guardianship in context: implications for burglary ... · crime-reduction implications for the...

33
\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 1 8-NOV-07 15:36 GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY VICTIMIZATION RISK AND PREVENTION* PAMELA WILCOX Division of Criminal Justice University of Cincinnati TAMARA D. MADENSEN Department of Criminal Justice University of Nevada, Las Vegas MARIE SKUBAK TILLYER Division of Criminal Justice University of Cincinnati KEYWORDS: burglary, guardianship, crime prevention Survey data from 4,227 Seattle residents nested within 100 “neighbor- hoods” (census tracts) were analyzed to discern interrelationships between various dimensions of individual-level and neighborhood-level guardianship. We focused on four dimensions of guardianship—physi- cal (target hardening), personal (home occupancy), social (informal control), and natural (surveillance through environmental design)—at both individual and neighborhood levels. A multilevel opportunity, the- oretical framework guided hypotheses, which suggests that each of the four dimensions of individual guardianship would be related more neg- atively to burglary as each of the four dimensions of aggregate guardi- anship increased. Multilevel logistic regression analysis revealed support for many of such hypothesized moderating effects of aggregate guardianship. More specifically, 6 of the 16 possible interaction effects were statistically significant at the .05 level and an additional 3 interac- tion effects were significant at the .10 level. In particular, individual- level target hardening, place management, and natural surveillance were related more negatively to burglary as neighborhood-level target * Direct correspondence to Pamela Wilcox, Division of Criminal Justice, University of Cincinnati, 600 Dyer Hall, PO Box 210389, Cincinnati, OH 45221- 0389 (e-mail: [email protected]). CRIMINOLOGY VOLUME 45 NUMBER 4 2007 771

Upload: others

Post on 16-Jul-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 1 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT:IMPLICATIONS FOR BURGLARYVICTIMIZATION RISK ANDPREVENTION*

PAMELA WILCOXDivision of Criminal JusticeUniversity of Cincinnati

TAMARA D. MADENSENDepartment of Criminal JusticeUniversity of Nevada, Las Vegas

MARIE SKUBAK TILLYERDivision of Criminal JusticeUniversity of Cincinnati

KEYWORDS: burglary, guardianship, crime prevention

Survey data from 4,227 Seattle residents nested within 100 “neighbor-hoods” (census tracts) were analyzed to discern interrelationshipsbetween various dimensions of individual-level and neighborhood-levelguardianship. We focused on four dimensions of guardianship—physi-cal (target hardening), personal (home occupancy), social (informalcontrol), and natural (surveillance through environmental design)—atboth individual and neighborhood levels. A multilevel opportunity, the-oretical framework guided hypotheses, which suggests that each of thefour dimensions of individual guardianship would be related more neg-atively to burglary as each of the four dimensions of aggregate guardi-anship increased. Multilevel logistic regression analysis revealedsupport for many of such hypothesized moderating effects of aggregateguardianship. More specifically, 6 of the 16 possible interaction effectswere statistically significant at the .05 level and an additional 3 interac-tion effects were significant at the .10 level. In particular, individual-level target hardening, place management, and natural surveillancewere related more negatively to burglary as neighborhood-level target

* Direct correspondence to Pamela Wilcox, Division of Criminal Justice,University of Cincinnati, 600 Dyer Hall, PO Box 210389, Cincinnati, OH 45221-0389 (e-mail: [email protected]).

CRIMINOLOGY VOLUME 45 NUMBER 4 2007 771

Page 2: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 2 8-NOV-07 15:36

772 WILCOX, MADENSEN & TILLYER

hardening increased, as neighborhood-level informal social controlincreased, and as neighborhood-level natural surveillance increased.

Many opportunity-based theories of crime, which include routine activi-ties theory, situational crime prevention, and defensible space/crime pre-vention through environmental design (CPTED), suggest that individual-level efforts to increase the security, surveillance, or guardianship pro-vided to one’s home should decrease burglary victimization risk. Empiricalsupport is often found for such linkages (e.g., see Kennedy and Forde,1990; Miethe and Meier, 1990, 1994; Miethe, Stafford, and Long, 1987;Mustaine and Tewksbury, 1998; Smith and Jarjoura, 1989). In addition, thecrime-reduction implications for the effects of guardianship on burglaryare assumed to be generalizable across contexts. In other words, previoustheory and research suggests that guardianship often does work to preventcrime, with an implicit assumption that this effect is similar, regardless ofcontext.

However, several multilevel studies over the past few decades have castdoubt on this assumption of cross-context generalizability in the effects ofindicators of criminal opportunity, including guardianship, on burglary vic-timization (e.g., Miethe and McDowall, 1993; Outlaw, Ruback, and Britt,2002; Wilcox Rountree and Land, 1996; Wilcox Rountree, Land, andMiethe, 1994). Micro–macro statistical interactions emerged in these stud-ies despite little systematic, a priori theory that guides hypotheses abouthow indicators of individual-level criminal opportunity would or shouldvary across social contexts. Rather, most interactions emerged in suchstudies from exploratory examination of the data using hierarchical linearand logistic regression techniques, which made them somewhat “ad hoc”and not readily interpreted within an existing integrative theoreticalframework or easy to apply. In addition to this limitation, most previousstudies of guardianship’s effects on burglary employed less-than-compre-hensive conceptualizations and imprecise measurement of guardianship(Mustaine and Tewksbury 1997, 1998; Tewksbury and Mustaine, 2003).Many previous studies focus on one or two types of guardianship—most atthe individual level—rather than examine ways in which a variety ofdimensions exist at both individual and aggregate levels (and interactacross those levels). In sum, whereas cross-level interaction effects havebeen estimated in previous burglary victimization research, studies of suchinteractions are limited in that 1) they have not focused on various dimen-sions of microlevel and macrolevel guardianship specifically and 2) theyhave not been guided by a theoretical framework suggestive of mul-tifaceted, multilevel, and contextualized guardianship.

In this study, we draw on recent advances in criminal opportunity theoryto offer the first formal test of specific multilevel hypotheses about how

Page 3: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 3 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 773

the effects of individual-level guardianship should vary across neighbor-hood context, depending on aggregate-level guardianship. In doing so, weprovide conceptual advancement by delineating several distinct dimen-sions of guardianship spanning physical, personal, social, and naturalspheres. Each of these dimensions of guardianship is measured at bothindividual and neighborhood levels, which provides a more comprehensiveconceptualization and operational definition of guardianship than previ-ous studies. In summary, our study clarifies ways that various dimensionsof individual guardianship work more or less effectively to prevent crime,depending on different dimensions of guardianship provided by the neigh-borhood in which the individual is embedded. In addition to testing theinteractive effects of individual and neighborhood guardianship—while inthe process of extending our understanding of the various dimensions ofguardianship—practical implications for a “contextualized crime preven-tion” are offered.

THEORETICAL BACKGROUND

For the purpose of this article, we define guardianship at individual andenvironmental levels and we draw on recent multicontextual criminalopportunity theory for specific definitions at each level. At the individuallevel, guardianship is defined as “possessing qualities that relate to socialties and interpersonal control,” with interpersonal control referring to “thedegree to which individuals and objects in a bounded locale can beobserved and impeded from experiencing criminal acts because they areproximate and exposed to agents of formal control, agents of informalcontrol, and non-human protection devices” (Wilcox, Land, and Hunt,2003: 62; see also Miethe and Meier, 1994). Environmental-level guardian-ship is “the collective degree to which individuals or objects in a boundedlocale possess qualities related to social ties and social control,” with socialcontrol encompassing, again, informal, formal, and nonhuman security(Wilcox, Land, and Hunt, 2003: 64). Given our conceptualization of guard-ianship, we turn now to a review of literature that employs at least oneaspect of our definition in understanding burglary.

GUARDIANSHIP AND BURGLARY VICTIMIZATION RISK

The importance of guardianship in understanding how opportunitystructures can facilitate or inhibit crime was delineated in Cohen and Fel-son’s (1979) routine activities theory, which argued that most crimesrequire the convergence in time and space of likely offenders, suitabletargets, and absence of capable guardians. Although this convergencedescribed a microlevel event, Cohen and Felson (1979) framed their the-ory within the context of the larger social structure, which argued that the

Page 4: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 4 8-NOV-07 15:36

774 WILCOX, MADENSEN & TILLYER

organization of society affected the routine activities of people. In particu-lar, they suggested that an increase in nonhousehold activities wouldincrease the likelihood that offenders would meet targets in the absence ofguardians. Their analysis determined that the national level ofnonhousehold activity was related positively to burglary rates as well as torape, assault, robbery, and homicide rates over time (Cohen and Felson,1979). Consistent with Cohen and Felson (1979), subsequent macrolevelempirical research has demonstrated that society’s structural organizationrelated to aggregate levels of guardianship explains variations in crimerates. Several measures of aggregate guardianship have been shown to berelated to macrolevel crime trends, including the proportion of the popula-tion living alone (Felson and Cohen, 1980), the percentage of single-parenthouseholds (Smith and Jarjoura, 1988), the dispersion of activities awayfrom the family and household (Cohen, Felson, and Land, 1980; Messnerand Blau, 1987), and the unemployment rate (Cohen, Felson, and Land,1980).

Although used initially to explain changes in crime rates over time, rou-tine activities theory and the notion of opportunity structures have beenextended to explain variations in individual victimization. Several mea-sures of individual-level guardianship related to personal “place manage-ment,” social ties, or informal social control have been linked to burglaryvictimization specifically, such as occupation of dwelling (Coupe andBlake, 2006; Titus, 1989), household composition (Cohen, Kluegel, andLand, 1981; Miethe and Meier, 1990), labor status (Cohen, Kluegel, andLand, 1981), and other-than-work routine activities that remove peoplefrom their homes, including school, sports, meetings, and movies (Cohenand Cantor, 1981; Kennedy and Forde, 1990; Miethe and Meier, 1994;Miethe, Stafford, and Long, 1987; Wilcox Rountree and Land, 1996, Wil-cox Rountree, Land, and Miethe, 1994). Additionally, routine activitiesthat provide physical protection (i.e., “hardening”) of homes, includinglocking doors, installing alarms, and light timer devices, have also beenshown to correlate with burglary victimization (Miethe and McDowall,1993; Miethe and Meier, 1990, 1994; Wilcox Rountree and Land, 1996,Wilcox Rountree, Land, and Miethe, 1994). Consistent with such findings,successful situational crime-reduction initiatives have included hardeningand surveillance-enhancing techniques that overlap with “guardianship,”including property marking (Forrester, Chatterton, and Pease, 1988; Lay-cock, 1985), improved street lighting (Painter and Farrington, 2001),neighborhood watch (Forrester, Chatterton, and Pease, 1988), burglaralarms (Buck, Hakim, and Rengert, 1993; Hakim et al., 1995), andimproved locks and doors (Forrester, Chatterton, and Pease, 1988; Tilleyand Webb, 1994).

Page 5: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 5 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 775

Defensible space/CPTED also provides concepts relevant to guardian-ship. In the defensible space/CPTED tradition, Jacobs (1961) initiallyintroduced the idea that environmental design can reduce crime byincreasing natural surveillance. The notion of natural surveillance acknowl-edges the role that the natural or built environment plays in structuringopportunities for residents and their immediate neighbors to observe anddefend the space around their homes (Cisneros, 1995; Newman, 1972,1996; Mawby, 1977; Taylor and Gottfredson, 1986; Taylor and Harrell,1996). Thus, informal social control is still key (as in routine activities the-ory), but environmental design is assumed to foster/impede such efforts.Building type, street layout, presence of tall shrubs or fences, or hiddenalcoves or alleyways (i.e., cover) are just some features of an environmentwith natural surveillance (e.g., Brantingham and Brantingham, 1975;Coupe and Blake, 2006; Donnelly and Kimble, 1997; Fisher and Nasar,1992; Merry, 1981; Newman, 1972, 1996; Taylor and Harrell, 1996; White,1990).

In sum, various conceptualizations of guardianship have been integratedinto opportunity-based theories of crime, and various measures thereofhave been shown empirically to be related inversely to burglary. All ofthese findings comport well with a rational choice perspective thatassumes a reasoning burglar who considers effort, risk, and reward associ-ated with targets (e.g., Bennett and Wright, 1984; Clarke, 1980, 1995; Cor-nish and Clarke, 2003; Coupe and Blake, 2006; Cromwell, Olson, andAvary, 1991; Palmer, Holmes, and Hollin, 2002; Rengert and Wasilchick,1985; Shover, 1991; Wright and Decker, 1994). Despite overall evidence ofthe multifaceted nature of guarding property, individual studies that esti-mate burglary tend to focus on one or two domains (e.g., social vs. physicalguardianship), thus minimizing our understanding of the relative effects ofvarious guardianship dimensions (see Garofalo and Clark, 1992; Mietheand Meier, 1990; Norris and Kaniasty, 1992 for exceptions). In addition,findings are inconsistent from the few studies that do incorporate multipledomains of guardianship. For instance, Miethe and Meier (1990) suggestedthat social guardianship (i.e., more than one person in the household)deterred burglary more than physical guardianship. In contrast, Norris andKaniasty (1992) found that none of their examined dimensions of guardi-anship—alertness, locks, informal cooperation of neighbors, and assis-tance from professionals—reduced victimization risk. Although theoriesthat advance the idea of guardianship’s effect on burglary have receivedoverall support in the literature, tests thereof have largely relied on amixed bag of one-dimensional measures, and as a result, we do not have aclear sense of whether some practices work better than others do. Contin-ued progress toward an integrative, multidimensional conceptualization

Page 6: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 6 8-NOV-07 15:36

776 WILCOX, MADENSEN & TILLYER

and measurement of guardianship would help to make the practical impli-cations of these theories most useful to potential victims.

MULTILEVEL ASPECTS OF GUARDIANSHIP

The idea that guardianship is multdimensional can be extended to incor-porate the idea that guardianship not only exists in multiple forms at anyone level of analysis, but also it exists in all its multiple forms across levelsof analysis. For instance, in their theoretical discussion of the role of thephysical environment on crime, Brantingham and Brantingham (1993) dis-cussed how target selection occurs within the context of an “environmen-tal backcloth,” which includes the broader physical setting in whichspecific crime sites or targets are situated. Similarly, Taylor and Gottfred-son (1986: 389) described target selection as a multilevel process (e.g.,region, street block, and site as important levels) whereby “particularareas are identified as affording an abundance of targets, and particularblocks or houses within those areas are selected as targets.” Likewise,Coupe and Blake (2006: 456) suggested that offenders, especially neweroffenders, “may actively evaluate different types of areas for potentialtargets,” with area selection that involves “the same sets of dwelling attrib-utes used for choosing targets once a strategy and area have been settledupon” (see also Clarke and Cornish, 1985). In addition, Groff andLaVigne (2001) noted the possibility of a “threshold effect,” which sug-gests that the effects of target selection variables on burglary victimizationmay differ depending on the characteristics of a neighborhood. They arguethat when criminals are in areas of adequate burglary opportunities, spe-cific target selection criteria may be less important because “one target isas good as the next” (Groff and LaVigne, 2001: 274). Finally, Wright andDecker’s (1994) ethnographic study of active burglars revealed that, in theabsence of information about the contents of a specific property, burglarsin their sample sometimes looked to environmental cues to assess aspectsof opportunity such as, target attractiveness and guardianship. Sometimes,these cues were located just beyond a house in question (i.e., wouldinclude the exterior of a particular property), but other times theyincluded an assessment of the neighborhood in which the property wassituated. Wright and Decker, for instance, suggest:

These offenders had a vested interest in being able to rely on predic-tions about the habits of the occupants of an intended target. Absentspecific information about the occupants, a general familiarity withthe sort of people who resided in the area represented the next bestmeans of making such predictions (1994: 88).

Consistent with the suggestion of multilevel opportunity structuresoffered by the research reviewed above, a few additional studies have

Page 7: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 7 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 777

incorporated measures of opportunity explicitly—and guardianship spe-cifically—at individual and environmental levels of analysis. Recognizingpotential linkages between individual and community dimensions of vic-timization risk, Sampson and Wooldredge (1987) examined the effects ofboth individual and community variables on burglary victimization, andthey found both sets of factors to be influential, which supports a “general(multilevel) opportunity” framework. Guided by Sampson and Wool-dredge’s (1987) multilevel conceptualization of victimization risk, severalother researchers have attempted to integrate individual and contextualfactors in empirical studies of burglary victimization (see, e.g., Miethe andMcDowall, 1993; Miethe and Meier, 1994; Outlaw, Ruback, and Britt,2002; Smith and Jarjoura, 1989; Wilcox Rountree and Land, 1996; WilcoxRountree, Land, and Miethe, 1994). These studies indicate that burglaryvictimization risk is influenced by both individual and contextual opportu-nity components, thus suggesting a need for multilevel models of victimi-zation. Furthermore, studies that examined the interaction betweenindividual and contextual variables indicate that the effects of individualguardianship are not uniform across contexts. For example, studies thatuse Miethe’s Seattle victimization survey data reported that the effect ofindividual level guardianship on burglary victimization was tempered indisadvantaged and/or disorderly areas, which implies that crime-preven-tion efforts by residents in these areas needed to be more extensive toachieve the same level of burglary reduction (Miethe and McDowall, 1993;Miethe and Meier, 1994; Wilcox Rountree, Land, and Miethe, 1994).1

Although researchers have considered the multilevel components ofopportunity and some significant micro–macro interactions have emerged,previous models have lacked a strong theoretical foundation from whichhypotheses can be derived in regard to interaction effects across levels. Inan effort to fill this void and to provide a theoretical framework for sucheffects, Wilcox, Land, and Hunt (2003) put forth a multicontextual crimi-nal opportunity theory, which specifies the components of opportunitystructures that operate at both the individual and the environmental levelsand that interact across levels. Most relevant to the focus of this study,they argued that the effect of guardianship is more negative as aggregate

1. In an examination of displacement and/or free-rider effects, Miethe (1991)explored the effects of individual neighbor guardianship on own burglary riskand, in a macro model, collective neighboring-area guardianship on own neigh-borhood rates of burglary victimization. His study implied that components ofcriminal opportunity existed at multiple levels, and it was measured at multiplelevels, but these levels were never combined into a single model. Miethe’s studybegs the question (the focus of this study) as to whether the protective actionstaken by the collective of one’s own neighborhood (as opposed to surroundingneighborhoods) play a role in individual burglary risk, net of individualguardianship.

Page 8: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 8 8-NOV-07 15:36

778 WILCOX, MADENSEN & TILLYER

guardianship increases. Conceptualizing aggregate units as “markets” forcrime, they suggested that aggregate-level guardianship increases the mar-ket costs of crime. The deterrence provided by the market serves toenhance the relationship between individual-level guardianship and vic-timization risk, making it more negative (Wilcox, Land, and Hunt, 2003:109).

THE CURRENT STUDY

We draw on the extant empirical and theoretical work reviewed aboveto test the multilevel effects of multiple dimensions of guardianship onburglary victimization. In doing so, we extend previous research on bur-glary victimization by exploring the effects of guardianship using 1) a vari-ety of guardianship domains, so that relative prevention effectiveness canbe assessed, and 2) a multilevel conceptualization and operational defini-tion of guardianship so that the extent to which individual and collectiveguardianship practices interact can be better understood. We focus specifi-cally on the interplay between individual guardianship and neighborhood-level guardianship. In doing so, we focus on four dimensions of guardian-ship at each level: physical, personal, social, and natural. More specifically,in terms of physical guardianship, we examine guardianship offered by tar-get hardening strategies. In the sphere of personal guardianship, we esti-mate the effects of homeowner occupancy. For social guardianship, weconsider neighbor-provided surveillance or informal social control. Finally,in terms of natural guardianship, we examine elements of environmentaldesign that foster surveillance or “defensible space.”

Traditional “single-level” conceptualizations of criminal opportunity areclear that each of these measures, either at the individual or aggregatelevel, should be related negatively to crime/victimization, and recent mul-tilevel conceptualizations suggest that simultaneous main effects at bothindividual and neighborhood levels are likely. However, until recently,theory has been lacking on how or why indicators of opportunity shouldinteract across levels. Drawing on the recent theoretical development inregard to such interactions, we posit the following hypotheses (Wilcox,Land, and Hunt, 2003):

Hypothesis 1: Individual-level target hardening will be related more neg-atively to burglary victimization as neighborhood-level target hardening,home occupancy, informal social control, and natural surveillanceincrease.

Hypothesis 2: Individual-level home occupancy will be related morenegatively to burglary victimization as neighborhood-level target harden-ing, home occupancy, informal social control, and natural surveillanceincrease.

Page 9: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 9 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 779

Hypothesis 3: Individual-level informal social control will be relatedmore negatively to burglary victimization as neighborhood-level targethardening, home occupancy, informal social control, and natural surveil-lance increase.

Hypothesis 4: Individual-level defensible space will be related more neg-atively to burglary victimization as neighborhood-level target hardening,home occupancy, informal social control, and natural surveillanceincrease.

Again, the context is risky in which aggregate-level target-hardening,home occupancy, informal social control, or defensible space is high.Therefore, such a context serves to buttress or to exacerbate the effects ofindividual guardianship behaviors. In contrast, a context in which aggre-gate-level target hardening, home occupancy, informal social control, ordefensible space are low provides an opportunistic market. Absent infor-mation about specific targets, reasoning offenders will generalize to alltargets within, thus diminishing the impact of individual-level guardianshipon victimization relative to the impact observed in high-cost markets(Clarke and Cornish, 1985; Groff and LaVigne, 2001; Wright and Decker,1994). We applied this “neighborhood-as-crime-market” rationale consist-ently to all 16 interactions suggested by hypotheses 1 through 4. We haveno a priori theoretical reason to suspect that the cross-level interactionprocess should operate differently across the various domains of guardian-ship. In short, each aggregate dimension measures some aspect of the cost-liness of the context, and market costliness (in any form) enhances thecrime-prevention benefits of strong individual guardianship (in any form)or, conversely, tempers the vulnerability associated with weak individualguardianship (in any form).

DATA

The above hypotheses were tested by a series of hierarchical logisticregression models of burglary victimization. The primary source of datafor these models was from a 1990 crime/victimization telephone survey ofSeattle residents (Miethe, 1992). Sample respondents were part of 300block-pairs, one block of which had experienced a burglary in the yearpreceding the survey.2 The block-pairs were nested within 100 of Seattle’s

2. A previous-year burglary on one block in each block-pair was required, as one ofthe original motives for the survey was to study the displacement among contigu-ous blocks. As such, the sample intentionally oversamples blocks with knownburglaries.

Page 10: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 10 8-NOV-07 15:36

780 WILCOX, MADENSEN & TILLYER

121 stable census tracts.3 Consistent with many neighborhood crime stud-ies, which include those that use the same Seattle data used here, we usedthe census tract as the “neighborhood” unit of analysis in this study (e.g.,Bellair, 2000; Warner and Wilcox Rountree, 1997).4

Across the 100 Seattle census tracts included in the sample, a total of12,303 telephone numbers were dialed for survey purposes, with 9,250household contacts obtained as a result. The eligible sample was 7,159households, after excluding those who could not participate because noadult was present or that adult was hearing-impaired or non-Englishspeaking. There was a 74.1 percent response rate among the 7,159 eligiblehousehold contacts, which resulted in 5,302 completed interviews.5Listwise deletion of cases with missing data on variables key to this studyresulted in a sample of 4,227 individuals and 100 census tracts for analysispurposes.6

3. Project PI, Terance Miethe, defined a census tract as “stable” if its boundarieshad not changed since 1960. He notes that this selection does limit the sample interms of its ability to make generalizations about the entire city of Seattle(Miethe, 1991: 426). The number of reported burglaries in the omitted unstablecensus tracts for the years 1989 and 1990 were, on average, lower than thosereported in the tracts used in this analysis (i.e., a 2-year average of 59.75 versus70.78 burglaries).

4. We recognize that census tracts may not correspond perfectly with residents’ideas about neighborhood boundaries, but this same criticism can be leveled atother possible units. For instance, the “block-pair” unit has been used in severalother studies using Miethe’s Seattle data (e.g., Miethe and McDowall, 1993; Wil-cox Rountree, Land, and Miethe, 1994). This unit, however, may be smaller thanwhat many residents consider as their neighborhood. Furthermore, using theblock-pair as a neighborhood unit for this analysis was not optimal given thesmall number of individuals per block-pair (14, on average, with some containingfewer than 10) in combination with the large number of level 2 variables. None-theless, in the supplemental analysis (not shown, but available from the firstauthor on request), we estimated similar models to those presented here, but weused block-pair as the unit of analysis as opposed to census tract as the unit ofanalysis. Substantive conclusions are very similar regardless of neighborhood unitemployed. Only two of the nine cross-level interactions revealed herein were notsupported in the analysis using block-pair as the neighborhood unit (individualplace management × aggregate informal social control, individual place manage-ment × aggregate defensible space).

5. In addition to its intentional oversampling of blocks with known burglaries,Miethe (1991) noted several limitations that characterize the sampling strategy.With use of a reverse telephone directory, it excluded the 28 percent of Seattlehouseholds with unlisted numbers. In addition, the resulting sample under-represented recent movers and overrepresented homeowners and individualswith some college education.

6. Although approximately 42 respondents exist per census tract after listwise dele-tion, variation occurred, with numbers of respondents per tract ranging from 30to 50.

Page 11: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 11 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 781

MEASURES OF VARIABLES

Metrics and descriptive statistics for all study variables are summarizedin table 1. The dependent variable for this study was burglary victimiza-tion, measured as a dichotomous variable that indicates whether therespondent (1 = yes; 0 = no) had experienced a burglary, including an

Table 1. Descriptive Statistics for Study Variablesa

Variables Metric Mean SD Minimum Maximum

Dependent VariablesBurglary victimization (1 = yes, 0 = no) .18 .38 .00 1.00Individual-Level Guardianship VariablesTarget hardening (# of safety precautions) 2.88 1.14 .00 6.00Home occupancy (# days/nights home occupied) 7.50 3.35 .00 14.00Informal social control (Whether neighbors watch home; .74 .44 .00 1.00

1 = yes, 0 = no)Defensible space (# of property characteristics) 5.78 1.12 .00 8.00Individual-Level Control VariablesHousehold goods (# of items owned) 2.62 1.40 .00 5.00Businesses nearby (# of nearby business) 1.79 1.65 .00 6.00School nearby (1 = yes, 0 = no) .17 .38 .00 1.00Park/playground nearby (1 = yes, 0 = no) .58 .49 .00 1.00Perceived disorder (# of problems within 4 blocks of .83 .95 .00 3.00

index residence)Age (1 = 10–19 years to 7 = 70 years 4.24 1.68 1.00 7.00

and older)Sex (male) (1 = male, 0 = female) .51 .50 .00 1.00Race (nonwhite) (1 = nonwhite, 0 = white) .14 .34 .00 1.00Family income (1 = < $10,000 to 7 = > $100,000) 3.42 1.35 1.00 7.00Live alone (1 = yes, 0 = no) .25 .43 .00 1.00Neighborhood-Level Guardianship VariablesTarget hardening (Mean # of safety precautions) 2.88 .35 1.89 3.63Home occupancy (Mean # days/nights home 7.51 .82 4.90 10.02

occupied)Informal social control (Proportion who have neighbors .74 .14 .26 .95

watch home)Defensible space (Mean # of property 5.78 .43 4.41 6.80

characteristics)Neighborhood-Level Control VariablesHousehold goods (Mean # of items owned) 2.62 .32 1.75 3.24Businesses nearby (Mean # of business places within 1.79 1.04 .16 4.57

three blocks)School nearby (Proportion living within three .17 .20 .00 .79

blocks of a school)Park/playground nearby (Proportion living within three .58 .23 .09 .98

blocks of park/playground)Perceived disorder (Mean # of problems within four .83 .45 .08 1.98

index blocks of residences)Age (Mean of ordinal scale) 4.25 .48 2.95 5.38Sex (Proportion male) .51 .08 .32 .70Concentrated (Summed index of tract-level .33 .27 .08 1.14

disadvantage census data)Population instability (Summed index of tract-level .00 2.03 –3.31 6.73

census data)Downtown location (Whether tract is located .04 .19 .00 1.00

downtown; 1 = yes, 0 = no)

a Descriptive statistics are based on 4,227 individuals within 100 neighborhoods (census tracts).

Page 12: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 12 8-NOV-07 15:36

782 WILCOX, MADENSEN & TILLYER

attempt, at their current home within 2 years preceding the survey. Astable 1 suggests, 18 percent of sampled respondents were burglary victims.

Key explanatory variables included different types of guardianship—physical guardianship (target hardening), personal guardianship (homeoccupancy), social guardianship (neighbor-based informal control), andnatural guardianship (defensible space)—measured at both individual andcensus tract levels. Individual-level physical guardianship, or target hard-ening, refers to the steps the respondent had taken to increase the physicaleffort or risk associated with burglarizing her/his property. It is measuredas an index that sums the number of safety precautions engaged in by therespondent 2 years before the survey, including 1) locked doors, 2) leavinglights on to make it seem as if someone is home, 3) installation of extralocks, 4) use of a burglar alarm, 5) having a dog in the home, and 6) keep-ing a weapon in the home. Respondents, on average, engaged in 2.88 tar-get-hardening strategies, with individual values that range from 0 to 6 (seetable 1). Individual-level personal guardianship, or home occupancy, ismeasured as the number of days and nights per week that the respondent’shome was occupied 2 years before the study began. On average, respon-dents indicated that their homes were occupied 7.5 days and nights, out of14 total days and nights. Individual-level social guardianship, or neighbor-based informal social control, is measured as a single item, which indicateswhether (1 = yes; 0 = no) the respondent had neighbors watch their prop-erty when away, with the time referent again as 2 years before the survey.Figures in table 1 show that 74 percent of the sample engaged in this formof guardianship. Individual-level natural guardianship/surveillance, ordefensible space, refers to characteristics of the house/property design thatmake it more or less able to be accessed and/or watched. Defensible spacewas measured as an index, which sums the number of the following envi-ronmental features that characterize the respondent’s house/property: 1)ground floor windows, 2) tall fence/hedge around the dwelling, 3) anempty lot next door, 4) a corner lot, 5) an alley behind the home, 6) a two-way (as opposed to one-way or dead-end) street, 7) trees/shrubs blockingthe front door, and 8) multiple units within the dwelling.7 Table 1 suggeststhat respondents reported, on average, 5.78 environmental design features,

7. Proximal land use is an aspect of the environmental design often associated withnatural surveillance as certain nonresidential land uses promote the convergencein time and space of offenders, suitable targets, and absence of guardianship(Brantingham and Brantingham, 1993, 1999; Kurtz, Koons, and Taylor, 1998;Roncek and Maier, 1991; Sherman, Gartin, and Buerger, 1989; Taylor and Har-rell, 1996; Taylor et al., 1995). We therefore recognize the theoretical importanceof such nonresidential land uses to criminal opportunity—and perhaps guardian-ship specifically—and we do include land use controls at both individual andneighborhood levels. We do not treat proximal nonresidential land use as a directmeasure of guardianship because it picks up land uses as far as three blocks away

Page 13: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 13 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 783

which is indicative of strong natural surveillance, although individualresponses ranged from 0 to 8.

Neighborhood target hardening, home occupancy, informal social con-trol, and defensible space are each measured as tract-level means of therespective individual-level guardianship variables. Table 1 indicates sub-stantial cross-neighborhood variation in such collective guardianship.Neighborhood average target hardening scores ranged from 1.89 precau-tions to 3.63 precautions; neighborhood home occupancy ranged from 4.90to 10.02; neighborhood informal social control ranged from .26 to .95; andneighborhood defensible space ranged from 4.41 to 6.80.

In examining the main versus interactive effects of individual-level andneighborhood-level guardianship, we also control for other variables sug-gested by previous literature to be important in understanding burglaryvictimization. At the individual level, we control for suitable/attractivehousehold goods, land uses that surround the respondent residence, per-ceived disorder, and respondent sociodemographic characteristics. Individ-ual-level household goods are measured through an index that sums fivedichotomous measures to indicate whether the respondent owned (1 = yes;0 = no) the following portable, expensive goods: a color television, a VCR,a 35-mm camera, a home computer, a bicycle or a motorcycle. Land usethat surrounds the respondent’s property was measured using three vari-ables, based on literature that indicates differential effects of differenttypes of nonresidential land uses on crime risk and fear (see, e.g., Kurtz,Koons, and Taylor, 1998; Wilcox et al., 2004). First, we control for busi-ness-oriented land use by summing individual responses to six surveyitems that indicate (1 = yes; 0 = no) whether the respondent lived withinthree blocks of a store or gas station, a bar or nightclub, a fast food restau-rant, a bank or office, a shopping center or mall, or a hotel/motel. In addi-tion, we control for whether the respondent lived within three blocks of aschool (1 = yes; 0 = no) or park/playground (1 = yes; 0 = no) with twoseparate single-item measures.8 Perceived disorder among respondentswas measured by summing responses to three dichotomous items, whichindicate whether respondents saw the following problems within fourblocks of their residences (1 = yes; 0 = no): litter or garbage on the street,abandoned houses or buildings, or presence of loitering teens. Finally, we

from the residential property in question. As such, it lacks the “target-specific”character of our other measures of guardianship.

8. Factor analysis of all eight land-use items resulted in two factors, with business-oriented places loading highly on one factor and schools and parks/playgroundsloading highly on a second factor. However, because of a low correlationbetween the schools and parks/playgrounds measures, these items were not com-bined. The business-oriented land use items, in contrast, showed high internalconsistency (a = .72).

Page 14: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 14 8-NOV-07 15:36

784 WILCOX, MADENSEN & TILLYER

controlled for respondent age (1 = 10–19 years. . . 7 = 70 years and older),sex (1 = male; 0 = female), race (1 = nonwhite; 0 = white), family income(1 = <$10,000. . . 7 = >$100,000), and whether the respondent lived alone(1 = yes; 0 = no).9

To discern properly individual, compositional versus contextual effects,we included analogous control variables at the neighborhood (tract) level.We measured tract-level household goods, tract-level presence of businessplaces, tract-level presence of schools, tract-level presence of parks/play-grounds, tract-level disorder, tract-level age structure, and gender compo-sition by averaging within census tracts values on those respectiveindividual-level measures. In addition, we controlled for neighborhooddisadvantage and instability by incorporating data from 1990 tract-levelU.S. Census files. Factor analysis of seven theoretically relevant socialstructural characteristics from census data revealed two factors. As such,we controlled for the first factor—concentrated disadvantage—with anindex summing tract-level proportion living in poverty, nonwhite popula-tion, female-headed households, and unemployment (a = .62). The secondfactor—population instability— was measured as the sum of standardizedvalues of tract-level population density, proportion of attached and/ormultiple housing units, and proportion of respondents living in the samehouse less than 5 years (a = .85).10 Values on census-based items weremultiplied by their respective squared factor loadings before beingsummed for index construction. Finally, in the analyses presented below,we controlled for downtown location (1 = yes; 0 = no) of each neighbor-hood.11 Bivariate correlations for all study variables can be found inappendix A.

ANALYSIS AND RESULTS

We test our study hypotheses using a hierarchical logistic modeling(HLM) approach and the HLM 6 software specifically (Raudenbush,Bryk, and Congdon, 2000).12 This strategy accounts appropriately for the

9. Although living alone is sometimes used as a proxy for lack of guardianship, wedo not include it in our operational definition of home occupancy because thenumber of days/nights the home is occupied provides a more direct measure ofpersonal guardianship provided. Living alone strictly is a control for householdstructure in this study.

10. The instability index was highly correlated with the aggregate persons livingalone. As such, aggregated “living alone” was not included as a level 2 variable.

11. Downtown location was determined by overlaying census tracts to the bounda-ries of Downtown Seattle—1 of 12 districts defined by the City of Seattle Plan-ning Department.

12. For all models shown, explanatory variables were grand–mean centered. PQLestimation was used, with robust standard errors reported.

Page 15: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 15 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 785

nonindependence among individuals clustered within neighborhood con-texts. Individuals and neighborhoods are represented with submodels inan HLM framework. These submodels, along with nested error terms,account for the variation in burglary victimization at each of the levels(see, e.g., Raudenbush and Bryk, 2002; Kreft and DeLeeuw, 1998). Thegeneral level 1 model estimated for each individual (i) in neighborhood (j)is of the following form:

Logit (VICTIMIZATION)ij = ß0j + ß1jX1ij + . . . ßkjXkij + eij (1)

where X1. . .Xk represent individual-level variables, the ßkj are logisticregression coefficients (or intercepts, as in the case of ß0j), and eij is a level1 error term assumed to be distributed binomially. The level 2, or neigh-borhood-level model, is a normal-errors regression model of the followinggeneral form:

ßkj = Qk0 + Qk1W1j . . .. . QkqWqj + ukj (2)

where W1 . . . .Wq represent neighborhood-level variables, the Qkq areregression coefficients (or intercepts) to be estimated, and ukj is a normallydistributed error term. These submodels are combined into an overall esti-mation of burglary victimization as follows:

Logit (VICTIMIZATION)ij = q00 + q10X1ij . . .+ q01W1j + q11X1ijW1j . . . +u0j + u1jX1ij + eij (3)

This combined model states that the logit of burglary victimization is afunction of an overall intercept (q00), main effects of individual-level fac-tors (i.e., q10), main effects of neighborhood-level characteristics (i.e., q01),cross-level interaction effects (i.e., q11), and random error (i.e., u0j + u1jX1ij

+ eij).In employing such multilevel logistic regression models, we began by

first estimating an intercept-only hierarchical logistic regression model toassess the variation in the mean risk of burglary across neighborhoods.This unconditional model (not shown) revealed significant variation inburglary victimization across the level 2 units (uij = .151, SD = .389, X2 =197.29, p < .001), thus confirming hierarchical modeling as an appropriatestatistical technique for the proposed analysis. In the next step of our anal-ysis, we incorporated individual-level and then individual-level and neigh-borhood-level explanatory variables into random-intercept hierarchicallogistic regression models. The model notation for this step of our analysisis shown below, and results from these models are presented in table 2.

Level 1( models 1 and 2, table 2): Logit (BURGLARY VICTIMIZA-TION)ij = ß0j + ß1j(target hardening)ij + ß2j(home occupancy)ij + ß3j(infor-mal social control)ij + ß4j(defensible space)ij + ß5j(household goods)ij + ß6j

Page 16: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 16 8-NOV-07 15:36

786 WILCOX, MADENSEN & TILLYER

(businesses nearby)ij + ß7j(school nearby)ij + ß8j(park nearby)ij + ß9j(per-ceived disorder)ij + ß10j(age)ij + ß11j(male)ij + ß12j(nonwhite)ij + ß13j(familyincome)ij + ß14j(live alone)ij + eij (4)

Level 2( model 1, table 2): ß0j = Q00 + u0j (5)

Level 2( model 2, table 2): ß0j = Q00 + Q01(neigh. target hardening)j + Q02

(neigh. home occupancy)j + Q03(neigh. informal social control)j + Q04

(neigh. defensible space)j + Q05(neigh. goods)j + Q06(neigh. businessesnearby)j + Q07(neigh. school nearby)j + Q08(neigh. park nearby)j + Q09

(neigh. disorder)j + Q010(neigh. age)j + Q011(neigh. sex)j + Q012(concen-trated disadvantage)j + Q013(population instability)j + Q014(downtown)j +u0j (6)

The individual-level and neighborhood-level fixed effects that corre-spond to the above model specifications are displayed in table 2 in termsof both logistic regression coefficients (with corresponding standarderrors) and odds ratios or exponentiated coefficients. A 1-unit increase inan explanatory variable x corresponds with a [(exp (bx) – 1) × 100] percentincrease or decrease in the odds of burglary victimization. Examination offindings for fixed effects across models 1 and 2 in table 2 reveals that, ofthe four individual-level guardianship variables, only target hardening anddefensible space features were related to the risk of burglary “on average”(i.e., with individuals pooled across neighborhoods). Both target harden-ing and defensible space were significantly related negatively to burglaryvictimization. Those who used a greater number of different types of“physical” safety precautions were significantly less likely to experience aburglary than those who used fewer. The odds ratio from model 2 in table2 indicates that the odds of burglary declined by 11 percent per incrementincrease in target hardening strategies. Also, properties with a greaternumber of environmental design features that promote defensible space(i.e., facilitate natural surveillance) were significantly less likely to be bur-gled than properties without these characteristics. More specifically, theexponentiated coefficient for the defensible space variable in model 2 sug-gests that each increment increase in defensible space was associated withan 18 percent decline in the odds of burglary victimization. On the otherhand, no significant relationship was found between burglary victimizationand either informal social control or home occupancy, though the effectfor home occupancy was very close to conventional levels of significanceand was in the expected negative direction. Model 2 also shows that net ofneighborhood-level characteristics, number of household goods, perceiveddisorder, and family income were related positively to burglary victimiza-tion risk, whereas being nonwhite was related negatively to burglary vic-timization. In terms of neighborhood-level main effects, inspection of

Page 17: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 17 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 787

results from model 2 of table 2 shows that all aggregate guardianship vari-ables were nonsignificant. However, neighborhood disadvantage wasrelated positively to burglary victimization, and downtown location wasrelated negatively to such victimization.13

Although these pooled or average effects are enlightening, the majorpremise of recent multilevel opportunity theory is that individual andaggregate guardianship—whether significant or not on average—interactwith one another such that the effects of individual guardianship are con-ditional on levels of aggregate guardianship, with the implication that indi-vidual guardianship practices may work better/worse in particularcontexts. Such conditional effects were posited in hypotheses 1 though 4.In the next step of our analysis, we build on the baseline random interceptmodel with neighborhood-level predictors provided by model 2 in table 2and examine each of the four stated hypotheses through a series of 16random coefficient regression models. Recall that each hypothesis impliedan interaction between a single individual-level dimension of guardianshipand four aggregate dimensions of guardianship. Collinearity among thefour interaction terms implied within each hypothesis precluded theirsimultaneous inclusion into a single model, and instead we elected to esti-mate the four interactions addressed in each hypothesis separately, for atotal of 16 models.

In the estimation of the 16 models necessary for testing our hypotheses,all coefficients were assumed to be constant across neighborhoods asopposed to random except for 1) the intercept and 2) the slope associatedwith the individual-level guardianship measure involved in themicro–macro interaction in question. In other words, we only estimatedvariation in slope coefficients for those individual-level predictors forwhich we had a priori theoretical rationale, as provided by our hypotheses.

13. When contiguous contextual units are employed, a concern exists that contextualeffects may be overestimated if spatial autocorrelation is not accounted for in themodeling process (e.g., Baller et al., 2001; Chaix, Merlo, and Chauvin 2005). Yet,scholars have noted the difficulty in estimating both spatial effects and multileveleffects simultaneously (Morenoff, 2003; Morenoff et al., 2001). In an attempt todiscern whether the neighborhood effects in our models were confoundedbecause of spatial autocorrelation, we estimated a three-level, random-interceptmodel identical to the two-level model presented in model 2, table 2, but witherror estimated at individual, census tract, and district levels of analysis. In 1990,the City of Seattle recognized 12 districts that were clusters of census tracts. Ourthree-level model, thus, controlled for correlation among contiguous units withindistricts. This three-level model (not shown) did suggest significant variation inburglary victimization at the district level, but the effects of explanatory variableswere not different from those presented in our two-level model (all effects pro-duced by the two-level model fell within the confidence intervals produced by thethree-level model). Therefore, we feel confident that the findings reported hereindo not overestimate contextual effects.

Page 18: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 18 8-NOV-07 15:36

788 WILCOX, MADENSEN & TILLYER

Table 2. Random-Intercept Models of BurglaryVictimizationa

Model 1 Model 2Fixed Effect Coefficient SE OR Coefficient SE OR

Target hardening, q10 −.106* (.044) .899 −.115* (.046) .892Home occupancy, q20 −.023 (.014) .995 −.025 (.013) .975Informal social control, q30 .008 (.098) 1.007 .035 (.101) 1.036Defensible space, q40 −.211* (.037) .810 −.199* (.039) .820Household goods, q50 .068* (.032) 1.07 .070* (.034) 1.073Businesses nearby, q60 −.005 (.032) .995 −.015 (.040) .985School nearby, q70 −.125 (.120) .883 −.167 (.127) .846Park/playground nearby, q80 .194* (.092) 1.214 .175 (.107) 1.191Perceived disorder index, q90 .192* (.039) 1.211 .128* (.044) 1.136Age, q100 −.023 (.027) .978 −.020 (.029) .980Sex (male), q110 −.008 (.084) .992 −.017 (.085) .983Race (nonwhite), q120 −.124 (.138) .883 −.290* (.146) .748Family income, q130 .075* (.035) 1.078 .097* (.036) 1.102Live alone, q140 .110 (.107) 1.117 .118 (.110) 1.125Intercept (mean burglary),q00 −1.607* (.053) — −1.616* (.047) —Neigh. target hardening, q01 — — — .190 (.273) 1.209Neigh. home occupancy, q02 — — — .058 (.093) 1.060Neigh. informal soc. control, q03 — — — −.472 (.572) .624Neigh. defensible space, q04 — — — −.110 (.138) .896Neigh. household goods, q05 — — — .009 (.237) 1.008Neigh. businesses nearby, q06 — — — .132 (.097) 1.142Neigh. school nearby, q07 — — — .084 (.266) 1.088Neigh. park nearby, q08 — — — .201 (.243) 1.223Neigh. disorder, q09 — — — .036 (.196) 1.037Neigh. age, q010 — — — −.001 (.142) .999Neigh. sex (male), q011 — — — −.044 (.671) .957Concentrated disadvantage, q012 — — — .979* (.221) 2.219Population instability, q013 — — — .003 (.046) 1.003Downtown location, q014 — — — −.643* (.321) .525

Random Effect Variance SD Chi-square Variance SD Chi-square

Mean burglary risk, uoj .097 .031 159.36* .060 .244 118.37*Level 1 extra binomial error, eij .966 .983 .980 .990

a Results are based on 4,227 individuals within 100 neighborhoods (census tracts).* p < .05.

The level 1 model specification associated with the 16 random coefficientregression models testing hypotheses 1 through 4 is the same as the level 1specification used for the models displayed in table 2 (see equation 4).Likewise, the level 2 equation for estimation of ß0j in this stage of theanalysis remains the same as that used for model 2, table 2 (see equation6). In addition, we add the following components to our level 2specification:

ßkj = Qk0 + Qk1(neigh. target hardening) + ukj (7a)ßkj = Qk0 + Qk2(neigh. home occupancy) + ukj (7b)ßkj = Qk0 + Qk3(neigh. informal social control) + ukj (7c)ßkj = Qk0 + Qk4(neigh. defensible space) + ukj (7d)

where ßkj represents the slope coefficients for target hardening (ß1j), home

Page 19: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 19 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 789

occupancy (ß2j), informal social control (ß3j), or defensible space (ß4j), andqk1. . . qk4 represent the coefficients for the effect of the four neighbor-hood-level guardianship measures on the random slope. Cross-level inter-actions involving individual-level guardianship measures andneighborhood-level guardianship measures result when equations 7a–7dare combined with the level 1 equation. To ease presentation of resultsfrom the 16 random-coefficient models estimated, we do not report thelevel 1 and level 2 fixed effects as no substantive changes in such effectsexisted from those reported in table 2.14 In table 3, however, we report thelogistic regression coefficient, the standard error, and the odds ratio foreach of the interaction effects discussed in hypotheses 1 through 4 anddepicted by the level 2 specification in equations 7a–7d.

Looking at the first column of table 3, we see that three of the fourestimated interactions with individual-level target hardening were signifi-cant and that each significant effect was consistent with hypothesis 1. Forinstance, we find that the negative effect of individual target hardeningbecame stronger (more negative) as neighborhood-level target hardeningincreased, which suggests that using more target hardening strategies, suchas locking doors or installing burglar alarms, is likely to be more effectivein neighborhoods where other residents also use a greater number of tar-get hardening techniques. More specifically, the odds ratio for the interac-tion effect (.745) indicates that for persons with neighborhood targethardening 1 unit above the mean, the change in odds of burglary victimiza-tion associated with a 1-unit increase in individual target hardening is .745times the change in odds of burglary victimization per unit increase in indi-vidual target hardening for someone at the mean level of neighborhoodtarget hardening. With exp(b) for individual target hardening equal to .890in the full model with this interaction term, the change in odds of victimi-zation per unit increase in individual target hardening is .890 for someoneat the mean level of neighborhood target hardening (and withgrand–mean centering, this mean level = 0). However, for someone at the“mean + 1” level of neighborhood target hardening (in this case, neighbor-hood target hardening = 1), the change in odds of victimization per unitincrease in individual target hardening is: .890 × .745 = .665. For illustrativepurposes, figure 1 displays the change in the odds of victimization associ-ated with a unit change in target hardening across different conditions.Each cluster of bars on figure 1 represents the change in the odds ratio

14. Although coefficients fluctuated slightly over the 16 models, significance levels(and thus substantive conclusions) remained consistent for all variables exceptfor “downtown location”—the p value for downtown moved from p < .05 to p <.10 in several models that incorporate random slopes and contextual predictorsthereof.

Page 20: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 20 8-NOV-07 15:36

790 WILCOX, MADENSEN & TILLYER

Tab

le 3

.C

ross

-Lev

el I

nter

acti

ons

betw

een

Indi

vidu

al a

nd N

eigh

borh

ood

Gua

rdia

nshi

p in

Est

imat

ing

Log

Odd

s of

Bur

glar

y V

icti

miz

atio

na

Indi

vidu

al L

evel

Gua

rdia

nshi

p

Tar

get

hard

enin

gH

ome

occu

panc

yIn

form

al s

ocia

l co

ntro

lD

efen

sibl

e sp

ace

Nei

ghbo

rhoo

d L

evel

Gua

rdia

nshi

pC

oeff

icie

ntSE

OR

Coe

ffic

ient

SEO

RC

oeff

icie

ntSE

OR

Coe

ffic

ient

SEO

R

Targ

et h

arde

ning

, q k

1–.

294*

(.10

5).7

45–.

097*

(.03

4).9

07–.

345

(.25

4).7

08–.

083

(.09

6).9

20H

ome

occu

panc

y, q

k2–.

029

(.04

7).9

71–.

021

(.01

5).9

78–.

133

(.09

1).8

76–.

042

(.04

4).9

59In

form

al s

ocia

l co

ntro

l, q k

3–.

751*

(.35

1).4

72–.

205*

(.09

0).8

14–1

.214

+(.

651)

.297

–.42

5(.

301)

.654

Def

ensi

ble

spac

e, q

k4–.

234*

(.09

2).7

92–.

063*

(.02

5).9

39–.

382+

(.22

1).6

83–.

146+

(.08

1).8

64

a Res

ults

are

bas

ed o

n 4,

227

indi

vidu

als

wit

hin

100

neig

hbor

hood

s (c

ensu

s tr

acts

).+p

< .

10;

* p

< .

05.

Page 21: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 21 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 791

associated with target hardening when the neighborhood level of guardi-anship is at the mean versus at the level “mean + 1.” The first cluster ofbars depicts the change in the odds of burglary victimization per unitincrease in target hardening associated with a neighborhood context inwhich aggregate target hardening is at the level of the sample mean versusa neighborhood context in which aggregate target hardening is at the levelof “mean + 1.”

Figure 1. Change in Odds of Burglary Victimization perUnit Change in Individual-Level TargetHardening across Varying NeighborhoodGuardianship Contexts.

Looking more at the interaction revealed in the first column of table 3,whereas neighborhood-level social guardianship did not influence burglaryrisk directly, greater levels of this aggregate form of informal social controlinfluenced the preventative effect of individual target hardening. The neg-ative effect of individual target hardening was stronger (more highly nega-tive) as neighborhood-level social guardianship, or the proportion ofneighbors asked to watch others’ property, increased. In fact, the odds

Page 22: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 22 8-NOV-07 15:36

792 WILCOX, MADENSEN & TILLYER

ratio for the interaction term suggests that the change in the odds of bur-glary victimization for someone experiencing the “mean +1” level ofneighborhood informal social control is .472 times the change in odds ofburglary victimization per unit increase in target hardening for someone atthe mean level of neighborhood informal social control (see third clusterof bars in figure 1 for visual display of this interaction effect).

In the first column of table 3, we also find that the negative effect ofindividual target hardening was strengthened as the neighborhood level ofdefensible space increased. This finding implies that the use of individualtarget hardening strategies may be more effective in neighborhoods thathave more environmental design features that facilitate natural surveil-lance and may be less effective in neighborhoods that impede this process.This interaction effect is illustrated in the last set of bars in figure 1. Theonly estimated interaction that was not consistent with hypothesis 1 wasthe conditioning effect of neighborhood home occupancy. The interactionbetween individual-level target hardening and neighborhood-level homeoccupancy, although in the expected direction, failed to reach significance.

To test hypothesis 2, we explored in similar fashion the interactionsbetween neighborhood-level guardianship and individual-level home occu-pancy. The coefficients in the second column of table 3 report these inter-action effects. Here, again, we find that three of the four estimatedinteractions were significant, and each significant effect was consistentwith our second hypothesis. The negative effect of home occupancy onburglary victimization was stronger (more negative) as neighborhoodlevels of target hardening, informal social control, or defensible spaceincreased. The odds ratios for these interaction effects suggest that asneighborhood levels of target hardening, informal social control, anddefensible space increase by 1 unit, the change in the odds of victimizationassociated with a 1-unit increase in individual home occupancy is multi-plied by .907, .814, and .939, respectively. In contrast, the effect of individ-ual home occupancy is not influenced by neighborhood home occupancy.

The coefficients in the third column of table 3 represent the condition-ing effects of neighborhood guardianship on the relationship between indi-vidual informal social control and burglary victimization. We hadhypothesized that individual-level informal social control would be relatedmore negatively to burglary victimization as neighborhood-level targethardening, home occupancy, informal social control, and natural surveil-lance increased. None of the interaction effects posited by hypothesis 3was significant at the p < .05 level, but several coefficients had associated pvalues that fall between .05 and .10. For instance, marginal evidenceshowed that, as hypothesized, informal social control was related morenegatively to burglary victimization as neighborhood informal social con-trol and aggregate defensible space increased. Despite being significant

Page 23: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 23 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 793

only at p < .10, these interaction effects reveal that as neighborhood infor-mal social control and aggregate defensible space increased by 1 unit, thechange in the odds of victimization per unit increase in individual informalsocial control was multiplied by .297 and .683, respectively.

Finally, in the last column of table 3, we see little evidence of a condi-tioning effect of neighborhood-level target hardening, home occupancy, orinformal social control on the relationship between individual defensiblespace and burglary victimization as proposed in hypothesis 4. However,consistent with hypothesis 4, we found modest evidence that greaterneighborhood levels of defensible space strengthened the negative effectof individual defensible space. Thus, properties with more defensible spacecharacteristics were least likely to experience burglaries in neighborhoodsthat had a greater number of properties with these same features. In fact,the odds decline in burglary associated with a 1-unit increase in individualdefensible space was .808 when neighborhood defensible space was at themean level but .708 when neighborhood defensible space was at a level 1-unit above the mean.

DISCUSSION AND CONCLUSIONS

The primary purpose of this article was to extend the literature thataddresses the relationship between guardianship and burglary victimiza-tion in two important ways. First, we aimed to provide a more comprehen-sive conceptualization of guardianship by incorporating multipledimensions of guardianship at both individual and aggregate levels into asingle study. Our study focused on four dimensions of guardianship—physical, personal, social, and natural—at both individual and neighbor-hood levels of analysis. In considering only “average” or “main” effects ofthese dimensions, individual-level physical guardianship (target harden-ing) and natural guardianship (defensible space) seemed to be the mostimportant in understanding burglary victimization.

However, relying on only fixed effects can foster misleading conclusionsin regard to the importance of other dimensions of guardianship. Thisissue gets at the second way we attempted to extend the literature regard-ing the effects of guardianship—through an emphasis on the interplaybetween individual and aggregate guardianship. Analysis of interactionsbetween individual and aggregate measures of guardianship revealed thatneighborhood-level physical, social, and natural guardianship moderatedeffects of individual guardianship significantly, especially individual-levelphysical and personal guardianship. All significant interactions were con-sistent with recent advances in multilevel opportunity theory, which sug-gests that individual guardianship is related more negatively to crime/victimization as neighborhood guardianship increases (see, e.g., Wilcox,

Page 24: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 24 8-NOV-07 15:36

794 WILCOX, MADENSEN & TILLYER

Land, and Hunt, 2003). Our study revealed that three of the four dimen-sions of guardianship at the neighborhood level served to condition certainmeasures of individual guardianship in the manner hypothesized. Thus,whereas neighborhood guardianship does not seem to play a major role inburglary victimization in terms of “main” effects, net of individual differ-ences in guardianship, clearly it plays an important role in terms of moder-ating effects. In short, aggregate patterns of guardianship seemed to forman important “backcloth” or context in which individual guardianshipefforts were situated. This backcloth impacts the extent to which individ-ual guardianship correlates with victimization. More specifically, contextswith higher levels of ambient guardianship tended to exacerbate the pro-tective nature of individual guardianship efforts in the form of target hard-ening or home occupancy, in particular. Conversely, the effects ofindividual guardianship were tempered in contexts with weaker collectiveguardianship. As such, we find support for the idea that the supply ofaggregate guardianship increases market costs associated with crime,which reinforces individual guardianship practices (Wilcox, Land, andHunt, 2003: 109).

Such conclusions have important crime-prevention implications. Mostcrime-prevention policy is “single level” in nature, focusing on either indi-vidual/situational prevention practices or community-based crime preven-tion. Our findings suggest that such strategies need to be considered inconjunction with one another. Individual or microsituational guardianshipmeasures will not yield the same crime-reduction benefit in all environ-mental contexts. Rather, individual efforts are most effective when carriedout in neighborhoods where many people are making similar efforts.Because collective effort has the potential to “make or break” individualeffort, crime-prevention policy must be multilevel, with a clear focus onenhancing environmental or collective guardianship. Site-specific targethardening or home occupancy, for instance, may produce disappointingresults if not combined with efforts to enhance target hardening, informalsocial control, or defensible space throughout the neighborhood. Suchpossibilities suggest the use of context-driven crime-prevention policy,despite the fact that most variation in burglary victimization is at the indi-vidual level. Although macrolevel opportunity may exhibit few directeffects on burglary, net of individual differences, its potential to alter theutility of microlevel efforts makes it an important target for preventionefforts.

We find moderately strong support for a multilevel opportunity theoryand context-driven multilevel crime-prevention policy, but important limi-tations of our study necessarily qualify our conclusions. For instance, inour test of multilevel opportunity theory, we focus only on micro–macrointeractions in regard to one aspect of criminal opportunity: guardianship.

Page 25: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 25 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 795

Although we control for certain dimensions of individual and aggregateexposure and target suitability, we do not incorporate micro–macro inter-actions regarding these other concepts. As such, the possibility for miss-pecification exists with the models presented herein. Furthermore, ourconceptualization and operational definition of guardianship is imperfect.We see value in integrating compatible concepts across theoretical per-spectives (i.e., target hardening, home occupancy, informal social control,or natural surveillance) under the umbrella of “guardianship,” but we real-ize that other scholars might oppose such a broad conceptualization. Onthe other hand, others might criticize our conceptualization and opera-tional definition of guardianship for being too narrow, perhaps omittingkey dimensions. Indeed, we recognize that our measurement of guardian-ship is not exhaustive, and we only suggest that it offers one of the morecomprehensive measurements in a single study to date.

Finally, we estimated our models using data from neighborhoods withinone city. We make no claims that the Seattle, Washington neighborhoodsexamined here are representative of neighborhoods in other cities or evenof neighborhoods within the city of Seattle, and therefore, we recognizethe potentially limited generalizability of our findings. Replication withmultilevel data from other locales and with greater attention tomicro–macro interactions in regard to other aspects of criminal opportu-nity, therefore, is needed before the theoretical and practical implicationsof our work gain greater clarity. Despite these important qualifications, wesee our findings as an important step in refining knowledge of how guardi-anship practices reduce burglary victimization.

REFERENCES

Baller, Robert D., Luc Anselin, Steven F. Messner, Glenn Deane, andDarnell F. Hawkins. 2001. Structural covariates of U.S. county homi-cide rates: Incorporating spatial effects. Criminology 39:561–90.

Bellair, Paul E. 2000. Informal surveillance and street crime: A complexrelationship. Criminology 38:137–70.

Bennett, Trevor, and Richard Wright. 1984. Burglars on Burglary. Alder-shot, U.K.: Gower Publishing.

Brantingham, Paul J., and Patricia L. Brantingham. 1975. The spatial pat-terning of burglary. Howard Journal of Criminal Justice 14:11–23.

Brantingham, Patricia L., and Paul J. Brantingham. 1993. Nodes, paths,and edges: Considerations on the complexity of crime and the physicalenvironment. Journal of Environmental Psychology 13:3–28.

Page 26: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 26 8-NOV-07 15:36

796 WILCOX, MADENSEN & TILLYER

Brantingham, Patricia L., and Paul J. Brantingham. 1999. A theoreticalmodel of crime hot spot generation. Studies on Crime and Crime Pre-vention 8:7–26.

Buck, Andrew J., Simon Hakim, and George F. Rengert. 1993. Burglaralarms and the choice behavior of burglars: A suburban phenomenon.Journal of Criminal Justice 21:497–507.

Chaix, Basile, Juan Merlo, and Pierre Chauvin. 2005. Comparison of a spa-tial approach with the multilevel approach for investigating placeeffects on health: The example of healthcare utilization in France. Jour-nal of Epidemiology and Community Health 59:517-26.

Cisneros, Henry G. 1995. Defensible space: Deterring crime and buildingcommunity. Washington, DC: U.S. Department of Housing and UrbanDevelopment.

Clarke, Ronald V. 1980. “Situational” crime prevention: Theory and prac-tice. British Journal of Criminology 20:136–47.

Clarke, Ronald V. 1995. Situational crime prevention. Crime and Justice19:91–150.

Clarke, Ronald V., and Derek B. Cornish. 1985. Modeling offenders’ deci-sions: A framework for research and policy. Crime and Justice6:147–85.

Cohen, Lawrence E., and David Cantor. 1981. Residential burglary in theUnited States: Lifestyle and demographic factors associated with theprobability of victimization. Journal of Research in Crime and Delin-quency 18:113–27.

Cohen, Lawrence E., and Marcus Felson. 1979. Social change and crimerate trends: A routine activity approach. American Sociological Review44:588–608.

Cohen, Lawrence E., Marcus Felson, and Kenneth C. Land. 1980. Prop-erty crime rates in the United States: A macrodynamic analysis, 1947-1977; with ex ante forecasts for the mid-1980s. The American Journalof Sociology 86:90–118.

Cohen, Lawrence E., James R. Kluegel, and Kenneth C. Land. 1981. Socialinequality and predatory criminal victimization: An exposition and testof a formal theory. American Sociological Review 46:505–24.

Cornish, Derek B., and Ronald V. Clarke. 2003. Opportunities, precip-itators, and criminal decisions: A reply to Wortley’s critique of situa-tional crime prevention. In Theory for Practice in Situational Crime

Page 27: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 27 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 797

Prevention, eds. Martha J. Smith, and Derek B. Cornish. Monsey, NY:Criminal Justice Press.

Coupe, Timothy, and Laurence Blake. 2006. Daylight and darkness target-ing strategies and the risks of being seen at residential burglaries.Criminology 44:431–64.

Cromwell, Paul F., James N. Olson, and D’Aunn Wester Avary. 1991.Breaking and entering: An ethnographic analysis of burglary. NewburyPark, CA: Sage.

Donnelly, Patrick G., and Charles E. Kimble. 1997. Community organiz-ing, environmental change, and neighborhood crime. Crime and Delin-quency 43:493–511.

Felson, Marcus, and Lawrence E. Cohen. 1980. Human ecology and crime:A routine activity approach. Human Ecology 8:389–406.

Fisher, Bonnie S., and Jack L. Nasar. 1992. Fear of crime in relation tothree exterior site features: Prospect, refuge, and escape. Environmentand Behavior 24:35–65.

Forrester, David, Mike Chatterton, and Ken Pease. 1988. The KirkholtBurglary Prevention Project, Rochdale. Crime Prevention Unit Paper13. London, U.K.: Home Office.

Garofalo, James, and David Clark. 1992. Guardianship and residentialburglary. Justice Quarterly 9:443–63.

Groff, Elizabeth R., and Nancy G. LaVigne. 2001. Mapping an opportu-nity surface of residential burglary. Journal of Research in Crime andDelinquency 38:257–78.

Hakim, Simon, Mary Ann Gaffney, George Rengert, and JohannanShachmurove. 1995. Costs and benefits of alarms to the community:Burglary patterns and security measures in Tredyffrin Township, Penn-sylvania. Security Journal 6:197–204.

Jacobs, Jane. 1961. The Death and Life of Great American Cities. NewYork: Random House.

Kennedy, Leslie W., and David R. Forde. 1990. Routine activities andcrime: An analysis of victimization in Canada. Criminology 28:137–51.

Kreft, Ita, and Jan DeLeeuw. 1998. Introducing Multilevel Modeling.Thousand Oaks, CA: Sage.

Page 28: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 28 8-NOV-07 15:36

798 WILCOX, MADENSEN & TILLYER

Kurtz, Ellen M., Barbara A. Koons, and Ralph B. Taylor. 1998. Land use,physical deterioration, resident-based control, and calls for service onurban street blocks. Justice Quarterly 15:121–49.

Laycock, Gloria. 1985. Property Marking: A Deterrent to Domestic Bur-glary? Crime Prevention Unit Paper 3. London, U.K.: Home Office.

Mawby, R. I. 1977. Defensible space: A theoretical and empiricalappraisal. Urban Studies 14:169–79.

Merry, Sally F. 1981. Defensible space undefended: Social factors in crimeprevention through environmental design. Urban Affairs Quarterly16:397–422.

Messner, Steven F., and Judith R. Blau. 1987. Routine leisure activitiesand rates of crime: A macro-level analysis. Social Forces 65:1035–52.

Miethe, Terance D. 1991. Citizen-based crime control activity and victimi-zation risks: An examination of displacement and free-rider effects.Criminology 29:419–39.

Miethe, Terance D. 1992. Testing theories of criminality and victimizationin Seattle, 1960-1990. Ann Arbor, MI: Inter-University Consortium forPolitical and Social Research.

Miethe, Terance D., and David McDowall. 1993. Contextual effects inmodels of criminal victimization. Social Forces 71:741–59.

Miethe, Terance D., and Robert F. Meier. 1990. Opportunity, choice, andcriminal victimization: A test of a theoretical model. Journal ofResearch in Crime and Delinquency 27:243–66.

Miethe, Terance D., and Robert F. Meier. 1994. Crime and its social con-text: Toward an integrated theory of offenders, victims, and situations.Albany: State University of New York Press.

Miethe, Terance D., Mark C. Stafford, and J. Scott Long. 1987. Social dif-ferentiation in criminal victimization: A test of routine activities/lifes-tyle theories. American Sociological Review 52:184–94.

Morenoff, Jeffrey D. 2003. Neighborhood mechanisms and the spatialdynamics of birth weight. American Journal of Sociology 5:976–1017.

Morenoff, Jeffrey D., Robert J. Sampson, and Stephen W. Raudenbush.2001. Neighborhood inequality, collective efficacy, and the spatialdynamics of urban violence. Criminology 39:517–60.

Mustaine, Elizabeth E., and Richard Tewksbury. 1997. Obstacles in theassessment of routine activity theory. Social Pathology 3:177–94.

Page 29: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 29 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 799

Mustaine, Elizabeth E., and Richard Tewksbury. 1998. Predicting risks oflarceny theft victimization: A routine activity analysis using refinedlifestyle measures. Criminology 36:829–58.

Newman, Oscar. 1972. Defensible Space: Crime Prevention through UrbanDesign. New York: Macmillan.

Newman, Oscar. 1996. Creating Defensible Space. Washington, DC: U.S.Department of Housing and Urban Development.

Norris, Fran H., and Krzysztof Kaniasty. 1992. A longitudinal study of theeffects of various crime prevention strategies on criminal victimization,fear of crime, and psychological distress. American Journal of Commu-nity Psychology 20:625–48.

Outlaw, Maureen, R. Barry Ruback, and Chester L. Britt. 2002. Repeatand multiple victimizations: The role of individual and contextual fac-tors. Violence and Victims 17:187–204.

Painter, Kate A., and David P. Farrington. 2001. Evaluating situationalcrime prevention using a young people’s survey. The British Journal ofCriminology 41:266–84.

Palmer, Emma J., Angela Holmes, and Clive R. Hollin. 2002. Investigatingburglars’ decisions: Factors influencing target choice, method of entry,reasons for offending, repeat victimization of a property and victimawareness. Security Journal 15:7–18.

Raudenbush, Stephen W., and Anthony S. Bryk. 2002. Hierarchical LinearModels: Applications and Data Analysis Methods, 2nd ed. ThousandOaks, CA: Sage.

Raudenbush, Stephen W., Anthony S. Bryk, and Richard Congdon. 2000.HLM 6 Hierarchical Linear and Nonlinear Modeling. Pleasanton, CA:Scientific Software International.

Rengert, George, and John Wasilchick. 1985. Suburban Burglary: A Timeand Place for Everything. Springfield, IL: Charles Thomas.

Roncek, Dennis W., and Pamela Maier. 1991. Bars, blocks, and crimesrevisited: Linking the theory of routine activities to the empiricism of‘hot spots.’ Criminology 29:725–53.

Sampson, Robert J., and John D. Wooldredge. 1987. Linking the micro-and macro-level dimensions of lifestyle-routine activity and opportu-nity models of predatory victimization. Journal of Quantitative Crimi-nology 3:371–93.

Page 30: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 30 8-NOV-07 15:36

800 WILCOX, MADENSEN & TILLYER

Sherman, Lawrence, Patrick Gartin, and Michael Buerger. 1989. Hotspotsof predatory crime: Routine activities and the criminology of place.Criminology 27:27–55.

Shover, Neal. 1991. Burglary. Crime and Justice 14:73–113.

Smith, Douglas A., and G. Roger Jarjoura. 1988. Social structure and crim-inal victimization. Journal of Research in Crime and Delinquency 25:27-52.

Smith, Douglas A., and G. Roger Jarjoura. 1989. Household characteris-tics, neighborhood composition and victimization risk. Social Forces68:621–40.

Taylor, Ralph B., and Stephen D. Gottfredson. 1986. Environmentaldesign, crime, and prevention: An examination of community dynam-ics. Crime and Justice 8:387–416.

Taylor, Ralph B., and Adele V. Harrell. 1996. Physical Environment andCrime. Washington, DC: U.S. Department of Justice, Office of JusticePrograms, National Institute of Justice. NCJ 157311.

Taylor, Ralph B., Barbara A. Koons, Ellen M. Kurtz, Jack R. Greene, andDouglas D. Perkins. 1995. Street blocks with more nonresidential landuse have more physical deterioration: Evidence from Baltimore andPhiladelphia. Urban Affairs Review 31:120–36.

Tewksbury, Richard, and Elizabeth Ehrhardt Mustaine. 2003. College stu-dents’ lifestyles and self-protective behaviors: Further considerationsof the guardianship concept in routine activity theory. Criminal Justiceand Behavior 30:302–27.

Tilley, Nicholas, and Janice Webb. 1994. Burglary Reduction: Findingsfrom Safer Cities Schemes. Crime Prevention Unit Paper 51. London,U.K.: Home Office.

Titus, Richard M. 1989. Factors Relating to Completed vs. Attempted Forci-ble Residential Burglary. Washington, DC: U.S. Department of Justice,National Institute of Justice.

Warner, Barbara D., and Pamela Wilcox Rountree. 1997. Local social tiesin community and crime model: Questioning the systemic nature ofinformal control. Social Problems 44:423–39.

White, Garland F. 1990. Neighborhood permeability and burglary rates.Justice Quarterly 7:57–67.

Wilcox Rountree, Pamela, and Kenneth C. Land. 1996. Burglary victimiza-tion, perceptions of crime risk, and routine activities: A multilevel

Page 31: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 31 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 801

analysis across Seattle neighborhoods and census tracts. Journal ofResearch in Crime and Delinquency 33:146–80.

Wilcox Rountree, Pamela, Kenneth C. Land, and Scott A. Hunt. 2003.Criminal Circumstance: A Dynamic Multicontextual Criminal Opportu-nity Theory. New York: Aldine de Gruyter.

Wilcox Rountree, Pamela, Kenneth C. Land, and Terance D. Miethe.1994. Macro-micro integration in the study of victimization: A hierar-chical logistic model analysis across Seattle neighborhoods. Criminol-ogy 32:387–414.

Wright, Richard T., and Scott H. Decker. 1994. Burglars on the Job: Street-life and Residential Break-Ins. Boston, MA: Northeastern UniversityPress.

Pamela Wilcox is an associate professor of criminal justice at the Uni-versity of Cincinnati. Her research interests include the continued devel-opment and application of a multilevel criminal opportunity perspective topredict offending, victimization, fear, and precautionary behavior in com-munity and school contexts.

Tamara D. Madensen is an assistant professor of criminal justice at theUniversity of Nevada, Las Vegas. Her research interests include theopportunity structures of drug markets, violence, white-collar crime, andcrowd disturbances. Her latest research focuses on developing a theoreti-cal model of place management that explains crime differences across sim-ilar types of facilities.

Marie Skubak Tillyer is a doctoral student in the Division of CriminalJustice at the University of Cincinnati. Her research interests includecrime prevention, criminological theory, and violence reduction. She iscurrently involved in a city-wide homicide reduction initiative in Cincin-nati, Ohio.

Page 32: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 32 8-NOV-07 15:36

802 WILCOX, MADENSEN & TILLYER

App

endi

x A

.B

ivar

iate

Cor

rela

tion

s

Biv

aria

te C

orre

lati

ons

Am

ong

Lev

el 1

Var

iabl

es

12

34

56

78

910

1112

1314

15

1.B

urgl

ary

vict

imiz

atio

n1.

002.

Targ

et h

arde

ning

–.05

1.00

3.H

ome

occu

panc

y–.

07.1

41.

004.

Info

rmal

soc

ial

cont

rol

–.03

.26

.12

1.00

5.D

efen

sibl

e sp

ace

–.12

.08

.13

.10

1.00

6.H

ouse

hold

goo

ds.0

4.0

9–.

17.0

7.0

31.

007.

Bus

ines

ses

near

by.0

5–.

15–.

13–.

14–.

19–.

061.

008.

Scho

ol n

earb

y–.

01.0

5.0

2.0

0.0

0.0

0.0

91.

009.

Par

k/pl

aygr

ound

nea

rby

.06

.00

–.07

–.01

–.05

.04

.09

.14

1.00

10.

Per

ceiv

ed d

isor

der

inde

x.1

1–.

01–.

14–.

05–.

17.0

0.2

5.0

8.1

31.

0011

.A

ge–.

07.1

2.3

8.1

6.1

1–.

28–.

17.0

1–.

13–.

201.

0012

.Se

x (m

ale)

.01

.04

–.07

–.01

–.05

.09

.02

.02

.06

–.02

–.04

1.00

13.

Rac

e (n

onw

hite

).0

0.0

3.0

3–.

06–.

06–.

02.0

4.0

9.0

6.0

5–.

05.2

21.

0014

.Fa

mily

inc

ome

.03

.11

–.13

.11

.04

.40

–.14

–.01

.06

–.08

–.15

.13

–.09

1.00

15.

Liv

e al

one

.02

–.16

–.08

–.12

–.08

–.27

.14

–.03

–.02

.03

.16

–.06

–.02

–.28

1.00

Page 33: GUARDIANSHIP IN CONTEXT: IMPLICATIONS FOR BURGLARY ... · crime-reduction implications for the effects of guardianship on burglary are assumed to be generalizable across contexts

\\server05\productn\C\CRY\45-4\CRY403.txt unknown Seq: 33 8-NOV-07 15:36

GUARDIANSHIP IN CONTEXT 803

App

endi

x A

.(c

onti

nued

)

Biv

aria

te C

orre

lati

ons

Am

ong

Lev

el 2

Var

iabl

es

12

34

56

78

910

1112

1314

15

1.B

urgl

ary

rate

1.00

2.Ta

rget

har

deni

ng–.

141.

003.

Hom

e oc

cupa

ncy

–.16

.69

1.00

4.In

form

al s

ocia

l co

ntro

l–.

34.6

9.5

61.

005.

Def

ensi

ble

spac

e–.

43.4

8.5

4.5

81.

006.

Hou

seho

ld g

oods

–.24

.49

.30

.61

.49

1.00

7.B

usin

esse

s ne

arby

.30

–.70

–.56

–.66

–.52

–.43

1.00

8.Sc

hool

nea

rby

.10

.16

.07

.12

.05

.18

–.03

1.00

9.P

ark/

play

grou

nd n

earb

y.2

2–.

05–.

05–.

09–.

12.0

2.0

1.2

31.

0010

.P

erce

ived

dis

orde

r in

dex

.52

–.24

–.27

–.51

–.52

–.53

.48

.10

.26

1.00

11.

Age

–.30

.49

.56

.58

.42

.10

–.49

.04

–.14

–.34

1.00

12.

Sex

(mal

e).0

9–.

14–.

38–.

20–.

32.0

1.1

8.0

3.0

2.1

8–.

401.

0013

.C

once

ntra

ted

disa

dvan

tage

.47

.16

.14

–.17

–.21

–.31

.05

.18

.17

.66

.00

–.10

1.00

14.

Popu

lati

on i

nsta

bilit

y.3

4–.

73–.

65–.

76–.

62–.

59.7

1–.

02.1

1.4

9–.

55.1

3.1

31.

0015

.D

ownt

own

loca

tion

.00

–.39

–.25

–.39

–.22

–.43

.41

–.13

.18

.30

.08

–.01

.09

.28

1.00