bonus

kaylenandpridemore2011.pdf

A Reassessment of the Association Between Social Disorganization and Youth Violence in Rural Areas∗

Maria T. Kaylen, Indiana University

William Alex Pridemore, Indiana University

Objective. To study the association between social disorganization and youth vio- lence rates in rural communities. Method. We employed rural Missouri counties (N = 106) as units of analysis, measured serious violent victimization data via hospi- tal records, and the same measures of social disorganization as Osgood and Chambers (2000). Controlling for spatial autocorrelation, the negative binomial estimator was used to estimate the effects of social disorganization on youth violence rates. Results. Unlike Osgood and Chambers, we found only one of five social disorganization measures, the proportion of female-headed households, to be associated with rural youth violent victimization rates. Conclusion. Although most research on social dis- organization theory has been undertaken on urban areas, a highly cited Osgood and Chambers (2000) study appeared to extend the generalizeability of social disorga- nization as an explanation of the distribution of youth violence to rural areas. Our results suggest otherwise. We provide several methodological and theoretical reasons why it may be too early to draw strong conclusions about the generalizeability of social disorganization to crime rates in rural communities.

Adolescent violent victimization represents a serious social problem in the United States. The Surgeon General officially recognized its importance in 1985 and has since encouraged efforts to address it (U.S. Surgeon General, 2001). Researchers have identified medical, financial, behavioral, educational, criminal, and cognitive and emotional short- and long-term impacts of victim- ization (Boney-McCoy and Finkelhor, 1995; Macmillan and Hagan, 2004; Menard, 2002; Stewart, Schreck, and Simons, 2006). Nationally in 2000, ado- lescents aged 12–15 and 16–19 were violently victimized at a rate of 60 and 64 per 1,000 persons, respectively (Rennison, 2001). These two age groups represent the highest violent victimization rates of any age group, and this

∗Direct correspondence to William Alex Pridemore, Indiana University 〈wpridemo@ indiana.edu〉. This article was originally presented at the 2008 Meetings of the American Society of Criminology in St. Louis and the 2009 Meetings of the American Sociological As- sociation in San Francisco. We thank Wayne Osgood and Mitch Chamlin for their insightful comments and helpful critiques of earlier drafts of this article, and Tony Grubesic for his help with the spatial models. We are happy to share our data, coding, etc. with anyone wishing to replicate our analyses.

Social Disorganization and Youth Violence in Rural Areas 979

victimization tends to be at the hands of other adolescents (Boney-McCoy and Finkelhor, 1995).

Adolescent victimization rates, however, are not evenly distributed across all communities. Social disorganization theory may offer a partial explanation for the distribution of adolescent victimization rates. The theory posits that communities whose members are well acquainted and on good terms with one another have lower delinquency rates since adults in such communities are more likely to positively influence youth through informal surveillance, supervision, and shaping values (e.g., adults criticizing unacceptable behavior of nonfamilial youth with whom they are acquainted) (Osgood and Cham- bers, 2000). Bursik (1988) also claims a theoretical connection between social disorganization and victimization. Similar to routine activities (Felson and Cohen, 1980), he argues that community social disorganization reflects the ability of the community to supervise the interaction between offenders and opportunities. When potential offenders and opportunities meet, victimiza- tion occurs. Sampson (1985) likewise says that when communities are densely populated, surveillance and guardianship are inhibited and thus victimization increases.

There is growing interest in the social science literature on rural areas, including crime in rural communities. In analyses of nonmetropolitan counties in various states, for example, Osgood and Chambers (2000) and Bouffard and Muftić (2006) found an association between social disorganization and the distribution of youth violent arrest rates. Building on Osgood and Chambers (O-C), the goal of the present study was to test the hypothesis that this effect is conditioned by community size and density in nonmetropolitan settings. Using a similar rural sample and the same measures of social disorganization as O-C, however, our findings were inconsistent with those of Osgood and Chambers, showing little support for the social disorganization model in rural counties. We conclude that while O-C’s study was rigorously undertaken and their conclusions warranted, others may have been too quick to draw strong conclusions about the association between social disorganization and rural youth violence based largely on this single study, especially given several remaining unanswered theoretical and methodological questions about social disorganization, social relations, and crime in rural communities.

Literature Review

There are reasons to believe that social structure generally and social disor- ganization specifically may operate similarly in communities of different sizes. The empirical literature on the relationship between changing community structure and community problems tends to find similar relationships in rural and urban settings. For example, de-industrialization in urban communities and farm job loss in rural communities resulted in comparable social and economic problems (Ginder, Stone, and Otto, 1985; Tickamyer and Duncan,

980 Social Science Quarterly

1990). Similarly, rapid population growth has negative social and economic implications for all types of communities (Bacigalupi and Freudenberg, 1983; Guillaume and Wenson, 1980). Given these similarities, we might expect the structural antecedents of social disorganization—residential mobility, ethnic heterogeneity, family disruption, poverty, and population density—to influ- ence violence in rural and urban settings alike.

Social Disorganization Theory

Shaw and McKay’s (1942) social disorganization theory posits that commu- nity social disorganization leads to higher rates of delinquency because some communities are unable to achieve shared values or solve common problems (Bursik, 1988). The structural antecedents mentioned above are thought to create poor social integration and thus low levels of organization. This social disorganization interferes with community members’ ability to work together to supervise and socialize youth, and thus is associated with high levels of crime and delinquency (Kornhauser, 1978; Sampson and Groves, 1989). More re- cent studies of urban settings have moved beyond these structural antecedents and now focus on the community characteristics of social disorganization suspected of mediating the relationship between social structure and crime. These studies have consistently found measures of informal control and social cohesion to be inversely related to crime rates (Bellair, 1997; Elliott et al., 1996; Markowitz et al., 2001; Sampson, 1997; Sampson and Groves, 1989). Largely due to a lack of data, however, studies of rural settings have not ad- vanced to this stage (but see, for a test of the systemic model with perceived incivility and burglary, respectively, as outcome variables, Reisig and Cancino, 2004; Cancino, 2003).

Social Disorganization in the Rural Setting

The combination of high urban crime rates and Shaw and McKay’s work in Chicago led social disorganization researchers to focus primarily on cities. The interest in large cities is common to the macro-level criminological literature, though recently some scholars have called for a closer examination of crime in rural settings (Lee and Bartkowski, 2004; Lee and Ousey, 2001; Osgood and Chambers, 2000; Weisheit, Falcone, and Wells, 2006; Wells and Weisheit, 2004). Osgood and Chambers, for example, pointed out that “if the study of communities and crime is to mature, it must expand to encompass the full variety of communities” (2000:82).

Consistent Support for the Generalizability of the Theory. Social dis- organization research that has included rural settings finds a few consistent relationships between the structural antecedents of social disorganization and

Social Disorganization and Youth Violence in Rural Areas 981

crime. Some studies examined only rural communities (Bouffard and Muftić, 2006; Osgood and Chambers, 2000; Petee and Kowalski, 1993), while others examined the rural-urban or metropolitan-nonmetropolitan dichotomy (Bar- nett and Mencken, 2002; Lee, Maume, and Ousey, 2003). In their study, Osgood and Chambers (2000) found residential instability, ethnic hetero- geneity, female-headed households, and population density to be positively related to juvenile arrests in nonmetropolitan counties. Bouffard and Muftić (2006) found ethnic diversity, female-headed households, and density to be associated with assaults in rural counties. Barnett and Mencken (2002) found percent of nonwhite residents and population change to be positively related to violent crime in nonmetropolitan counties. Petee and Kowalski (1993) found that residential mobility, racial heterogeneity, and percent single parents are all positively associated with robbery and assault in rural counties. Finally, Lee, Maume, and Ousey (2003) found that residential mobility and percent divorced are positively associated with homicide in rural communities. Thus, the positive association of crime with residential mobility, ethnic heterogene- ity, and family disruption is largely consistent in the empirical literature on social disorganization and crime in rural areas.

The similar rural and urban findings are not surprising. Rapid population change is often associated with social problems in communities of all sizes, as are ethnic diversity and percent of single-parent households (Bacigalupi and Freudenburg, 1983; Freudenburg and Jones, 1991; Guillaume and Wenson, 1980; Rephann, 1999). Major economic changes also generally cause disor- ganization in rural and urban settings. For example, the farm crisis of the 1980s had major economic and social ramifications on rural communities: population declined, retail and service sales decreased, and social institutions such as schools and churches suffered. Although rural residents did not move out of their communities at extremely high rates, those who remained began shopping in metropolitan areas where they could buy goods at cheaper prices (Ginder et al., 1985). Although not exactly mirroring each other, these eco- nomic and social problems from the farm crisis are similar to the economic and social problems attributed to de-industrialization in urban areas. Resi- dents who could afford to do so moved out of the central city, thus residents, businesses, and jobs went to the suburbs, taking tax revenue with them and leaving behind an increasing concentration of poverty and declining social institutions and infrastructure (Wilson, 1996). Given these similarities, it is reasonable to expect that the structural antecedents of social disorganization might be similarly associated with crime in both urban and rural settings.

Inconsistent Support for the Generalizability of the Theory. Despite these similarities, considerable inconsistencies remain in the relationships be- tween the structural antecedents of social disorganization and violence not only between the rural and urban literatures but also within the rural liter- ature. Osgood and Chambers (2000), Petee and Kowalski (1993), and Lee, Maume, and Ousey (2003) found that poverty, percent low income, and

982 Social Science Quarterly

poverty concentration were not significantly related to various measures of violence in rural counties. In contrast, Bouffard and Muftić (2006) found a strong and significant negative association between poverty and assault, rob- bery, and rape. On the other hand, Barnett and Mencken’s (2002) finding that resource disadvantage has a positive association with violent crime is consistent with the urban literature. Unlike the rest of the rural literature, Lee, Maume, and Ousey (2003) found that racial dissimilarity is negatively associated with homicide. Also noteworthy is Bouffard and Muftić’s (2006) finding that the diversity index was only significantly associated with as- sault when population density and proximity to urban areas were controlled. Finally, Weisheit and Wells’s exploratory analysis of homicide patterns in metropolitan and nonmetropolitan counties led them to conclude “that mod- els with very high predictive power in the largest metropolitan areas have much less success in accounting for homicide differences in the most rural areas” (2005:55).

Do Population Size and Density Condition the Effects of Social Disorga- nization on Crime? The very notion of testing the association between social disorganization and crime in rural areas suggests that the relationship may be conditioned by population size and/or density. After all, it is not uncommon to see in the literature arguments that smaller and more rural communities are more cohesive than larger and more urban communities. Further, Bouffard and Muftić’s (2006) finding that crime rates are much lower in rural counties with small towns than in rural counties with “large” towns, and that town size is a better predictor of crime than adjacency to a metropolitan area, suggests that population size and density may moderate the effects of social struc- ture on crime rates. Thus, even if the social disorganization-crime association operates in rural areas, it may still be weaker in smaller rural communities. Although comparisons are drawn between rural and urban cultural differ- ences, there are cultural differences between different rural communities as well. Given differences in population size and density, for example, it may be that large nonmetropolitan communities behave more like urban communi- ties (generally, and in any association between social structure and crime) than small nonmetropolitan communities. The close social scrutiny characteristic of rural life “clearly has implications for crime and justice in the rural setting, and for theories of crime and methods for studying it” (Weisheit and Wells, 1996:384).

Lending empirical support to the idea that population size and density are related to crime in rural communities is Osgood and Chambers’s (2000) finding that the size of the juvenile population (which they used as a proxy for population density) has a curvilinear relationship with juvenile arrests. Juvenile population was correlated with arrests only when the former was less than 4,000. Communities with juvenile populations of 4,000 or more did not exhibit a correlation between increased juvenile population size and arrests. In a footnote, the authors briefly mentioned testing for interactions between

Social Disorganization and Youth Violence in Rural Areas 983

the structural antecedents of social disorganization and population size, but did not discuss their findings in depth. Given that population size and density may differentially affect the nature of social relations in rural communities, our initial goal in the present study was to explore this possibility in greater detail. Specifically, our aim was first to estimate the same O-C model in our sample of nonmetropolitan counties and then to test the hypothesis that the strength of the association between social disorganization and crime is conditioned by population size and density. As we will show, however, our basic findings for the direct effects of measures of social disorganization on rural youth violence were very different from O-C, leading us to reconsider what we assumed to be true.

Data and Method

Sample

Following Osgood and Chambers (2000), we used nonmetropolitan coun- ties as our unit of analysis. Local government, and even social and economic activity, is often closely organized along county lines in nonmetropolitan set- tings, making them a common unit of analysis in political and social research on rural areas. Despite the convenience of using counties, however, there are potential drawbacks. As discussed above, in the United States social disorga- nization theory was developed mainly to explain crime rates in communities and neighborhoods. Even counties with very small populations often contain multiple communities, and thus the values assigned to counties on our inde- pendent and dependent variables will in actuality vary within our units. Yet the same limitation can easily be true of studies that use urban Census tracts as their unit of analysis. Further, Land, McCall, and Cohen’s (1990) research provides evidence that the structural covariates of violence rates are consistent across varying levels of aggregation.

In our cross-sectional study, we employed the 106 Missouri counties with a population of less than 100,000. Total populations ranged in size from 2,382 to 93,807, which is consistent with Osgood and Chambers (2000), whose counties ranged in size from 560 to 98,000 residents.

Data

Dependent Variable. Although Osgood and Chambers used arrest data from the Uniform Crime Reports (UCR) to measure rural youth violence, they were candid in their assessment of the threats to the validity of such data in rural jurisdictions. They stated that “there have been no studies of the validity of arrest statistics in rural jurisdictions,” that “recent reviews of research on rural crime have not addressed this topic,” and that “direct assessment of the

984 Social Science Quarterly

validity of this measure for communities awaits further research” (2000:92). Following the publication of the O-C article, the validity of county-level UCR data came under scrutiny. Most notably, Maltz and Targonski (2002, 2003) documented a number of serious errors in these data, and concluded that “county-level crime data cannot be used with any degree of confidence” (Maltz and Targonski, 2002:316). Lott and Whitley (2003) also noted that the largest errors in county-level UCR data are in counties with small populations. Further, though we reasonably expect violent crimes to be better measured than less serious offenses, Wiersema, Loftin, and McDowall’s (2000) study of county-level homicide data led them to conclude that even homicide estimates from smaller counties (specifically, counties with fewer than 100,000 residents) may be unreliable.

Given these limitations of the UCR crime and arrest data in rural coun- ties, we employed hospital data as our measure of serious assaultive violence among adolescents and young adults. We measured violent victimization as all outpatient hospitalizations for injuries coded as assaults for adolescents aged 10–17 and adolescents and young adults aged 15–24. O-C used adolescents aged 11–17 in their analysis, but the closest disaggregated age group from the hospital data was those aged 10–17. We also employed the 15–24 age group due to its high levels of victimization and offending and in order to see if our findings were sensitive to the age group under study. Outpatients are those who receive hospital care (including emergency room treatment) but are not admitted to the hospital for overnight or prolonged care. Although we discuss this at greater length in the “Discussion” section below, we mention for now that hospital records are a valid source of information about serious violent victimization rates since most people who are seriously injured seek medical treatment. Further, unlike official crime statistics, these counts are not dependent on whether the victimization was reported to the police or an arrest was made. For example, data from the National Crime Victimization Survey indicate that in 2000 only 48 percent of violent victimizations were reported to the police (Rennison, 2001), and only about 55 percent of all aggravated assaults reported to the police in Missouri in 2000 were cleared (Doyle and Johnson, n.d.).

Every hospital in Missouri is required to report inpatient and outpatient data to the Missouri Department of Health and Senior Services (DHSS). A Patient Abstract System is kept by Missouri’s DHSS and includes information about each visit, including patient age, sex, race, and county of residence.1

Injuries from assaults are coded by hospital staff in accordance with the World Health Organization’s International Classification of Diseases version 10 (ICD-10) external cause of injury codes X85–Y09 (see Appendix A for

1Unlike the police data available from the UCR, there is no single, centralized, national database of county-level hospital data. Each state has its own reporting system, sometimes run by state health departments and sometimes by state hospital associations. Missouri was chosen because of (1) its large number of nonmetropolitan counties, (2) mandatory reporting by all hospitals, and (3) the relative ease of data accessibility.

Social Disorganization and Youth Violence in Rural Areas 985

TABLE 1

Descriptive Statistics for Rural Counties in Missouri

Mean SD

Assault victimization rate, 10–17 369.5 202.3 Assault victimization rate, 15–24 778.6 365.7 Residential instability 0.43 0.06 Ethnic heterogeneity 0.11 0.08 Female-headed households 0.18 0.04 Poverty rate 0.14 0.05 Unemployment rate 0.03 0.01 Adjacent to metro area 0.57 0.50 Population at risk, 10–17 2740.9 2225.3 Log population at risk, 10–17 7.64 0.75 Population at risk, 15–24 3189.9 2843.6 Log population at risk, 15–24 7.73 0.84

a description of the codes). For the purposes of this study, counts of violent victimizations for those aged 10–17 and 15–24 were created by summing the counts for all types of assault injuries. To smooth the data and reduce the influence of random variation in any one year, which is especially important to do for rare events in small populations, counts for the years 1999–2001 were summed. The counts were based on the victim’s county of residence, which in some cases was different from where the victim presented to the emergency room. That is, a victim may have traveled to another county and been victimized there but the victimization count is recorded for the county of residence. Of course, the victim may also have been victimized in his or her home county and traveled to a nearby county for medical attention. Table 1 provides the descriptive statistics for all dependent and independent variables.

Explanatory Variables. Due to our desire to replicate as closely as possible O-C’s analysis of social disorganization and crime in rural communities, and consistent with prior studies of social disorganization theory more generally, our measures of social disorganization included residential instability, ethnic heterogeneity, family disruption, poverty, and population density. All data on these and the control variables were obtained from the 2000 Census (Missouri Census Data Center, 2008). Residential instability was measured as the pro- portion of households occupied by people who moved from another dwelling in the preceding five years. Ethnic heterogeneity was measured as a diversity index. The index was calculated as 1−(�pi

2), with pi being the proportion of households in a specific ethnic group. This index reflects the probability of two randomly chosen individuals being from different ethnic groups. Pos- sible values range from 0 to 1, with lower scores representing lower ethnic

986 Social Science Quarterly

heterogeneity.2 Family disruption was measured as the ratio of female-headed households to all households with children. Osgood and Chambers (2000) argued that the responsibility of monitoring youth falls on adults (primarily women) who have children living at home. Thus, the ratio of female-headed households to all households with children is thought to be more closely related to youth delinquency and victimization than the proportion of female-headed households to all households. Poverty was measured as the percent of resi- dents living under the poverty level. Finally, consistent with O-C, we used the population at risk for violent victimization (in our case, county populations of those aged 10–17 and 15–24) as a proxy for population density.3

Our control variables were the same as those used by O-C. Specifically, we included in our models controls for the unemployment rate and for each county’s proximity to a metropolitan county using a dummy-coded variable from the Beale Code (U.S. Department of Agriculture, 2003), with 1 being adjacent to a metropolitan statistical area and 0 being not adjacent.

Method

The distribution of all the continuous variables in our model were positively skewed, with skew statistics greater than twice their standard errors for each. In an attempt to normalize the distributions of these variables, we took the natural logarithm of each. Subsequent examination of the distributions of the transformed variables revealed substantially decreased skews, and thus the logarithmically transformed independent variables were employed in model estimation.

Consistent with Osgood and Chambers (2000), we used the negative bino- mial estimator. One would normally use victimization rates as the dependent variable and employ ordinary least squares regression. In our analysis, however, we have rare events (violent victimization) in small populations. This results in an increasingly less precise rate and an increasingly skewed distribution. Poisson models are well suited to these conditions. A normal Poisson model, however, assumes that the model explains all the variation in victimization rates in the sample. Another member of the Poisson family, the negative bi- nomial model, relaxes this constraint by adding a term for residual variance in the underlying victimization rates. It was the Osgood and Chambers (2000; see also Osgood, 2000) study, in fact, that helped popularize the use of the

2With one exception, the measures of all independent variables were exactly the same as in Osgood and Chambers (2000). The single minor exception is that O-C included only white and nonwhite in their measure of ethnic heterogeneity, whereas we included all ethnic groups in calculating this index.

3As with Osgood and Chambers (2000), in our sample we found the size of the populations at risk to be highly correlated with population density (r = 0.90 and r = 0.86 for density’s correlation with population 10–17 and 15–24, respectively), and thus used the population size for these age groups as proxies for density in our models.

Social Disorganization and Youth Violence in Rural Areas 987

negative binomial estimator with macro-level criminological data. A negative binomial model employs counts as the dependent variable, but since we are interested in rates, we standardized the model by adding the natural logarithm of the size of the population at risk (10–17- and 15–24-year-olds in our two models, respectively) as the offset variable (Cameron and Trivedi, 1998; Gard- ner, Mulvey, and Shaw, 1995). We used STATA/SE 10 to carry out our model estimation.

Results

Table 2 shows the results when assaults are regressed on the explanatory variables. The first column shows the results for assault victimization among 10–17-year-olds, the second column the results for 15–24-year-olds. As with Osgood and Chambers, we included multiple measures for the population at risk (log, log squared, and log cubed) to control for any variation in victim- ization rates by size of the population at risk (which, as discussed above and similarly to O-C, we also used as a proxy for density given its high correlation with size of population at risk). The coefficients for population size at risk were estimated rather than fixed (again, this is in concordance with the approach taken by O-C).

The results for both 10–17- and 15–24-year-olds are similar. Specifically, only one of the five measures of social disorganization was associated with violent victimization among adolescents and young adults. The nonsignificant finding for poverty among 10–17-year-olds, and even the significant negative association for poverty with assault victimization among 15–24-year-olds is not especially surprising. These results for poverty mirror those from Osgood and Chambers (2000) and are consistent with other evidence that shows poverty to be differentially associated with a host of other factors in rural relative to urban areas.

The findings for the other social disorganization variables were unexpected. Unlike O-C—who found residential instability, ethnic heterogeneity, fam- ily disruption, and the proxy for density to be positively and significantly associated with arrest rates for assault (both aggravated and simple) among youth—we found only family disruption to be positively and significantly associated with assault victimization. The inferences drawn by Osgood and Chambers led them to conclude that their “[f]indings support the gener- ality of social disorganization theory” to youth violence in rural settings (2000:81). The inferences drawn from our results, on the other hand, sug- gest that the association between traditional social disorganization variables and youth violence may not generalize to rural areas. Thus, social disorga- nization may not be as robust an explanation for assaultive violence in rural areas as many have come to believe it is given the findings from Osgood and Chambers.

988 Social Science Quarterly

TABLE 2

Serious Assault Victimizations Regressed on Explanatory Variables (N = 106)

Serious Assault Serious Assault Victimizations, 10–17 Victimizations, 15–24

Ln residential instability b −0.512 −0.798 s.e. 0.438 0.388 p 0.243 0.040

Ln diversity b −0.076 −0.050 s.e. 0.094 0.077 p 0.417 0.515

Ln female-headed households b 1.742 1.887 s.e. 0.331 0.270 p <0.001 <0.001

Ln poverty b −0.239 −0.276 s.e. 0.174 0.139 p 0.170 0.047

Ln unemployment b −0.076 −0.082 s.e. 0.186 0.153 p 0.685 0.592

Adjacent to metro area b −0.075 0.079 s.e. 0.094 0.078 p 0.422 0.313

Population at Risk Log

b 0.228 0.147 s.e. 0.128 0.100 p 0.074 0.141

Log squared b −0.071 −0.042 s.e. 0.071 0.045 p 0.315 0.353

Log cubed b −0.012 −0.044 s.e. 0.066 0.042 p 0.846 0.299

Constant b −8.821 −6.735 s.e. 1.862 1.531 p <0.001 <0.001

NOTE: In both models, p < 0.001 for α (i.e., the unexplained residual variance beyond that expected from a simple Poisson process).

Social Disorganization and Youth Violence in Rural Areas 989

Regression Diagnostics And Sensitivity Analysis

Several common regression diagnostics were considered. First, examination of the leverage values suggested Pulaski County to be a potential outlier on the x-axis. Similarly, the values on the studentized deleted residuals revealed Reynolds and St. Francois Counties to be potential outliers on the y-axis. The dfBetas also showed Reynolds and Pulaski Counties to be possible influential cases on various regression coefficients. When these cases were excluded in- dividually and in concert, there were no changes to the inferences described above. Regression diagnostics showed no problems in these analyses associated with heteroskedastic error variance or nonnormal distribution of errors. We undertook multiple checks to determine if multicollinearity might be playing a role. Using the negative binomial estimator does not allow for the computa- tion of variance inflation factors (VIFs). We reestimated the model using OLS regression, however, and examined the VIFs generated by this approach. None were close to critical values. Looking at the correlation matrix in Appendix B, we can see also that the strongest correlation among the explanatory variables is 0.63, and there are only a few other correlations at or above 0.50. These cor- relations are similar to those found in other important aggregate-level studies of social disorganization in urban areas and are very similar to those recorded by O-C.

To gauge the sensitivity of our results to various alternatives, several further models were estimated. First, models were reestimated with multiple sub- samples based on county population size, which resulted in no changes to any of the inferences drawn. This is largely because the majority of county populations in Missouri are small, and thus these samples remain largely sim- ilar in terms of counties included. For example, of the 106 counties with a population of fewer than 100,000 residents, over 90 percent have fewer than 50,000 residents, nearly three-quarters have fewer than 25,000 residents, and about half have fewer than 15,000 residents. Second, we estimated models with a sample that was selected based on exactly the same definition used by Osgood and Chambers (i.e., not only fewer than 100,000 residents, but also not containing a city of 50,000 or more nor having 50 percent of their population residing in a metropolitan area of 100,000 or more). Again, this resulted in no changes to the inferences drawn, and again this was because the composition of this restricted sample was essentially the same as the sample we initially used.

Finally, we used GeoDa (Anselin, 2003) to discover and account for any spatial autocorrelation in the models by examining the residuals from the negative binomial estimates for both 10–17- and 15–24-year-olds. Opera- tionalized with a Euclidean spatial weights matrix and a global Moran’s I statistic (Moran, 1950), the residuals for the 15–24 cohort displayed no spa- tial autocorrelation (I = 0.0399; p = 0.2148), while the residuals for the 10–17 cohort were significant and positive (I = 0.1591; p = 0.0048). A Moran’s I of this size is modest and unlikely to have an impact on the inferences drawn. To

990 Social Science Quarterly

be sure, however, we decomposed the global statistic for the 10–17-year-olds model using a local test for spatial association (Anselin, 1995). The results indicated that five geographically contiguous counties (Crawford, Iron, Madi- son, St. Francois, and Washington) displayed significant levels of local spatial association, resulting in systematic underprediction of counts of violent vic- timization in these counties by the negative binomial model. The 10–17-year- olds model was reestimated with the inclusion of a dummy variable (coded 1 for these five counties; 0 otherwise). The dummy variable was significant, but the inferences for the other explanatory variables did not change, thus revealing little impact of spatial autocorrelation on the results provided in Table 2.

Discussion

Based mostly on the work of Osgood and Chambers (2000; see also Bouffard and Muftić, 2006; Petee and Kowalski, 1993), many have concluded that social disorganization theory generalizes to rural areas. We assumed this as well, and the initial goal of our study was to extend O-C’s research by testing the hypothesis that the strength of the association between social disorganization and rural youth violence is conditioned by community size and density even among rural communities.4 However, in our first step of estimating the same model as O-C before testing this hypothesis, we found very different results for the effects of social disorganization on youth violence in rural areas. That is, while O-C found positive and significant effects on rural youth violence for four of the five elements of social disorganization—residential instability, ethnic heterogeneity, female-headed households, and population at risk (as a proxy for density)—we found only one of these (female-headed households) to be positively and significantly associated with rural youth violence.

The Osgood and Chambers (2000) study has been highly cited and is deserving of the attention it received. It not only appeared to extend the gen- eralizeability of a popular structural theory to rural areas, but helped introduce innovative methods (i.e., Poisson-based regression, and especially the nega- tive binomial estimator; see also Osgood, 2000) to macro-level criminological studies. We see no reason to doubt the theoretical care or methodological rigor in O-C’s study, and in fact our original desire was to build on the foun- dation of their findings. Nevertheless, that our results largely fail to replicate theirs leads us to believe that the strong conclusions drawn by many about

4Osgood and Chambers also appear to have tested this hypothesis, but they mention it only briefly. Specifically, in their footnote 7 they state that they “tested for potential interactions in which [the association between social disorganization and youth violence] would vary . . . with population size. Only chance levels of interactive relationships emerged” (Osgood and Chambers, 2000:99).

Social Disorganization and Youth Violence in Rural Areas 991

the applicability of social disorganization to rural youth violence rates may have been premature. In the discussion that follows, we first provide possible methodological reasons for these inconsistent results. We then draw from the theoretical and empirical literature on rural communities to suggest why so- cial disorganization may operate differently in its association with violence in these areas relative to urban communities.

Methodological Explanations

There are a few potential methodological explanations for why Osgood and Chambers’s (2000) findings do not replicate in our study. These include a different measure of the dependent variable, a different sample, spatial autocorrelation, and model misspecification (in both studies).5

Measurement of the Dependent Variable. We measured the dependent variable, youth violence, differently than O-C. The latter measured youth vio- lent offending using county arrest rates obtained from the UCR. We measured youth violent victimization using data from hospital records. Although it could be true that social disorganization theory might be applicable to adolescent of- fending but not victimization, we believe this unlikely. First, O-C’s results for simple and aggravated assault were similar to each other, and the latter offense coincides closely with our measure of violent crime victims who present to emergency rooms. Second, most nonfamily aggravated assaults against youth are committed by peers within three years the age of the victim (Boney-McCoy and Finkelhor, 1995), so our victimization and O-C’s arrest variables should be closely related. That our results for those aged 15–24 are the same as our results for those aged 10–17 provides us with further assurance that the victim- ization data for 10–17-year-olds are not unique. Third, as discussed elsewhere in our article, several scholars have made reasonable theoretical connections between community disorganization and victimization rates. Finally, several key studies of social disorganization theory, including Sampson and Groves (1989), used victimization data.

Another issue related to measurement of the dependent variable is the va- lidity of the UCR data for rural counties. As discussed earlier, Osgood and Chambers were candid in their statements concerning UCR arrest rates for violence in rural areas. At the time of their study, there had been little system- atic assessment of these data. Indeed, O-C even expressed some reservations, noting that systematic error might be likely due to things such as informal arrest practices in rural areas.6 Following the publication of their study, how- ever, there have been multiple indications that county-level UCR data are

5In a companion piece to the present article, we systematically address and test each of these potential methodological explanations.

6There are also theoretical reasons to believe that reporting and recording of assaults in rural areas may vary systematically. In smaller areas, for example, social networks and relationships

992 Social Science Quarterly

unreliable, especially for smaller counties (Lott and Whitley, 2003; Maltz and Targonski, 2002, 2003; see Wiersema, Loftin, and McDowall, 2000, for a discussion of the potential problems associated with homicide data in small counties). These reservations about the county-level UCR data led us to employ victimization data from hospital records.

Hospital data have their own set of strengths and weaknesses. Although rarely used by criminologists, these data are commonly used by epidemiologists and others interested in injury and violence. This includes studies examining topics familiar to criminologists, such as the county-level association between racial segregation and violence rates (Fabio et al., 2009) and ecological models revealing the association between alcohol outlets and assaults (Gruenewald et al., 2006). The main concern with such data is underenumeration due to victims of violence who do not present to an emergency room (Rosenfeld, 2007). This limitation is attenuated in our study because we restricted our analyses to serious assaults, which are more likely to show up in an emergency room and create a medical record. Importantly, hospital data “provide infor- mation on assault victims that may not be reflected in the UCR” (Rosenfeld, 2007:827), and are recognized as providing similar estimates of serious assault to those from the NCVS (Cohen and Lynch, 2007), which is known to pro- duce higher assault estimates relative to the UCR. Other potential sources of error with the hospital data include the failure to assign an E-code or misclassi- fication of the coding. Yet recent analyses reveal essentially universal coverage, with greater than 90 percent agreement when the initial injury records are sub- sequently reviewed by experts for specificity, sensitivity, and intent (LeMier, Cummings, and West, 2001; see also Gruenewald et al., 2006).

Sample Composition. A second possible methodological explanation for our conflicting results centers on the size and composition of the two samples. Although Osgood and Chambers (2000) used nonmetropolitan counties in four states (Florida, Georgia, South Carolina, and Nebraska), we used non- metropolitan counties in one state, Missouri. However, while we used one- fourth the number of states, our sample size was 40 percent that of O-C (106 counties compared to 264 counties) and larger than many other macro-level studies of social disorganization.

Given the hypothesis that social disorganization theory should apply to nonmetropolitan communities in general, there are no obvious and compelling reasons why the state(s) included should yield different results. Nevertheless, we cannot rule out that there might be something unique about Missouri causing the conflicting results. Of course, the same could be said of the Osgood and Chambers (2000) sample. Still, one-third of the nonmetropolitan counties

may reduce the victim’s willingness to report and the officer’s willingness to record an incident, both of which would bias estimates in favor of the social disorganization hypothesis. This could also vary by the size of the community, which is important given the relatively broad distribution of county population size included as “rural” in our and the Osgood and Chambers (2000) study.

Social Disorganization and Youth Violence in Rural Areas 993

in O-C’s sample came from a neighboring state of Missouri, and Bouffard and Muftić’s (2006) sample consisted of rural counties in upper midwestern states (North Dakota, South Dakota, Minnesota, and Wisconsin), suggesting that nonmetropolitan counties in Missouri should not be unique.

Spatial Autocorrelation. None of the prior studies of social disorganiza- tion and crime in rural areas cited here appears to have accounted for spatial autocorrelation. If present, spatial autocorrelation underestimates the standard errors of the estimates, thus increasing the possibility of finding a significant effect when none exists. Recent evidence reveals that assaults (Bell, Schuur- man, and Hameed, 2008) and homicide (Baller et al., 2001; Mencken and Barnett, 1999) are spatially clustered, and the work of Baller et al. (2001) shows that the effects of social structure on serious assaultive violence can vary by geographic location. Unlike earlier analyses of social disorganization and crime in rural counties, we controlled for spatial autocorrelation.

We were not overly concerned that the differences in our results relative to Osgood and Chambers were due to spatial effects. This is because (1) the inconsistencies between our and O-C’s results existed before we controlled for spatial autocorrelation, (2) spatial autocorrelation was present in only one of our models, and (3) this autocorrelation did not have an effect on the inferences drawn. Nevertheless, given the spatial patterning in assaultive violence and the potential of spatial autocorrelation to influence prediction errors, it would be wise for future studies of this type to test for these effects and to model them if they exist.

Model Misspecification. The final methodological issue, model misspeci- fication, is also related to theory. As detailed in our review of studies of social disorganization in rural communities, the theory has not been fully supported empirically. Although findings for residential instability and family disrup- tion are similar, diversity and poverty findings are conflicting (Barnett and Mencken, 2002; Bouffard and Muftić, 2006; Lee, Maume, and Ousey, 2003; Osgood and Chambers, 2000; Petee and Kowalski, 1993). These inconsisten- cies, together with our own very different findings, may result from model misspecification. Sampson and Groves (1989) argued that model misspeci- fication was the cause of similar inconsistent findings in research on social disorganization and crime in urban areas. At that time with urban studies, and today with rural studies, scholars were carrying out only partial tests of the social disorganization model. Specifically, they were examining only the direct effects on crime of the structural antecedents of social disorganization, as shown in Figure 1. As Sampson and Groves (1989) pointed out, the the- oretical model is more nuanced. While it may be true that these structural factors have direct effects on crime, the theory argues their more important effect is to reduce social integration and organization within communities.

994 Social Science Quarterly

FIGURE 1

Partial Model of Social Disorganization and Youth Violence as Tested in This Study and by Osgood and Chambers (2000)

Mobility

Ethnic heterogeneity

Family disruption

Poverty

Offending and victimization rates

Population density

What should be tested is a model similar to that shown in Figure 2, where measures of social disorganization mediate the effects of the structural factors on offending and victimization within communities. Although subsequent empirical tests on data from urban areas have provided substantial evidence supporting this mediating model, the problem of model misspecification con- tinues to plague social disorganization research in rural areas, largely due to a lack of contextual data.

Like O-C and others before us, we were unable to test directly the ef- fects of community-level social disorganization. Bursik (1988) points out that the original theory does not assume a direct relationship between social structure and delinquency. He said that the prevalence of these structural factors does not necessitate social disorganization and that “the degree to which these ecological processes are associated with the ability of a commu- nity to regulate itself is an empirical question” (Bursik, 1988:531). If rural communities really are as cohesive as Freudenburg (1986), Weisheit, Falcone, and Wells (2006), Weisheit and Wells (1996), and other rural scholars argue, then these structural factors might not be expected to influence levels of inte- gration and organization similarly in rural communities. It may be, though, that the stronger community integration and cohesion moderates the effects of social structure on rural crime.

This model misspecification in studies of the effects of social disorganization on crime in rural areas presents a serious challenge to scholars. The next steps

Social Disorganization and Youth Violence in Rural Areas 995

FIGURE 2

Properly Specified Social Disorganization Model

Mobility

Ethnic heterogeneity

Family disruption

Poverty

Offending and victimization rates

Population density

Social disorganization

Poor community integration

Disrupted informal controls

Anonymity

Superficial relationships between families

Low participation in community orgs

in this line of research must (1) use existing or new theoretically appropriate measures of social integration and community organization in rural areas, (2) test to see if these constructs vary between rural communities and, if so, if they are influenced by the levels of the structural antecedents of social disor- ganization, and (3) determine if these same constructs mediate or moderate the effect of social structure on criminal offending and victimization in rural areas.

Theoeretical Explanations

There are also potential theoretical explanations for the differences between our and O-C’s findings. Perhaps, contrary to O-C’s findings, the effect of social disorganization on youth violence in rural areas is not truly consistent. The association between social disorganization and crime simply may not generalize to rural areas. The theoretical and empirical literature provides a number of plausible reasons for this.

The nature of social structure and its impact on social relations in rural com- munities may be different than in urban communities. For example, there is evidence that collective efficacy moderates the effects of social structure on crime in urban areas (Sampson, Raudenbush, and Earls, 1997). Similarly, it could be that any potential effects of the structural antecedents of social dis- organization on violence are nearly completely moderated by the higher levels of community cohesion in rural areas. If so, however, this would mean that levels of cohesion in most rural communities (Freudenburg, 1986; Weisheit, Falcone, and Wells, 2006; Weisheit and Wells, 1996) would need to be much higher than in urban areas in order to completely offset the effects of social

996 Social Science Quarterly

structure. This set of related possibilities was discussed above in the subsection on model misspecification.

Differences in the nature of social relations in rural communities relative to urban communities are also possible. In a critique of the applicability of social disorganization theory to rural crime, Donnermeyer (2006) points out that some forms of community organization in rural areas may actually facilitate crime. He and his colleagues (Barclay, Donnermeyer, and Jobes, 2004; Donnermeyer and Barclay, 2005) found that in rural Australia, livestock theft is often not reported or police do not follow up about complaints because of a fear of disrupting community cohesion. Likewise, Donnermeyer (2006) summarizes DeKeseredy et al.’s (2006) findings that community organization in rural areas can actually support violence against women. Men in some of these communities share stories and techniques about violence with each other in public places, and support for this type of violence has become a part of the mainstream culture in these areas. Thus, rather than community organization inhibiting certain types of crime in rural communities, it might actually facilitate it.

It may also be that social disorganization has an influence on crime rates in rural communities, but that the structural antecedents of disorganization commonly used in urban studies may be less applicable. Traditional measures like mobility, ethnic heterogeneity, and population density may not capture disorganization in rural communities. Instead, scholars may wish to consider measuring the decline in small and local businesses—like drug, grocery, and hardware stores—and the rise of big-box stores, which can wreak havoc on local businesses and downtown areas. This is not dissimilar to the effects of farm job loss in the rural United States during the 1980s, accounts of which cannot be read without recognizing elements of social disorganization and its debilitating impact on these communities, including crime. After all, while rural communities are certainly subject to larger regional, national, and global influences, they must be understood in terms of their own social organization (Weisheit, Falcone, and Wells, 2006), which will not always mirror urban organizational patterns.

Whatever the reason or mechanism, it may be that social disorganization theory as an explanation for the distribution of violence rates does not gener- alize to rural areas. Based on our results and those of others, we are not ready to come to this conclusion. On the other hand, we do believe that many have drawn strong conclusions about the generalizeability of social disorganization to rural areas based largely on the results of a single study, when in fact further research is required to answer the theoretical questions under scrutiny.

Conclusion

Our initial aim in examining social disorganization theory in nonmetropoli- tan areas was to extend Osgood and Chambers’s (2000) research by testing

Social Disorganization and Youth Violence in Rural Areas 997

the hypothesis that the strength of the effect of social disorganization on rural youth violence is moderated by community size and density. We were unable to carry out this innovation, however, because O-C’s results largely failed to replicate in our sample.

Replication is a hallmark of science. This should be true especially with macro-level studies, where units of analysis and measures of explanatory vari- ables are often based more on convenience and availability than on theory. We see no reason to criticize the rigor of Osgood and Chambers’s study, nor, therefore, the conclusions they draw. We were convinced by their results and hoped to extend their research. Our own findings, however, remind us that the results of a single study should not be accepted as the definitive answer to a scientific question, as we present conflicting evidence of the generalizeability of the association between social disorganization and youth violence to rural settings. We have suggested potential substantive and methodological reasons for these inconsistent results, and we invite others to carry out further studies that will aid us in answering these important questions as they relate to social disorganization and to youth violence. It should not simply be the case that we rely on findings about crime in urban areas to direct our efforts in rural communities; instead, research on crime in rural communities can also aid us in modifying and updating what we know about social organization and crime more generally.

REFERENCES

Anselin, L. 1995. “Local Indicators of Spatial Association: LISA.” Geographical Analysis 27:93– 115.

———. 2003. GeoDa 0.9 User’s Guide. Urbana-Champaign, IL: Spatial Analysis Laboratory, University of Illinois.

Bacigalupi, L. M., and W. R. Freudenberg. 1983. “Increased Mental Health Caseloads in an Energy Boomtown.” Administration and Policy in Mental Health and Mental Health Services Research 10:306–22.

Baller, R. D., L. Anselin, S. F. Messner, G. Deane, and D. F. Hawkins. 2001. “Structural Covari- ates of U.S. County Homicide Rates: Incorporating Spatial Effects.” Criminology 39:561–90.

Barclay, E., J. F. Donnermeyer, and P. C. Jobes. 2004. “The Dark Side of Gemeinschaft: Crimi- nality Within Rural Communities.” Crime Prevention and Community Safety: An International Journal 6:7–22.

Barnett, C., and F. C. Mencken. 2002. “Social Disorganization Theory and the Contextual Nature of Crime in Nonmetropolitan Counties.” Rural Sociology 67:372–93.

Bell, N., N. Schuurman, and S. M. Hameed. 2008. “Are Injuries Spatially Related? Join-Count Spatial Autocorrelation for Small-Area Injury Analysis.” Injury Prevention 14:346–53.

Bellair, P. E. 1997. “Social Interaction and Community Crime: Examining the Importance of Neighbor Networks.” Criminology 35:677–704.

Boney-McCoy, S., and D. Finkelhor. 1995. “Psychosocial Sequelae of Violent Victimization in a National Youth Sample.” Journal of Consulting and Clinic Psychology 63:726–36.

998 Social Science Quarterly

Bouffard, L. A., and L. R. Muftić. 2006. “The ‘Rural Mystic’: Social Disorganization and Violence Beyond Urban Communities.” Western Criminology Review 7:56–66.

Bursik, R. J. 1988. “Social Disorganization and Theories of Crime and Delinquency: Problems and Prospects.” Criminology 26:519–51.

Cameron, A. C., and P. K. Trivedi. 1998. Regression Analysis of Count Data. New York: Cambridge University Press.

Cancino, J. M. 2003. “Breaking from Orthodoxy: The Effects of Social Disorganization on Perceived Burglary in Nonmetropolitan Communities.” American Journal of Criminal Justice 28:125–45.

Cohen, J., and J. P. Lynch. 2007. “Exploring Differences in Estimates of Visits to Emergency Rooms for Injuries from Assaults Using the NCVS and NHAMCS.” Pp. 183–222 in J. P. Lynch and L. A. Addington, eds., Understanding Crime Statistics: Revisiting the Divergence of the NCVS and UCR. New York: Cambridge University Press.

DeKeseredy, W. S., M. D. Schwartz, D. Fagin, and M. Hall. 2006. “Separation/Divorce Sexual Assault: The Contribution of Male Support.” Feminist Criminology 1:1–23.

Donnermeyer, J. F. 2006. “Rural Crime: Roots and Restoration.” Paper presented at the International Conference on Rural Crime. Available at 〈http://www.ruralfutures. une.edu.au/rurcrime/resources/proceedings/Donnermeyer.pdf〉. Donnermeyer, J. F., and E. Barclay. 2005. “The Policing of Farm Crime.” Police Practice and Research 6:3–17.

Doyle, J., and T. Johnson. n.d. Crime in Missouri, 2001. Jefferson City, MO: Mis- souri State Highway Patrol. Available at 〈http://www.mshp.dps.missouri.gov/MSHPWeb/ SAC/pdf/2001CrimeInMO.pdf〉. Elliott, D. S., W. J. Wilson, D. Huizinga, R. J. Sampson, A. Elliott, and B. Rankin. 1996. “The Effects of Neighborhood Disadvantage on Adolescent Development.” Journal of Research in Crime and Delinquency 33:389–426.

Fabio, A., E. K. Sauber-Schatz, K. E. Barbour, and W. Li. 2009. “The Association Between County-Level Injury Rates and Racial Segregation Revisited: A Multilevel Analysis.” American Journal of Public Health 99:748–53.

Felson, M., and L. E. Cohen. 1980. “Human Ecology and Crime: A Routine Activity Ap- proach.” Human Ecology 8:389–406.

Freudenburg, W. R. 1986. “The Density of Acquaintanceship: An Overlooked Variable in Community Research.” American Journal of Sociology 92:27–43.

Freudenburg, W. R., and R. E. Jones. 1991. “Criminal Behavior and Rapid Community Growth: Examining the Evidence.” Rural Sociology 56:619–45.

Gardner, W., E. P. Mulvey, and E. C. Shaw. 1995. “Regression Analysis of Counts and Rates: Poisson, Overdispersed Poisson, and Negative Binomial.” Psychological Bulletin 118:392–405.

Ginder, R. G., K. E. Stone, and D. Otto. 1985. “Impact of the Farm Financial Crisis on Agribusiness Firms and Rural Communities.” American Journal of Agricultural Economics 67:1184–90.

Gruenewald, P. J., B. Freisthler, L. Remer, E. A. LaScala, and A. Treno. 2006. “Ecological Models of Alcohol Outlets and Violent Assaults: Crime Potentials and Geospatial Analysis.” Addiction 101:666–77.

Guillaume, C. L., and S. R. Wenson. 1980. “Child Abuse in Impact Communities.” Pp. 85–98 in J. Davenport III and J. A. Davenport, eds., The Boom Town: Problems and Promises of the Energy Vortex. Laramie, WY: Department of Social Work, University of Wyoming.

Social Disorganization and Youth Violence in Rural Areas 999

Kornhauser, R. R. 1978. Social Sources of Delinquency. Chicago, IL: University of Chicago Press.

Land, K. C., P. L. McCall, and L. E. Cohen. 1990. “Structural Covariates of Homicide Rates: Are There Any Invariances Across Time and Social Space?” American Journal of Sociology 95:922–63.

Lee, M. R., and J. P. Bartkowski. 2004. “Love Thy Neighbor? Moral Communities, Civic Engagement, and Juvenile Homicide in Rural Areas.” Social Forces 82:1001–35.

Lee, M. R., M. O. Maume, and G. C. Ousey. 2003. “Social Isolation and Lethal Violence Across the Metro/Nonmetro Divide: The Effects of Socioeconomic Disadvantage and Poverty Concentration on Homicide.” Rural Sociology 68:107–31.

Lee, M. R., and G. C. Ousey. 2001. “Size Matters: Examining the Link Between Small Manu- facturing, Socioeconomic Deprivation, and Crime Rates in Non-Metropolitan Communities.” Sociological Quarterly 42:581–602.

LeMier, M., P. Cummings, and R. A. West. 2001. “Accuracy of External Cause of Injury Codes Reported in Washington State Hospital Discharge Records.” Injury Prevention 7:334–38.

Lott, J. R. Jr., and J. Whitley. 2003. “Measurement Error in County-Level UCR Data.” Journal of Quantitative Criminology 19:185–98.

Macmillan, R., and J. Hagan. 2004. “Violence in the Transition to Adulthood: Adolescent Victimization, Education, and Socioeconomic Attainment in Later Life.” Journal of Research on Adolescence 14:127–58.

Maltz, M. D., and J. Targonski. 2002. “A Note on the Use of County-Level UCR Data.” Journal of Quantitative Criminology 18:297–318.

———. 2003. “Measurement and Other Errors in County-Level UCR Data: A Reply to Lott and Whitley.” Journal of Quantitative Criminology 19:199–206.

Markowitz, F. E., P. E. Bellair, A. E. Liska, and J. Liu. 2001. “Extending Social Disorganization Theory: Modeling the Relationships Between Cohesion, Disorder, and Fear.” Criminology 39:293–319.

Menard, S. 2002. Short- and Long-Term Consequences of Adolescent Victimization. Washington, DC: Office of Juvenile Justice and Delinquency Prevention and the Centers for Disease Control and Prevention.

Mencken, F. C., and C. Barnett. 1999. “Murder, Non-Negligent Manslaughter, and Spatial Autocorrelation in Mid-South Counties.” Journal of Quantitative Criminology 15:407–22.

Missouri Census Data Center. 2008. MCDC Demographic Profile 3 Menu Page for Counties. Available at 〈http://mcdc2.missouri.edu/websas/dp3_2kmenus/mo/counties.html〉. Moran, P. A. P. 1950. “Notes on Continuous Stochastic Phenomena.” Biometrika 37:17–23.

Osgood, D. W. 2000. “Poisson-Based Regression Analysis of Aggregate Crime Rates.” Journal of Quantitative Criminology 16:21–43.

Osgood, D. W., and J. M. Chambers. 2000. “Social Disorganization Outside the Metropolis: An Analysis of Rural Youth Violence.” Criminology 38:81–115.

Petee, T. A., and G. S. Kowalski. 1993. “Modeling Rural Violent Crime Rates: A Test of Social Disorganization Theory.” Sociological Focus 26:237–89.

Reisig, M. D., and J. M. Cancino. 2004. “Incivilities in Nonmetropolitan Communities: The Effects of Structural Constraints, Social Conditions, and Crime.” Journal of Criminal Justice 32:15–29.

1000 Social Science Quarterly

Rennison, C. M. 2001. Criminal Victimization 2000: Changes 1999–2000 with Trends 1993– 2000. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

Rephann, T. J. 1999. “Links Between Rural Development and Crime.” Papers in Regional Science 78:365–86.

Rosenfeld, R. 2007. “Transfer the Uniform Crime Reporting Program from the FBI to the Bureau of Justice Statistics.” Criminology & Public Policy 6:825–33.

Sampson, R. J. 1985. “Neighborhood and Crime: The Structural Determinants of Personal Victimization.” Journal of Research in Crime and Delinquency 22:7–40.

———. 1997. “Collective Regulation of Adolescent Misbehavior: Validation Results from Eighty Chicago Neighborhoods.” Journal of Adolescent Research 12:227–44.

Sampson, R. J., and B. Groves. 1989. “Community Structure and Crime: Testing Social- Disorganization Theory.” American Journal of Sociology 94:774–802.

Sampson, R. J., S. W. Raudenbush, and F. Earls. 1997. “Neighborhoods and Violent Crime: A Multilevel Study of Collective Efficacy.” Science 15:918–24.

Shaw, C. R., and H. D. McKay. 1942. Juvenile Delinquency and Urban Areas: A Study of Rates of Delinquency in Relation to Differential Characteristics of Local Communities in American Cities. Chicago, IL: University of Chicago Press.

Stewart, E. A., C. J. Schreck, and R. Simons. 2006. “‘I Ain’t Gonna Let No One Disrespect Me’: Does the Code of the Street Reduce or Increase Violent Victimization Among African American Adolescents.” Journal of Research in Crime and Delinquency 43:427–58.

Tickamyer, A. R., and C. M. Duncan. 1990. “Poverty and Opportunity Structure in Rural America.” Annual Review of Sociology 16:67–86.

U.S. Department of Agriculture. 2003. 2003 Rural Urban Continuum Codes for MO. Available at 〈http://www.ers.usda.gov/Data/RuralUrbanContinuumCodes/2003/LookUp RUCC.asp?C=R&ST=MO〉. U.S. Surgeon General. 2001. Youth Violence: A Report of the Surgeon General. Washington, DC: U.S. Department of Health and Human Services.

Weisheit, R. A., D. N. Falcone, and L. E. Wells. 2006. Crime and Policing in Rural and Small-Town America, 3rd ed. Prospect Heights, IL: Waveland Press.

Weisheit, R. A., and L. E. Wells. 1996. “Rural Crime and Justice: Implications for Theory and Research.” Crime & Delinquency 42:379–97.

———. 2005. “Deadly Violence in the Heartland: Comparing Homicide Patterns in Non- metropolitan and Metropolitan Counties.” Homicide Studies 9:55–80.

Wells, L. E., and R. A. Weisheit. 2004. “Patterns of Rural and Urban Crimes: A County-Level Comparison.” Criminal Justice Review 29:1–22.

Wiersema, B., C. Loftin, and D. McDowall. 2000. “A Comparison of Supplementary Homicide Reports and National Vital Statistics System Homicide Estimates for U.S. Counties.” Homicide Studies 4:317–40.

Wilson, W. J. 1996. When Work Disappears: The World of the New Urban Poor. New York: Alfred A. Knopf.

World Health Organization. 2007. Tabular List of Inclusions and Four-Character Subcategories. Available at 〈http://www.who.int/classifications/apps/icd/icd10online〉.

Social Disorganization and Youth Violence in Rural Areas 1001

Appendix A: ICD-10 External Cause of Injury Classification Codes

Code Cause of Injury Description

X85 Assault by drugs, medicaments, and biological substances X86 Assault by corrosive substance X87 Assault by pesticides X88 Assault by gases and vapors X89 Assault by other specified chemicals and noxious substances X90 Assault by unspecified chemical or noxious substance X91 Assault by hanging, strangulation, and suffocation X92 Assault by drowning and submersion X93 Assault by handgun discharge X94 Assault by rifle, shotgun, and larger firearm discharge X95 Assault by other and unspecified firearm discharge X96 Assault by explosive material X97 Assault by smoke, fire, and flames X98 Assault by steam, hot vapors, and hot objects X99 Assault by sharp object Y00 Assault by blunt object Y01 Assault by pushing from high place Y02 Assault by pushing or placing victim before moving object Y03 Assault by crashing of motor vehicle Y04 Assault by bodily force Y05 Sexual assault by bodily force Y06 Neglect and abandonment Y07 Other maltreatment syndromes Y08 Assault by other specified means Y09 Assault by unspecified means

SOURCE: World Health Organization (2007).

Appendix B: Partial Correlations Among Explanatory Variables

1 2 3 4 5 6 7

1. Ln residential instability 1.00 2. Ln diversity 0.53 1.00 3. Ln female-headed households 0.35 0.61 1.00 4. Ln poverty −0.10 0.01 0.37 1.00 5. Ln unemployment 0.16 0.17 0.45 0.51 1.00 6. Adjacent to metro area 0.21 0.07 −0.18 −0.42 −0.30 1.00 7. Ln population at risk 0.63 0.51 0.35 −0.35 0.08 0.34 1.00