reduce crime. Though not explicitly examined here, some studies suggest an objective dangerousness hypothesis and a contrasting self-defense hypothesis for delineating the relationship between guns and crime (Stolzenberg & D Alessio, 2000). In attempting to replicate the findings of Stolzenberg and D Alessio (2000), this study presents evidence that may help address some of the issues that surround the gun crime debate. The analysis of univariate statistics (Table 1) and mapping efforts (Figures 1-5) as well as the multivariate models (Table 2) exhibit clear evidence refuting the position that there is no relationship between the availability of guns and crime. However, the findings do appear to support the notion that guns are related to elevations in violent crime and that guns do not lead to lower crime rates. Both the legal and illegal gun measures were positively and significantly related to each of the three offenses examined in this study violent crimes, gun crimes, and knife crimes. The significant relationships held even after controlling for various other community structural factors. Yet, it is not entirely clear that the evidence is as definitive as these results might suggest. First, it is important to note that both the legal and illegal gun measures are only proxies (see Kleck, 2004). Second, in addition to this measurement issue, efforts to draw definitive conclusions are complicated by the relative strength of the regression coefficients and the presence of two competing explanations for the relationship between guns and crime. Specifically, the larger magnitudes of the coefficients for the illegal gun measure in the multivariate modeling seem to support the objective dangerousness hypothesis. On the other hand, the results could also be interpreted as support for the self-defense argument that the availability of legal guns serves as a deterrent to criminals. This line of logic draws support from two perspectives. First, our results indicate that the regression coefficients for both legal and illegal guns were significant across all three models (contrary to Stolzenberg and D Allessio s findings). The finding that the availability of legal guns is a significant predictor of these crimes lends some support to the self-defense hypothesis. Yet, the regression coefficients for legal guns were considerably smaller than the coefficients obtained for the measure of illegal guns. Second, it could be argued that the illegal gun measure employed in this study is likely to underrepresent the true level of illegal gun activity present in a community (see footnote 3), thereby suggesting that the disparity in illegal versus legal gun effects may be much greater than discovered in the current analysis. As a result, the quantitative findings offer some support to both contentions, but the magnitude of the estimates are comparatively more supportive of the objective dangerousness hypotheses. Our analysis, like Stolzenberg and D Allesio s (2000), supports this general conclusion, but we caution that these relationships are most pronounced in violent and gun crimes and less evident in knife crimes. While this may seem to some as inconsequential, this is at the center of the argument for many concerned with the gun crime debate.

18 156 JUSTICE RESEARCH AND POLICY These findings have implications for many current programs and policies. It is evident that strategic planning elements of many firearm crime reduction initiatives (e.g., Project Safe Neighborhoods) can be informed by such efforts to map and identify hot spots for gun crime through spatial analysis. Specifically, recall that the geographic analyses revealed that gun crimes are not uniformly distributed across the state. Rather, these crimes are significantly clustered in only a few counties. This suggests that any policies, programs, or practices designed to target such criminal incidents are likely to be best launched in and around these counties rather than applied statewide. Such analyses are likely to be of particular importance to states with small population centers such as West Virginia, where counties rather than cities are often the units targeted for crime reduction initiatives. In such states, many programmatic and policy decisions are formulated at the state level and directed toward counties rather than specific urban centers. Finally, while our analysis supports the notion that both legal and illegal guns contribute in specific ways to crime problems in West Virginia, the statistical models do not reveal the underlying spatial distributions for the geographic concentrations of violence and gun crime in the state. For example, some efforts may be more likely to produce the intended effects by successfully confronting violence and gun crime in counties such as Kanawha (i.e., Charleston) and Mercer (i.e., Bluefield) than if they were applied in Harrison (i.e., Clarksburg) and Ohio (i.e., Wheeling) counties. Nonetheless, the promise of NIBRS data to support crime analysis and criminal justice policy questions should not be understated. Through the use of NIBRS data, this study was able to systematically identify potential areas for future crime reduction initiatives and provide evidence to better inform the debate over the relationship between gun availability and crime. It is important to note that some limitations persist in exploring the guncrime question further. Though the present study as well as Stolzenberg and D Alessio s (2000) research in South Carolina were able to be conducted, this type of analysis is still not possible for many jurisdictions across the country given that law enforcement agencies do not yet participate in NIBRS-type crime data collections. While West Virginia is a NIBRS-certified state, until more agencies from larger urban localities begin submitting data in the NIBRS format, analyses such as this cannot be done in many jurisdictions. In addition, better measures of illegal and legal gun availability are needed for analysis. Currently, such measures are difficult to obtain and verify. This study relied upon metrics used in previous work that reflect some aspects of gun prevalence. However, the development of more reliable and valid data sources would yield more precise analyses and further contribute to the current debate (see Kleck, 2004). For instance, even NIBRS is not able to identify specific types of firearms (e.g.,.45,.22,.38 caliber, etc.) involved in gun crimes or the exact usage of the weapon in criminal incidents (i.e., the number of shots fired, locations of injuries, etc.). Although these data elements are unlikely to

19 Gun Availability and Crime 157 ever become part of NIBRS, they may be important in the crime and public health discussions relating to not only guns but other weapons that are common in criminal behavior. Similarly, other data sources may need to be mined to provide additional insight into the gun-crime connection. Although there is not an easily accessible and widely agreed upon metric for measuring gun availability (National Research Council, 2004), newly developed data sources might include elements pertaining to the points of origin and migration patterns of guns involved in crime. These data may come from traces conducted by the Bureau of Alcohol, Tobacco, Firearms and Explosives as well as information gained from cooperative suspects. Lastly, research activities can also be directed toward examining the impact of public safety initiatives such as Project Safe Neighborhoods and other efforts using different metrics than simply incidents of gun violence. Outcome measures such as nonviolent crimes that are gun-related and firearm-related homicides may introduce additional dynamics worth investigating as part of the gun-crime debate.

20 158 JUSTICE RESEARCH AND POLICY * References Blumstein, A. (1995). Youth violence, guns, and the illicit-drug industry. Journal of Criminal Law and Criminology, 86, Blumstein, A. (2002). Youth, guns, and violent crime. Children, Youth, andgun Violence, 12(2), Chilton, R., & Jarvis, J. (1999). Using National Incident-Based Reporting System (NIBRS) to test estimates of arrestee and offender characteristics. Journal of Quantitative Criminology, 15(2), Cook, P. J., & Ludwig, J. (2004). Does gun prevalence affect teen gun carrying after all? Criminology, 42(1), Cork, D. (1999). Examining space-time interactions in city-level homicide data: Crack markets and the diffusion of guns among youth. Journal of Quantitative Criminology, 15(4), Faggiani, D., & McLaughlin, C. (1999). Using National Incident-Based Reporting System data for strategic crime analysis. Journal of Quantitative Criminology, 15(2), Federal Bureau of Investigation. (1992). National Incident Based Reporting Handbook. Washington, DC: Government Printing Office. Federal Bureau of Investigation. (2007). Crime in the United States Washington, DC: Government Printing Office. Fingerhut, L. A., & Christoffel, K. K. (2002). Firearm-related death and injury among children and adolescents. Children, Youth, and Gun Violence, 12(2), Griffiths, E., & Chavez, J. M. (2004). Communities, street guns and homicide trajectories in Chicago, : Merging methods for examining homicide trends across space and time. Criminology, 42(4), Hepburn, L. M., & Hemenway, D. (2004). Firearm availability and homicide: A review of the literature. Aggression and Violent Behavior, 9, Hoskin, A. W. (2001). Armed Americans: The impact of firearm availability on national homicide rates. Justice Quarterly, 18(3), Kleck, G. (2004). Measures of gun ownership levels for macro-level crime and violence research. Journal of Research in Crime and Delinquency, 41(1), 3 36.

21 Gun Availability and Crime 159 Lattimore, P. K., Trudeau, J., Riley, K. J., Leiter, J., & Edwards, S. (1996). Homicide in eight U.S. cities: Trends, context, and policy implications. Washington, DC: National Institute of Justice. Lott, J.R. (1998). More guns, less crime. Chicago: University of Chicago Press. Moody, C. E., & Marvell, T. B. (2005). Guns and crime. Southern Economic Journal, 71(4), Morenoff, J. D., Sampson, R. J., & Raudenbush, S. W. (2001). Neighborhood inequality, collective efficacy, and the spatial dynamics of urban violence. Criminology, 39, National Research Council. (2004). Firearms and violence: A critical review. Washington, DC: National Academies Press. Odland, J. (1988). Spatial autocorrelation. (Scientific Geography Series, Vol. 9.) Beverely Hills, CA: Sage Publications. Rennison, C. (2002). Criminal victimization 2001: Changes with trends Washington, DC: Bureau of Justice Statistics. Stolzenberg, L., & D Alessio, S. J. (2000). Gun availability and violent crime: New evidence from the National Incident-Based Reporting System. Social Forces, 78(4), Turley, E. & Haas, S. (2006, April). Patterns of violent crime and weapon use, Charleston, WV: Criminal Justice Statistical Analysis Center, Division of Criminal Justice Services, Department of Military Affairs and Public Safety. Available online at: U.S. Department of Justice. (2001). Project Safe Neighborhoods: America s Network Against Gun Violence. Available at Wells, W., & Horney, J. (2002). Weapon effects and individual intent to do harm: Influences on the escalation of violence. Criminology, 40(2), Zawitz, M. W. (July 1995). Guns used in crime. Washington, DC: Bureau of Justice Statistics.