Commentary on de Vocht et al. (2020): The value of local causal inference models for alcohol policy debates.

Abstract

de Vocht et al. [1] demonstrate novel methods that researchers may be used to counter two common complaints policymakers express within debates about alcohol outlet density: (1) that alcohol licensing restrictions are ‘anti-business’ and (2) that the harms arise not from alcohol outlets themselves, but rather from the bad neighborhoods that often surround them. There is a key disconnect in policy debate: despite a proliferation of alcohol outlet density research [2, 3], those in positions to use that research—alcohol policymakers, community coalitions, civil servants and so on—remain largely unaffected by it [4]. Oddly, from a research perspective, at the same time that the literature is getting closer to establishing a causal role for alcohol outlets in neighborhood harms, governments around the globe are instead rolling back regulations on them [5, 6]. At least in part, this disconnect may stem from alcohol outlet density researchers’ failure to address some of the most pressing concerns from those who are not swayed by the research. News media coverage often focuses on pocketbook aspects of policies, such as whether they threaten jobs or profit margins, and alcohol availability policies are no exception [7]. However, de Vocht et al. demonstrate methods that permit the identification of harmful neighborhood impacts of particular high-risk retailers. This type of evidence could readily feed into a framing of restrictions on alcohol availability as being ‘pro-community’ policies. Survey data have also demonstrated that customers patronize businesses less often if they fear being harmed by drinkers [8]. Merging such data with a study such as that of de Vocht et al.'s could, in turn, help to frame alcohol availability restrictions as pro-business in the long term. The neighborhood hypothesis suggests that the link between alcohol outlets and harms is not causal, but is instead explained by the type of neighborhood in which outlets are located. Although current data do not support this hypothesis as the most important determinant of findings between alcohol outlets and harms, explaining why this is the case has been challenging, especially when one considers that alcohol outlets often distribute along racial and economic lines [9]. The underlying story is complex; neighborhood contexts are probably one of many contributors or possibly effect measure modifiers of the association between outlets and harms [10, 11]. Further, multi-level models have partitioned variation between and within census tracts to identify particular outlet characteristics—such as significant externally visible retail signage—that are associated with harms over and above the neighborhood context [12]. These explanations are hard to grasp. de Vocht et al.'s key contribution is to use synthetic controls to approximate a randomized controlled trial that holds neighborhood context ‘constant’, thereby providing a more intuitive example of how the harms could arise from the outlets themselves. Methods such as those modeled by de Vocht et al. that focus at the outlet level could additionally help audiences to understand the significance of outlet density research, because phenomena become more recognizable at the local level. This is especially true when they have a name and an address. In the face of nuisance outlets, while researchers build local models they could also collaborate with community partners to photograph the types of behaviors and harms in and around the outlets over time. Such images, side-by-side with local causal inference models, could send a more convincing message. In contrast to de Vocht et al.'s work, interpreting the results of previous ecological studies involves a type of mental gymnastics that requires abstract reasoning and an appreciation of the ecological fallacy; these are higher-order skills that may not be common practice to many with a stake in alcohol outlet licensing decisions. Finally, de Vocht et al. demonstrate methods for how to estimate the public health benefits of closing high-risk alcohol outlets using locally available data at the micro level. When performed in a timely way to evaluate licensing decisions, this type of analysis could give policymakers and community members the confidence they need to know that revoking licenses from high-risk outlets will have a positive outcome for their neighborhoods and business environments. None.