Too Close for Comfort? Registered Sex Offender Spatial Clustering and Recidivistic Sex Crime Arrest Rates

Abstract

This study examined whether three measures of the spatial distribution of registered sex offenders (RSOs) in September 2010 were associated with differences in county-level rates of recidivistic sex crime arrests over the following year in 52 upstate New York counties. Results indicate that RSO clustering was positively associated with modest increases of recidivistic sex crime arrest rates, but results were significant only for adult victim sex crimes and only for certain types of RSO clustering. Under no circumstances, however, was increased RSO clustering associated with decreased rates of recidivistic sex crime arrests. The results of this study, combined with the limited prior research, suggest that RSO clustering has only a limited association with recidivistic sex crime arrest rates. This implies that housing policies such as residence restrictions may be useful in mitigating risk from some types of recidivistic sex crimes only to the extent that they result in more equitable distributions of RSOs within a county.

Keywords

child sexual abuse sex offenses sexual recidivism reentry residence restrictions sex offender policy

Each year over a half-million individuals are released from incarceration into communities throughout the United States (Anderson-Facile, 2009; Council of State Governments, 2005; Hughes & Wilson, 2003). Within three years of release, two out of three of these individuals will be rearrested after committing new crimes or violating their conditions of parole (Langan, Schmitt, & Durose, 2003), resulting in many of these ex-offenders returning to jail or prison (Beck & Shipley, 1987, p. 9). In addition to the obvious drain on corrections resources, this recidivism can have negative consequences on the communities where ex-offenders are most likely to live upon release, particularly when this recidivism involves violent victimization. A number of criminal justice policies have been implemented in an attempt to influence the reentry of these ex-offenders to reduce instances of recidivism, and in doing so, increase the safety of community members.

In particular, both policymakers and the public have become concerned about where registered sex offenders (RSOs) are living in the community. In fact, some polls indicate that residents want policymakers to make sex offender management a top concern (CSOM, 2008; Levenson, Brannon, Fortney, & Baker, 2007; Mears, Mancini, Gertz, & Bratton, 2008). This is not surprising, as media reports have identified high fear levels for residents who lived near “clusters” of such RSOs (Bain & German, 2006; Kilgannon, 2008). These anecdotal stories have also been supported by research evidence (e.g., Beck & Travis, 2004; Kernsmith, Craun, & Foster, 2009; Zevitz, 2004; but see Beck, Clingermayer, Ramsey, & Travis, 2004). Further, proximity to RSOs has been related to economic consequences, such as decreased housing values (Larsen, Lowrey, & Coleman, 2003; Linden & Rockoff, 2008; Pope, 2008). Perhaps most concerning, some initial research suggests that proximity to RSOs may increase adults’ risk of sexual victimization (Tewksbury, Mustaine, & Covington, 2010).

To address these concerns, policymakers have introduced a number of policies over the past two decades. These include registration and community notification policies that inform residents of where RSOs are living (for reviews, see Adams, 2002; Levenson & D’Amora, 2007; Socia & Stamatel, 2010), clustering restrictions that prohibit RSOs from living with or near one another (Bain & German, 2006; Gustavson, 2011), and residence restrictions that prohibit RSOs from living near places with large groups of potential victims, such as schools, daycares, and parks (Burchfield, 2011; Levenson, 2009; Socia, 2011).

Influencing where RSOs are living, and where they are clustering, has become a particular focus of recent legislative action. Indeed, some policymakers are explicit in their belief that RSO clustering is an important issue that can have consequences for community members. For example, in proposing a new residence restriction in the town of Brookhaven, New York, Councilwoman Connie Kepert said that “these code changes are a very important step to take for the protection of all of the children of Brookhaven, particularly those areas that have been over saturated with sex offenders” (Brookhaven.org, 2008, p. 1). In justifying this concern, Kepert said that “our primary goal is to amend our code to limit the number of sex offenders who can be dumped into one particular place because it deteriorates the community” (Bain & German, 2006, p. 2). Clearly, concern with RSO clustering was a driving force behind Kepert’s policy proposal.

The typical residence restriction policy prohibits RSOs from living within a given distance of “child congregation locations,” which can be defined as any place where children might gather in a community. These policies are based on the assumption that limiting housing near these locations will make it harder for RSOs to find and approach young children that they could sexually assault. These policies also assume that the “suitable targets” of RSOs are children who gather around public locations such as schools and daycares, which does not appear to be supported by existing literature (e.g., Duwe, Donnay, & Tewksbury, 2008; Zandbergen, Levenson, & Hart, 2010).

However, in order for these policies to work as intended, two relationships must be present. First, a residence restriction must affect where RSOs are actually living. If RSOs are not moving in response to the housing limitations from a residence restriction, then it is unlikely that these policies will be effective at reducing the routine activities (and associated criminal opportunities) of RSOs, and thus should not affect the overall rates of recidivistic sex crimes. In this study, the terms “sex crimes” and “sex crime arrests” are used interchangeably.

While some research has indicated that residence restrictions may cause RSOs to become less clustered over time (e.g., Morgan, 2008; Socia, in press; Youstin & Nobles, 2009), other research suggests that residence restrictions may result in RSOs being forced to cluster in a limited number of neighborhoods, potentially increasing RSO clustering over time (e.g., Grubesic, Murray, & Mack, 2011; Socia, 2011). However, high rates of noncompliance also appear to be the norm (e.g., Berenson & Appelbaum, 2011; Schaible & Hughes, 2008; Tewksbury & Mustaine, 2006; Youstin & Nobles, 2009). Compliance with residence restrictions depends on a number of factors that are not easily measured, including the extent that local authorities enforce the policy and that grandfathering provisions exempt certain RSOs from compliance. Further, employment or loitering restrictions may also influence where RSOs are likely to live and, as a result, are likely to cluster in the community. Thus, measuring the extent that residence restrictions affect RSO clustering remains a difficult task, especially when examining large study areas covering multiple jurisdictions.

For instance, one study in upstate New York found that the presence of relatively newer residence restrictions were associated with RSOs living farther apart from each other (i.e., less clustering), while older residence restrictions were associated with RSOs living closer together (i.e., more clustering; Socia, in press). While still cross-sectional, the results of that study suggest that residence restrictions can have varying effects on RSO clustering depending on how long they are in place. That same study found that RSOs lived closer to each other in areas that were more disadvantaged and residentially unstable, had more available and affordable housing, and had higher population densities (Socia, in press). Thus RSOs are more likely to be clustered in the most urban, socially disorganized areas offering available and affordable housing options to RSOs (see also Socia & Stamatel, 2012). More information regarding existing county-level residence restriction policies can be found in the recent work of Socia (2012a, 2012b). While examining the link between these policies and RSO clustering is outside the scope of the present study, it must be kept in mind when considering how these policies may indirectly affect sex crime rates.

Second, and more importantly, where RSOs are living (or clustering) must somehow be associated with rates of recidivistic sex crimes

Despite the rhetoric used by politicians and newspaper columnists to suggest that the increased clustering of RSOs results in increased risk of recidivism (e.g., Bain & German, 2006; Kilgannon, 2008; U.S. States News, 2008), the actual relationship between RSO clustering and recidivistic sex crimes is largely unknown. For instance, some existing research shows that the proximity of RSOs to “child congregation locations” has little direct effect on sexual recidivism (e.g., Colombino, Mercado, Levenson, & Jeglic, 2011; Duwe et al., 2008; Maguire & Singer, 2011; Minnesota Department of Corrections, 2007, p. 28; Zandbergen et al., 2010), while other research has focused on the indirect relationship involving the hardships RSOs face in the community (e.g., Barnes, Dukes, Tewksbury, & De Troye, 2009; Mustaine, Tewksbury, & Stengel, 2006; Socia, 2011; Zandbergen & Hart, 2006). Yet surprisingly little research has focused directly on RSO clustering and recidivistic sex crimes.

In fact, only a single study has found that the presence of RSOs was associated with higher rates of sex crimes against adult victims, but not against child victims (Tewksbury et al., 2010). However, this study examined only a single county (Jefferson County, KY), did not specifically examine recidivistic sex crimes, and examined the number of RSOs in an area, rather than their spatial clustering. Thus, more research is needed that explores how RSO clustering in the community is (or is not) related to recidivistic sex crimes. Without confirming the presence of this second relationship between clustering and sex crime rates, even if RSOs are moving in response to a residence restriction, this movement would be unlikely to affect their ability or likelihood to recidivate. Exploring the connection between RSO clustering and recidivistic sex crimes is the focus of the current study.

County-Level Characteristics and Recidivistic Sex Crime Arrests

Regardless of whether a relationship exists between RSO clustering and rates of recidivistic sex crime arrests, in order to isolate this association, other county-level characteristics associated with RSO clustering and sex crime arrests must be controlled for. In other words, a number of factors unrelated to residence restrictions can influence RSO clustering and/or sex crimes. For instance, counties with more RSOs may have more recidivistic sex crimes (and therefore more recidivistic sex crime arrests) compared to other counties, simply because they contain more potential recidivists. Thus, examining the rate of these arrests compared to the number of RSOs in a county allows for comparisons between counties that have different RSO population sizes.

Prior research also finds that neighborhoods with certain characteristics experience increased numbers of RSOs, as well as increased crime and arrest rates (and reduced reentry success of ex-offenders). These characteristics include indicators of structural social disorganization, such as concentrated disadvantage, residential instability, and ethnic heterogeneity (Hipp, Turner, & Jannetta, 2010; Holzer, Raphael, & Stoll, 2003; Kubrin & Stewart, 2006a; Mulford, Wilson, & Parmley, 2009; Peterson, Krivo, & Harris, 2000; Pratt & Cullen, 2005; Sampson, 1985; Smith & Jarjoura, 1988; Tewksbury et al., 2010), and population density (Sampson, 1985; Smith & Jarjoura, 1988; but see Walker & Ervin-McLarty, 2000). A larger population of residents may yield more suitable targets, or more routine interactions with such targets, and thus more opportunities to commit crime (Cohen & Felson, 1979; Tewksbury, Mustaine, & Stengel, 2008). Finally, controlling for rates of arrest per resident for crime that is unrelated to the clustering of RSOs (e.g., burglary) can help control for some of the county-level crime policy differences that could otherwise affect the rate of recidivistic sex crime arrests, such as differences in policing strategies.

Implications for Policy

Research on whether RSO clustering is associated with rates of recidivistic sex crime arrests can have important implications for policy. For instance, an association could mean that policies and neighborhood characteristics that affect RSO clustering, such as residence restrictions, the placement of half-way houses or treatment centers, or access to public transportation, could indirectly affect recidivistic sex crime arrests and the reentry success of RSOs. Conversely, no association would mean that policies that influence the spatial placement of RSOs would be unlikely to live up to their expectations of reducing sex crimes. In either case, providing policymakers with evidence that a relationship does or does not exist between the spatial distribution of RSOs and rates of recidivistic sex crime arrests can help promote evidence-based policy decisions. This evidence-based decision making may also help to reduce the difficulties that RSOs experience when attempting to reenter a community after their incarceration, and lead to the implementation of policies with the greatest chance of successfully protecting residents from future sexual assault by RSOs.

This study explores whether RSO clustering is associated with rates of county-wide recidivistic sex crime arrests involving either child or adult victims across 52 counties in Upstate New York.¹ Upstate New York provides a particularly useful setting as it contains counties with a wide variation of demographic, social, geographic, and crime characteristics, has a mixture of counties (and local-level jurisdictions) with and without residence restriction policies, does not have an official state-wide residence restriction policy, and excludes counties in and around the New York City area that are not reflective of the rest of the state.² Further, this study controls for other county-level characteristics that may influence rates of recidivistic sex crime arrests.

The Current Study

Drawing on the prior literature regarding RSO residences, reentry hardships, and recidivism, the current study examined whether RSO clustering within a county was associated with county rates of recidivistic sex crime arrests. In doing so, this study controlled for other demographic and social indicators potentially related to rates of recidivistic sex crime arrests.

This study considered two research questions: (a) Are measures of RSO clustering associated with county-level rates of recidivistic sex crime arrests involving child victims, controlling for other demographic, social, and crime indicators? (b) Are measures of RSO clustering associated with county-level rates of recidivistic sex crime arrests involving adult victims, controlling for other demographic, social, and crime indicators?

Method

Data and Sample

The data consisted of county-level data for 52 counties in the upstate New York area. The dataset originally included neighborhoods from 53 counties, but one county did not have any RSOs with valid addresses as of early September 2010, and was thus excluded from the dataset. As of September 2011, this same county only had one RSO listed on the public registry, and thus clustering measures were unable to be computed.

These data came from the 2000 U.S. Census (2002), the 2010 U.S. Census Population Estimates Program (2010), the New York State Office of Sex Offender Management (NYS OSOM; 2010), the New York State Division of Criminal Justice Services Computerized Criminal History Database (NYS DCJS CCH; 2012), and from county-level analysis (not shown) using Geoda (Anselin, Syabri, & Kho, 2006) and ClusterSeer (TerraSeer, 2010). While the use of 2000 Census data was not ideal, 2010 Census data was not available for all measures of interest at the time of this study. While it is possible that some of the demographic characteristics changed between 2000 and 2010, it was expected that any such changes would not be sizable enough at the county level to change overall conclusions. Still, this must be kept in mind when interpreting results.

Dependent Variables

The dependent variables were two measures of annual county-level rates of recidivistic sex crime arrests for the twelve months following the month in which RSO residences were measured. These two measures were calculated as the count of sex crime arrests for crimes committed by offenders with a prior sexual conviction on their record (i.e., RSOs), between October 2010 and September 2011, which involved either (a) child victims, or (b) adult victims.

Note that an arrest was considered as a proxy for commission of the crime, and thus crimes were considered “recidivistic” if the arrested offender had a prior sexual conviction on their official record. Aggregated data on annual recidivistic criminal incidents for each county, separated by victim type, were obtained from the NYS DCJS CCH (2012). If the arrested offender had no prior sex crime conviction on their official record, they were not considered a sexual recidivist, and thus that sex crime was not included in the final dataset.

Separating arrests for recidivistic sex crimes involving child victims from those involving adult victims was important given potential differences in the etiology of these crimes. For example, while sex crimes are more likely to involve acquaintances rather than strangers, the probability of a stranger committing a sex crime against an adult victim is greater than it is against a child victim (Greenfield, 1997). Conversely, the probability of a family member committing a sex crime against a child victim is much greater than it is against an adult victim (Greenfield, 1997). This difference may lead to different relationships between RSO clustering and recidivistic sex crime arrests based on victim type.

The dependent variables were each examined in separate analyses; the first involving the count of recidivistic sex crime arrests involving child victims, and the second involving the count of recidivistic sex crime arrests involving adult victims. Each analysis involved three models, with each model examining a unique measurement of RSO clustering. In each model, an exposure variable was included to account for the population of RSOs in the county (in 100s) as of September 2010. This method essentially converted the recidivistic sex crime arrest counts into rates of recidivistic sex crime arrests per 100 RSOs in a county. Using the population of residents in a county as of September 2010 as an exposure term (combined with adding the population of RSOs as a control variable) did not influence final conclusions in any of the models (results available from the author). This was not surprising, given that the number of RSOs and the population of residents in a county were highly correlated (r = .90).

Primary Independent Variables

The primary independent variables were three county-level measures of RSO clustering as of September 2010. To calculate these variables, RSO residential locations (as listed on the NYS OSOM [2010] public sex offender registry) were first Geocoded using a combination of parcel and point address databases. This yielded a successful match in over 98% of cases. More information on the Geocoding process is available from the author upon request.

As noted earlier, the three variables included one measure of between-offender clustering and two measures of between-neighborhood clustering. Between-offender RSO clustering was measured using a nearest neighbor analysis, averaged to the county level, while between-neighborhood RSO clustering was measured using the revised index of isolation (RII), as well as the Oden’s I*_pop value. Each of the three measures is explained in more detail below, and a graphical example of what each of these measures would look like at the county level is presented in Figure 1.

Figure 1.

Example standardized RSO clustering values at the county level.

Mean county nearest neighbor analysis

The first measure of RSO clustering was a measure of between-offender RSO clustering based on a nearest neighbor analysis (NNA). The NNA technique measured the average distance for each RSO to their five nearest RSO neighbors. Thus, NNA essentially measured the spatial distance (in feet) between nearby RSOs, regardless of what neighborhood they lived in. Restricting the NNA measure to the five nearest RSO neighbors helped ensure that RSO clustering was measured in relation to RSOs in nearby proximity (i.e., a “cluster” of RSOs). This limited the extent that variations in the overall physical county size could have affected this measure. This measure was aggregated to the county level by taking the average NNA measure for all RSOs living in the county. For comparison purposes, this variable was reverse coded and standardized to have a mean of 0 and a standard deviation of 1. Higher NNA values indicated RSOs were living closer together (i.e., relatively more RSO clustering), and lower NNA values indicated RSOs were living farther apart (i.e., relatively less RSO clustering).

Revised index of isolation

The second measure of RSO clustering was RII. This was a measure of between-neighborhood RSO clustering that measured how isolated (or conversely, how clustered) RSOs were between neighborhoods in a county. The measure compared the probable interaction of RSOs within a neighborhood, given the current distribution of RSOs and residents, to the probable interaction of RSOs within a neighborhood if RSOs were homogenously distributed across neighborhoods within the county based on the underlying distribution of residents. In essence, this measure indicated whether RSOs were unequally distributed between neighborhoods in a county, regardless of the spatial locations of these neighborhoods, based on the actual distribution of residents among these neighborhoods. Similar to the mean county NNA, this measure was standardized to have a mean of 0 and a standard deviation of 1. Higher RII values indicated RSOs were more unevenly distributed across neighborhoods in a county (i.e., relatively more RSO clustering), and lower RII values indicated RSOs were more evenly distributed in a county (i.e., relatively less RSO clustering). For more information on how RII was calculated, see Appendix.

Oden’s I*_pop

The third measure of RSO clustering was an alternative measure of between-neighborhood RSO clustering (Oden, 1995). Unlike RII, Oden’s I*_pop measured the spatial clustering of neighborhoods based on rates of RSOs. As such, this measure was similar to the Moran’s I statistic, but accounted for the differences in the underlying resident population across all neighborhoods in a county, which could have ultimately affected the number of RSO residences contained in those neighborhoods. This measure was standardized to have a mean of 0 and a standard deviation of 1. Higher Oden’s I*_pop values indicated that neighborhoods with similar RSO rates were relatively closer together in a county (i.e., more/positive spatial clustering of neighborhoods based on RSO rates), lower Oden’s I*_pop values indicated that neighborhoods with similar RSO rates were evenly distributed across the neighborhood (i.e., less/negative spatial clustering), and Oden’s I*_pop values near zero indicated no spatial pattern based on RSO rates (i.e., random spatial clustering).

Control Variables

Each model included demographic and social indicators potentially associated with the rate of recidivistic sex crime arrests. These included the three previously mentioned indicators of social disorganization (i.e., concentrated disadvantage, residential instability, and ethnic heterogeneity), population density, physical county size, and the burglary arrest rate.

Concentrated disadvantage was measured as a factor score based on the following: Percent living in poverty, percent unemployed, percent female heads of household with children, and percent non-Hispanic Black residents. Residential instability was measured as a factor score based on the following: Percent owner-occupied homes and percent residents five years and older who have lived in the same house for at least five years, and was reverse coded. Ethnic heterogeneity was measured as a factor score based on the following: Percent Hispanic residents and percent foreign born residents. Population density was measured as the number of residents per 100 square miles of land area. Physical county size was measured as square miles of land area. The burglary arrest rate was measured as the number of burglaries, from October 2010 through September 2011, per 100,000 residents.

To account for higher rates of recidivistic sex crime arrests in a county stemming from the greater availability of potential victims, the models originally included a measure of the annual population of residents in the entire county. However, since this measure was highly correlated (r = .84) with the population density, it was excluded from the final models. Substituting population of residents in place of population density did not influence conclusions.

Analytical Model

Because the dependent variables were abnormally distributed counts, the analysis used multivariate Poisson regression models to evaluate the relationship between each of the three different measures of county-level RSO clustering in September 2010 and the two counts of recidivistic sex crime arrests for the subsequent 12 months. As noted earlier, an exposure term controlled for differences in the RSO population in each county (in 100s), and essentially converted counts of arrests into rates of arrests per 100 RSOs. Control variables accounted for other demographic, social, and crime indicators potentially related to RSO clustering and/or recidivistic sex crime rates. A likelihood-ratio test of each model (analysis not shown) indicated that a Poisson model was preferred over a Negative Binomial model. Further, a Vuong test indicated the use of a Zero Inflated Poisson model was not appropriate.

Results

Descriptive statistics are presented in Table 1. An examination of the correlation matrix (available from the author) indicated that while some of the independent variables have strong bivariate correlations (r > .60), there were no independent variable correlations at or above .80 after exclusion of the population of residents from the models. RII was highly collinear with Oden’s I*_pop (r = .96). However, as these variables were not included within the same model, this did not present a problem for the individual models. Additionally, the fact that these two measures were highly collinear suggests they are both measuring a similar construct of between-neighborhood RSO clustering. It was also not surprising that neither was highly collinear with mean county NNA, as it measures between-offender RSO clustering, which is an inherently different construct. Further, examinations of the correlation matrix of the estimated coefficients and the (uncentered) variance inflation factors of each model suggested no multicollinearity concerns. As such, it is unlikely that multicollinearity presented a problem for interpreting model estimates. Tests of model specification indicated no concerns with omitted variable bias.

Table 1.

Descriptive Statistics.

Variable	N	Mean	SD	Minimum	Maximum
Rate of recidivistic sex crime arrests involving child victims per 100 RSOs^a	52	0.49	0.69	0	3.33
Rate of recidivistic sex crime arrests involving adult victims per 100 RSOs^a	52	1.40	1.38	0	4.84
Mean county NNA^b	52	0	1	−3.15	1.44
Revised index of isolation^b	52	0	1	−0.61	5.09
Oden’s I_pop*^b	52	0	1	−0.68	3.94
Concentrated disadvantage	52	0.02	0.81	−2.59	1.57
Residential instability	52	0.03	0.82	−1.50	3.40
Ethnic heterogeneity	52	0.02	0.84	−0.96	2.78
Population density (residents per 100 square miles of land area)	52	1.92	2.26	0.21	11.17
County size (100 square miles of land area)	52	8.34	4.51	2.06	26.86
Burglary arrest rate (per 100,000 residents)	52	89.17	28.76	26.19	161.86

Note: NNA = nearest neighbor analysis; RSO = registered sex offenders.

Variable has been converted from a count into a rate per 100 RSOs in the county to account for the use of an exposure term in the model. Crimes are measured from October 2010 through September 2011. The average number of RSOs in a county was 136.56 (SD = 129.64).

Variable has been standardized to have a mean of 0 and a standard deviation of 1, with higher values indicating more RSO clustering.

Results are presented in Tables 2 and 3 as unstandardized and exponentiated coefficients (i.e., incidence rate ratios) for each model. For ease of interpretation, the rest of the study interprets model coefficients as incidence rate ratios. Robust standard errors were used to account for any unforeseen issues with model specification and assumption violations.

Table 2.

County Characteristics Associated with Recidivistic Sex Crime Arrests Involving Child Victims.

	Rate of recidivistic sex crime arrests involving child victims
	1		2		3
Variable	b	e ^b	b	e ^b	b	e ^b
Mean county NNA^a	0.14	1.14
	(0.21)	(0.24)
Revised index of isolation^a			0.07	1.07
			(0.11)	(0.12)
Oden’s I_pop*^a					0.09	1.09
					(0.14)	(0.15)
Concentrated disadvantage	0.28	1.33	0.28	1.32	0.27	1.31
	(0.34)	(0.44)	(0.32)	(0.43)	(0.33)	(0.43)
Residential instability	0.11	1.11	0.08	1.09	0.08	1.08
	(0.21)	(0.24)	(0.21)	(0.23)	(0.21)	(0.23)
Ethnic heterogeneity	−0.26	0.77	−0.22	0.81	−0.21	0.81
	(0.30)	(0.23)	(0.30)	(0.24)	(0.31)	(0.25)
Population density	−0.09	0.92	−0.08	0.92	−0.09	0.92
	(0.07)	(0.07)	(0.07)	(0.06)	(0.07)	(0.06)
County size	−0.06	0.95	−0.07*	0.94*	−0.07*	0.94*
	(0.03)	(0.03)	(0.03)	(0.02)	(0.03)	(0.02)
Burglary arrest rate	0.00	1.00	−0.01	1.00	−0.01	0.99
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Constant	0.29	1.33	0.51	4.66	0.53	1.69
	(0.77)	(1.03)	(0.75)	(1.24)	(0.75)	(1.26)
N	52		52		52

Note: All models include an exposure term measuring the county RSO population (in 100s) as of September 2010. Robust standard errors are in parentheses. NNA = nearest neighbor analysis.

Variable has been standardized to have a mean of 0 and a standard deviation of 1, with higher values indicating more RSO clustering.

p < .05 (two-tailed).

Table 3.

County Characteristics Associated with Recidivistic Sex Crime Arrests Involving Adult Victims.

	Rate of recidivistic sex crime arrests involving adult victims
	1		2		3
Variable	b	e ^b	b	e ^b	b	e ^b
Mean county NNA^a	0.02	1.02
	(0.11)	(0.12)
Revised index of isolation^a			0.10*	1.11*
			(0.04)	(0.04)
Oden’s I_pop*^a					0.10*	1.10*
					(0.05)	(0.05)
Concentrated disadvantage	0.37*	1.45*	0.41*	1.50*	0.38*	1.47*
	(0.18)	(0.26)	(0.19)	(0.28)	(0.18)	(0.27)
Residential instability	0.06	1.06	0.01	1.01	0.02	1.02
	(0.10)	(0.11)	(0.11)	(0.11)	(0.11)	(0.11)
Ethnic heterogeneity	0.03	1.03	0.07	1.07	0.07	1.07
	(0.11)	(0.11)	(0.11)	(0.12)	(0.11)	(0.12)
Population density	−0.12*	0.89*	−0.13*	0.87*	−0.14*	0.87*
	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)	(0.03)
County size	−0.11*	0.89*	−0.12*	0.89*	−0.12*	0.89*
	(0.04)	(0.03)	(0.04)	(0.03)	(0.04)	(0.03)
Burglary arrest rate	0.00	1.00	0.00	1.00	0.00	1.00
	(<0.01)	(<0.01)	(<0.01)	(<0.01)	(<0.01)	(<0.01)
Constant	1.33*	3.79*	1.61*	4.99*	1.56*	4.74*
	(0.32)	(1.21)	(0.33)	(1.65)	(0.34)	(1.59)
N	52		52		52

Note: All models include an exposure term measuring the county RSO population (in 100s) as of September 2010. Robust standard errors are in parentheses. NNA = nearest neighbor analysis.

Variable has been standardized to have a mean of 0 and a standard deviation of 1, with higher values indicating more RSO clustering.

p < .05 (two-tailed).

Results are described in two sections. The first section describes results of the models involving the rate of recidivistic sex crime arrests involving child victims (Table 2). The second section describes results of the models involving the rate of recidivistic sex crime arrests involving adult victims (Table 3).

Recidivistic Sex Crimes Involving Child Victims

The relationship between RSO clustering and rates of recidivistic sex crime arrests involving child victims was nonsignificant (p > .05) for all three measures of RSO clustering (Table 2). However, all relationships with the variables of interest were positive. Specifically, a one standard deviation increase in mean county NNA, RII, or Oden’s I*_pop was associated with an increase of 14, 7, and 9 percent, respectively, in the rate of recidivistic sex crimes committed against child victims (Table 2). In other words, when RSOs were one standard deviation more clustered in a county, rates of recidivistic sex crimes were expected to be between 7% and 14% higher, depending on the type of RSO clustering. While all three of these increases were nonsignificant, given the relative rarity of these crimes and the small number of observations, it is not surprising that the results did not achieve statistical significance. However, while this does not confirm a positive association, it does suggest that there is not a negative relationship. In other words, the data suggest that increased RSO clustering would not yield fewer recidivistic sex crimes.

Results also suggest that physically larger counties were associated with lower rates of recidivistic sex crimes against children. That is, every additional 100 square miles of land area in a county’s size was associated with about 6% fewer recidivistic sex crimes against children, controlling for the number of RSOs living in the county. While other control variables had large associations with recidivistic sex crime rates against children (e.g., concentrated disadvantage, ethnic heterogeneity), they did not achieve statistical significance. The fact that there are almost no significant findings in all three models involving child victims may suggest the presence of Type II error due to a low number of observations, or may suggest problems with the low base rate of such crimes, which results in little variance in the dependent variable.

Recidivistic Sex Crimes Involving Adult Victims

As shown in Table 3, the relationship between RSO clustering and the rate of recidivistic sex crime arrests involving adult victims was non-significant for the between-offender RSO clustering measure, but was significant (p < .05) for both of the between-neighborhood measures. Similar to the results of the child victim models, all relationships with the variables of interest were positive. Specifically, a one standard deviation increase in mean county NNA, RII, or Oden’s I*_pop was associated with an increase of 2%, 11%, and 10%, respectively, in the rate of recidivistic sex crimes committed against adult victims (Table 3). In other words, when RSOs lived one standard deviation closer together in a county, recidivistic sex crime rates against adult victims were expected to be about 2% higher (a nonsignificant increase). However, when more RSOs were clustered into a county, or RSOs were more clustered into spatially proximate parts of a county, recidivistic sex crime rates against adult victims were expected to be between 10% and 11% higher, which were both significant increases.

Other control variables achieved statistical significance in the adult victim models. Specifically, a one standard deviation increase in concentrated disadvantage was associated with between 45% and 50% higher rates of recidivistic sex crimes against adult victims. Disadvantage has frequently been associated with violent crime rates as well as offender recidivism (Kubrin & Stewart, 2006b), and so this finding is not surprising (and also supports the positive but nonsignificant association found in all three child victim models). Further, increased population density was associated with lower rates of recidivistic sex crimes against adults. While this finding may seem surprising, remember that increased population density may also increase the overall number of RSOs living in an area, thus potentially reducing the ‘rate’ of recidivistic sex crimes when controlling for the number of RSOs. Similar to the child victim models, counties with more physical land area were expected to have significantly lower rates of recidivistic sex crimes against adult victims.

Overall Results

Overall these results suggest that RSO clustering, regardless of how it was measured, was not significantly associated with rates of recidivistic sex crimes involving child victims. At best, these results may indicate a weakly positive association, depending on the presence of Type II error due to the low number of observations in the models. Further, between-offender RSO clustering was not associated with rates of recidivistic sex crime arrests involving adult victims. Thus, when RSOs lived closer together, rates of adult victim recidivistic sex crimes were not expected to be significantly higher (or lower).

However, between-neighborhood RSO clustering was significantly and positively associated with rates of recidivistic sex crimes against adult victims. This suggests that the mechanisms that involve between-offender clustering are likely different than those that involve between-neighborhood clustering, particularly in terms of the association with recidivistic sex crimes against adult victims. This also suggests that the mechanisms influencing recidivistic sex crimes involving child victims may differ from those involving adult victims.

Still, all of the increases in sex crime rates were relatively modest, between 2% and 14% higher per standard deviation increase in RSO clustering, regardless of how RSO clustering was measured or the type of victim involved. Further, only two of the six associations achieved statistical significance. Thus RSO clustering, at most, had only a weak positive relationship with rates of recidivistic sex crimes against adult victims.

Discussion

This study was concerned with the underexplored relationship between RSO clustering and rates of recidivistic sex crime arrests involving either child or adult victims. Results indicated that all three measures of RSO clustering (see Figure 1) were positively associated with rates of recidivistic sex crime arrests involving child victims, but all three were also nonsignificant (Table 2).Thus, when controlling for the number of RSOs in an area, all three types of RSO clustering did not appear to have significant associations with rates of recidivistic sex crimes against child victims. Similarly, between-offender RSO clustering was also positively but nonsignificantly associated with rates of recidivistic sex crime arrests involving adult victims. Thus, how close RSOs live to each other (NNA, Figure 1) did not have a significant association with rates of recidivistic sex crimes against adults.

However, both measures of between-neighborhood RSO clustering (RII and Oden’s I*_pop, Figure 1) were positively and significantly associated with rates of recidivistic sex crime arrests involving adult victims. Thus, counties with more uneven distributions of RSOs between neighborhoods, or with neighborhoods that are more spatially clustered based on RSO rates, were expected to have significantly higher rates of recidivistic sex crime arrests involving adult victims. However, the expected increase is relatively modest in size, with a standard deviation increase in between-neighborhood RSO clustering yielding only a 10% to 11% increase in recidivistic sex crimes against adult victims, controlling for the number of RSOs in the area.

There are a few different explanations that may account for the two significant and (albeit weakly) positive findings. First, formal or informal social control measures within a neighborhood may become overwhelmed by the vast number of RSOs concentrated into relatively small areas of a county. This could reduce the extent that RSOs are subject to social control mechanisms, increasing their likelihood of sexual recidivism involving adult victims. Second, when RSOs are relegated to limited neighborhoods within a county, the hardships associated with this relegation (e.g., lack of employment opportunities, access to treatment facilities, supportive family members) may increase the likelihood of sexual recidivism involving adult victims (Bonnar-Kidd, 2010; Levenson & Cotter, 2005; Mercado, Alvarez, & Levenson, 2008). In addition, other policies, such as employment and loitering restrictions, may also aggravate these hardships. Third, increased clustering could increase the interactions between RSOs, which may reinforce attitudes conducive to future criminal behavior, as has been suggested by numerous criminological theories (see Akers, 1973; Burgess & Akers, 1966; Shaw & McKay, 1942; Sutherland, 1947). Finally, the informal social control mechanisms of neighborhoods may become overwhelmed when there are many potential “predators” that need to be monitored.

In any case, these tentatively suggest an indirect relationship between where RSOs are living (and clustering), and rates of recidivistic sex crimes against adult victims. Thus, policies or other circumstances that severely limit where RSOs can live and work in the community and/or increase RSO clustering may be counterproductive at protecting adults from recidivistic sex crimes, as well as ineffective at protecting children from recidivistic sex crimes.

The relative differences between the child and adult victim models make sense when one considers the differences between offender and victim relationships. As noted earlier, sex crimes involving child victims are less likely to involve strangers, and more likely to involve acquaintances or family members (Greenfield, 1997; Maguire & Singer, 2011; MNDOC, 2007). Thus, it may be that increasing RSO clustering increases the likelihood of RSOs interacting with adult strangers who are potential victims, but does not increase their ability to become acquainted with potential child victims. More research on the individual-level mechanisms influencing RSO recidivism, particularly when separated both by victim type and relationship to victim, would help support these hypotheses.

However, one must also consider the relatively high likelihood of Type II error with such a small number of observations (N = 52) and the relatively rare occurrence of recidivistic sex crimes. Setting aside the issue of statistical significance for the moment, in all six models increased RSO clustering was associated with modest (2%-14%) increases in rates of recidivistic sex crimes. Thus, all instances of RSO clustering may be positively associated with recidivistic sex crime rates, but the current data are limited in finding significance in these associations, especially for the models involving child victims. Further, even if the results suffered from Type II error, the overall positive relationships are still quite modest in size across all six models. In other words, even with a significant relationship, it would require a large change in RSO clustering levels to have even a small effect on recidivistic sex crime rates.

It is also important to address what the results do not suggest. That is, despite their nonsignificance, none of the models suggest that increased RSO clustering was associated with decreased recidivistic sex crime arrests involving either child or adult victims. Thus, even if there were Type II error, the results would almost certainly indicate a positive relationship (i.e., that increased RSO clustering is associated with increased recidivistic sex crime arrests), rather than a negative one.

As such, when policies such as residence restrictions increase RSO clustering, these results suggest they would not have the crime-reducing effect policymakers and residents are expecting. At worst, they may indirectly increase recidivistic sex crimes, albeit modestly. This also supports the concerns of residents and policymakers regarding RSO clusters, as noted earlier. However, these results also suggest that policies that lead to decreased RSO clustering might indirectly lead to modest decreases in recidivistic sex crimes, at least in certain instances. The effort required to effect these decreases, however, would likely be substantial.

Limitations

This study has a number of limitations that should be noted. First, as this study used data only from counties in upstate New York, the generalizability of these findings are obviously limited to areas similar to upstate New York. As such, future research on other states, or that examined the New York City area, would be useful to support the conclusions of this study.

Additionally, given the relatively low number of observations (52) and the low variation in the dependent variables, Type II error is a concern, particularly given the finding of positive (but frequently nonsignificant) effects across all six models. This is important, especially since RSO clustering achieved significance in only two of the six models. That is, if Type II error were present, it would support the finding that increased RSO clustering is weakly associated with increased rates of recidivistic sex crime arrests regardless of victim type.

Another limitation is the assumption that RSOs will commit future sex crimes in their county of residence. If RSOs were likely to commit sex crimes (or be arrested for such crimes) in other counties, then the association between the spatial distribution of RSOs and recidivistic sex crime arrest rates in a county may be unduly influenced by RSOs in neighboring counties. Future studies of individual-level offender recidivism data would help determine the extent that RSOs commit subsequent sex crimes, and are arrested for such crimes, in other counties.³

Measuring sex crimes using arrests may also be a limitation. The relatively low rates of both reporting and conviction for sex crimes represent problems when attempting to measure the “real” rate of such crimes in a community. As such, using the rate of arrests was considered a reasonable middle ground between high rates of unreported sex crimes (for which data does not exist at the county level) and low rates of conviction. Thus, the arrest rate was expected to be a reasonable proxy of the rate of actual recidivistic sex crimes.

The extent that residence restrictions influence where RSOs are actually living is another concern. That is, some RSOs may be grandfathered from compliance with these restrictions if they had established their residence prior to the implementation of the law and/or if they owned their homes. Further, the enforcement of residence restrictions, or other policies such as employment restrictions, may vary between areas, which can influence how RSOs cluster in the community. However, these influences on RSO clustering, while important for determining why and how RSOs cluster in a community (see Socia, in press), are less relevant for the present study as it explores the association between the existing residential clustering of RSOs and recidivistic sex crime rates. While the “why” of RSO clustering is outside the scope of this study, if one wants to extrapolate results to inform on the efficacy of residence restrictions, one must consider the actual influence these policies have on RSO clustering.

Perhaps the most important limitation involves the lack of historical/longitudinal data. Unfortunately, historical data on RSO residences were unavailable, and thus this study was limited to examining cross-sectional associations with a 1 year lag period between residence and crime rate measurements, rather than a causal relationship. Further, a longer follow up period may provide more recidivistic sex crimes to study, which is particularly important given the relatively low rate of RSO recidivism. However, the low base rate of recidivism makes it unlikely that even substantially increased recidivistic sex crime arrests would affect the overall spatial distribution of RSOs through the removal of RSOs from the community (and into incarceration). Thus, were a causal relationship to exist, it would likely involve the spatial distribution of RSOs affecting the rate of recidivistic sex crime arrests, rather than the reverse. Future research using longitudinal data would help to support this contention.

On a related note, while the use of 2000 Census data was not ideal, 2010 Census data was not available for all measures of interest at the time of this study. While it is possible that some of the demographic characteristics changed between 2000 and 2010, it was expected that any such changes would not be sizable enough at the county level to change overall conclusions. Still, this must be kept in mind when interpreting results.

Implications for Policy and Future Research

The findings of the present study are important given policymakers’ attempts to legally control where RSOs are able to live in a community through the passage of policies like residence restrictions. That is, if policies result in increased RSO clustering, they will at best be ineffective, and at worst, may indirectly increase such crimes. However, if these policies are able to more equitably distribute RSOs throughout a county (i.e., reduce RSO clustering), then they may be indirectly decreasing the rates of recidivistic sex crime arrests (and associated sex crimes), albeit modestly.⁴ This means that policies and neighborhood characteristics that encourage RSOs to find housing across larger areas of a community, such as widespread access to public transportation, availability of employment opportunities, or equitable distributions of halfway houses and treatment facilities across different neighborhoods, may help indirectly decrease recidivistic sex crimes.

Further, modeling techniques that help explore more equitable distributions of RSOs, such as those described by Grubesic and Murray (2008), may be useful to policymakers seeking to address existing (and future) clusters of RSOs while still maintaining the safety of residents. These policy decisions would be valuable despite evidence that the basic assumption of such laws—that RSOs living near schools and daycares are more likely to recidivate—has been largely disproven (MNDOC, 2007; Colombino et al., 2011; Duwe et al., 2008; Maguire & Singer, 2011; Zandbergen et al., 2010). That is, even if RSOs do not recidivate more often as a result of increased or decreased clustering, increased RSO clustering may still disproportionately expose certain residents (i.e., those living near RSO clusters) to higher risks of sexual victimization.

This results in a balancing act for policymakers. On one hand, clustering RSOs into only a few areas may help contain some of the risk of sexual victimization, at least for those recidivistic crimes committed by RSOs. The downside is that some residents (i.e., those living in high-concentration RSO areas) may be subject to much higher risks of sexual victimization. On the other hand, stopping RSOs from clustering in any particular area can help to more equitably distribute risk among all residents. In either case, if future research does not consider the indirect relationships between RSO clustering and recidivistic sex crimes aside from the spatial proximity to schools and daycares, an important connection to policy and practice will be overlooked.

This study has also provided a number of avenues for future research. One such avenue involves exploring the spatial location of individuals RSOs in a community, and how that is associated with their propensity to recidivate. Such research can help further examine whether the spatial distribution of RSOs is, in aggregate, related to the rate of recidivistic sex crimes or arrests, and to what extent RSOs recidivate outside of their local community or county.

A second avenue for future research involves replicating this study using data from other states and/or longitudinal data on RSO residences. As noted in the limitations section, the findings of this study are limited in their generalizability and overall power. Thus, research findings involving other states and/or longitudinal data would both help support the findings and generalizability of the current study.

This study also provided some best practices for future research as it relates to measuring the spatial distribution of RSOs at the county level. These measures (mean county NNA, RII, and Oden’s I*_pop) provide future researchers with a blueprint for measuring between-offender and between-neighborhood clustering in ways that are both methodologically sound and conceptually distinct. Further, as shown by this study, how RSO clustering is measured can have important implications on results.

Conclusions

Overall, this study found only partial support for the assumption of policymakers and residents that sex offender clustering is related to sexual recidivism (e.g., Kilgannon, 2008; U.S. States News, 2008). That is, the present study found that more RSO clustering was weakly related to higher rates of recidivistic sex crimes, but this relationship was statistically significant only for the two measures of between-neighborhood RSO clustering, and only when considering recidivistic sex crime arrests involving adult victims.

The most important policy implication from this study concerns the future of policies and practices that result in the relegation of sex offenders to certain areas of a community. Specifically, the results of this study, combined with the limited prior research, imply that policies such as residence restrictions may not be appropriate methods to protect citizens from recidivistic sex crimes if they result in the increased clustering of RSOs. That is, both the weak significant findings and the nonsignificant findings suggest that increased RSO clustering does not result in decreased sex crime rates. Even if policies resulted in decreased RSO clustering, results suggest that recidivistic sex crimes involving adult victims would only experience modest decreases at best.

However, if these policies and practices can more equitably distribute RSOs across a county, city, or town, then even without a decrease in recidivistic sex crimes, the risk of sex crimes may be more equitably distributed among residents. One must decide whether it is better to have RSOs living in clusters, where the risk of sexual victimization is geographically constrained but potentially higher, or to have RSOs distributed more evenly around a community, resulting in a more equitable distribution of the risk of sexual victimization.

Thus, while laws such as residence restrictions are obviously popular with both politicians and residents, the overall results suggest that they are simply not very useful at reducing the risk of sexual victimization by RSOs. Further, if they lead to increased RSO clustering, then not only would residents not be protected from future sex crimes, but they may actually face increased risks. In any event, it may be more beneficial to focus on other ways of promoting successful reentry of RSOs, such as the spatial placement of half-way houses, treatment centers, or access to public transportation, rather than on ways to legally constrain where RSOs are living.

While the call for more research is the standard bearer of social sciences, in this case it is not a mindless cliché. Where RSOs are living, how this affects residents’ health and safety, and how this influences RSO reentry success (and recidivism risk) are all important questions that can have implications for communities in terms of crime control and public safety. But such research must also be paired with a change in the ways public policies are typically implemented. In terms of sex offender laws, these policies typically come as knee-jerk reactions to tragic but extremely rare events, are based on little existing research, and (at least thus far) are rarely revised or ultimately overturned by policymakers as the result of future research evidence. Without a closer link between research evidence and the passage and modification of reentry policies, particularly for sex offenders, it seems unlikely that the policies most likely to be implemented will also be the most likely to succeed.

Footnotes

Appendix

Acknowledgements

The author would like to thank Sam DeWitt, Christopher Dum, Andrew Wheeler, and Drs. Robert Apel, Jill Levenson, Alan Lizotte, Steven Messner, Greg Pogarsky, Allison Redlich, and Janet Stamatel for their helpful comments on and readings of earlier versions of this study. In addition, the author thanks David van Alstyne and the New York Division of Criminal Justice Services for providing much of the data used in the study. Three anonymous reviewers, Dr. Michael Seto, and Dr. James Cantor also provided helpful suggestions in revising this article for publication. The author is probably responsible for any and all errors.

Declaration of Conflicting Interests

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

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This project was supported by Award No. 2010-IJ-CX-0004, awarded by the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. Further, data on the New York State Sex Offender Registry as well as county-level crime statistics were supplied by the New York State Division of Criminal Justice Services. The opinions, findings, and conclusions or recommendations expressed in this publication are those of the author and do not necessarily reflect those of the Department of Justice, the New York State Division of Criminal Justice Services, or anyone else of importance and stature. Neither New York State nor DCJS assumes liability for its contents or use thereof.

Notes

Author Biography

Kelly M. Socia is an assistant professor in the School of Criminology and Justice Studies at the University of Massachusetts, Lowell. He received his PhD in criminal justice from the School of Criminal Justice at the University at Albany, SUNY. His research interests include offender reentry and recidivism, registered sex offenders, public policymaking, geographic information systems, and spatial analyses. His recent publications have appeared in journals such as Crime and Delinquency, Criminology and Public Policy, and Psychology, Public Policy, and Law.

References

Adams

D. B.

(2002). Summary of state sex offender registries, 2001 (Bureau of Justice Statistics, Office of Justice Programs, Trans.): Washington, DC: Department of Justice.

Akers

R. L.

(1973). Deviant behavior—A social learning approach. Belmont, CA: Wadsworth.

Anderson-Facile

(2009). Basic challenges to prisoner reentry. Sociology Compass, 3, 183-195. doi:10.1111/j.1751-9020.2009.00198.x

Anselin

Syabri

Kho

(2006). GeoDa: An introduction to spatial data analysis. Geographical Analysis, 38(1), 5-22. doi:10.1111/j.0016-7363.2005.00671.x

Bain

German

(2006, September 26). How a cluster grew so large: Low rents, willing landlords and politics play roles. Newsday. Retrieved from http://find.galegroup.com/gtx/infomark.do?&contentSet=IAC-Documents&type=retrieve&tabID=T004&prodId=SPN.SP04&docId=CJ151896742&source=gale&srcprod=SP04&userGroupName=albanyu&version=1.0

Barnes

J. C.

Dukes

Tewksbury

De Troye

T. M.

(2009). Analyzing the impact of a statewide residence restriction law on South Carolina sex offenders. Criminal Justice Policy Review, 20(1), 21-43. doi:10.1177/0887403408320842

Beck

A. J.

Shipley

B. E.

(1987). Recidivism of young parolees. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

Beck

V. S.

Clingermayer

Ramsey

R. J.

Travis

L. F.

(2004). Community response to sex offenders. Journal of Psychiatry & Law, 32, 141-168.

Beck

V. S.

Travis

L. F.

(2004). Sex offender notification and fear of victimization. Journal of Criminal Justice, 32, 455-463. doi:10.1016/j.jcrimjus.2004.06.007

10.

Berenson

J. A.

Appelbaum

P. S.

(2011). A geospatial analysis of the impact of sex offender residency restrictions in two New York counties. Law and Human Behavior, 35, 235-246. doi:10.1007/s10979-010-9235-3

11.

Bonnar-Kidd

K. K.

(2010). Sexual offender laws and prevention of sexual violence or recidivism. American Journal of Public Health, 100, 412-419. doi:10.2105/ajph.2008.153254

12.

Brookhaven.org. (2008, June 10). Brookhaven Child Protection Act passes with 7-0 vote. Retrieved from http://www.brookhaven.org/PressRoom/tabid/56/newsid970/363/Brookhaven-Child-Protection-Act-Passes-with-7-0-Vote/Default.aspx

13.

Burchfield

K. B.

(2011). Residence restrictions. Criminology & Public Policy, 10, 411-419. doi:10.1111/j.1745-9133.2011.00716.x

14.

Burgess

R. L.

Akers

R. L.

(1966). A Differential Association-Reinforcement Theory of criminal behavior. Social Problems, 14(2), 128-147.

15.

Cohen

L. E.

Felson

(1979). Social change and crime rate trends: A routine activities approach. American Sociological Review, 44, 588-608. doi:10.2307/2094589

16.

Colombino

Mercado

C. C.

Levenson

J. S.

Jeglic

(2011). Preventing sexual violence: Can examination of offense location inform sex crime policy? International Journal of Law and Psychiatry, 34(3), 160-167. doi:10.1016/j.ijlp.2011.04.002

17.

Council of State Governments. (2005). Report of the re-entry policy council: Charting the safe and successful return of prisoners to the community. New York: Author, Reentry Policy Council.

18.

Center for Sex Offender Management (CSOM). (2008). Legislative trends in sex offender management. New York: Author.

19.

Duwe

Donnay

Tewksbury

(2008). Does residential proximity matter? A geographic analysis of sex offense recidivism. Criminal Justice and Behavior, 35, 484-504. doi:10.1177/0093854807313690

20.

Greenfield

L. A.

(1997). Sex offenses and offenders: An analysis of data on rape and sexual assault (Bureau of Justice Statistics, Trans.). Washington, DC: National Institute of Justice.

21.

Grubesic

T. H.

Murray

A. T.

(2008). Sex offender residency and spatial equity. Applied Spatial Analysis and Policy, 1(3), 175-192. doi:10.1007/s12061-008-9013-5

22.

Grubesic

T. H.

Murray

A. T.

Mack

(2011). Sex offenders, residence restrictions, housing, and urban morphology: A review and synthesis. Cityscape: A Journal of Policy Development and Research, 13(3), 7-31.

23.

Gustavson

(2011, July 5). Brookhaven investigating Coram Homes housing multiple sex offenders. North Shore Sun. Retrieved from http://northshoresun.timesreview.com/2011/07/13784/brookhaven-investigating-coram-homes-housing-multiple-sex-offenders/

24.

Hipp

J. R.

Turner

Jannetta

(2010). Are sex offenders moving into social disorganization? Analyzing the residential mobility of California parolees. Journal of Research in Crime and Delinquency, 47, 558-590. doi:10.1177/0022427810381093

25.

Holzer

Raphael

Stoll

M. A.

(2003). Employment barriers facing ex-offenders. Washington, DC: Urban Institute.

26.

Hughes

Wilson

D. J.

(2003). Reentry trends in the United States. Washington, DC: U.S. Department of Justice, Bureau of Justice Statistics.

27.

Kernsmith

P. D.

Craun

S. W.

Foster

(2009). Public attitudes toward sexual offenders and sex offender registration. Journal of Child Sexual Abuse, 18, 290-301. doi:10.1080/10538710902901663

28.

Kilgannon

(2008, January 24). Fears anew in an area rife with sex offenders. New York Times, p. B3.

29.

Kubrin

C. E.

Stewart

E. A.

(2006a). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44(1), 165-197. doi:10.1111/j.1745-9125.2006.00046.x

30.

Kubrin

C. E.

Stewart

E. A.

(2006b). Predicting who reoffends: The neglected role of neighborhood context in recidivism studies. Criminology, 44(1), 165-197. doi:10.1111/j.1745-9125.2006.00046.x

31.

Langan

P. A.

Schmitt

E. L.

Durose

M. R.

(2003). Recidivism of sex offenders released from prison in 1994. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Bureau of Justice Statistics.

32.

Larsen

J. E.

Lowrey

K. J.

Coleman

J. W.

(2003). The effect of proximity to a registered sex offender’s residence on single-family house selling price. Appraisal Journal, 71, 253-265.

33.

Levenson

J. S.

(2009). Sex offender residence restrictions. In Wright

R. G.

(Ed.), Sex offender laws: Failed policies, new directions (pp. 267-290). New York: Springer.

34.

Levenson

J. S.

Brannon

Y. N.

Fortney

Baker

(2007). Public perceptions about sex offenders and community protection policies. Analysis of Social Issues and Public Policy, 7(1), 1-25.

35.

Levenson

J. S.

Cotter

L. P.

(2005). The impact of sex offender residence restrictions: 1,000 feet from danger, or one step from absurd? International Journal of Offender Therapy and Comparative Criminology, 49(2), 168-178. doi:10.1177/0306624X04271304

36.

Levenson

J. S.

D’Amora

D. A.

(2007). Social policies designed to prevent sexual violence: The emperor’s new clothes. Criminal Justice Policy Review, 13(8), 168-199. doi:10.1177/0887403406295309

37.

Linden

L. L.

Rockoff

J. E.

(2008). Estimates of the impact of crime risk on property values from Megan’s Law. American Economic Review, 98, 1103-1127. doi:10.1257/aer.98.3.1103

38.

Maguire

Singer

(2011). A false sense of security: Moral panic driven sex offender legislation. Critical Criminology, 19, 301-312. doi:10.1007/s10612-010-9127-3

39.

Massey

D. S.

Denton

N. A.

(1985). Spatial assimilation as a socioeconomic outcome. American Sociological Review, 50(1), 94-106. doi:10.2307/2095343

40.

Mears

D. P.

Mancini

Gertz

Bratton

(2008). Sex crimes, children, and pornography: Public views and public policy. Crime & Delinquency, 54, 532-559. doi:10.1177/0011128707308160

41.

Mercado

C. C.

Alvarez

Levenson

J. S.

(2008). The impact of specialized sex offender legislation on community reentry. Sexual Abuse: A Journal of Research and Treatment, 20(2), 188-205. doi:10.1177/1079063208317540

42.

Minnesota Department of Corrections. (2007). Residential proximity & sex offender recidivism in Minnesota. St. Paul, MN: Author.

43.

Morgan

J. K.

(2008). Spatial implications of municipal sex offender residency restrictions in Bay County, Florida, 2005-2007. Master of Science, Northwest Missouri State University, Maryville, MO. Retrieved from http://www.nwmissouri.edu/LIBRARY/theses/MorganJenniferK/jmorgan_thesis.pdf

44.

Mulford

C. F.

Wilson

R. E.

Parmley

A. M.

(2009). Geographic aspects of sex offender residency restrictions: Policy and research. Criminal Justice Policy Review, 20(1), 3-12. doi:10.1177/0887403408327683

45.

Mustaine

E. E.

Tewksbury

Stengel

K. M.

(2006). Social disorganization and residential locations of registered sex offenders: Is this a collateral consequence? Deviant Behavior, 27, 329-350. doi:10.1080/01639620600605606

46.

New York State Division of Criminal Justice Services. (2012). Computerized criminal history database. Albany, NY: Author.

47.

New York State Office of Sex Offender Management. (2010). Sex offender registry. Retrieved from http://criminaljustice.state.ny.us/nsor/

48.

Oden

(1995). Adjusting Moran’s I for population density. Statistics in Medicine, 14(1), 17-26. doi:10.1002/sim.4780140104

49.

Peterson

R. D.

Krivo

L. J.

Harris

M. A.

(2000). Disadvantage and neighborhood violent crime: Do local institutions matter? Journal of Research in Crime and Delinquency, 37(1), 31-63. doi:10.1177/0022427800037001002

50.

Pope

J. C.

(2008). Fear of crime and housing prices: Household reactions to sex offender registries. Journal of Urban Economics, 64, 601-614. doi:10.1016/j.jue.2008.07.001

51.

Poston

D. L.

Micklin

(2006). Handbook of population. New York, NY: Springer Science+Business Media.

52.

Pratt

T. C.

Cullen

F. T.

(2005). Assessing macro-level predictors and theories of crime: A meta-analysis. Crime and Justice, 32, 373-449.

53.

Sampson

R. J.

(1985). Neighborhood and crime: The structural determinants of personal victimization. Journal of Research in Crime and Delinquency, 22(1), 7-40. doi:10.1177/0022427885022001002

54.

Schaible

L. M.

Hughes

L. A.

(2008). Neighborhood disadvantage and reliance on the police. Crime & Delinquency. doi:10.1177/0011128708322531

55.

Shaw

C. R.

McKay

H. D.

(1942). Juvenile delinquency and urban areas. Chicago: University of Chicago Press.

56.

Smith

D. A.

Jarjoura

G. R.

(1988). Social structure and criminal victimization. Journal of Research in Crime and Delinquency, 25(1), 27-52. doi:10.1177/0022427888025001003

57.

Socia

K. M.

(2011). The policy implications of residence restrictions on sex offender housing in upstate, NY. Criminology & Public Policy, 10, 351-389. doi:10.1111/j.1745-9133.2011.00713.x

58.

Socia

K. M.

(2012a). The efficacy of county-level sex offender residence restrictions in New York. Crime & Delinquency, 58, 612-642.

59.

Socia

K. M.

(2012b). The implementation of county residence restrictions in New York. Psychology, Public Policy, and Law, 18(2), 206-230. doi:10.1037/a0024993

60.

Socia

K. M.

(in press). Residence restrictions and the association with registered sex offender clustering. Criminal Justice Policy Review. doi:10.1177/0887403412445613

61.

Socia

K. M.

Stamatel

J. P.

(2010). Assumptions and evidence behind sex offender laws: Registration, community notification, and residency restrictions. Sociology Compass, 4(1), 1-20. doi:10.1111/j.1751-9020.2009.00251.x

62.

Socia

K. M.

Stamatel

J. P.

(2012). Neighborhood characteristics and the social control of registered sex offenders. Crime & Delinquency, 58, 565-587. doi:10.1177/001112871142011

63.

Sutherland

E. H.

(1947). Principles of criminology (4th ed.). Philadelphia, PA: J. B. Loppincott.

64.

TerraSeer

(2010). ClusterSeer 2: TerraSeer, Inc. Retrieved from http://www.terraseer.com/products_clusterseer.php

65.

Tewksbury

Mustaine

E. E.

(2006). Where to find sex offenders: An examination of residential locations and neighborhood conditions. Criminal Justice Studies, 19(1), 61-75. doi:10.1080/14786010600615991

66.

Tewksbury

Mustaine

E. E.

Covington

(2010). Offender presence, available victims, social disorganization and sex offense rates. American Journal of Criminal Justice, 35(1), 1-14. doi:10.1007/s12103-010-9070-6

67.

Tewksbury

Mustaine

E. E.

Stengel

K. M.

(2008). Examining rates of sexual offenses from a routine activities perspective. Victims & Offenders, 3(1), 75-85. doi:10.1080/15564880701752306

68.

U.S. Census. (2002). U.S. summary: 2000, Census 2000 profile (Economics and Statistics Administration, Trans.). Washington, DC: Author.

69.

U.S. Census. (2010). Population estimates program (U.S. Department of Commerce). Retrieved from http://factfinder.census.gov/home/en/official_estimates_2008.html

70.

U.S. States News. (2008, April 15). Assemblyman Thiele calls on assembly to vote on legislation to protect communities from sex offenders. Press release. U.S. States News.

71.

Walker

J. T.

Ervin-McLarty

(2000). Sex offenders in Arkansas: Characteristics of offenders and enforcement of sex offender laws. Little Rock, AR: Arkansas Crime Information Center.

72.

Waren

(2008). Spatial attainment trends of racial and ethnic groups in Houston, Texas, 1970 to 2000. PhD dissertation, Texas A&M University, College Station, TX.

73.

Youstin

T. J.

Nobles

M. R.

(2009). Residency restrictions: A geospatial analysis of sex offender movement over time. Crime Mapping: A Journal of Research and Practice, 1(1), 55-76.

74.

Zandbergen

P. A.

Hart

T. C.

(2006). Reducing housing options for convicted sex offenders: Investigating the impact of residency restriction laws using GIS. Justice Research and Policy, 8(2), 1-24. doi:10.3818/JRP.8.2.2006.1

75.

Zandbergen

P. A.

Levenson

J. S.

Hart

T. C.

(2010). Residential proximity to schools and daycares: An empirical analysis of sex offense recidivism. Criminal Justice and Behavior, 37, 482-502. doi:10.1177/0093854810363549

76.

Zevitz

R. G.

(2004). Sex offender placement and neighborhood social integration: The making of a scarlet letter community. Criminal Justice Studies, 17(2), 203-222.

Too Close for Comfort? Registered Sex Offender Spatial Clustering and Recidivistic Sex Crime Arrest Rates

Abstract

Keywords

Second, and more importantly, where RSOs are living (or clustering) must somehow be associated with rates of recidivistic sex crimes

County-Level Characteristics and Recidivistic Sex Crime Arrests

Implications for Policy

The Current Study

Method

Data and Sample

Dependent Variables

Primary Independent Variables

Mean county nearest neighbor analysis

Revised index of isolation

Oden’s I*pop

Control Variables

Analytical Model

Results

Recidivistic Sex Crimes Involving Child Victims

Recidivistic Sex Crimes Involving Adult Victims

Overall Results

Discussion

Limitations

Implications for Policy and Future Research

Conclusions

Footnotes

Appendix

Acknowledgements

Declaration of Conflicting Interests

Funding

Notes

Author Biography

References

Oden’s I*_pop