Loss prevention is one of the most neglected areas in criminology; in fact, searching the most prestigious journal, Criminology, for “loss prevention” returns zero articles. Likewise, a search of Justice Quarterly returns four articles; the Journal of Criminal Justice returns four articles; and the Journal of Research on Crime and Delinquency returns just two articles—and these journals have been around for decades. This is not to say that high-quality loss prevention research does not exist; in fact, our colleagues at the Loss Prevention Research Council (LPRC) have published many articles, and there are many studies in criminology and other fields that are directly applicable to loss prevention. Nevertheless, the point remains that loss prevention is an immensely neglected aspect of research.
This lack of research is problematic because retail crimes are a serious threat to retailers and communities. Retailers need to know what does and does not work to reduce shrink and protect their assets while increasing sales. Furthermore, criminal justice policies have a tremendous impact on retailers, and we need to know how these policies affect theft and other forms of shrink. If we are going to make informed, evidence-based decisions, we must conduct rigorous studies using the appropriate data. Unfortunately, even though retailers generate a wealth of data on a daily basis, many decisions are still based on anecdotal evidence—that is, on things that “we all know.”
Rigorous, well-designed loss prevention studies would not only give criminologists new insights to the causes and nature of retail losses but also help retailers and the public make more informed decisions. For example, it is common to hear loss prevention leaders suggest that raising felony shoplifting thresholds is associated with increases in shoplifting. If anyone understands trends in retail theft throughout the industry, it is loss prevention leaders who examine these trends on a regular basis; however, existing research suggests that raising felony thresholds has little to no effect on crime. Of course, the primary problem with these studies is that they often rely on official statistics to determine whether policy changes result in a change in crime.
For example, a widely discussed study by Bartos and Kubrin (2018) used data from the Uniform Crime Reports and found that California’s enactment of Proposition 47 had no detectable effect on violent crime, and possibly a small increase in some property crimes, if any. This study has been cited by policy makers and think tanks to justify raising felony thresholds elsewhere; unfortunately, studies such as these cannot account for what criminologists refer to as the “dark figure” of crime. The dark figure of crime is the number of crimes that go unreported or undiscovered and, therefore, never show up in the official statistics. If most retail theft goes undetected and/or unreported, then studies that rely on official statistics cannot establish whether increasing thresholds are associated with changes in theft.
Of course, there are several reasons to believe retail theft is underreported. First, many retail thefts are not discovered until products are inventoried, and once a shortage is discovered, it is difficult (if not impossible) to determine whether this is due to theft or some other source of shrink. Second, one of the determinants of whether crime is reported is whether the victims believe something will be done about the crime. If individuals believe the criminal justice system will not treat the crime seriously, then victims may not bother reporting it to the police.
If raising the felony theft threshold leads to a reduction in crime reporting among retailers, then studies that rely on official crime reports may show an apparent decrease in retail crime even if it is, in fact, increasing. Studies that rely on official statistics may not find the “true” effect of these policy changes because the increase in crime is offset by a decrease in reporting. This issue reveals two important criminological questions: (1) is an increase in the felony threshold for theft associated with greater shrink, and (2) is an increase in the felony theft threshold associated with reductions in crime reporting by retailers? These types of questions can only be answered if retailers cooperate with researchers to conduct the necessary research. However, there are many more topics in loss prevention that require additional research, including the effectiveness of loss prevention solutions, the factors that promote effective collaboration among ORC investigators, and many other topics.
The Problems with “We All Know Thinking”
Of course, we all know that increasing felony thresholds is associated with increases in crime; however, there are several problems with “we all know” thinking in loss prevention, just as there are problems with “we all know” thinking in criminology. “We all know thinking” is the tendency for people to unquestionably stick with what they and others in their group think they know. First, this type of thinking is problematic because in many cases, we are lying to ourselves. It can be unpleasant to think about and admit, but there is a lot we do not know. Second, this type of thinking is problematic because it is contrary to innovative thinking—it does not require an individual to think critically or question what they think they know, because they “know” that they are right about what they know. Finally, this type of thinking is prone to cognitive biases; that is, defects in thought processes that are likely to produce flawed decisions. In fact, “we all know” thinking represents a cognitive bias known as bandwagon bias or the bandwagon effect. The bandwagon effect refers to the tendency of people to adopt the widely held attitudes and beliefs of others within their social groups.
For example, at one point, police all agreed that rapid response to calls for service, random patrol, and reactive investigations were, unquestionably, the best policing practices. This lasted until research suggested that: (1) in most cases, victims waited too long to call police, so rapidly responding to these calls would not solve crimes; (2) random patrol did not seem to reduce crime in well-designed studies; and (3) only a small subset of cases could be solved via investigations, and those that could be solved were largely solved because of victim cooperation and evidence collection at the scene of the crime, not anything done in a crime lab. All these studies challenged policing and led to more innovative, proactive policing strategies.
Establishing Causal Relationships in Loss Prevention Science
The discussion of the relationship between felony thresholds and crime reveals the complexity of establishing causal relationships. It is important to establish whether relationships are causal; otherwise, we might come to believe that a supposed cause is more important than it actually is. For example, did you know that increases in US spending on science, space, and technology is associated with increased suicides by hanging, strangulation, and suffocation from 1999 through 2009? Based on this information, we might decide that we should reduce spending on science and technology. Of course, it is more likely that a third factor is related to both of these and that there is no causal relationship between US science and technology and suicide. This highlights the importance of establishing causal relationships—if we believe a cause produces an effect, we will make decisions based on that belief.
There are at least three criteria that must be satisfied to establish a causal relationship between a presumed cause (such as the adoption of a product protection solution) and a presumed effect (such as a change in crime or shrink). First, there must be a systematic relationship between the presumed “cause” and the “effect”; in other words, when one event occurs the other event must consistently co-occur, which is called correlation. Second, the “cause” must always precede the “effect” in time; there cannot be times when the “effect” precedes the “cause” (temporal precedence). Finally, all other explanations for the relationship must be ruled out (isolation of cause). The studies on the relationship between felony thresholds satisfy the first two criteria; however, they cannot satisfy the third because they cannot determine the extent to which the policy changes affect changes in reporting.
Different types of studies and different types of evidence are better equipped to establish causal relationships. While speaking with loss prevention professionals, it has become clear that most loss prevention decisions are made based on anecdotal evidence, at worst, or on the basis of results from a descriptive study, at best. Furthermore, even when retailers conduct an internal study, they are often rushed and are not conducted with the rigor necessary to provide the best evidence. Of course, this is the nature of our fast-paced industry—we need solutions that reduce shrink and improve sales, and we need them fast.
The Hierarchy of Evidence in Loss Prevention Research
The hierarchy of evidence in loss prevention research, which has been adapted from evidence-based nursing, shows the different types of studies and the “quality” of evidence they tend to provide. At the top of the hierarchy are systematic reviews and meta-analyses of well-conducted research, while anecdotal evidence and expert opinions are at the bottom. Types of studies near the peak of the hierarchy tend to satisfy all three criteria for establishing a causal relationship (correlation, temporal precedence, and isolation of cause) when they are well designed and well conducted, while types of evidence at the bottom cannot even begin to establish a causal relationship. However, the types of research at the bottom are an incredibly important foundation upon which more rigorous studies can be built.
To return to the example of locked cases and cabinets, anecdotal evidence (level 7) and research that describes the problem (levels 5 and 6) suggest that locked cases are associated with a reduction in sales. This is presumably because they introduce friction into the shopping experience. However, unless we conduct a well-designed study, we will never know whether this is true because there are many other possible explanations. For example, it is possible that retailers installed these cases and cabinets in locations that were experiencing a simultaneous increase in crime and local economic decline. This would produce a combination of reduced sales and increased theft that would have encouraged retailers to install cases to protect these stores’ diminishing profitability. Later, when management examines the sales reports, they may find that sales decreased after the cases were installed. This establishes that there is a correlation between cases and reductions in sales, and establishing a correlation is a key feature of case control and cohort studies. However, to ensure the trend started after the locking cases were installed (and not before), retailers must examine the trends in theft before and after the installation. Examining the trends before and after installing the cases will allow managers to ensure there is a correlation and that the cause preceded the effect—two of the basic criteria for establishing causal relationships. Examining the “before and after” effects of a solution is a key aspect of many quasi-experimental studies (level 4).
Of course, this is preposterous because “it is accepted wisdom, and we know for certain” that cabinets and cases reduce sales specifically because of shopper friction. This is a joke in “scare quotes” because we cannot be sure our explanation is correct. Cases and cabinets may not affect sales because they slow down the shopping experience or reduce impulse sales, but rather they may just drive customers away because they signal that the store is crime prone. In this case, removing thousands of dollars’ worth of cases may not be the right, cost-effective solution; instead, stores may need to take other actions to buffer the negative effect of the cases on customer perceptions by making the store locations safer, or at least making them seem to be safer.
If we want to produce the best knowledge regarding the effect of locking products in cases on shrink and sales, we should conduct randomized controlled trials (level 2 in the hierarchy). In this type of study, several stores would be randomly assigned to a control and treatment group, and glass cases would be installed in the treatment stores to determine whether these stores experienced reductions in sales, theft, and customer traffic, and whether installing the cases was associated with increased fear of crime. All of these factors (sales, shrink, customer traffic, and fear) would need to be measured before and after the installation of the cases in order to establish a correlation between cases and these outcomes, and to ensure that cause led to the effect. However, random assignment of stores allows us to rule out other explanations because it ensures that any differences between the groups is due to random chance only; in other words, the control group and the treatment stores should be very similar with regard to other characteristics. This is important because it means there cannot be any other factor that explains differences in outcomes between the stores. This study would enable us to understand the effects on sales and shrink and would provide evidence of whether this was due to decreased store traffic and/or increased fear of crime at the store.
Finally, at the peak of the hierarchy are systematic reviews and meta-analyses of well-designed and well-conducted studies. Systematic reviews are highly structured summaries of research on a topic, while meta-analyses provide a quantitative summary of the effect of one or more phenomena on others. For example, there are meta-analyses of the effects of drugs across many studies; these provide a summary of the average effect of a drug. However, to conduct meta-analyses and systematic reviews, there must be a large amount of research on a topic. Unfortunately, we have not conducted enough well-designed and well-conducted studies to even begin to conduct a meta-analysis. In other words, we don’t know enough.
Improving the Science and Practice of Loss Prevention Research
The industry can do many things to promote evidence-based research. The first step is not to be too ambitious—there are many things that we do not know that we do not know because we have not conducted basic research. For example, qualitative and observational studies need to be conducted on topics such as: (1) what works to encourage compliance with mask policies, (2) what is the safest and most efficient way to conduct curbside delivery, and (3) what factors contribute to the successful prosecution of ORC offenders. Heck, we don’t even have a good, industry-wide definition of organized retail crime. Addressing these types of issues and answering these types of questions are the foundational first step to promoting evidence-based practice.
Of course, making loss prevention evidence-based will require that retailers invest in research, but there are many opportunities for retailers to do this. First, there are many young (and old) researchers at universities who must produce research to advance their careers. Many of these scholars need data and would love to conduct surveys with your customers or access your existing data. Furthermore, they are accustomed to doing research with humans and protecting their privacy and confidentiality. Of course, all of this is also true of organizations like the LPRC that work with retailers and LP solution providers every week to conduct impactful, relevant research. A large amount of research needs to take place, and it will take time to develop loss prevention into a fully fledged science with a large body of research findings. Nevertheless, if we work together as an industry, then one day we will be able to honestly and confidently say what we all know and what we do not know.
