Understanding the Criminogenic Properties of Vacant Housing: A Mixed Methods Approach

Vacant Housing and Crime

In their formulation of crime pattern theory, Brantingham and Brantingham (1993) distinguish between crime attractors and generators. While crime attractors boast crime reputations or present criminal opportunities (e.g., items to steal), crime generators draw large amounts of people, some of whom may be potential offenders or victims (e.g., stadiums). Following this theoretical reasoning, an abandoned house could attract and/or generate crime. On the one hand, abandoned houses may provide direct, tangible opportunities for crime. They are less likely than other places to “have effective locks and other security measures” and may “reduce costs of crime by increasing the supply of easy theft and victimization opportunities” (MacDonald 2015:359). Skogan (1990) suggests these structures are targets for vandalism and the stripping of copper. Further, Spelman (1993) finds that 83 percent of abandoned buildings in a neighborhood showed evidence of use for various types of crime, including prostitution and drug dealing, but he also argues that abandoned buildings can be used as “staging areas or gathering areas” for crimes committed elsewhere. Teixera (2016) observes in a recent study that youth perceive vacant houses as ideal locations to conceal drug use.

If an abandoned house attracts offenders for the abovementioned reasons, it may generate additional crimes as well. For example, two unassociated individuals may go into an abandoned house separately to use drugs. If they cross paths inside, this could lead to an unexpected altercation. Abandoned housing may also lead to crime more indirectly as well. The presence of dilapidated housing could serve as a cue to potential offenders that residents and police are not invested in the area and that the neighborhood is in decline (Mummolo and Brubaker 2008; Skogan 1990; Wilson and Kelling 1982). Thus, offenders may offend in and around abandoned housing, perceiving these locations to be “fair game” for a range of illicit behavior.

A relatively small literature examines the salience of abandoned housing for crime, consistently finding a positive association between the presence or level of abandoned housing and crime. However, research is somewhat mixed as to whether abandoned housing is relevant for all crimes or only certain types of crime. Consistent with the tenets of situational crime prevention (Clarke 1995), the removal of abandoned housing may only be relevant for the specific criminal opportunities afforded by these structures. For example, Kim (2016) finds that the percentage of occupied homes on a street has a robust negative association with property crimes, but an inconsistent association with violent crimes. Similarly, Raleigh and Galster (2014) find that the proportion of vacant housing units on a street is positively related to burglary and drug crimes, but negatively associated with larceny and unrelated to violent crimes. Boessen and Chamberlain (2017) also find that the presence of abandoned housing is positively correlated with property, but not violent crime. As a departure from these studies, Branas and colleagues (2012) find that increases in vacant housing are associated with increases in aggravated assault and Wheeler et al. (2018) find evidence that demolishing abandoned houses in Buffalo, New York, was associated with reductions for both violent and nonviolent calls at the microgeographic level. However, effects did not consistently hold for census tracts. Wheeler and colleagues (2018) note that their inconsistent findings suggest a need for future research, especially using “measures of citizen perceptions” on whether removing vacant housing is actually connected with greater collective efficacy (p. 418).

We agree that more research is needed on the relationship between abandoned housing and crime. Research to date provides a somewhat limited, inconsistent, and incomplete picture of the dynamics underlying the relationship. Moreover, the use of citizen perceptions to better understand the meanings and connections between abandoned housing and crime is a fruitful direction. In addition, we argue that the perceptions of police and ex-offenders can be especially helpful here. Police are valuable informants about local crime dynamics. Also, little criminological research examines the processes by which offenders choose targets or make decisions about where and when to commit their crimes (Brantingham 2013; Jacobs, Topalli, and Wright 2003; Jacobs and Wright 1999; Wright et al. 1995). This more qualitative approach to understanding offender decision-making is essential to understanding how aspects of the environment, such as abandoned housing, are used in the commission of crime.

In short, our study builds on this area of research by investigating two related questions:

How does crime change at the street level after the removal of abandoned housing? and (2) what is the role of abandoned houses in generating or attracting crime? Given prior theory and research, we expect that streets where abandoned housing is torn down will account for a lower portion of total neighborhood crime and that a lower portion of CFS from those streets will pertain to serious crime. Based on prior research, we also expect that the removal of vacant properties may be more relevant for property versus violent crime. This is a possibility we explore both through our qualitative and quantitative data. ¹

Study Site

We focus our study on a high-crime neighborhood, “Hope Gardens,” which is located in a midsized Ohio city that had a violent crime rate nearly triple that of the national level and a property crime rate twice as high in 2012 (Uniform Crime Reports). ² Our research was carried out in cooperation with the local police department, who wanted to learn more about crime hot spots. Hope Gardens had a crime rate almost double the city rate. It is also the most diverse and youngest neighborhood in the city. As of 2000 decennial census, the neighborhood was 43 percent Black and 78 percent of all residents were under the age of 45. About 37 percent of residents were living in poverty and 40 percent of residents who were 25 years or older had not completed high school. As of 2015, 39 percent of the lots in Hope Gardens were vacant, giving the neighborhood a “checkered” appearance (Western Reserve Land Conservancy Thriving Communities Institute 2015). The neighborhood also has the highest percentage (10 percent) of abandoned properties in the city (Western Reserve Land Conservancy Thriving Communities Institute 2015). Some of our participants referred to Hope Gardens as a “forgotten” neighborhood, feeling as though people were moving out of the neighborhood in droves and that resources to revitalize the community were lacking. However, the neighborhood is home to seven community gardens—an effort among local outreach programs to convert vacant lots into positive spaces for residents. Hope Gardens is bordered on one side by a lake and on two other sides by a highway or industrial and commercial properties. These provide clear neighborhood boundaries and were consistent with those identified by participants. Thus, we use them to approximate our study area.

Data

Geonarratives

We carried out 35 ride-alongs with ex-offenders, police, and residents to collect geonarratives—environmentally cued narratives where place is used to stimulate discussion about fine-scale neighborhood characteristics (see Curtis et al. 2016). This is a simple yet powerful tool to collect perceptions and experiences of places. Geonarratives have been used to understand environmental contexts for outcomes such as health, crime, and postdisaster recovery (Curtis et al. 2015). As Curtis and colleagues (2015:22) note, “the need for personal context is compelling,” given that local knowledge may be essential to understanding relationships between people and places. Moreover, asking individuals to comment on an environment without being situated within that environment may yield less reliable or valid data. For example, cognitive mapping typically requires participants to identify places on a map that are meaningful to the research (e.g., “unsafe” places). However, maps may be confounded by factors such as spatial memory, familiarity with street names, or even the size and shape of the map provided (Pocock 1976). Placing an individual in the environment of interest and capturing “on the ground” reactions to places should thus produce more valid and reliable information.

In our study, participants provided comments about Hope Gardens while guiding two members of the research team through the area. These ride-alongs lasted generally 1 to 1.5 hours. Participants were instructed to provide open-ended comments about the neighborhood (specific to crime) while moving through the environment. Each route taken was influenced by what the participant felt was important for us to see, since the intention was to learn what was problematic or criminogenic about the neighborhood from their perspectives.

We used a mixture of nonrandom sampling strategies to recruit our participants. Police participants were selected based on their working relationship with Hope Gardens and were paid overtime by their department to participate. We recruited ex-offenders from local reentry support groups and meetings, where ex-offenders obtain information about employment, education, legal, or health concerns. We also employed purposive sampling by giving preference to ex-offenders and police officers over community members, since their insights may be more germane to understanding crime dynamics. Although the insights of community members may be equally valuable for understanding where crime occurs, police and ex-offenders should know more about the “why” and the “what” with respect to abandoned houses. We located community members through a popular local ministry and café, where a diverse crowd of individuals eat, socialize, hold local business meetings, and attend services. With community members and ex-offenders, we also used snowball sampling and met future participants through past participants. Community members and ex-offenders were provided a US$20 gift card to Family Dollar in exchange for their participation.

Table 1 shows demographic characteristics for our participants. There is a fair amount of variation across groups in terms of race and sex. Police officers were all male and mostly White. These demographics mirror the sampling frame for the city police department, which is only 10 percent female and mostly White (81 percent). On the other hand, our sample for community members and ex-offenders is more diverse. Notably, we only had one participant who did not identify as non-Hispanic Black or non-Hispanic White. This was a Hispanic male ex-offender.

OPEN IN VIEWER

Table 1.

Participant Demographics.

Group	Mean Age	Percent Male	Percent Black
Police officers (N = 13)	44.5	100	30.1
Ex-offenders (N = 14)	43.4	64	64.2
Community (N = 8)	45	62.5	50

Analytic Strategy

To investigate our first research question, the locations of all cleared lots from the August 2015 spatial video collection were compared to all other spatial video collected for the neighborhood. “New” clearances were then mapped at each wave. A total of 23 houses on 18 streets were torn down. We created 25-meter block face buffers extending around the street segment on which the house sat. If the house was on a corner, both the facing and side segments were buffered together and counted as one. Figure 1 shows “treatment” buffers (streets where houses were razed during our study period) and crime as of April 2014.

OPEN IN VIEWER

Figure 1.

Buffers for analysis and calls for service hot spots in April 2014.

Using our spatial video data, we marked the first wave where a house was no longer observed as the beginning of the “posttest” period for that street. Pretest and posttest periods vary across the 18 streets depending on when a house was removed. ³ Thus, rather than crime being observed for all street segments at one point in time (e.g., April 2014) and then again after 17 months (e.g., September 2015), our approach takes into account the timing of demolitions. For example, Figure 2 shows three hypothetical houses observed over our six time points. House A is standing at each wave. Houses B and C are torn down at different points during our study. For house B, the pretest would consist of waves 1 to 3 and the posttest of waves 4 to 6. Alternatively, house C was removed between time points 4 and 5, meaning the pretest would consist of waves 1 to 4 and the posttest of waves 5 and 6.

OPEN IN VIEWER

Figure 2.

Observations of houses.

Of course, some houses in our study were torn down between the same waves. Specifically, two houses were torn down between waves 1 and 2 (spring 2014), two more between waves 2 and 3 (summer 2014), eight were torn down between waves 3 and 4 (fall 2014), another five were demolished between waves 4 and 5 (winter 2015), and one more between waves 5 and 6 (spring 2015). Essentially, this gives us five pre- and posttest periods across streets. We use these groupings to present changes graphically. First, we examine whether treatment buffers (i.e., streets where houses were torn down) accounted for a lower portion of total neighborhood crime after houses were razed. To be clear, this is an analysis of changes in distribution, not number of incidents. We use this approach because (1) our microgeographic lens is focused on whether crime moved away from these streets rather than if crime in general reduced and (2) analyzing overall levels would be more problematic considering differing times of exposure and seasonal changes across our study. For example, for a street where a house was torn down in fall 2014, we may observe that CFS reduced. However, this reduction could be due to the winter time period immediately following demolition.

We calculate the proportion of neighborhood crime on streets as shown by equation (1), where subscript i represents a group of street segments that had houses removed at the same time and j denotes the time period. The denominator represents total CFS for the neighborhood for that same time period. We expect that street segments will account for a lower portion of total crime in the neighborhood after the removal of abandoned homes.

∑ CFS i , j CFS j .

1

Rather than using all CFS in the equation above, we only calculate the portion of serious crime calls including property, violent, and drug crimes. In other words, we are interested in the proportion of crime-related calls originating from these streets rather than calls for more minor matters such as stray animals, suspicious persons, and traffic violations. In addition, we do not include calls related to more “social disorder” types of items including calls for drunk and disorderly and loitering (Sampson and Raudenbush 1999).

Our second phase of quantitative analysis is limited to the treatment buffers themselves. For this analysis, we compute changes in the proportion of total CFS on each street that related to serious crime (using abovementioned crime types). For example, for house B in Figure 2, we would compare the proportion of CFS related to serious crime from that street during time points 1 to 3 to the proportion of CFS related to serious crime from that street during time points 4 to 6. Further, we present these results for violent and property crime separately since we anticipate greater changes for property relative to violent crimes after houses are removed. For example, equation (2) shows how this would be calculated for the proportion of calls related to violent crime.

∑ Violent CFS i , j All CFS i , j .

2

As before, we sum CFS across streets that had the same cut point for pre- and posttest periods. However, there is interesting variation within these groups that we explore these as well. We also draw on our spatial video and geonarrative data to better understand some of the patterns we observe in our quantitative data.

Notably, our quantitative analytic strategy presents both advantages and limitations. ⁴ Houses were not randomly selected for demolition, and we do not use a random sample of houses, limiting our ability to make causal inferences or to generalize results. Houses torn down were condemned properties which had already been ordered demolished prior to Moving Ohio Forward. They were condemned through a variety of means—including neighbor complaints, police reports, or housing inspections. Once a house was condemned, it was reviewed by the housing appeals board of the city and typically recommended for demolition. The Moving Ohio Forward Program first targeted houses that had been condemned for longer periods of time. However, a house may be moved to the front of the line if it presented a safety or health threat to the community, such as in the case of a fire. As noted, we also observe the total number of houses that were razed in our study area. This raises questions about the use of inferential statistics to interpret our findings, since we are not working with a sample, let alone a random sample. Therefore, all of our findings should be interpreted as descriptive and not generalizable.

To investigate our second research question, we analyze geonarrative data. Narratives were recorded using an audio digital recorder and then transcribed using Express Scribe 7.01 software. Coding and ride-alongs were an iterative process. It became apparent early in the study that abandoned houses were an area of concern for many of our participants. Therefore, we started asking participants who raised the issue of abandoned housing follow-up questions such as “What do you think about abandoned houses?” and “Why do you think they are a problem?” After all rides and transcriptions were completed, each transcription was read in its entirety to identify excerpts about abandoned housing. We also used open-coding to identify emerging themes within this category related to why vacant houses were perceived as problematic. Word searches in NVivo 10 were carried out for “abandoned” and “vacant” as well to be certain no references were overlooked. Abandoned houses were identified as problematic by all but two participants who simply did not comment on abandoned houses. On average, the amount of text related to vacant housing across interviews was 14 percent. Other than the general issue of drugs in the neighborhood, abandoned housing was the most consistent concern raised by our participants.

Housing Demolitions and Crime Change

We first investigate whether there were changes in the percentage of neighborhood crimes originating from street segments where houses were torn down (“treatment” buffers). Figure 3 shows changes across four groups of street segments depending on the timing of housing demolitions. The vertical line cutting through each trend line marks the period when houses were torn down. Notably, we do not show changes for the street where an abandoned house was removed during summer 2015 because this street yielded zero CFS in the three months before and after the house was removed. In total, the street produced nine calls for serious crime between April 2014 and April 2015, but none thereafter during our study period.

OPEN IN VIEWER

Figure 3.

Changes in the percentage of neighborhood calls from “treatment” buffers. Vertical cut point represents when abandoned houses were torn down.

As shown in Figure 3, each group of streets experienced a reduction in the proportion of neighborhood calls for serious crime. The dotted lines across graphs indicate the average proportions for pretest and posttest periods, showing that each group experienced a reduction, although notably these changes are rather small. Taken together, the average change across all streets in the proportion of neighborhood calls was 0.5 percent. In other words, after houses were torn down, each street accounted for 0.5 percent less neighborhood crime, on average, than when the houses were standing.

There is also interesting variation across individual streets. Taken individually, two thirds of streets experienced a decrease in the percentage of neighborhood calls originating from them after a house was demolished and the remaining one third experienced slight increases. The largest reductions across streets were by 3.4 and 2.7 percent, respectively. In the latter case, the reduction was observable across all types of crime observed (violent, drugs, and property). In the former case, the reduction is attributable to violent crime specifically, which changed from 8.9 to 1.8 percent. This street had a house torn down relatively early in the study compared to other cases (by July 2014), meaning the pretest observation was much shorter than the posttest. Consequently, the denominator is understandably smaller in the pretest than in the posttest (225 vs. 804). This raises some concern, since a smaller denominator would give more weight to crime occurring during that period. Looking more in-depth at this locale, there were 13 CFS during the 3-month pretest (mostly violent) and 19 calls in the 15 months afterward (mostly property). Indeed, in the three months prior to the house being removed there were eight calls for violent crime, and only five calls after the house was removed during the following 15 months. Therefore, regardless of the shifting denominators, this seems to be a notable change in criminal activity on this street.

We draw on our spatial video and geonarrative data to better understand the context of this street. First, we observe that 31 percent of lots on this street were vacant and 15 percent of houses were abandoned at the start of our study. In essence, this equates to 12 vacant lots, 4 abandoned houses, and 23 occupied dwellings. While the percentage of housing that is abandoned is equal to the average across our study buffers, the percentage of lots that are vacant is actually lower than the average (48 percent). This raises an interesting possibility. One reason a street may produce relatively more calls than others could relate to the percentage or presence of occupied housing. Similar to the old adage, “if a tree falls but no one is around to hear it…,” if no one is present to observe a crime, a crime may not be reported. In addition, a unique feature of this street is that it is the only one in our sample with a community garden. Thus, perhaps there were rejuvenation efforts on this street that contributed to the positive changes observed. For example, one community member remarked that he believed the community gardens in the neighborhood give “a little bit of a sense of connection and community” and an ex-offender commented that he believed the gardens send a “positive message, like everybody in the community, can get out and help and garden, and that will feed the community.”

We call attention to two streets which experienced slight increases as well. Two adjacent buffers accounted for 2.8 and 5 percent of neighborhood crime before abandoned housing was razed. Afterward, they experienced increases of 0.4 and 0.2 percent. To be sure, these are relatively small changes, but we look more in-depth at these streets because they did not experience the anticipated decrease. In examining our spatial and narrative data, we observe that these streets are located across from an affordable housing apartment complex that was perceived by participants to be problematic. As one officer relayed, “We get some drug activity running in and out of there. We get a lot of fights, a lot of alcohol infused fights in this particular complex.” There is also a convenience store on the connecting street which draws a lot of foot traffic. One possibility is that the removal of vacant housing in this case may have facilitated travel to and from the store and may have provided more open space for residents to interact. ⁵ Figure 4 shows one of the abandoned houses before its removal and a zoomed-in map of this area. The abandoned house shown is located on the west side of the map directly across the street from the affordable housing complex.

OPEN IN VIEWER

Figure 4.

Microgeographic context of crime hot spot, 2014.

Moving to the second phase of our quantitative analysis, Figure 5 shows how the nature of calls within treatment buffers changed after the removal of abandoned houses. Among the calls originating from these streets, we investigate whether there was a shift in the types of calls. In general, we find that the percentage of calls related to serious crime decreased across groups after abandoned housing was razed. The average reduction across segments was 7 percent. In other words, among calls originating from these streets, there was an average reduction of 7 percent in the proportion of calls that related to serious crimes such as property, drug, and violent offenses.

OPEN IN VIEWER

Figure 5.

Changes in the percentage of calls from “treatment” buffers related to serious crime.

As with our first set of findings, there is some variation across buffers, with four experiencing increases in the proportion of calls related to serious crimes. However, the modal experience was a decrease, and we focus on one buffer in particular, which experienced the largest decrease (57 percent to 33 percent). As shown in Figure 6, this street has a liquor store on the one end and a car wash on the other. These were known crime generators or attractors in the area according to our participants, who commonly remarked on the problem of men loitering outside of the liquor store. This car wash was also suspected by police to be a front for drug activity. One police officer explains, “We would see a lot of people coming and going from there. But when you have a business, especially a car wash where that’s kind of what you see, it’s hard to get a handle of what was going on there.” A resident confirmed to us that this car wash was in fact a hub for drug sales, stating, “People with legitimate [reasons] coming in and out kind of is a smoke screen for whatever else that goes on behind the scenes, and there’s some drug dealing because the people who run those little independent car washes like that, they can make more money selling drugs.”

OPEN IN VIEWER

Figure 6.

Street with largest reduction in percentage of serious calls, 2014.

Similar to the ideas of Spelman (1993), who noted that abandoned houses may be used as staging areas for other crimes, we argue that in order to understand the influence of abandoned housing, scholars should consider the surrounding environment as well (see also Branas et al. 2012). Indeed, the presence of abandoned housing alone may not be as criminogenic as abandoned housing located nearby other attractors and generators such as this liquor store and car wash. For example, people may buy drugs at the liquor store or the car wash and then use the drugs in nearby abandoned housing. By demolishing that house, associated crime may be reduced or pushed elsewhere. Another aspect of the environment we note in this figure is the sheer amount of vacant land. To be sure, Hope Gardens may be a unique context in which to study the relationship between housing demolitions and crime. A large portion of the neighborhood is empty—either represented by vacant housing or lots—which may affect crime patterns we observe.

Finally, Figure 7 illustrates changes in the proportion of calls related to property and violent crime specifically. When observing these crimes separately, it does not provide a clear picture of whether reductions are more pronounced for violent or property crime, respectively. The proportion of calls related to violence decreased for streets where houses were torn down in spring 2014 or winter 2015. Alternatively, streets where houses were torn down in summer 2014 or fall 2014 experienced no discernible decline. Similarly, the results for property crime are inconsistent. We thus conclude that we do not observe any evidence of changes being crime-specific. While some of these patterns may be seasonal, our study is also not suited to investigate that possibility. ⁶

OPEN IN VIEWER

Figure 7.

Changes in the percentage of calls from “treatment” buffers related to serious crime, property crime, and violent crime.

Up to this point, we have provided a descriptive account of crime change following housing demolitions in Hope Gardens. On average, we find that the percentage of neighborhood crime attributable to each buffer where a house was torn down decreased. Additionally, we find that a lower percentage of calls originating from these streets pertained to serious crimes. In the following section, we turn more squarely to our narratives to explore the ways in which abandoned housing is connected to crime in Hope Gardens.

Insights from Ex-offenders, Police, and Residents

Before delving into the different criminogenic roles played by vacant housing in Hope Gardens, we highlight a seeming contradiction between our call data and interview data regarding drug crime. As suggested by prior theory and research, the presence of abandoned housing may be especially pertinent for drug activity since offenders may use the houses to use or deal drugs. However, there were no drug calls in 72 percent of our observations across streets, which mirrors the larger tendency in the neighborhood. Due to the rarity of calls pertaining to drug dealing or drug use, we were not able to assess changes over time in these crimes, but our geonarratives suggest that drugs are perceived as a pervasive issue in Hope Gardens and that abandoned housing plays an instrumental role in drug activity. We use our geonarratives to help to contextualize this contradiction between our data sources.

In short, participants commented that residents are reluctant to call the police sometimes. As one community member told us, he would only call the police if “it’s something real, real bad,” and another resident said he would call the police in the case of a missing child or “if I see somebody being beat up.” Reasons related to impressions that response time was slow or that the police “are just not fair.” One ex-offender noted that he had called the police a couple of times in the past when witnessing violent fights but stopped calling because “you call the police—by the time they come, they done beat the brakes off the dude.” In fact, one former burglar told us that he did not worry about alarms for this reason. As he stated, “That’s why we don’t care about alarms. We feel like, you know, police don’t care.” In this ex-offender’s case, he actually hid nearby and watched how long it took police to arrive after he burglarized a house with an alarm. Seeing that more than 30 minutes lapsed before police were on the scene, he chose to burglarize that house again and to spend more time inside gathering items. In short, perceiving that police response is inadequate may lead residents to tolerate certain types of crime and report more serious crimes, like violence.

Aside from drugs, the issue of abandoned housing was the most frequently raised concern among participants. In fact, these issues were often raised in tandem, since abandoned houses were frequently identified as former drug houses. For example, the following excerpt with a police officer shows a typical reaction as we would drive down a street with abandoned housing:

Yes, here, here, here (pointing left and right) all used to be drug houses. Yes, I mean all this vacant stuff all used to be drug houses. You know, unfortunately like this house back here…oh not this one, next one. That was a crack house. Right there, we did a search warrant on that [house]. Right here all the way back there. Um 10–12 years ago and it’s just like it is after we left it. I mean we kicked everybody out and they haven’t done anything with it.

Discussion

In this study, we provide a descriptive exploration of a housing demolition program in a high-crime neighborhood in Ohio. We find that streets where abandoned houses were removed accounted for a lower portion of neighborhood crime after removal and that the nature of calls changed, with a lower portion of calls pertaining to serious crime. Our findings are generally supportive of prior work finding a link between abandoned housing and crime; however, our results should be interpreted as descriptive rather than as evidence of this particular demolition program being an effective crime control strategy. We also examined whether changes were specific to certain types of crimes, but we do not find evidence supporting this possibility. Additionally, we draw attention to the variation we observed across streets where abandoned housing was razed. In particular, a handful of streets became more crime-saturated after a house was razed or experienced slight increases in the proportion of calls related to serious crime. Looking more in-depth at the microgeographic contexts of our buffer areas, we surmise that changes spurred by housing demolitions may be contingent on the surrounding environment. For example, the proximity of other crime attractors may be relevant, as may the percentage of occupied residences or the proximity of potentially protective places such as churches and community gardens. Future research should continue to drill down into these spaces, examining how interactions between places and the larger neighborhood context shape crime patterns.

Our narratives also help to contextualize our quantitative findings. For example, taking CFS data at face value we may have concluded that drug crimes are relatively rare in Hope Gardens. However, drugs were the most common issue discussed by participants. The rarity of calls may be due to residents’ reluctance to call the police in certain situations. Our participants reported that slow response times dissuaded them from calling the police unless crimes were perceived as “serious,” such as in the case of violence. Investigating the nature of call data is an interesting avenue for future research. While calls are often considered the most valid and unfiltered source of crime data across geographic areas (Warner and Pierce 1993), they still suffer from measurement error. Future research should further explore the factors which influence crime reporting with more of a lens on context rather than individual characteristics.

Our narratives also reveal important insights into how abandoned housing is connected to crime in Hope Gardens. Specifically, three main themes emerged across participants, which were that abandoned houses are connected to crime as hiding spots, available spaces, and as targets. Notably, our qualitative results align more with crime pattern theory (Brantingham and Brantingham 1993) and routine activities theory (Cohen and Felson 1979) than with broken windows theory (Wilson and Kelling 1982). Our participants explained that abandoned houses are problematic for crime because they represent tangible and direct opportunities than because they signal neighborhood decline. Similar to qualitative work on burglary decision-making (Rengert and Wasilchick 2000; Wright and Decker 1994), participants noted that abandoned houses were attractive because of factors related to opportunity; they are easy targets, relatively accessible, lack guardianship, and provide concealment. Perceived criminal opportunities also ranged across many offense types, including prostitution, robbery, burglary, drug use, sexual assault, and even homicide (as described by one of our police officers, they are a “free for all”).

Our work also contributes to qualitative work on offender decision-making. As noted by Brantingham (2013), “remarkably few criminologists have actually asked offenders how they find targets” (p. 548). Indeed, a few of our ex-offender participants reported being strategic in their use or selection of abandoned houses—picking ones that met a specific set of criteria (e.g., isolation and foot traffic). This comports with prior work showing that some offenders engage in rational and deliberate decision-making processes in the commission of their crimes (Deslauriers-Varin and Beauregard 2010; Jacobs and Miller 1998). For example, Wright and Decker (1994) find that many burglars in their sample choose targets based on the perceived worth of possessions inside the residence, but also weighted that possible gain against perceptions of risk, as signaled by factors such as alarm systems, guard dogs, and familiarity with the neighborhood. Importantly, studying the ways in which abandoned housing is used by offenders also offers unique insights since these structures are not as often selected because of their attractiveness as “targets” per se. Rather, offenders in our sample used these structures more often as instruments than as targets. Abandoned houses can be used as escape havens following a crime, as a space to conduct or plan criminal activity, or as a place to hide criminally obtained materials. Thus, future research should continue to explore offender decision-making with qualitative approaches that take into account target search as well as other ways offenders interact with their built environments to commit crime.

Also consistent with crime pattern theory and routine activities theory, abandoned houses were perceived as crime generators. While offenders may be attracted to abandoned houses, abandoned houses are also spaces where potential victims and motivated offenders are likely to converge in space and time and without guardianship. Thus, our results support the notion that abandoned houses are also crime generators (Brantingham and Brantingham 1993). Moreover, a concerning theme that emerged from our narratives was the tendency for potential victims to be youth who venture into the abandoned houses. Once inside an abandoned house, youths may stumble across homeless persons, active offenders, drug paraphernalia, condoms, or dead bodies. Thus, abandoned housing may carry implications for child well-being, increasing exposure to violence, infectious disease, accidental injury, or victimization.

Largely based on the tenets of routine activities theory and rational choice, situational crime prevention approaches postulate that the built environment influences “opportunity structures that lead motivated offenders to decide whether to commit a crime” (MacDonald 2015:341). As a result, altering the built environment should alter the associated opportunity structures for crime (Braga and Bond 2008; Clarke 1995). Indeed, our participants strongly leaned toward removing abandoned houses for precisely this reason. Interestingly, only two participants felt that the houses should be renovated and restored, an approach that would be more consistent with a broken windows perspective. Further, participants felt boarding up the houses was largely ineffective, contrary to some research finding positive results associated with this measure (Kondo et al. 2015). Additionally, situational crime prevention suggests that the effects of environmental alterations may be crime-specific (e.g., removing a payphone used by drug dealers may not affect theft). Despite theory and research suggesting that the removal of abandoned houses may be more consequential for property versus violent crime, we do not find evidence of that outcome. However, we caution again that our study is descriptive and that future studies should examine this possibility with rigorous designs—preferably a random control trial.

We conclude with a comment about the utility of drilling down into high-crime neighborhoods and using insiders to learn more about high-crime environments. Just as a top-down approach may be useful for testing the relationship between abandoned housing and crime, a bottom-up approach can be useful for understanding the factors driving any relationship (see Wright, Jacques, and Stein 2015). Our study suggests that there may be a myriad of dynamics—including spatial, physical, and social—shaping crime (as well as crime reporting) at the street segment and neighborhood levels in Hope Gardens. The role of abandoned housing is merely a small piece of these complex dynamics. Future research should continue to delve into high-crime neighborhoods to examine how offenders interact with the environment to commit crime and how the alteration of these environments may be consequential for crime. Understanding this complexity can offer key insights into theory as well as practical insights into crime reduction and prevention efforts.