Domestic violence is of growing public health concern, impacting the health and safety of up to 1 in 3 men and more than 1 in 3 women in the US (Smith et al., 2019). Domestic violence is a leading cause of severe physical injury and death, and can result in long-term chronic health consequences including depression, gastrointestinal disorders, reproductive disorders, heart disease, and post-traumatic stress disorder (Campbell, 2002). Domestic violence crime, which can take many forms including physical violence, sexual violence, psychological abuse, stalking, and more, represents 21% of all violent crime (Truman, 2014). Furthermore, domestic crime has been declining at a far slower rate than total violent crime over the past few decades (Catalano, 2012) and, in fact, rates of intimate partner homicide have increased in the US in recent years (Fridel and Fox, 2019). Preventive interventions for domestic violence with the potential for population-wide effects are needed. Despite the fact that domestic violence often occurs behind closed doors, theoretical and empirical evidence suggests that domestic violence is driven, in part, by environmental factors (Beyer et al., 2015; Pinchevsky and Wright, 2012; Voith, 2019).

Informed by social disorganization and other theories, prior research demonstrates that key environmental factors contribute to concentrated neighborhood disadvantage as well as to social norms that promote or condone violence and aggression, such as open air drug markets, alcohol availability, and littering (Branas et al., 2018; Cunradi et al., 2014; Keizer et al., 2008; Pinchevsky and Wright, 2012; Voith, 2019). Studies have suggested a link between neighborhood socioeconomic status and domestic violence (Beyer et al., 2015; Coulton et al., 2007; Cunradi et al., 2000; Grisso et al., 1999; Li et al., 2010; Nadan et al., 2015), including intimate partner violence (IPV) and child maltreatment, and have also revealed protective environmental factors in the neighborhood such as collective efficacy (Browning, 2002). IPV is associated with neighborhood conditions such as unemployment (O’Campo et al., 1995), deprivation/disadvantage (DeJong et al., 2011; Li et al., 2010), and poverty (Cunradi et al., 2000; Pearlman et al., 2003). Neighborhood disorder may weaken ties among residents (Kim, 2010; Packard et al., 2013; Shaw and McKay, 1942), potentially leaving survivors more vulnerable to violence from their partners (Stets, 1991; Wright and Skubak Tillyer, 2017) or leading to less intervention on their behalf by neighbors (Browning, 2002). It may also intensify stress among couples (Ross and Mirowsky, 2009), which may increase the likelihood of violence within the partnership (Wright and Benson, 2011). Socially disordered neighborhoods, including those with poor housing quality or vacant housing, have also been linked to higher rates of child injuries due to accidents (Reading et al., 2005), and higher rates of child maltreatment (Coulton et al., 1999; Freisthler, 2004; Freisthler et al., 2005). With respect to child maltreatment, parents living in disordered or disadvantaged neighborhoods may have less opportunities to develop social networks and work together to enforce shared norms (Sampson et al., 1999).

In recent years, there has been increased recognition of the widespread public health benefits of place-based violence interventions (Hohl et al., 2019). Place-based interventions move beyond individual-level drivers of violence and crime to address the underlying environmental and social contexts that foster violence. One class of place-based interventions are those focused on urban blighted property remediation. Blighted properties are abandoned structures or vacant lots that result from systematic disinvestment in specific neighborhoods, often reflecting historical and ongoing racist and discriminatory practices such as redlining and other racialized patterns of structural neighborhood disinvestment (Bieretz and Schilling, 2019). The presence of these properties results from disinvestment and contributes to ongoing cycles of disinvestment wherein a lack of future investment is justified by the presence of blighted properties (Bieretz and Schilling, 2019). Additionally, research shows that a high density of blighted properties is associated with increased neighborhood crime (Curtis and Mills, 2011). Blighted property remediation interventions are interventions that aim to interrupt such cycles of disinvestment and prevent crime through a variety of strategies such as property maintenance and rehabilitation, demolition and greening, and/or partnerships with government land banks to increase local ownership of rehabilitated properties (Carpenter et al., 2015).

There is an expanding body of evidence demonstrating that blighted property remediation interventions can reduce neighborhood crime. Observational studies have been complemented by recent quasi-experimental studies and randomized controlled trials in urban areas across the United States that have demonstrated consistent evidence of the causal effects of blighted property remediation on crime (Branas et al., 2011, 2016, 2018; Garvin et al., 2013; Kondo et al., 2018a; Moyer et al., 2018). Blighted property remediation is hypothesized to reduce crime and contribute to improved public health through a number of different social, economic, and biological pathways, including strengthening of social cohesion and social ties; reducing strain on local economies, property values, and social services systems; and removing place-based chronic stressors like environmental disorder (Hohl et al., 2019). Notably, the majority of research concerning the role of blighted property remediation in reducing crime has been focused on firearm violence, drug crime, and property crime. To date, the evidence concerning the impacts of neighborhood intervention on domestic crime is inconclusive (Hohl et al., 2019; Kondo et al., 2018a).

While blighted properties are strongly correlated with socioeconomic conditions as well as neighborhood disorder and collective efficacy, studies examining their relationship with domestic crime are limited. Blighted properties may have an impact on domestic crime, such as child maltreatment risk (Freisthler et al., 2006) and IPV (Browning, 2002), potentially through social cohesion and control. Neighborhoods with fewer boarded up and abandoned buildings have been correlated with lower child maltreatment rates (McDonell and Skosireva, 2009). Studies linking child maltreatment with a diminished social ecology, including a scarcity of neighborhood resources, run-down housing, and parents’ own perceptions of their neighborhood being a poor environment for raising children, are not new (Garbarino and Sherman, 1980), but such studies are rare. In this way, the link between blighted properties in a neighborhood and domestic violence, particularly IPV, is inconsistent (Waller et al., 2012) and the studies that have been conducted to date are cross-sectional and unable to disentangle the interdependence between the emergence of the blighted properties and abandonment and the occurrence of domestic violence. It is very possible that an increase in neighborhood domestic violence may have led to a cascade of family disruption, relocation, and eventual property abandonment and blight, the opposite temporal direction of what prior cross-sectional studies have assumed. New research that better measures and accounts for the temporal direction of these associations is needed.

One of the more commonly studied environmental factors related to domestic violence is alcohol availability. Alcohol consumption is a strong predictor of domestic violence and aggression at the individual level (Graham et al., 2011; McKinney et al., 2010). Substance use, including alcohol, is strongly associated with more physically severe incidents of IPV and is highly predictive of the likelihood that incidents will be reported to law enforcement (Novisky and Peralta, 2015). Emerging research assessing the neighborhood-level drivers of domestic crime has documented associations between neighborhood alcohol outlet density and domestic crime (Cunradi et al., 2014), with much of the existing research focused on IPV. Specifically, higher alcohol outlet density is associated with increased IPV-related police phone calls, IPV-related crime reports, IPV-related emergency department visits, and self-reported IPV perpetration (Cunradi et al., 2012, 2011; Waller et al., 2013). Of note, studies demonstrated distinct effects by type of alcohol outlet: namely, off-premise outlets and bars demonstrated more consistent relationships with domestic crime as compared with other on-premise outlets, such as restaurants and hotels (Cunradi et al., 2012, 2011; Livingston, 2010; McKinney et al., 2009; Roman and Reid, 2012; Waller et al., 2013). The relationship between off-premise alcohol outlet density and domestic crime has been shown to peak during weekends and lessen during weekdays, suggesting that the impact of office-premise alcohol availability may depend on patterns of social engagement and routine activity (Roman and Reid, 2012). Together, these associations are consistent with the hypothesis that alcohol availability contributes to increased alcohol consumption and increased alcohol-related harm (Babor et al., 2010).

Further, there is evidence that supports associations between neighborhood physical disorder, neighborhood income, population size, and alcohol outlet density. Evidence suggests that the relationship between neighborhood physical disorder (e.g., blighted properties, evidence of vandalism) and alcohol availability is complex and likely cyclical, both reflecting consequences of concentrated disadvantage and disinvestment (Milam et al., 2014, 2013). Alcohol outlets are more likely to be concentrated in lower income areas, despite lower alcohol consumption in these areas (Morrison et al., 2015). Research suggests that higher income areas have increased ability to exclude alcohol outlets through higher land and structure rents and through resistance from residents, compared to lower income areas (Morrison et al., 2015). Meanwhile, population size has a positive association with alcohol outlet density: higher population size is associated with higher alcohol outlet density, likely representing a response to market demands (Jin et al., 2018; Morrison et al., 2015). Notably, bar density is linked to market demands over a larger geographic area compared with restaurants and off-premise outlets, reflecting potential differences in alcohol outlet distribution by outlet type (Morrison et al., 2015). These varying relationships between alcohol outlet density and neighborhood income and population size suggest differences in alcohol availability in relation to the presence of blighted property.

The present study builds upon the extant literature through the use of a quasi-experimental study design with clear before/after temporal measures of dependent and independent variables to answer an understudied question: the association between a place-based blighted property remediation and domestic violence. The primary theoretical framework that guides this analysis is an application of the social ecological model toward the study of domestic violence (Whitaker et al., 2009), which recognizes that while the incidence of domestic violence is shaped by an individual’s behavior, it is also influenced by a number of complex, interacting social and environmental factors that can be intervened upon at the macro-level toward the prevention of domestic violence.

The aim of this study was to assess the impact of a blighted property remediation and vacant lot abatement program on the incidence of domestic crime. Given the importance of alcohol availability as a determinant of domestic crime, and the potential for alcohol outlets to be disproportionately located in lower income areas often with greater concentrations of blighted properties (Jin et al., 2018), we also assessed possible interactions between this place-based intervention and alcohol outlet density. Using data from a blighted property abatement initiative of vacant land in New Orleans following Hurricane Katrina, we employed a difference-in-differences analysis to examine the program’s impact on domestic crime and a triple-difference analysis to determine if alcohol outlet density modifies the relationship between the intervention and domestic crime.

This study focuses on a blighted property remediation intervention called New Orleans Chapter 66 Vacant Lot Maintenance Program, or Fight the Blight, the details of which are described elsewhere (Kondo et al., 2018b). In brief, the city of New Orleans launched the program in 2014 to remediate vacant lots across the city, many of which had become blighted as a result of the damage caused by Hurricane Katrina in 2005. Lot remediation consisted of inspection, notice, removal of debris, and mowing of vegetation. All vegetation in these lots was mowed at least once, and in some cases multiple times, with a frequency of no more than every 3 weeks. Mowing occurred with more frequency during summer months. Remediation and routine maintenance were funded by the City on behalf of property owners, with the costs of services recorded on the owner’s tax bill. From October 2014 to July 2016, 1,614 lots were remediated as part of the Fight the Blight program (Kondo et al., 2018b).

Of the 1,614 remediated lots, we selected 204 lots that received removal of debris and mowing of vegetation under the Flight the Blight program between October 2014 and July 2016. We selected only the 204 that had violated the property ordinance for weed and plant overgrowth or for sanitation (most received citation for both) prior to remediation. Of these 204 lots, 64 lots received one treatment (i.e., debris and vegetation removal), and the remaining 140 lots received two or more treatments. These 204 remediated lots were then randomly matched in a 1:3 ratio with 612 control lots that received citations for weed and plant overgrowth or sanitation and were eligible for the Flight the Blight program but did not receive the intervention. Matched control lots were selected from the same neighborhood statistical area as their remediated counterparts but were over 250 feet away (the average length of a city block) to minimize spillover. The matching process is also described in more detail elsewhere (Kondo et al., 2018b). The matched control lots were at most 1,581 feet away from each remediated lot and the average distance between a remediated lot and control lot was 476 feet.

The unit of analysis was the lot-month over a 7-year study period (2011–2017), resulting in a total of 68,544 lot-months for analysis.

To assess whether the blighted property abatement intervention impacted rates of domestic crime, we conducted a quasi-experimental difference-in-differences analysis (Model 1) using Tobit regression. Tobit models were used to account for zero truncation at the lower bound (Greene, 2003; Grogan-Kaylor and Otis, 2003; Tobin, 1958). Model 1 served as a test for consistency with previously published findings using these data, which demonstrated no significant effect of the intervention on domestic crime (Kondo et al., 2018b). In the present analysis, compared with the prior analysis, we additionally adjusted for population density and entertainment venue density, and employed a longer study period (2011–2017 vs. 2014–2016). Model 1 was specified as in Equation 1.

To assess whether alcohol outlet density modified the effectiveness of the blighted property abatement intervention, we conducted a quasi-experimental triple-difference analysis, also using Tobit regression models. In these models, the independent variable of interest was the triple-difference term indicating, for each lot-month: 1) whether the lot was a remediated lot or a control lot; 2) whether the month was before or after the intervention date for remediated lots and their matched control lots; and 3) the density of alcohol outlets (per square mile) at the centroid of each lot. The first of these models included the density of all alcohol outlets (Model 2), the second included bars only (Model 3), and the final model including off-premise alcohol outlets only (Model 4). These models were specified as in Equation 2. We did not include both bar density and off-premise alcohol outlet density in the same model due to their collinearity.

This triple-difference analytic approach is consistent with prior literature in epidemiology and econometrics (Cornelissen and Sonderhof, 2009; Jaccard and Turrisi, 2003).

All regression models included all potential confounders described above. We performed all data management and analysis using R (version 3.6) and ArcMap (version 10.6).