Placing the Neighborhood Accessibility–Burglary Link in Social-Structural Context

Abstract

Foundational research on the link between neighborhood accessibility and burglary has consistently shown a positive association. However, recent research has found that less accessible neighborhoods have higher burglary rates. Geographically referenced data from 401 neighborhoods in Jacksonville, Florida, are used to determine whether these inconsistencies can be explained by a conditioning effect of neighborhood social-structural context. Results from spatially lagged regression models indicate that neighborhood accessibility fails to have a direct effect on burglary rates after social-structural variables are controlled; rather, the effect of neighborhood accessibility on burglary rates is conditioned by the level of concentrated disadvantage of the neighborhood. Two potential explanations for the empirical findings are offered, and implications of the results for “designing out” crime are discussed.

Keywords

spatial regression neighborhood accessibility concentrated disadvantage burglary

It is generally accepted that both spatial and social-structural characteristics of neighborhoods are important independent contributors to neighborhood crime rates, yet the typical way in which crime rates are accounted for in neighborhood-level research is rather limited. Macro-level studies that examine the effects of social-structural context on crime rates often neglect spatial context, and studies that examine the effects of spatial context on crime rates fail to contextualize these effects within the social-structural environment. Although we know that the spatial environment plays an important role in influencing human behavior in general (e.g., Altman, 1975; Handy, Boarnet, Ewing, & Killingsworth, 2002; Jacobs, 1961), more criminological research is needed to specify exactly how the spatial environment influences criminal behavior specifically (Weisburd, 1997) and how social-structural factors may interact with spatial factors to influence crime (see Taylor, 2002). Research into the fundamental spatial characteristics of crime is important because empirical findings can further develop the relationships between space and social-structural variables. This can assist in the formulation or revision of theories, resulting in a greater capability to explain criminal behavior and the resulting crime patterns (Rengert, Piquero, & Jones, 1999).

Of the research that has examined the effect of the spatial environment on crime, considerable attention has been given to the concept of neighborhood accessibility, which has two related dimensions. Neighborhood permeability refers to the ease of penetration into a neighborhood, whereas neighborhood connectivity refers to the ease of movement once inside a neighborhood. Foundational research in this area suggests a consistent positive empirical association between neighborhood accessibility and burglary (Beavon, Brantingham, & Brantingham, 1994; Bevis & Nutter, 1977; Brantingham & Brantingham, 1975; Frisbie, Fishbine, Hintz, Joelsons, & Nutter, 1977; Greenberg, Rohe, & Williams, 1982; White, 1990). Taylor (2002) succinctly states the findings of this first generation of empirical research: “neighborhoods with smaller streets or more one-way streets or fewer entrance streets or with [fewer] turnings have lower property crime rates” (p. 419). Although foundational research on the link between neighborhood accessibility and burglary reveals a clear and consistent positive relationship, a second generation of research employing space syntax analysis has suggested that the opposite may be true (Shu, 2000; see also Hillier, 2004; Hillier & Shu, 1999). More specifically, recent studies suggest that burglary may actually cluster in neighborhoods, or on streets, that are less accessible.

The emerging inconsistencies in the literature could potentially be an artifact of divergent methodologies employed to measure the influence of neighborhood accessibility on burglary rates. However, not all second-generation studies employing space syntax methodology have resulted in similar conclusions (Baran, Smith, & Toker, 2007), and resultantly, this explanation alone seems unsatisfactory. Moreover, differences in empirical findings attributed solely to methodology overly simplify a complex issue, because it is highly likely that the association between spatial design and crime is conditioned by social, cultural, and economic contextual factors (Taylor, 2002). Although there are a number of possible moderating constructs, very little empirical research systematically assesses how neighborhood social-structural context interacts with neighborhood accessibility to influence burglary rates. One study finds that whereas more integrated neighborhoods are less susceptible to residential burglary overall, the observed effect is greatest for lower income neighborhoods (Shu & Huang, 2003). This suggests the potential importance of neighborhood socioeconomic factors in moderating the effect of neighborhood accessibility on burglary rates, but this notion has not yet been fully explicated nor adequately tested.

The current study seeks to extend prior empirical work on the accessibility–burglary link by determining whether neighborhood social-structural context conditions the effect of neighborhood accessibility on burglary. In other words, the current study attempts to reconcile the emerging inconsistencies in the literature by explaining why neighborhood accessibility increases burglary rates in some contexts but decreases burglary rates in other contexts. We use geographically referenced census data from Jacksonville, Florida, and employ spatially lagged regression models to determine whether a prominent social-structural factor (i.e., concentrated disadvantage) moderates the effect of neighborhood accessibility on burglary, while controlling for traditional social-structural covariates and spatial autocorrelation in neighborhood burglary rates.

Two Conflicting Generations of Neighborhood Accessibility–Burglary Research

Research on the accessibility–burglary link has been empirically grounded in several perspectives, two of which are the rational offender and behavioral geography perspectives. The rational offender perspective views target selection as driven by a process of rational calculation of the costs and benefits of committing a crime (Cornish & Clarke, 1986, 1987), and research suggests that criminals are indeed rational actors, albeit to a limited degree (Biron & Ladouceur, 1991; Cromwell, Olson, & Avary, 1991; Feeney, 1986; Kube, 1988; Maguire, 1988; Rengert & Wasilchick, 1990). The rational selection of a target is highly nuanced, including factors at the site, block, and neighborhood level (Taylor & Gottfredson, 1986). Characteristics of neighborhoods are important because the target selection process begins at the neighborhood level (Brantingham & Brantingham, 1978; Brown & Altman, 1991; Cornish & Clarke, 1986; Taylor & Gottfredson, 1986). With respect to neighborhood accessibility, offenders find neighborhoods with higher accessibility attractive because of decreased entry and exit time, which in turn decreases the potential for detection and apprehension. Moreover, in the event of detection, multiple routes and exits may decrease the chance of apprehension.

The behavioral geography perspective views target selection as a function of an offender’s familiarity with a place (Brantingham & Brantingham, 1981). Supporting this theoretical perspective, Rengert and Wasilchick (1985) find that offenders select targets for burglary along familiar routes, such as the path they take to work. Hence, neighborhoods that are more accessible have more potential offenders passing through on their way to other destinations, which increases the chances of a dwelling in a highly traveled area being targeted for criminal activity. In fact, dwellings in less accessible neighborhoods, despite being less target-hardened, are less likely to be burglarized than more target-hardened dwellings in more accessible neighborhoods (White, 1990).

Most of the early studies examining neighborhood accessibility find that it is a significant predictor of burglary rates (Taylor, 2002). Dwellings located on busy streets provide easy access and a quick escape to potential burglars and tend to be burglarized at higher rates (Rengert & Hakim, 1998). Brantingham and Brantingham (1975) found that homes on the border of a neighborhood were more likely to be burglarized than homes on the interior of a neighborhood. The authors speculated that offenders might be reluctant to deeply penetrate areas with which they were unfamiliar. Research by White (1990) found that neighborhoods with more access lanes leading from larger transit arteries significantly predicted higher burglary rates even when controlling for some social-structural neighborhood-level correlates of crime. Street segments that have a higher number of turnings are more likely to have higher rates of burglary (Beavon et al., 1994). Similarly, houses located on dead end streets are less likely to be burglarized, whereas houses located at intersections are more likely to be burglarized (Hakim, Rengert, & Shachuamurove, 2001). Bevis and Nutter (1977) concluded that cul-de-sacs are the safest type of street design because dead ends reduce the accessibility to nonresidents, eliminating these types of streets from offenders’ routine patterns of movement. Finally, as most previous research had analyzed burglary in neighborhoods located in larger cities, Yang (2006) extended this research to a smaller city and found that neighborhood accessibility is a significant positive predictor of burglary rates.

Despite the overwhelming number of studies that show a positive association between neighborhood accessibility and burglary rates, some evidence employing space syntax methodology suggests that areas that are less accessible may actually be at a higher risk for victimization (Hillier, 2004; Hillier & Shu, 1999; Shu, 2000; Shu & Huang, 2003). Hillier (2004) finds that traditional street layouts (e.g., gridiron) are overall safer places than modern street layouts (e.g., loops and lollipops) and suggests that this is the case for both lower and higher status neighborhoods. Moreover, there is evidence to suggest that burglaries cluster at cul-de-sac sites (not on street corners), because these environments reduce both the number of available observers and natural surveillance (Hillier & Shu, 1999; Shu, 2000). Taken together, recent research suggests neighborhood accessibility and burglary may actually have a negative relationship as opposed to the positive relationship that has been established by traditional criminological research in this area.

Contextualizing the Neighborhood Accessibility–Burglary Link

The contradictory empirical findings on the accessibility–burglary link may not be all that surprising given there are two theoretical notions that hypothesize opposite relationships between neighborhood accessibility and burglary rates.¹ On the one hand, Jacobs (1961) contends that increasing neighborhood accessibility decreases crime because permeable and well-connected neighborhoods increase the number of “eyes on the street.” Thus, although there are more potential offenders in accessible neighborhoods, active neighborhoods also have more noncriminals who can serve as watchmen and help prevent crime. On the other hand, Newman (1972) argues for restricting movement and access to neighborhoods in order to create “defensible space,” which leads to less crime. The contradiction of these notions can most clearly be seen when examining these ideas in light of routine activities theory. Cohen and Felson (1979) maintain that crime occurs when a motivated offender, a suitable target, and the absence of capable guardians collide in time and space. Thus, Jacobs’s (1961) approach involves increasing the number of capable guardians by making neighborhoods more accessible, whereas Newman’s (1972) approach involves decreasing the suitability of targets by making neighborhoods less accessible.

It is plausible that Newman’s and Jacobs’s ideas are both correct, and there is some initial evidence that neighborhood accessibility does operate differently in divergent social-structural contexts. For example, Shu and Huang (2003) find that neighborhoods in Taiwan that have higher levels of accessibility tend to have lower burglary rates but that the income level of the neighborhood appears to condition this relationship such that the effect is more pronounced in moderate and lower income neighborhoods. These neighborhoods may have limited resources to employ target hardening devices and may be reliant on increased pedestrian traffic and activity to increase neighborhood safety (Shu & Huang, 2003). Thus, there may be important crime reduction benefits that stem from increased neighborhood accessibility, specifically in poor neighborhoods. However, increased neighborhood accessibility in less disadvantaged neighborhood contexts may have no crime reduction benefits.

Although we acknowledge that several relevant contextual factors may moderate the link between the physical environment and crime (see Taylor, 2002), we choose to assess the potential conditioning effect of concentrated disadvantage on the association between neighborhood accessibility and burglary for two reasons. First, we seek to build directly on the empirical work of Shu and Huang (2003) by assessing the moderating effect of concentrated disadvantage. If neighborhood income level conditions the effect of neighborhood connectivity on burglary rates, it may be the case that this effect is more pronounced when considering neighborhood concentrated disadvantage, a prominent social-structural factor in macro-level criminological research. Second, the choice to examine the potential interaction between concentrated disadvantage and neighborhood accessibility was further guided by direct consultation with local law enforcement analysts in Jacksonville. They noted that some predominately minority and low socioeconomic status neighborhoods, which featured cul-de-sacs and wooded border areas on one or more sides, presented special challenges for both routine police patrols as well as targeted interventions due to the relative inaccessibility of these neighborhoods. In short, the current study aims to shed light on the inconsistencies in the literature that are presently clouding the relationship between neighborhood accessibility and burglary rates by placing the neighborhood accessibility–burglary link in social-structural context.

Data and Method

Data

The research site, Jacksonville, Florida, has a merged city/county government, meaning that there are no jurisdictional issues affecting local ordinances, police practices, zoning restrictions, or code enforcement, any of which could hypothetically affect neighborhood composition or burglary rates. Moreover, Jacksonville is the largest city by land area in the 48 contiguous states. Collectively, these features make it an ideal study location for assessing the effects of neighborhood connectivity due to the high spatial variability in neighborhood layouts.

In an effort to measure both spatial and social-structural variables relevant to neighborhood burglary rates, the current study employs geographically referenced data from two primary sources. First, a geographically referenced database file was obtained from the Jacksonville Sheriff’s Office containing 21 months worth of crime incidents from January 2006 through September 2007. Incidents, distinguished here from calls for service, are defined as events in which an official police report was filed by the Jacksonville Sheriff’s Office. In addition to the incidents layer, the database contained a road layer that was used to calculate the neighborhood accessibility (i.e., neighborhood connectivity) variable. Second, the 2000 US TIGER Census Block Group geographically referenced data file provided the social-structural variables as well as defined the geographic boundaries for the neighborhoods. Only census block groups with valid data located within Jacksonville Sheriffs Office’s patrol districts were included. Consequently, the areas along the Atlantic coast patrolled by Jacksonville Beach, Neptune Beach, and Atlantic Beach police departments are not included in the analysis. This exclusion brings the total number of neighborhoods used in the analysis to 401.

Measures

Burglary

Whereas crime counts are generally standardized by taking into account numbers of individuals in a given area, burglary counts are more appropriately standardized by taking into account the number of potential burglary sites (i.e., dwellings; White, 1990). The dependent variable used in this study is the residential burglary rate, defined as the number of residential burglaries committed per 1,000 households during the study period. The distribution of the dependent variable had both skewness and kurtosis values indicating the potential for unreliable significance tests. To minimize the potential problems associated with nonnormally distributed residuals, the burglary rate underwent a natural logarithmic transformation.²

Neighborhood connectivity

The beta index, used as the neighborhood connectivity measure, is defined as the number of links divided by the number of nodes within a given unit of analysis (Steiner, Bond, Miller, & Shad, 2004; Yang, 2006). In other words, the beta index is the ratio of the number of street segments (links) to the number of intersections or cul-de-sacs (nodes). Higher scores indicate greater network connectivity and therefore higher neighborhood accessibility. One advantage of this measure is that the entire street network is analyzed collectively in determining the neighborhood connectivity score for a given census block group. The roads layer and the U.S. census block group layer were both used to perform the necessary spatial join function, which was executed in ESRI’s ArcGIS 9.2.

Concentrated disadvantage

An index of concentrated disadvantage (α = .78) was calculated by summing the standardized scores of three constructs including the following: percent below the poverty line, percent unemployed (i.e., number of unemployed laborers more than 18 years of age), and percent Black. Results of a principal components analysis reveal that a single factor should be retained. The scale is standardized; higher scores indicate higher concentrated disadvantage.

Social–spatial interaction

Although measures of neighborhood connectivity and concentrated disadvantage are used to assess their direct effects on burglary rates, the independent variable of primary interest in the current study is a social–spatial interaction variable that assesses the extent to which the effect of neighborhood connectivity is conditioned by the level of concentrated disadvantage of the neighborhood. An interaction term was calculated by multiplying the standardized neighborhood connectivity score by the standardized concentrated disadvantage index (Agresti & Finlay, 1997; Aiken & West, 1991).

Covariates

Burglary rates have been associated with a number of neighborhood-level factors that must be included in the models so as not to confound the effects of neighborhood accessibility with established social-structural-level predictors. Population density is defined as the total population per square mile of land area and is included to account for the number of potential burglars in a given neighborhood.

The age structure of the block groups is calculated by dividing all persons between the ages of 5 and 21 years by the total population in that block group; higher scores indicate higher proportion of youth and young adults.³ Including age structures for neighborhoods is important because the age–crime curve reveals that criminal behavior peaks during late adolescence, both in general terms (Blumstein, Cohen, & Farrington, 1988) and specifically for burglary (Steffensmeier, Allan, Harer, & Streifel, 1989). Thus, neighborhoods with a larger proportion of youth and young adults are likely to have higher rates of burglary.

A measure of structural density is employed to reflect the physical inaccessibility of dwellings within a given neighborhood (see Bernasco & Nieuwbeerta, 2005). For example, second story apartments are harder targets than single family homes. Structural density is calculated by dividing the number of dwellings with 5 or more units by the total number of dwellings; higher scores indicate higher structural density.

Racial heterogeneity is a measure of how racially mixed a given census block group is and has been associated with higher rates of burglary (Markowitz, Bellair, Liska, & Liu, 2001). Racial heterogeneity is calculated using the following formula: $1 - \sum p_{i}^{2}$ , where p_i denotes the proportion of each racial group; higher scores indicate higher racial heterogeneity (see Blau, 1977; Gibbs & Martin, 1962; see also Sampson, 1983).

An index of residential instability (α = .80) was calculated by summing the standardized scores of two constructs including percentage of renters and percentage of short-term residents (i.e., proportion of individuals moving into the neighborhood within the past 5 years). The summated scale is standardized; higher scores indicate higher residential instability.

Finally, Yang (2006) maintains that when measuring connectivity using the beta index, one should take into account the average length of the street segments as an important control. Thus, a measure of average block length is included as a control in the model.

Spatially lagged burglary

Neighborhoods border one another, and Tobler’s (1970) first law of geography states that neighborhoods that are nearer to one another are more alike than neighborhoods that are farther apart. That is, spatial autocorrelation is an important consideration in neighborhood research; spatial autocorrelation must be accounted for to correctly specify the multivariate regression models when spatial autocorrelation is statistically significant (see Anselin & Bera, 1998; Browning, Cagney, & Wen, 2003; Legendre, 1993). Neighborhood researchers tend to overlook this influence and simply treat neighborhoods as independent (Morenoff, 2003), albeit this technique has been used somewhat more frequently in recent years.

Because the phenomenon under study (burglary) has an implicit theoretical impetus related to the independent variable of interest (neighborhood connectivity), a spatial lag model is regarded as most appropriate and parsimonious for the analysis.⁴ As Anselin (2002) explains, the spatial lag model is a mixed regressive, spatial autoregressive model taking the following general form:

y = ρ W y + X β + ε,

where y represents the dependent variable, W represents a spatial weights matrix, ρ_. is the spatial autoregressive parameter, X represents the matrix of exogenous variables with an associated regression coefficient vector β, and ε represents the error term. Thus, the spatial lag term can be interpreted substantively as an independent variable representing the effect of burglary in surrounding neighborhoods on any given neighborhood, and it can also be viewed technically as a statistical correction for the violation of the independence of errors assumption.

Descriptive Statistics

During the 21-month study period, a total of 11,002 burglaries occurred in Jacksonville with an average of 41.89 burglaries per 1,000 households for each neighborhood.⁵ The mean connectivity score is 1.64, which indicates the neighborhood average ratio of street segments (links) to nodes. The average block length for all neighborhoods is 575 feet. The typical neighborhood has approximately 3,000 residents with just less than one quarter of the total population being less than 18 years of age. On average, 13% of all dwellings are multiunit, and neighborhoods have a racial heterogeneity score equal to .29. Because the standardized measures of concentrated disadvantage and residential instability do not have direct practical meaning, descriptive statistics for the subcomponents of these constructs are reported. With respect to residential instability, the typical neighborhood has 20% of its residents move into the neighborhood within the past 5 years and 31% of its residents rent. With respect to concentrated disadvantage, the average neighborhood has 35% of its residents more than the age of 18 years unemployed, and 16% of its residents below the poverty line. Additionally, 36% of residents in a typical neighborhood are Black (Table 1).

Table 1.

Descriptive Statistics.

	M	SD
Burglary rate	41.89	29.72
Neighborhood connectivity	1.64	0.31
Block length	575.14	244.09
Population density	3057.47	2208.64
Youth population	0.24	0.07
Structural density	0.13	0.19
Racial heterogeneity	0.29	0.19
Residential instability
Renter	0.31	0.20
New resident	0.20	0.08
Concentrated disadvantage
Unemployed	0.35	0.12
Below poverty line	0.16	0.14
Black	0.36	0.34

Analytic Strategy

To demonstrate the need for the inclusion of a spatially lagged dependent variable term, Moran’s I is estimated, and the spatial concentration of neighborhood burglary rates is illustrated with a Local Indicators of Spatial Association (LISA) cluster map (Anselin, 1995; Anselin, Syabri, & Kho, 2006). The application of LISA also permits identification of statistically significant areas of spatial autocorrelation in the dependent variable, compared with a null hypothesis of spatial randomness, by comparing values in each specific location to values in neighboring locations. In this way, the LISA cluster map establishes which areas are spatially autocorrelated and provides a means of visualizing the phenomena explicated in the multivariate analysis. Each block group in the area under study (N = 401) is categorized into one of four LISA types according to calculated local Moran’s I values—high–high, low–low, high–low, or low–high—depending on the value for the variable of interest for that block group as well as the range of values for all of the block groups that share a common border. Thus, the LISA map in this analysis not only identifies block groups that have significantly high or low values for the dependent variable, it also depicts the spatial concentration of those block groups relative to the range of values for neighboring block groups.

To assess the direct and conditioned effects of neighborhood connectivity on burglary rates, we estimate two spatially lagged regression models using Stata 10.1 MP.⁶ First, we estimate whether neighborhood connectivity influences burglary rates while controlling for important social-structural correlates of burglary (i.e., population density, youth population, structural density, racial heterogeneity, residential instability, and concentrated disadvantage). Second, the interaction term (i.e., connectivity * concentrated disadvantage) is added to the model to assess the conditioning effect of concentrated disadvantage on the relationship between neighborhood connectivity and burglary. All variables included in the model were mean-centered to minimize potential colinearity issues when including multiplicative terms in the regression model (see Cronbach, 1987; Smith & Sasaki, 1979). The coefficients of statistical interactions are often difficult to interpret directly; therefore, postestimation plots are used to graphically display and clarify the moderating effect of concentrated disadvantage on the link between neighborhood connectivity and burglary rates.

Results

Moran’s I, a global measure for spatial autocorrelation, indicates that there is statistically significant spatial autocorrelation in the dependent variable (I = .367, p < .001) necessitating the use of a spatial regression model. The spatial distribution of the dependent variable is illustrated in Figure 1. Specifically, the majority of high–high block groups (those that have a statistically significant local Moran’s statistic neighbored by other statistically significant groups) are clustered in the urban core of Jacksonville, where population density is highest. However, there are additional high–high block groups to the north and southeast, indicating that not all block groups featuring a high spatial concentration of burglary are geographically contiguous. Also, many block groups on the east coast of the county in the Jacksonville Beach area are in the low–low classification, indicating that burglary is infrequent and spatially disbursed in these areas of the county. As expected from Moran’s I statistic and the LISA map, we note that the spatially lagged burglary coefficient is statistically significant across both models. The inclusion of this term yields a more correctly specified statistical model and increases our confidence in the findings.

Figure 1.

Localized indicators of spatial association (LISA) cluster map for burglary.

Table 2 presents the zero-order correlations. With the exception of residential instability, all of the spatial and social-structural variables are significantly correlated with neighborhood burglary rates in the expected direction. The finding that residential instability is not significantly associated with neighborhood burglary rates is consistent with some prior research (e.g., Bernasco & Nieuwbeerta, 2005). The level of concentrated disadvantage and the proportion of the youth population have the strongest bivariate relationships with neighborhood burglary rates. It is worth noting that there are a few moderate to strong correlations among some of the independent variables. Not surprisingly, structural density and residential instability are highly correlated (r = 0.755), which raises some concerns for the possibility of multicolinearity. However, colinearity diagnostics for the multivariate models did not reveal any significant problems.⁷

Table 2.

Zero-Order Correlations.

	1	2	3	4	5	6	7	8
Burglary rate (ln)	1
Neighborhood connectivity	.152*	1
Block length	−.131*	−.363*	1
Population density	.193*	.470*	−.471*	1
Youth population	.325*	−.104*	−.037	.106*	1
Structural density	−.105*	−.059	−.121*	.222*	−.106*	1
Racial heterogeneity	.158*	−.081*	−.095	.137*	.122*	.347*	1
Residential instability	−.010	.122*	−.149*	.246*	−.090	.755*	.333*	1
Concentrated disadvantage	.376*	.341	−.292*	.354*	.290*	.165*	−.109*	.154*

p < 0.05.

Table 3 contains the results of the spatially lagged multivariate regression models. When social-structural variables are controlled, the association between neighborhood connectivity and burglary rates is no longer statistically significant (see Model 1). There are a number of important neighborhood social-structural variables that reach statistical significance. Particularly, burglary rates tend to be higher in areas that have more concentrated disadvantage and less structural density, which is consistent with previous research (White, 1990). Areas that have higher concentrations of youth and young adults tend to suffer more burglaries. Neighborhoods that have higher levels of racial heterogeneity also experienced more burglary. Overall, Model 1 explained approximately 34% of the variance in neighborhood burglary rates.

Table 3.

Spatial Lag Regression Models Predicting Neighborhood Burglary Rates With Robust Standard Errors.

	Model 1		Model 2
	b	RSE	b	RSE
Constant	2.903	0.131*	2.966	0.139*
Spatial variables
Neighborhood connectivity	.016	.039	−.001	.044
Block length	.026	.046	.037	.046
Social-structural variables
Population density	.080	.064	.083	.063
Youth population	.107	.041*	.114	.040*
Structural density	−.201	.078*	−.191	.075*
Racial heterogeneity	.151	.038*	.149	.038*
Residential instability	.041	.091	.040	.091
Concentrated disadvantage	.215	.048*	.252	.050*
Social–spatial interaction
Neighborhood connectivity * Concentrated disadvantage			−.102	.034*
Spatial autocorrelation
Spatially lagged burglary	.028	.005*	.026	.005*
Variance ratio (pseudo-R²)	.339		.356
Sigma	.610		.610

p < .05.

To determine if the relationship between neighborhood connectivity and burglary is conditioned by neighborhood social-structural context, we add the interaction term to the regression model (see Model 2). Importantly, the social-spatial interaction term is statistically significant (p < .05), indicating that the effect of neighborhood connectivity on neighborhood burglary rates is conditioned by the level of concentrated disadvantage of the neighborhood. Overall, Model 2 explains approximately 36% of the variance, and the inclusion of the interaction term yields a significantly better overall fit than the exclusion of the interaction (as in Model 1).

Figure 2 graphically displays the significant interaction; the three simple slopes represent the effect of neighborhood connectivity on neighborhood burglary rates at different levels of concentrated disadvantage while holding all other variables at their mean. The relationship between neighborhood connectivity and burglary is shown for low concentrated disadvantage (i.e., one standard deviation below the mean), mean concentrated disadvantage, and high concentrated disadvantage (i.e., one standard deviation above the mean) neighborhoods. In low concentrated disadvantaged neighborhoods, neighborhood connectivity is positively associated with burglary rates. In these neighborhoods, greater street network connectivity leads to higher neighborhood burglary rates. Conversely, in high concentrated disadvantaged neighborhoods, neighborhood connectivity is negatively associated with burglary rates. Thus, the more connected a neighborhood is in a high concentrated disadvantage area, the lower the neighborhood burglary rate. It is important to note that both the simple slopes for high and low concentrated disadvantage neighborhoods are not only in opposite directions but are also significantly different from zero (see Figure 2).

Figure 2.

Partial regression slopes illustrating the effect of neighborhood connectivity on burglary rates conditional on concentrated disadvantage.

Discussion and Conclusions

This study attempted to quantify and contextualize the relationship between neighborhood connectivity and one specific type of property crime (burglary). Exploratory spatial data analysis showed that burglary in Duval County was mostly concentrated in the urban core of Jacksonville, though not necessarily in geographically contiguous areas. Results from a robust and correctly specified spatial lag regression model found that neighborhood connectivity failed to have an independent effect on burglary rates when a host of social-structural controls were included in the model. Instead, the effect of neighborhood connectivity on burglary rates in Jacksonville operates differently in divergent neighborhood contexts. Specifically, we find that in high concentrated disadvantage neighborhoods, the more connected a neighborhood’s street network is the less likely burglary is to occur. In low concentrated disadvantage neighborhoods, however, increased neighborhood connectivity leads to higher rates of burglary.

There are several possible explanations for the observed empirical findings of the current study. One explanation stems from notions central to the rational offender perspective. Burglary might be the aftermath of a more systematic rational calculation by the offender (Bennett & Wright, 1986) when compared with other types of crimes. It follows that burglars are likely to take into account physical cues that indicate risks and benefits when selecting a target. Research suggests that this target selection process begins with the selection of a neighborhood (Brantingham & Brantingham, 1978; Brown & Altman, 1991; Cornish & Clarke, 1986; Taylor & Gottfredson, 1986). Neighborhoods with lower concentrated disadvantage levels (higher social status) may give an offender the outward impression that residents care and are more willing or more likely to intervene either directly or by calling the police (i.e., collective efficacy; see Sampson, Raudenbush, & Earls, 1997).

It is conceivable that concentrated disadvantage serves as an environmental indicator of the willingness of residents to take an active role in crime prevention.⁸ This line of reasoning rests on the assumption that concentrated disadvantage is tangible and can be physically seen at some level. It follows, then, that under conditions in which a burglar perceives that residents are willing to intervene, such as in low concentrated disadvantage areas, burglars may take into account access and escape routes when planning a burglary and subsequently target neighborhoods that are highly accessible. If a burglar perceives that residents are not willing to intervene, as may be the case in high concentrated disadvantage neighborhoods, then his or her main concern may be detection by the routine patrol of police. Hence, burglars may choose neighborhoods that are less accessible in higher concentrated disadvantage areas. Then, it may be rational to commit burglary in more or less accessible neighborhoods depending on the level of concentrated disadvantage.

Another potential explanation combines notions from the rational offender and behavioral geography perspectives. The behavioral geography perspective holds that places close to where offenders live, for example, are at an increased risk of victimization (see Taylor, 2002). Research suggests that offenders select targets for burglary along familiar routes (Rengert & Wasilchick, 1985). It is possible that individuals’ modes of transportation play an important role in determining their routes and, subsequently, planning their burglaries. Individuals residing in high concentrated disadvantage areas may be more likely than those in low concentrated disadvantage areas to be reliant on a combination of walking and public transportation for travel. Using pedestrian routes (e.g., shortcuts, footpaths) as opposed to streets may save significant amounts of travel time. This is especially true for neighborhoods with low street network connectivity providing increased motivation to use pedestrian routes in these areas. In high concentrated disadvantage neighborhoods, low street connectivity may foster burglary because individuals may have access to potential targets with minimal chances of detection. However, neighborhoods with high street connectivity may help keep pedestrians using streets, where they can more easily be monitored by both residents and passersby.

On the other hand, individuals residing in low concentrated disadvantage neighborhoods are more likely reliant on vehicular transportation for their daily activities. Driving a car into a neighborhood that has low street network connectivity to commit a burglary may raise suspicion. That is, residents in neighborhoods that do not receive much traffic because they are less connected may be more likely to identify a suspicious vehicle. In addition to increased risk of detection, neighborhoods with low connectivity may increase the risk of apprehension because there are fewer ways to elude police. For both these reasons, it may be more likely for burglars traveling in vehicles to commit burglaries in low concentrated disadvantage neighborhoods that have high street network connectivity. This explanation rests on two key assumptions. First, a burglar from a low concentrated disadvantage area targets neighborhoods that are relatively close to where he or she lives (see Taylor, 2002) but far enough away to warrant the use of a vehicle. Second, nearby concentrated disadvantage levels are relatively similar to concentrated disadvantage levels of an offender’s own neighborhood (see Tobler, 1970).⁹ Although this explanation seems plausible, we do not have data on pedestrian connectivity, and therefore, future research is needed to confirm or refute this explanation.

Our results suggest that we may not be able to “design out” crime with a simple “one-size-fits all” approach. In high concentrated disadvantage neighborhoods, it may be beneficial to increase neighborhood accessibility to augment the number of “eyes on the street,” which is consistent with the ideas put forth by Jacobs (1961). In low concentrated disadvantage neighborhoods, it may be advantageous to decrease neighborhood accessibility to restrict movement and access to reduce residential burglaries, which is consistent with the notions developed by Newman (1972). In any event, guidance for urban planners and neighborhood developers is desirable when crime is an a priori concern or problem to be addressed. Because new neighborhoods target individuals of a certain socioeconomic status depending on both the sales price and whether the housing development is designated low or moderate income, urban planners and neighborhood developers can use this information and “design out” crime accordingly. Thus, this analysis supports the idea of taking into account social-structural characteristics when considering how to employ spatial design and environmental features to reduce crime.

This study has certain limitations that lead us to accept the conclusions with caution. First, we relied on U.S. Census data (U.S. Census Bureau, 2000) for the creation of our neighborhood units. The geographical boundaries of census block groups may be somewhat arbitrarily defined; thus, it would be worthwhile to replicate this analysis with neighborhoods that may be more naturally defined (e.g., identified by residents). Along these lines, the present study assessed the moderating effect of concentrated disadvantage on the relationship between one dimension of neighborhood accessibility (neighborhood connectivity) and burglary. Future research should examine the extent to which neighborhood connectivity and neighborhood permeability both independently and interactively influence crime rates, and employing more naturally defined neighborhoods should aid greatly in this endeavor. Second, given that we use incident data as opposed to victimization survey data, differential reporting of burglary across geographical areas with differential levels of concentrated disadvantage is a concern. Although burglary is one of the most reliably measured crimes by the Uniform Crime Report and National Crime Victimization Survey (Blumstein, Cohen, & Rosenfeld, 1991; 1992), some research suggests that for some crimes there is differential reporting across economically disadvantaged neighborhoods. However, this differential reporting is such that disadvantaged neighborhoods are less likely to report crime (see Baumer, 2002; Goudriaan, Wittebrood, & Nieuwbeerta, 2006). In other words, to the extent that underreporting of burglary is a problem it would systematically bias the results toward not finding a statistically significant direct effect of concentrated disadvantage on neighborhood burglary rates. Despite this possible hazard, concentrated disadvantage maintains a direct effect on neighborhood burglary rates in our analysis (see Model 2). Then, despite these concerns, the present study takes an important step forward by examining the moderating role of an important social-structural factor and offers an explanation for the emerging inconsistencies in the literature on the relationship between neighborhood accessibility and burglary.

There are a number of important areas for future research examining the intersecting roles the social and spatial environments play in causing crime. Given that a number of social-structural variables may serve as moderating influences (Nubani & Wineman, 2005; Taylor, 2002), future studies should examine the extent to which other relevant neighborhood social-structural constructs influence the design–crime link. Perhaps, these interaction effects will be even larger than those observed in the present study. Future research should also examine the moderating role of concentrated disadvantage (or other social-structural variables) on the relationship between neighborhood accessibility and other types of property and violent crime, with an eye toward the generalizability of these findings for other crime types. Additionally, given there are potentially important differences between urban and rural environments with respect to social-structural and spatial characteristics, future research should be carried out in both contexts.

Future research in this area should continue to address the methodological issue of spatial autocorrelation by using a spatially lagged version of the dependent variable, when appropriate, to ensure unbiased parameter estimates. Spatial autocorrelation violates a key assumption of ordinary least squares regression and may result in the unreliability of significance tests. Therefore, it is important that future research address this limitation to provide sound conclusions about the role of neighborhood-level structural indicators in predicting crime and related outcomes. Because our findings hold after correcting for spatial autocorrelation in the data, it should be determined whether neighborhood context conditions the effect of neighborhood connectivity on burglary rates in other cities. In addition to dealing with spatial autocorrelation in a methodological sense, it may be fruitful to view spatial autocorrelation not simply as a nuisance but rather something that can be modeled and explained (see, e.g., Akers & Jensen, 2006). It is likely that burglars are not generally confined to the neighborhoods in which they reside, which could account for the origins of spatial autocorrelation in burglary data. Although we currently lack the data to examine the sources of spatial autocorrelation, efforts to explain this common empirical occurrence theoretically may result in a greater capacity to account for variation in neighborhood crime rates.

Footnotes

Acknowledgements

We extend a special thanks to Matt White and Chief Justin Hill from the Jacksonville Sheriff’s Office for their cooperation and continued support. We also would like to thank two anonymous reviewers for their perceptive comments and Marvin D. Krohn for his insightful suggestions on earlier versions of this article.

Declaration of Conflicting Interests

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

Funding

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

Notes

Author Biographies

Jeffrey T. Ward is an assistant professor in the Department of Criminal Justice at the University of Texas at San Antonio and a Harry Frank Guggenheim Dissertation Fellow. His research interests include quantitative methodology; sanction effects, violence, and gangs; developmental and life-course criminology; and space, place, and crime. His work has appeared in Biodemography and Social Biology and Journal of Criminal Justice Education and is forthcoming in Criminal Justice and Behavior and Journal of Drug Issues.

Matt R. Nobles is an assistant professor of criminal justice at Sam Houston State University. His research interests include violent and interpersonal crimes, gun policy, geographical information system and spatial econometrics, criminological theory, and quantitative methods. His work has appeared in Justice Quarterly, Criminal Justice and Behavior, Crime Mapping, and the Security Journal and is forthcoming in the Journal of Criminal Justice.

Tasha J. Youstin is a doctoral student at John Jay College, the Graduate Center, City University of New York. Her general research interests include theory testing, sexual offending and offender policies, quantitative methods, and environmental criminology. Her research has been published in Criminal Justice and Behavior and Crime Mapping: A Journal of Research and Practice.

Carrie L. Cook, PhD, is an assistant professor in the Department of Government and Sociology at Georgia College and State University. Her research interests include corrections, sentencing policy, victimization, and program evaluation. Her work has appeared in Criminal Justice Policy Review and is forthcoming in the Journal of Criminal Justice.

References

Agresti

Finlay

(1997). Statistical methods for the social sciences. Upper Saddle River, NJ: Prentice Hall.

Aiken

L. S.

West

S. G.

(1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage.

Akers

R. L.

Jensen

G. F.

(2006). Empirical status of social learning theory: Past, present, and future. In Cullen

F. T.

Wright

J. P.

Blevins

K. R.

(Eds.), Taking stock: The status of criminological theory (Vol. 15, pp. 37-76). New Brunswick, NJ: Transaction.

Altman

(1975). The environment and social behavior: Privacy, personal space, territory, and crowding. Monterey, CA: Brooks/Cole.

Anselin

(1988). Spatial econometrics: Methods and models. Dordrecht, Netherlands: Kluwer Academic.

Anselin

(1995). Local indicators of spatial association—LISA. Geographical Analysis, 27, 93-115.

Anselin

(2002). Under the hood: Issues in the specification and interpretation of spatial regression models. Agricultural Economics, 27, 247-267.

Anselin

Bera

A. K.

(1998). Spatial dependence in linear models with an introduction to spatial econometrics. In Ullah

Giles

(Eds.), Handbook of applied econometrics statistics (pp. 237-289). New York: Marcel Dekker.

Anselin

Syabri

Kho

(2006). GeoDa: An introduction to spatial data analysis. Geographical Analysis, 38, 5-22.

10.

Baran

P. K.

Smith

W. R.

Toker

(2007, June). The space syntax and crime: Evidence from a suburban community. Paper presented at the 6th International Space Syntax Symposium, Istanbul, Turkey.

11.

Baumer

E. P.

(2002). Neighborhood disadvantage and police notification by victims of violence. Criminology, 40, 579-616.

12.

Beavon

D. J.

Brantingham

P. L.

Brantingham

P. J.

(1994). The influence of street networks on the patterning of property offenses. In Clarke

R. V.

(Ed.), Crime prevention studies (Vol. 2, pp. 115-148). Monsey, NY: Willow Tree Press.

13.

Bennett

Wright

(1986). Burglars on burglary. Brookfield, VT: Gower.

14.

Bernasco

Nieuwbeerta

(2005). How do residential burglars select target areas? British Journal of Criminology, 44, 296-315.

15.

Bevis

Nutter

J. B.

(1977, November). Changing streets layouts to reduce residential burglary. Paper presented at the American Society of Criminology Annual Meeting, Atlanta, GA.

16.

Biron

L. L.

Ladouceur

(1991). The boy next door: Local teen-age burglars in Montreal. Security Journal, 2, 200-204.

17.

Blau

P. M.

(1977). Inequality and heterogeneity. New York: Free Press.

18.

Blumstein

Cohen

Farrington

D. P.

(1988). Criminal career research: Its value for criminology. Criminology, 26, 1-35.

19.

Blumstein

Cohen

Rosenfeld

(1991). Trend and deviation in crime rates: A comparison of UCR and NCS data for burglary and robbery. Criminology, 29, 237-263.

20.

Blumstein

Cohen

Rosenfeld

(1992). The UCR-NCS relationship revisited: A reply to Menard. Criminology, 30, 115-124.

21.

Brantingham

P. L.

Brantingham

P. J.

(1975). Residential burglary and urban form. Urban Studies, 12, 273-284.

22.

Brantingham

P. J.

Brantingham

P. L.

(1978). A theoretical model of crime site selection. In Krohn

M. D.

Akers

R. L.

(Eds.), Crime, law, and sanctions (pp. 105-118). Beverly Hills, CA: Sage.

23.

Brantingham

P. J.

Brantingham

P. L.

(1981). Mobility, notoriety, and crime: A study in the crime patterns of urban nodal points. Journal of Environmental Systems, 11, 89-99.

24.

Brown

B. B.

Altman

(1991). Territoriality and residential crime: A conceptual framework. In Brantingham

P. J.

Brantingham

P. L.

(Eds.), Environmental criminology (pp. 55-76). Prospect Heights, IL: Waveland Press.

25.

Browning

C. R.

Cagney

K. A.

Wen

(2003). Explaining variation in health status across space and time: Implications for racial and ethnic disparities in self-rated health. Social Science & Medicine, 57, 1221-1235.

26.

Cohen

L. E.

Felson

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

27.

Cornish

D. B.

Clarke

R. V.

(1986). The reasoning criminal: Rational choice perspectives on offending. Secaucus, NJ: Springer-Verlag.

28.

Cornish

D. B.

Clarke

R.V.

(1987). Understanding crime displacement: An application of rational choice theory. Criminology, 25, 933-948.

29.

Cromwell

P. F.

Olson

J. N.

Avary

D. W.

(1991). Breaking and entering: An ethnographic analysis of burglary. Newbury Park, CA: Sage.

30.

Cronbach

L. J.

(1987). Statistical tests for moderator variables: Flaws in analyses recently proposed. Psychological Bulletin, 102, 414-417.

31.

Feeney

(1986). Robbers as decision-makers. In Cornish

D. B

Clarke

R. V.

(Eds.), The reasoning criminal: Rational choice perspectives on offending (pp. 53-71). New York: Springer-Verlag.

32.

Frisbie

Fishbine

Hintz

Joelsons

Nutter

J. B.

(1977). Crime in Minneapolis: Proposals for prevention. St. Paul, MN: Governor’s Commission on Crime Prevention and Control.

33.

Gibbs

J. P.

Martin

W. T.

(1962). Urbanization, technology, and the division of labor. American Sociological Review, 26, 667-677.

34.

Goldstein

(1992). srd9: Box–Cox statistics for help in choosing transformations. Stata Technical Bulletin, 5, 22-25.

35.

Goudriaan

Wittebrood

Nieuwbeerta

(2006). Neighbourhood characteristics and reporting crime: Effects of social cohesion, confidence in police effectiveness and social-economic disadvantage. British Journal of Criminology, 46, 719-742.

36.

Greenberg

S. W.

Rohe

W. M.

Williams

J. R.

(1982). Safety in urban neighborhoods: A comparison of physical characteristics and informal territorial control in high and low crime neighborhoods. Population and Environment, 5, 141-165.

37.

Hakim

Rengert

G. F.

Shachuamurove

(2001). Target search of burglars: A revised economic model. Paper in Regional Science, 80, 121-217.

38.

Handy

S. L.

Boarnet

M. G.

Ewing

Killingsworth

R. E.

(2002). How the built environment affects physical activity: Views from urban planning. American Journal of Preventive Medicine, 23, 64-73.

39.

Hillier

(2004). Can streets be made safe? Urban Design International, 9, 31-45.

40.

Hillier

Shu

(1999). Designing for secure spaces. Planning in London, 29, 31-45.

41.

Jacobs

(1961). The death and life of great American cities. New York: Random House.

42.

Kube

(1988). Preventing bank robbery: Lessons from interviewing robbers. Journal of Security Administration, 11, 78-83.

43.

Legendre

(1993). Spatial autocorrelation: Trouble or new paradigm? Ecology, 74, 1659-1673.

44.

Maguire

(1988). Searchers and opportunists: Offender behavior and burglary prevention. Journal of Security Administration, 11, 70-77.

45.

Markowitz

F. E.

Bellair

P. E.

Liska

A. E.

Liu

(2001). Extending social disorganization theory: Modeling the relationships between cohesion, disorder, and fear. Criminology, 39, 293-319.

46.

Morenoff

J. D.

(2003). Neighborhood mechanisms and the spatial dynamics of birth weight. American Journal of Sociology, 108, 976-1017.

47.

Newman

(1972). Defensible space: Crime prevention through urban design. New York: Macmillan.

48.

Nubani

Wineman

(2005, June). The role of space syntax in identifying the relationship between space and crim. Paper presented at the Proceedings of the 5th International Space Syntax Symposium, Delft, Holland.

49.

Rengert

G. F.

Hakim

(1998). Burglary in affluent communities: A planning perspective. In Felson

Peiser

R. B.

(Eds.), Reducing crime through real estate development and management (pp. 39-52). Washington, DC: Urban Land Institute.

50.

Rengert

G. F.

Piquero

A. R.

Jones

P. R.

(1999). Distance decay reexamined. Criminology, 37, 427-446.

51.

Rengert

G. F.

Wasilchick

(1985). Suburban burglary: A time and a place for everything. Springfield, IL: Thomas.

52.

Rengert

G. F.

Wasilchick

(1990). Space, time and crime: Ethnography insights into residential burglary. Report submitted to U.S. Department of Justice, National Institute of Justice, Office of Justice Programs.

53.

Sampson

R. J.

(1983). Group size, heterogeneity, and intergroup conflict: A test of Blau’s inequality and heterogeneity. Social Forces, 62, 618-639.

54.

Sampson

R. J.

Raudenbush

S. W.

Earls

(1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918-924.

55.

Shu

(2000). Housing layout and crime vulnerability. Urban Design International, 5, 177-188.

56.

Shu

S. C.

Huang

J. N.

(2003, June). Spatial configuration and vulnerability of residential burglary: A case study of a city in Taiwan. Paper presented at the 4th International Space Syntax Symposium, London.

57.

Smith

K. W.

Sasaki

M. S.

(1979). Decreasing multicollinearity: A method for models with multiplicative functions. Sociological Methods and Research, 8, 35-56.

58.

Steffensmeier

D. J.

Allan

E. A.

Harer

M. D.

Streifel

(1989). Age and the distribution of crime. American Journal of Sociology, 94, 803-831.

59.

Steiner

R. L.

Bond

Miller

Shad

(2004). Future directions for multimodal areawide level of service handbook research and development. Tallahassee, FL: Florida Department of Transportation.

60.

Taylor

R. B.

(2002). Crime prevention through environmental design (CPTED): Yes, no, maybe, unknowable, and all of the above. In Bechtel

R. B.

Churchman

(Eds.), Handbook on environmental psychology (pp. 413-426). New York: Wiley.

61.

Taylor

R. B.

Gottfredson

(1986). Environmental design, crime, and prevention: An examination of community dynamics. Crime and Justice, 8, 387-416.

62.

Tobler

W. R.

(1970). A computer movie simulating urban growth in the Detroit region. Economic Geography, 46, 234-240.

63.

U.S. Census Bureau. (2000). Summary social, economic, and housing characteristics. Retrieved January 4, 2008, from http://www.census.gov/prod/cen2000/phc-2-a.pdf

64.

Weisburd

D. L.

(1997). Reorienting crime prevention research and policy: From the causes of criminality to the context of crime (No. NCJ165041). Washington, DC: National Institute of Justice.

65.

White

G. F.

(1990). Neighborhood permeability and burglary rates. Justice Quarterly, 7, 57-67.

66.

White

(1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817-838.

67.

Yang

(2006). Exploring the influence of environmental features on residential burglary using spatial-temporal pattern analysis. Unpublished PhD dissertation. University of Florida.