Abstract Although crime prevention tactics are designed to reduce offending, some studies have revealed instances where reported crime actually increases after introducing lower intensity interventions. An analogous trend—characterized by low-dose stimulation and high-dose inhibition—called hormesis has already been observed in the natural sciences. We argue that this phenomenon is theoretically applicable to crime prevention. Findings suggest that researchers should test varying intensities of interventions to avoid rejecting ones that would be otherwise effective at higher levels. Research using dose–response techniques and simulation models should be explored to determine whether a weak intervention backfire effect occurred or is possible. Knowledge of such information could lead to more effective crime prevention strategies and better specified analytic models for evaluation. Introduction Although many crime prevention programs have been tested in a variety of contexts, most investigate the effectiveness of a single strategy without considering the varying intensities that could be applied (see Ekblom and Pease 1995; Bowers et al. 2004). In other fields, such as the medical sciences, dose–response analyses are commonly used to establish optimal health outcomes by examining how incremental increases in drugs/treatments impact test subjects (Tallarida and Jacob 1979). Despite the successes of such trials in other disciplines, and the more recent recommendations for their use in criminological research (Farrington and Welsh 2002), few studies have investigated these dose–response relationships and their potential use for policy (for exceptions, see Bowers et al. 2004; 2005). Knowledge of such dosage levels can inform two important policy concerns, namely what will have the greatest impact in reducing crime and what is the most cost-effective strategy to apply (see Bowers et al. 2005; Welsh et al. 2015). Despite these potential benefits, such research can also be challenging and come at a price. Depending on the study, the evaluation of various doses can be costly, difficult to measure and sometimes even unethical (see Ekblom and Pease 1995; Eck and Liu 2008; Haberman 2016). These may be some of the primary reasons why we do not see many studies exploring dose–response relationships. However, this can cause a problem in our understanding of crime prevention strategies, namely through the elevation of type II errors. In fact, there is a strong possibility that certain failed crime prevention programs were simply implemented at an incorrect level of intensity. In analogous terms, while a 50 mg dose of Tylenol may not be sufficient to curb a headache, the drug would not be considered ineffective. It was simply administered at an insufficient dosage (a regular strength capsule contains 325 mg). That said, we also believe that certain dosages of an otherwise effective strategy might actually exacerbate a crime problem. Here, we are not referring to interventions that fail at any level of intensity. Instead, we focus on interventions that work but were administered at the wrong strength when implemented. We have found a number of empirical studies and developed some theoretically informed scenarios that demonstrate instances where crime prevention interventions actually increase crime at low dosages but appear to generate desired results at higher ones. Thus, the intent of this paper is to address this issue and propose a new criminological concept: weak intervention backfire. Although this idea may be new to the crime prevention literature, it is not novel in other fields such as biology, toxicology or medicine (Mattson and Calabrese 2009). In fact, for over 70 years, biologists have been studying a process called hormesis whereby exposure to toxic substances at low dosages can actually stimulate resistant responses as opposed to the eradication of an organism’s ability to function at higher doses (Zhang et al. 2009). Calabrese and Baldwin (2003) have since argued that hormesis appears to be generalizable because of the hormetic-like relationships that have been discovered in an array of research disciplines (e.g. ecology, toxicology, experimental psychology and epidemiology). However, because the relationship has been studied independently across so many fields, no one had previously consolidated the findings and attributed them to a single phenomenon. There is a long tradition of applying ideas from the natural sciences to criminology. Examples range from Shaw and McKay’s (1942/1969) use of Burgess’ (1925) concentric zone model, adapted from ecology, to applying optimal foraging theory from the biological sciences to predict offending patterns (Johnson 2014). The emergence of crime science has also created a multi-disciplinary field emphasizing that the ‘physical, social, biological and computer sciences are all seen as relevant to crime control’ (Laycock 2005: 6). Much of the research stemming from this area has provided two important take-aways. First, it is quite common to observe similar phenomena across multiple academic disciplines. Second, knowing the solutions found in these fields can help practitioners effectively address problems. This is important because it allows researchers to apply established techniques to solve a problem without having to design new ones. Given the apparent prevalence of hormesis in numerous fields, we believe that hormetic-like relationships are also possible in crime prevention. Humans constantly adapt to their environments, and small intentional disruptions to certain daily activities may result in unanticipated consequences. In the case of offenders, disturbing particular criminal processes in certain ways could actually elicit escalation (Grabosky 1996), retaliation (Ratcliffe and Rengert 2008), defiance (Sherman 1993), or create new opportunities thus resulting in a surge in offending. Although these outcomes have already been demonstrated in various contexts, we focus on how they might specifically apply to low doses or intensities of implementation. Our examination of the literature suggests that researchers need to further explore how individuals will respond to varying strengths of crime prevention strategies and the respective dosages that should be administered. Background What is hormesis? In the biological literature, hormesis is commonly defined as a ‘biphasic1 dose-response phenomenon characterized by low-dose stimulation and high-dose inhibition’ (Calabrese 2002: 1451). More specifically, because environmental conditions on earth are rife with toxins and hazards, some scientists believe that certain internal mechanisms have evolved in order to provide organisms the ability to survive milder threats and avoid extinction (Mattson 2008). This implies that certain levels of exposure to adverse conditions can propel them into a heightened state of resistance or tolerance to that stressor. In other words, it can be thought of as an overcompensation effect to milder threats. However, exposure to higher levels of stress or toxicity is often too great for an organism to combat and it ultimately dies. A number of scientific studies—particularly focused in toxicology, pharmacology and medicine—demonstrate examples of this phenomenon. In fact, although hormesis is not always present, a systematic review of the literature since the 1960s has found over 3,000 examples of the hormesis phenomenon indicating that it is not merely an anomaly (see Calabrese and Baldwin 2001). Some of the earliest biological research assumed that dose–responses to an adverse agent would possess a linear or threshold relationship (Calabrese 2002). As seen in Figure 1, while a linear relationship assumes constant decreases in survival as levels of toxicity administered increase (Figure 1a), the threshold relationship assumes that changes in mortality are not observed until toxicity levels reach a certain point (Figure 1b). However, the hormetic relationship is a third scenario where an opposite response is observed until a particular threshold is reached and expected observations become subsequently apparent. Figure 1c illustrates a hormetic response curve whereby mild stressors actually increase resistance and allow organisms to flourish until those stressors raise to more harmful levels. Fig. 1 View largeDownload slide (a–c) Linear, threshold and hormetic curves Fig. 1 View largeDownload slide (a–c) Linear, threshold and hormetic curves This discovery has been crucial in toxicological research. For example, in the field of agriculture, there is a constant struggle between protecting the environment from the exposure to chemicals and administering sufficient levels of insecticides to reduce pest populations for increased crop yield (van der Werf 1996). Understanding hormetic dose responses are critical to finding this balance. Some have argued that government regulators, such as the Environmental Protection Agency in the United States, do not consider this biphasic trend and advocate for the use of the most minimal doses of toxins possible (Mattson and Calabrese 2009). However, some studies have suggested that exceedingly mild doses of insecticides actually incite insects to lay more eggs onto plants in anticipation that they will perish in the immediate future. This may be an evolutionary response by the insects to increase their fitness by producing more offspring (Cohen 2006; Ayyanath et al. 2013). This approach typically backfires and leads to larger pest populations than were observed prior to the insecticide application. However, proponents of higher doses must also be aware of a pesticide’s impact on the environment. Such chemicals are not typically ‘fully selective to target organisms’ and can adversely affect so-called beneficial insects as well as those being targeted (van der Werf 1996: 85). Therefore, many researchers in the field are calling for both field and greenhouse trials that test multiple doses of insecticides and their residual impacts in order to find the optimal balance (Calabrese 2002). As proposed above, we believe this biological research issue is analogous to phenomena reported in the criminological literature.2 Smaller-scale crime prevention tactics may not be sufficient to entirely deter criminals from offending. For instance, changes in landscaping, such as the installation of planter boxes to create an increased sense of territoriality can actually supply offenders with more hiding places for their drugs (this is discussed in a later section). It is also possible that weaker doses of a crime prevention initiative could backfire and provide additional opportunities for offenders to act. For example, the removal of too few key members in a gang network will not outright dismantle the gang and will likely create promotion opportunities for lower ranking members that are typically attained through violent or criminal behaviour (this will also be discussed in further detail below). It should be noted that hormesis can also be expressed in a reciprocal manner depending on what or how an outcome of interest is being measured. Inversed curves are often observed in the medical sciences when researchers use substances or procedures designed to improve health outcomes (Calabrese and Baldwin 2001; 2003). In other words, the inverted u-shaped3 hormetic curve shown in Figure 1c would simply be flipped and exhibit a u-shaped curve instead. While the former typically reflects the effects of a toxic substance, the latter often measures the impact of imposing something designed to be beneficial. As such, criminologists should keep the possibility of a reciprocal curve in mind with analogous dose responses. Figure 2 provides an example of how the viewpoint, or outcome of interest, can flip the response curve. From the point of view of the offender, opportunities for crime are disrupted and generate an inverted u-shaped response curve (Figure 2a). However, if we were to evaluate the same strategy from the viewpoint of a community’s perceived level of safety, a u-shaped curve would likely result (Figure 2b). Both versions should be considered in the design phase of an intervention. However, in this paper, we focus mainly on how interventions impact offenders and their opportunities for crime. In the following sections, we provide examples and evidence as to why we believe this phenomenon is present in crime prevention as well as why it is imperative to consider when designing prevention strategies. Fig. 2 View largeDownload slide Reciprocal hormetic curves depending on the outcome being measured. (a) From the perspective of offenders and (b) from the perspective of others Fig. 2 View largeDownload slide Reciprocal hormetic curves depending on the outcome being measured. (a) From the perspective of offenders and (b) from the perspective of others Why are hormetic/biphasic relationships possible in crime prevention? Theories of environmental criminology often avoid speculation about criminal motivation and instead focus on the opportunity structures needed to successfully carry out offenses (Clarke 1995). One of the most influential theories, routine activities, states that an understanding of the spatio-temporal convergence of a motivated offender, suitable target and lack of a capable guardian are required in order for a crime to occur (Cohen and Felson 1979). A reconfiguring of the theory has since taken place to also accommodate the people or organizations who have control over each of the elements in the theory (Clarke and Eck 2005). Research shows that crime can be reduced when a strategy disrupts at least one of the three elements and that working with these controllers can provide a practical way to curb criminal offending. In more recent years, these developments have been expressed in what is now referred to as the problem analysis (or crime) triangle seen in Figure 3 (Clarke and Eck 2005). Below, we will demonstrate that certain tactics for each of the elements can experience weak intervention backfire (WIB), thus providing reasons to consider additional dimensions of offending and the intensity of prevention strategies when designing a particular program. Fig. 3 View largeDownload slide The crime problem triangle Fig. 3 View largeDownload slide The crime problem triangle Target/victim guardians First, guardians are considered the protectors of victims or targets that discourage crime through various forms of supervision over people or property (Felson and Cohen 1980). In other words, their mere presence can be enough to deter offenders from breaking into a house or robbing someone in the street (Felson 1995). However, the simple addition of a guardian into certain situations may not be sufficient to disrupt the crime triangle and reduce offending. In fact, low doses of this strategy may worsen the problem. For instance, on a specific street corner, a spike in robberies might occur because its physical layout provides multiple escape routes for motivated offenders. If we were to situate a street vendor such as a food truck at this location with hopes to increase guardianship, its mere presence may not necessarily be sufficient to reduce offending. In this example, motivated offenders might actually perceive this lone vendor as a suitable target who likely has money and few people around to intervene if they were to rob him. Or, the number of patrons for the food truck are insufficient to provide mutual guardianship and instead provide more targets. Thus, in this scenario, the so-called guardian could actually become a victim. However, if we were to modify our example and have multiple food trucks simultaneously locate themselves on this same street corner then there would be a number of potential guardians at that location. Both the vendors themselves and the customers that their businesses attract could take on this role. Suddenly, there would be far more individuals whose presence would likely deter motivated offenders from robberies on that street corner. That is to say, this increased dosage of guardians increases the risk of carrying out the crime because there are more people present to intervene or call the police. Some empirical studies also demonstrate how formal guardianship exerted by the police can experience weak intervention backfire. For instance, Koper (1995) investigated the optimal time police should spend patrolling hotspots. He broke the data down into five groups to represent all police stops that lasted up to 20 minutes, namely: 1–5, 6–10, 11–15 and 16–20 minutes. His parametric event history models revealed varying positive impacts on crime in all groups except for the group of shortest duration. More specifically, the locations that only received one to five minutes of police presence actually saw significantly worse (i.e. shorter) survival times until disorder following police intervention. As the police time at hot spots increased in 5-minute intervals, the data showed longer time periods without crime with an optimal police stop time of 11–15 minutes (Koper 1995). As such, the data appear to exhibit a hormetic relationship whereby short police presence may actually backfire in attempts to reduce disorder, but increasing the duration of their presence will generate more desired effects. Although Koper (1995) provides the coefficients generated at all dosages, the focus of his paper was to determine the optimal time period police should spend at hotspots. Thus, little attention was paid to the counterintuitive parameter generated for the one- to five-minute group. Given our hypothesis regarding weak intervention backfire, we offer a potential explanation as to why this relationship may have occurred. In line with Sherman’s (1990) theory of police crackdowns and residual deterrence at crime hot spots after a police intervention, it is possible that offenders adjust their behaviour both during and for a period after an intervention. If anything, the key to success in these strategies lies in how offenders perceive the actions of the police and how quickly they will return. For instance, if the police have not recently been to a location, offenders will always be uncertain as to whether they will show up at any given moment. However, it is possible that if the police arrive at a hotspot and spend a short period of time there, offenders will believe that there is a very low likelihood that they will return in the near future. Similar explanations of offender decision-making have been reported in other research. For example, in their interviews with vehicle burglars in Texas, Summers and Rossmo (2016) found that some offenders reported the following strategy: I would wait and see if [the police] just leave. If they’re driving around [rather than just passing], I’ll go to a different area (Summers and Rossmo 2016: n.p.). In other words, if the police do not stay long enough, offenders may be willing to simply wait until the police leave thinking that they will have the freedom to offend as soon as the authorities are gone. Thus, by knowing the police were just there, and are unlikely to return soon, their uncertainty about the police goes down. As a result, they feel more comfortable committing more crimes than they would have if the police had not shown up. Conversely, when the police spend 11–15 minutes, a reduction in disorder may have been observed because the offenders simply gave up and left the location. If the police were showing no signs of departure, they may have believed that the location was being staked out or waiting them out was not worth their time. Therefore, while the police tactic used at a problem location must be strategically chosen, the length of time that is spent there may be of equal importance because low doses of officer presence could actually backfire. Offender handlers The next side of the crime triangle considered is concerned with those who are actually able to exert control over criminals, namely offender handlers. Felson (1995) has argued that these individuals possess the ability to modify an offender’s behaviours because of their knowledge of or proximity to them. One can think of the control parents have over a child because of their direct supervision or awareness of the child’s whereabouts. In the case of adult offenders, the level of involvement from handlers may dictate how often they actually offend. We believe that a hormetic dose–response to this level of involvement is possible and that it can be administered either informally or formally. From an informal perspective, we turn to some of the work conducted on life course criminology, more specifically Sampson and Laub’s (1993) notion of desistance from crime as a result of structural turning points. Suppose you have a young adult male who frequently commits various offenses when sufficient opportunity presents itself. If one day he was to meet and begin dating a woman, her inclusion in his life may essentially change some of the ways that he spends his time. Initially, they may only see one another sporadically throughout the week and she likely would not have much awareness of, nor control over his other activities. Moreover, his new found interest in her could motivate him to actually commit more crime with hopes of obtaining money to take her out on dates and buy her gifts.4 However, if the relationship becomes increasingly serious and they were to begin living together, she may have a greater awareness of his activities. She might encourage him to engage in activities with her and socialize more often with her prosocial friends thus leading to an eventual desistance from crime because he no longer has the time (or maybe even desire) to offend anymore (example derived from Laub and Sampson 2014). In this hypothetical scenario, the introduction of the girlfriend as a handler into the offender’s life did not have a positive or even null effect. In fact, it prompted the offender to commit additional crimes in the beginning of their relationship as a means to please or impress her. However, as her involvement in his life increased beyond that initial period, we would expect to see a decline in offending thereafter. In a more formal context, some academic literature concerned with the post-release supervision of sexual offenders has also shown indications of weak intervention backfire. For instance, while the introduction of electronic monitoring for general offenders has received relatively positive reviews from the public (Brown and Elrod 1995), some scholars have raised the potential for unanticipated consequences when used specifically with sexual offenders (DeMichele et al. 2008). Critics argue that electronic monitoring might encourage agencies to reduce the amount of face-to-face time sexual offenders spend with their probation or parole officers and increase their level of reliance on GPS tracking. They considered this extremely problematic because research has consistently shown that sexual offenders are far more likely to offend close to or within their homes against someone they know, such as a younger family member (Payne and DeMichele 2011). As a consequence, a lower intensity of supervision might permit offenders to carry out their crimes while appearing to simultaneously comply with their court ordered sanctions. Place managers Lastly, we consider how crime prevention tactics that target the disruption of problem places through collaboration with their managers can exhibit weak intervention backfire. Place managers were first introduced into the literature by J. E. Eck (unpublished material) who found that problems concerning drug dealing within particular apartments over others could be largely attributed to how the owners/landlords controlled the activities and behaviours of residents. Numerous works have since argued that police partnership with place managers can be highly effective because the owners of property have certain legal rights that allow them to exert control over the places that they own (Eck 2002). Oftentimes these legalities give them an even greater ability to regulate behaviour than police officers. Place managers are able to control who is and is not permitted on their property as well as how they are permitted to behave (Madensen and Eck 2012). Reducing crime by changing places is not a new concept in criminology. In fact, there are a number of studies that have assessed the impacts of gentrification on crime but these have yielded mixed results (Covington and Taylor 1989; Kreager et al. 2011). The logic of gentrification is that when place managers upgrade problem places into desirable ones, it encourages more effective guardians to begin using the space thus reducing crime (Papachristos et al. 2011). From a place-based perspective, it would make sense for a hormetic relationship to exist as gentrification progresses. This process is usually very time consuming and can lead to a great deal of social disruption throughout its development. In the beginning, small changes are made that modify specific places. Investors purchase properties and begin building or renovating their businesses or residences. Because the ultimate goal is to begin attracting wealthier people, place managers attempt to make these places as appealing as possible to that demographic. As these individuals begin to migrate to these places, it provides an initial influx in the number of suitable targets at these locations. More specifically, places can become both crime attractors and crime generators (Brantingham and Brantingham 1995) providing motivated offenders with a wealth of attractive targets and few guardians to protect them. Therefore, we would actually predict an increase in crime at and around these areas when fewer places have been redeveloped. However, as the gentrification process continues and more place managers transform their properties, the number of capable guardians who live in, frequent, or run businesses in these places begin to overtake the area. Thus, the change in density of effective guardians and place managers would lead to the expected decline in crime, but only after the initial surge. All strategies intended to reduce crime can be thought of as disruptions to current ecological systems. All organisms within these systems, offenders included, will always adapt to the situations they find themselves in. Moreover, if sufficiently motivated, offenders will often find ways that crime can be of benefit to them, if possible. When change is introduced into an environment, there is always an adjustment period needed for organisms to reorient themselves. If environments are conducive to crime, offenders will try to find ways to maintain the status quo. Thus, less effective crime prevention initiatives may simply allow criminals to successfully innovate in how they carry out their misdeeds. Clearly, such an outcome is undesirable. Therefore, the following section examines the potential causes of a hormetic trend in criminal offending so that they can be considered when designing effective prevention programs in the future. What might cause weak intervention backfire? Just as cells or organisms can produce a hormetic dose response to the exposure of a strenuous condition or substance, offenders can exhibit differential responses to analogous threats. While the former demonstrates resistance to low doses of strain imposed upon them, there is no reason to believe that offenders will not react in similar ways. Individuals often repeat the same actions once they determine the most effortless way to accomplish a task (Zipf 1949). Therefore, the disruption of routines forces people to re-evaluate how or whether to keep engaging in certain behaviours. In some cases, and if the disruptions are low enough, there may be two principle types of responses that might generate a weak intervention backfire effect. First, offenders might opt to rebound from the intervention, thinking of it as more of an imposition than a complete deterrent or because they wish to retaliate against conditions they feel are imposed upon them. Second, low doses of an intervention might actually provide offenders with new opportunities to continue similar patterns of crime. Despite the two seemingly plausible explanations that we provide below, we must also discuss the possibility of a third scenario, false positives, namely instances where an evaluation appears to demonstrate a hormetic relationship, but is simply an artefact of statistical analyses, research design or presentation of results. Figure 4 provides a breakdown of the possible types of true and false positives. Specific explanations and examples are provided below. Fig. 4 View largeDownload slide True and false positives of weak intervention backfire Fig. 4 View largeDownload slide True and false positives of weak intervention backfire True positives Rebound Some key works in the criminological literature have documented the outright failure of crime prevention strategies and many of them are attributed to resistance or retaliation by offenders. For instance, McCord (2003) reviewed evaluations of the Scared Straight projects, a program where young or potential offenders are shown ‘what it would be like to be imprisoned’ (p. 26). Various sources not only confirmed the failure of the programs, but also found out that some of the teenagers admitted to committing offenses specifically to prove that they were not scared of imprisonment after their experiences in the program (Miller and Hoelter 1979). Similarly, some argue that coercive policing tactics can heighten hostility, aggravation and resistance when dealing with offenders (see Grabosky 1994; 1996). Such responses align with much of Clarke’s (1995) concept of provocation that has been a cornerstone in situational crime prevention suggesting that the perception of a crime prevention strategy is critical. Offender perceptions of the effectiveness of a given strategy will have a substantial impact on how they will react to any new constraints imposed upon them. For instance, Ekblom (1999) has suggested that the ‘credibility of the message’ (p. 43) is important when individuals evaluate their surrounding environments and decide how to behave. He argues that offenders can often tell or find out whether a tactic is fake or ineffective. However, perceptions of crime prevention strategies have also been known to work in favour of crime reduction via means such as publicity prior to implementation (Smith et al. 2002; Bowers and Johnson 2005). This is not surprising given Clarke’s (1980) principles of situational crime prevention and how changing offenders’ perceptions of the crime opportunities available to them or their risk of detection can have a substantial impact on their decision to offend (Clarke 2005). Thus just as anticipatory benefits (Smith et al. 2002) showed a decline in crime in expectation of a seemingly effective strategy being implemented, there is no reason to believe that offenders would not react in an opposite way if they perceived an intervention as badly chosen, poorly designed or ineptly executed. New opportunities Another possible explanation for weak intervention backfire could be that offenders are able to discover that their adaptations to prevention efforts can facilitate their criminal activity. For example, a Cincinnati neighbourhood has recently been undergoing a gentrification process via the investment and renovation of a number of properties in its central district. In addition to the complete redevelopment of business and residential areas, the redevelopment organization has also been trying to improve the area by streetscaping many public spaces such as walkways, parks and empty lots. The organization’s executive director explained that in one instance, there was a walkway that known drug dealers would use for passage to and from locations where they would commonly make transactions. While the organization was committed to transform that space in ways that would discourage this behaviour, they knew it would take them time to make all of the necessary changes. They first installed a number of large flower boxes with hopes that it would signal to the offenders that the space was being overtaken by effective guardians. Instead, the director explained that it only made the problem worse because the drug dealers began hiding their drugs inside the planters. Not only did this provide them a place to store larger quantities of drugs, but it also allowed offenders more time spent without them in their possession. When the police would stop them, they did not have any drugs on their person (K. Wright, personal communication). As a result, the offenders were able to carry out their crimes more easily and with less risk. It is also possible for weak intervention backfire to occur when crime prevention strategies target groups of offenders. For instance, Klein (1993) has discussed how the arrest of certain members in a gang can lead to intense recruitment in response to the loss of a member. As an example, the Cincinnati Police Department applied social network analysis to identify the key players in a large gang that had been the contributors to a high amount of offending in one neighbourhood. The police had turned to social network analysis after experience showed that disrupting gang networks was not overly effective in the long run because the gangs would eventually regroup. Instead, the police recognized that they had to dismantle the networks completely so that the offenders could not reassemble. To do this they had to select the most important members of these gangs to focus their attention on (Engel 2009; D. Gerard, personal communication). Social network analysis proved incredibly helpful in this processes because the police department could not only identify the key players in the network, but also the number of ties between the remaining individuals should key players be taken into custody (D. Gerard, personal communication). This is illustrated in Figure 5. Figure 5a shows the entire network. Figure 5b shows how the network would fracture if the top 20 members were removed. This generated several large sub-groups that could continue to offend even in the absence of the primary members of the gang. Figure 5c shows how the gang structure would look if the top 25 members were removed. Increasing the dosage in this way was enough to completely dismantle the offender network. When asked what outcome the police thought targeting the 20 key players would have produced, a lead officer in the investigation speculated that the initiative would not simply have been less effective, it actually would have worsened the violent crime problem. More specifically, he explained that when a network is disrupted and not completely dismantled, it often leaves a number of younger gang members at large and in charge. Their inexperience leads them to retaliate and act more violently in attempts to prove themselves on the streets and gain power within the gang network hierarchy (D. Gerard, personal communication). Such an outcome is not unheard of in gangs or drug markets (Ekblom and Pease 1995) and some scholars have even likened this relationship to military structures whereby subordinate officers take the place of commanding ones if the latter are killed in battle (Ekblom 1999). Fig. 5 View largeDownload slide (a–c) CIRV gang network analysis. (a) Illustration of entire gang network. (b) Gang network structure when the top 20 key players have been removed. (c) Gang network structure when the top 25 key players have been removed. Source: Institute of Crime Science, University of Cincinnati. Special thanks to retired Cincinnati Police Captain Daniel Gerard. Fig. 5 View largeDownload slide (a–c) CIRV gang network analysis. (a) Illustration of entire gang network. (b) Gang network structure when the top 20 key players have been removed. (c) Gang network structure when the top 25 key players have been removed. Source: Institute of Crime Science, University of Cincinnati. Special thanks to retired Cincinnati Police Captain Daniel Gerard. False positives Data artefacts Although we believe that weak intervention backfire is theoretically possible and appears to exist in some empirical studies, in some cases a relationship that appears to be hormetic may simply be an artefact of the data, research design or how the results of a study are presented. Just as statistical significance can be achieved by virtue of analyzing large samples (Lin et al. 2013), a biphasic curve may be a statistical artefact. For instance, in their evaluation of the Philadelphia foot patrol experiment, Ratcliffe et al. (2011) separated their data for analysis into percentiles based on the amount of violent crime present prior to the implementation of treatment. They then assessed the difference in slopes between the target and control areas and found that places with fewer incidents of pre-treatment violent crime actually yielded more crime in the target versus control areas.5 However, as the number of violent crimes at places increased they began to see a decline in crime in the target areas in comparison to the controls. This is an instance where we might suspect that the results that were consistent with a hormetic relationship may simply be a consequence of analyzing low-frequency spatial units. Due to the small sample sizes (the first percentile was defined as three pre-treatment violent crimes per unit), it is difficult to determine how reliable the measure is should one wish to test for a hormetic curve. Research design The way that we collect and analyze data may also give the false impression that weak intervention backfire is present. For instance, while the incarceration of offenders reduces crime via incapacitation, some studies of recidivism have indicated that smaller doses (i.e. shorter as opposed to longer in duration) of incarceration time may actually increase the odds of re-arrest among offenders (Meade et al. 2013). Some have speculated that fewer instances of recidivism among offenders who are held longer is simply an artefact of the research design (Loughran et al. 2009). This often occurs because studies can only have finite data collection periods thus allowing offenders who serve shorter sentences to be observed longer after their release. For example, if a study lasts for 60 months, then the offenders who received longer sentences, such as 48 months, will only be observed for a 12-month follow-up period. Conversely, those who receive shorter sentences, such as 24 months, are then afforded a follow-up period that is three times longer. If they offend at the same rate (crimes per month), it will appear that the short sentence offenders are more criminogenic than the long sentence offenders. This research limitation is commonly referred to as right-censoring in the corrections literature and is dealt with using survival analysis in order to account for varying follow-up periods among cases (Allison 1984). If the wrong statistical design was used, the findings might appear to be hormetic, when in fact they are the consequence of research design. Thus, researchers must be cognizant of how outcomes are measured to accurately assess whether weak intervention backfire actually occurred. True and false negatives The absence of hormetic dose–response relationships will likely be more common in the literature than its presence. However, in some cases, we do believe that such absences are the result of how data were collected or presented in empirical works. If too few observations are made when an insufficient dosage of an intervention is applied, hormesis might exist but not be seen. This continues to be an issue in the biological literature despite the fact that this phenomenon has been identified and studied within the field (Calabrese 2008). We hope that by introducing this potential research outcome to the literature that it will be considered and tested in future hypotheses and research designs. There are also some rare instances in the criminological literature where a hormetic response was actually expected but not found. For instance, Sherman et al. (1991) conducted a randomized experiment of arrest strategies in cases of domestic violence with the Milwaukee Police Department. Each domestic violence call that came to the police was randomly assigned to either a control (i.e. warning, no arrest), short- or long-term time in custody with mean durations of 2.8 and 11.1 hours, respectively. The authors hypothesized that police response to domestic violence via shorter time in custody would actually increase the risk of re-victimization within the same day because it was suspected that in jurisdictions where the suspect is released within an hour or two of arrest (unlike the minimum overnight custody in the Minneapolis experiment), the likelihood of intoxication or anger persisting from the original even would be very high. Under such conditions…the odds of immediate recidivism would be very high: The suspect could return to the victim and perhaps commit even more serious violence (Sherman et al. 1991: 825). Despite the plausibility of their initial arguments, their data collected over a 14-month period generated no evidence to support a weak intervention backfire hypothesis. In fact, the authors found nearly identical prevalence levels of repeat violence among offenders who were held for short and long time periods (Sherman et al. 1991). Thus, depending on the relationship being studied, there will likely be some instances where a weak intervention backfire effect will not be observed. However, whenever possible, its presence should be checked against data. Discussion Although weak intervention backfire will not always materialize in crime prevention initiatives, we do not believe that crime prevention research to date has been able to demonstrate its true prevalence. Many of the studies that may have possessed this relationship either failed to identify it or only discussed the phenomenon in passing. We suspect this occurs for a number of reasons. First, researchers simply may not have had an explanation for the spike in crime with low doses of the intervention other than assuming it was part of the random variation in the data. A second possibility may be that the presentation of results obscured the presence of this phenomenon. More specifically, if too few observations are recorded or if the results are averaged, it can be very difficult to determine whether weak intervention backfire occurred. In fact, it is not uncommon in the literature to see an emphasis on determining overall success or failure as opposed to assessing the intricacies of varied responses (see Weisburd et al. 2003). Lastly, due to limited resources, testing interventions of varying intensities is oftentimes infeasible and thus not done (Telep et al. 2014). Despite these challenges, the identification and better understanding of this phenomenon in various contexts could have highly beneficial policy implications. The only way to determine the pervasiveness of this relationship is to invest in additional resources in future studies that could evaluate the various doses of a program. With this information we can better inform police officials of how to actually design and implement such initiatives. That said, testing for a weak intervention backfire effect may be very difficult or even unethical. For instance, in the above-mentioned CIRV example, the low dose–response incapacitation of 20 (as opposed to 25) key players in the gang network was believed to worsen the gang violence problem in Cincinnati. Designing a study to rigorously test this conjecture would require careful thought and no little ingenuity. However, there may be some alternatives to circumvent this problem. Some recent proposals have been made to use simulation modelling as a cost-effective means to test crime prevention strategies (Groff and Birks 2008). These models have recently succeeded in demonstrating not only the explanatory power of criminological theories (Birks et al. 2012) and offender behaviours (Groff 2007), but they could allow researchers to test a number of ethically untestable hypotheses, such as the effects of varying dosages of police rapid response strategies (Eck and Liu 2008). Similar to the use of social network analysis in the CIRV gang network example, simulation modelling could likely provide invaluable information to practitioners when deciding on the optimal dosage of a program to implement. Use of Bayesian techniques are also becoming more popular in the criminological literature and have been used to circumvent many limitations of frequentist approaches (Law et al. 2014). Researchers could begin incorporating Bayesian methods that predict a strategy’s impact on crime given the known probability that weak intervention backfire will occur (Hacking 2001). As a consequence, evaluations using this approach would require specific a priori—as opposed to null—hypotheses. Thus, research would need to first determine the prevalence of backfire using field experiments or simulation modelling. Evaluations of interventions could also incorporate principles from psychophysics, namely the study of sensation and perception of physical stimuli by humans (Leek 2001). Psychophysics evaluates sensation detection via four dimensions: intensity, quality, extension and duration (Gescheider 1997). This field’s method of limits—a technique whereby sub-threshold stimuli are presented to subjects and increased incrementally until they detect it—has been used in disciplines such as audiometry (i.e. the measurement of hearing) (Gescheider 1997). This incremental process can also be continued in order to determine the just noticeable difference (JND) threshold. This approach demonstrates the amount a stimulus needs to change in order for someone to detect it (Luce and Edwards 1958). Analogous JND trials may be of use to crime prevention studies in order to determine which doses are most useful to test. This paper addressed backfire considerations analogous to the intensity and quality dimensions of psychophysics. However, researchers may also be interested in the duration dimension, namely how long detection of a sensation lasts. If lower doses of a strategy are unsuccessful, researchers could investigate whether these failures are permanent. As illustrated in Figure 6, a backfire effect immediately following an intervention could theoretically lead to one of three eventual outcomes. First, the intervention could backfire completely leading to an increase in crime that persists over time (complete backfire). In contrast, an initial spike could occur followed by a return to approximately the same level of offending as before (zero-gains backfire). Lastly, a desired outcome of crime reduction could be achieved but not before an initial backfire is experienced (initial backfire). This final possible outcome shares some similarities with findings in other fields. For instance, some research in the behavioural psychology literature have observed similar responses referred to as extinction bursts. They are characterized as ‘temporary increase[s] in the frequency, intensity, or duration of the target response’ that they wish to correct (Lerman and Iwata 1995: 93). If future research finds that the last scenario occurs, policy makers may still want to consider the intervention for future application. More specifically, anticipatory measures could be taken to compensate for a short initial surge in crime until the residual benefits begin to materialize. However, practitioners should consider whether the net gains over time outweigh the initial increase in offending. Multiple doses, or increasing the intensity, of a strategy might also sustain desired impacts or create an accumulation of benefits over time. If this is the case, starting an intervention at the lowest possible intensity would be most cost-effective. Fig. 6 View largeDownload slide The three possible crime prevention backfire outcomes Fig. 6 View largeDownload slide The three possible crime prevention backfire outcomes Lastly, researchers should also consider the various types of buffers that could exist. For instance, since the introduction of crime displacement and the diffusion of crime control benefits (Clarke and Weisburd 1994), spatial buffers have been used to study any residual impacts radiating from crime hotspots (see Bowers and Johnson 2003). Similarly, Smith et al.’s (2002) concept of anticipatory benefits showed how a temporal buffering effect could occur prior to the implementation of a crime prevention strategy. In a similar vein, researchers should explore the long-run implications of low-dose interventions. In other words, if an initial backfire trend is hypothesized, practitioners will want to know whether the initial undesired effects will occur before or after that actual intervention. To illustrate our point, Figure 7a–7d provide a graphical depiction of Smith et al.’s (2002) anticipatory benefits in comparison to a so-called temporal extension of weak intervention backfire: initial backfire. Figure 7a can be seen as a representation of a backward-looking buffer: some noticeable impact occurs for a certain period of time prior to policy implementation. In Figure 7b, we apply Sherman’s (1990) concept of residual deterrence using a forward-looking buffer: some impact occurs after the intervention. Our proposed temporal extension of weak intervention backfire, displayed in Figure 7c and 7d, illustrates forward- and backward-looking buffers associated with backfires. Figure 7c shows a type of temporal backfire whereby crime increases during early stages of an intervention until desired effects are observed later on. Figure 7d combines this with the notion of anticipation: a hybrid result. In contrast to Smith et al.’s (2002) model, our variation simply implies a reverse impact on crime that could result from similar influences such as pre-intervention publicity or creeping implementation. Fig. 7 View largeDownload slide Forward- and backward-looking buffers. (a) Backward-looking buffer (anticipatory benefits—Smith et al. 2002). (b) Forward-looking buffer (residual deterrence—Sherman 1990). (c) Forward-looking buffer (initial backfire). (d) Backward- and forward-looking buffer (anticipatory backfire) Fig. 7 View largeDownload slide Forward- and backward-looking buffers. (a) Backward-looking buffer (anticipatory benefits—Smith et al. 2002). (b) Forward-looking buffer (residual deterrence—Sherman 1990). (c) Forward-looking buffer (initial backfire). (d) Backward- and forward-looking buffer (anticipatory backfire) Given the presence of these temporary intervention-based crime reductions—such as anticipatory benefits and residual deterrence—and our proposed backfire effect, a future synthesis of all relevant research would also be of benefit to practitioners. Protocols from systematic review frameworks, such as the Campbell Collaboration (see Farrington and Petrosino 2001) and the EMMIE scale (see Johnson et al. 2015), should be continually updated to monitor the distribution of results that reveal these outcomes, or absence of them. If weak intervention backfire is present, beneficial intervention impacts over time might actually be underestimated. Moreover, monitoring studies that explore the mechanisms and contexts thought to cause backfire could facilitate stronger and more cost-effective intervention designs. Now that the concept of weak intervention backfire has been identified, it is hoped that future research will consider and be better able to explain seemingly unexpected increases in crime despite the implementation of an initiative designed to reduce it. Scholars now have the above set of theoretical and empirical examples that can inform future hypotheses and theories during the design and testing stages of crime prevention research. At the same time, the argument for this phenomenon can retroactively inform explanations of crime surges. If we are correct, researchers now have a framework that can guide the design of more efficient crime prevention strategies and have a better understanding of which strategies yield more beneficial outcomes. Acknowledgements This research was supported by the Social Sciences and Humanities Council of Canada (SSHRC). The authors would like to thank the anonymous reviewers for their insightful comments as well as J. C. Barnes, Kate Bowers, Daniel Gerard, Jordan Papp, Lucia Summers, Douglas Wiebe, Kevin Wright and Cheryl Zurowski for their suggestions on earlier drafts of this manuscript. References Allison P. D. ( 1984), Event History Analysis: Regression for Longitudinal Event Data . Sage Publications. Anderson E. ( 1999), Code of the Street: Decency, Violence, and the Moral Life of the Inner City . WW Norton & Company. Ayyanath M. M. Cutler G. C. Scott-Dupree C. D. and Sibley P. K. ( 2013), ‘ Transgenerational Shifts in Reproduction Hormesis in Green Peach Aphids Exposed to Low Concentrations of Imidacloprid’, Public Library of Science (PLoS) One , 8: e74532. Birks D. Townsley M. and Stewart A. ( 2012), ‘ Generative Explanations of Crime: Using Simulation to Test Criminological Theory’, Criminology , 50: 221– 54. Google Scholar CrossRef Search ADS Bowers K. J. and Johnson S. D. ( 2003), ‘ Measuring the Geographical Displacement and Diffusion of Benefit Effects of Crime Prevention Activity’, Journal of Quantitative Criminology , 19: 275– 301. Google Scholar CrossRef Search ADS Bowers K. J. and Johnson S. D. ( 2005), ‘ Using Publicity for Preventive Purposes’, in N. Tilley, ed., Handbook of Crime Prevention and Community Safety , 329– 54. Willian Publishing. Bowers K. J. Johnson S. D. and Hirschfield A. ( 2004), ‘ The Measurement of Crime Prevention Intensity and its Impact on Levels of Crime’, British Journal of Criminology , 44: 419– 40. Google Scholar CrossRef Search ADS Bowers K. J. Johnson S. D. and Hirschfield A. ( 2005), ‘ Closing Off Opportunities for Crime: An Evaluation of Alley-Gating’, European Journal on Criminal Policy and Research , 10: 285– 308. Google Scholar CrossRef Search ADS Brantingham P. L. and Brantingham P. J. ( 1995), ‘ Criminality of Place: Crime Generators and Crime Attractors’, European Journal on Criminal Policy and Research , 3: 5– 26. Google Scholar CrossRef Search ADS Brown M. P. and Elrod P. ( 1995), ‘ Electronic House Arrest: An Examination of Citizen Attitudes’, Crime & Delinquency , 41: 332– 46. Google Scholar CrossRef Search ADS Burgess E. W. ( 1925), ‘ The Growth of a City: An Introduction to a Research Project’, in R. E. Park E. W. Burgess and R. D. McKenzie, eds, The City , 47– 62. University of Chicago Press. Google Scholar CrossRef Search ADS Calabrese E. J. ( 2002), ‘ Hormesis: Changing View of the Dose-Response, A Personal Account of the History and Current Status’, Mutation Research , 511: 181– 9. Google Scholar CrossRef Search ADS PubMed Calabrese E. J. ( 2008), ‘ Hormesis: Why It Is Important to Toxicology and Toxicologists’, Environmental Toxicology and Chemistry , 27: 1451– 74. Google Scholar CrossRef Search ADS PubMed Calabrese E. J. and Baldwin L. A. ( 2001), ‘ Hormesis: U-Shaped Dose Responses and Their Centrality in Toxicology’, Trends in Pharmacological Sciences , 22: 285– 91. Google Scholar CrossRef Search ADS PubMed Calabrese E. J. and Baldwin L. A. ( 2002), ‘ Defining Hormesis’, Human & Experimental Toxicology , 21: 91– 7. Google Scholar CrossRef Search ADS PubMed Calabrese E. J. and Baldwin L. A. ( 2003), ‘ Hormesis: The Dose-Response Revolution’, Annual Review of Pharmacology and Toxicology , 43: 175– 97. Google Scholar CrossRef Search ADS PubMed Clarke R. V. ( 1980), ‘ “Situational” Crime Prevention: Theory and Practice’, The British Journal of Criminology , 20: 136– 47. Google Scholar CrossRef Search ADS Clarke R. V. ( 1995), ‘ Situational Crime Prevention’, in M. Tonry and D. P. Farrington, eds, Building a Safer Society: Strategic Approaches to Crime Prevention , Vol. 19, 91– 150. University of Chicago Press. Clarke R. V. ( 2005), ‘ Seven Misconceptions of Situational Crime Prevention’, in N. Tilley, ed., Handbook of Crime Prevention and Community Safety , 39– 70. Willian Publishing. Clarke R. V. and Eck J. E. ( 2005), Crime Analysis for Problem Solvers: In 60 Small Steps . Center for Problem Oriented Policing. Clarke R. V. and Weisburd D. ( 1994), ‘ Diffusion of Crime Control Benefits: Observations on the Reverse of Displacement’, Crime Prevention Studies , 2: 165– 84. Cohen E. ( 2006), ‘ Pesticide-Mediated Homeostatic Modulation in Arthropods’, Pesticide Biochemistry and Physiology , 85: 21– 7. Google Scholar CrossRef Search ADS Cohen L. E. and Felson M. ( 1979), ‘ Social Change and Crime Rate Trends: A Routine Activity Approach’, American Sociological Review , 44: 588– 608. Google Scholar CrossRef Search ADS Covington J. and Taylor R. B. ( 1989), ‘ Gentrification and Crime Robbery and Larceny Changes in Appreciating Baltimore Neighborhoods During the 1970s’, Urban Affairs Review , 25: 142– 72. Google Scholar CrossRef Search ADS DeMichele M. Payne B. K. and Button D. M. ( 2008), ‘ Electronic Monitoring of Sex Offenders: Identifying Unanticipated Consequences and Implications’, Journal of Offender Rehabilitation , 46: 119– 35. Google Scholar CrossRef Search ADS Eck J. E. ( 2002), ‘ Preventing Crime at Places’, in L. W. Sherman D. P. Farrington B. C. Welsh and D. L. MacKenzie, eds, Evidence-Based Crime Prevention , 241– 94. Routledge. Google Scholar CrossRef Search ADS Eck J. E. and Liu L. ( 2008), ‘ Contrasting Simulated and Empirical Experiments in Crime Prevention’, Journal of Experimental Criminology , 4: 195– 213. Google Scholar CrossRef Search ADS Ekblom P. ( 1999), ‘ Can We Make Crime Prevention Adaptive by Learning from Other Evolutionary Struggles?’, Studies on Crime and Crime Prevention , 8: 27– 51. Ekblom P. and Pease K. ( 1995), ‘ Evaluating Crime Prevention’, in M. Tonry and D. Farrington, eds, Crime and Justice: Building a Safer Society , Vol. 19, 585– 662. University of Chicago Press. Google Scholar CrossRef Search ADS Engel R. S. ( 2009), ‘ Taliband Network Connections ’, Technical Report Prepared for the Cincinnati Police Department, University of Cincinnati Policing Institute. Farrington D. P. and Petrosino A. ( 2001), ‘ The Campbell Collaboration Crime and Justice Group’, The Annals of the American Academy of Political and Social Science , 578: 35– 49. Google Scholar CrossRef Search ADS Farrington D. P. and Welsh B. C. ( 2002), ‘ Improved Street Lighting and Crime Prevention’, Justice Quarterly , 19: 313– 42. Google Scholar CrossRef Search ADS Felson M. ( 1995), ‘ Those Who Discourage Crime’, in J. E. Eck and D. Weisburd, eds, Crime and Place: Crime Prevention Studies , Vol. 4, 53– 66. Criminal Justice Press and Police Executive Research Forum. Felson M. and Cohen L. E. ( 1980), ‘ Human Ecology and Crime: A Routine Activity Approach’, Human Ecology , 8: 389– 406. Google Scholar CrossRef Search ADS Gescheider G. A. ( 1997), Psychophysics: The Fundamentals . Lawrence Erlbaum Associates, Inc. Grabosky P. ( 1994), ‘ Counterproductive Crime Prevention’, in Crime Prevention Conference 1994 , 151– 65. Centre for Crime Policy and Public Safety, Griffith University. Grabosky P. N. ( 1996), ‘ Unintended Consequences of Crime Prevention’, in R. Homel, ed., Crime Prevention Studies , Vol. 5, 25– 56. Criminal Justice Press. Groff E. R. ( 2007), ‘ “Situating” Simulation to Model Human Spatio-Temporal Interactions: An Example Using Crime Events’, Transactions in GIS , 11: 507– 30. Google Scholar CrossRef Search ADS Groff E. and Birks D. ( 2008), ‘ Simulating Crime Prevention Strategies: A Look at the Possibilities’, Policing , 2: 175– 84. Google Scholar CrossRef Search ADS Haberman C. P. ( 2016), ‘ A View Inside the “Black Box” of Hot Spots Policing from a Sample of Police Commanders’, Police Quarterly , 19: 488– 517. Google Scholar CrossRef Search ADS Hacking I. ( 2001), An Introduction to Probability and Inductive Logic . Cambridge University Press. Google Scholar CrossRef Search ADS Johnson S. D. ( 2014), ‘ How Do Offenders Choose Where to Offend? Perspectives from Animal Foraging’, Legal and Criminological Psychology , 19: 193– 210. Google Scholar CrossRef Search ADS Johnson S. D. Tilley N. and Bowers K. J. ( 2015), ‘ Introducing EMMIE: An Evidence Rating Scale to Encourage Mixed-Method Crime Prevention Synthesis Reviews’, Journal of Experimental Criminology , 11: 459– 73. Google Scholar CrossRef Search ADS Klein M. W. ( 1993), ‘ Attempting Gang Control by Suppression: The Misuse of Deterrence Principles’, Studies on Crime & Crime Prevention , 2: 88– 111. Koper C. S. ( 1995), ‘ Just Enough Police Presence: Reducing Crime and Disorderly Behavior by Optimizing Patrol Time in Crime Hot Spots’, Justice Quarterly , 12: 649– 72. Google Scholar CrossRef Search ADS Kreager D. A. Lyons C. J. and Hays Z. R. ( 2011), ‘ Urban Revitalization and Seattle Crime, 1982–2000’, Social Problems , 58: 615– 39. Google Scholar CrossRef Search ADS PubMed Laub J. H. Nagin D. S. and Sampson R. J. ( 1998), Trajectories of Change in Criminal Offending: Good Marriages and the Desistance Process. American Sociological Review , 63: 225– 38. Google Scholar CrossRef Search ADS Laub J. H. and Sampson R. J. ( 2014), ‘ A Theory of Persistent Offending and Desistance from Crime’, in F. T. Cullen R. Agnew and P. Wilcox, eds., Criminological Theory: Past to Present – Essential Readings , 545– 51. Oxford University Press. Law J. Quick M. and Chan P. ( 2014), ‘ Bayesian Spatio-Temporal Modeling for Analysing Local Patterns of Crime Over Time at the Small-Area Level’, Journal of Quantitative Criminology , 30: 57– 78. Google Scholar CrossRef Search ADS Laycock G. ( 2005), ‘ Defining Crime Science’, in M. J. Smith and N. Tilley, eds., Crime Science: New Approaches to Preventing and Detecting Crime , 3– 26. Willan Publishing. Leek M. R. ( 2001), ‘ Adaptive Procedures in Psychophysical Research’, Perceptions & Psychophysics , 63: 1279– 92. Google Scholar CrossRef Search ADS Lerman D. C. and Iwata B. A. ( 1995), ‘ Prevalence of the Extinction Burst and its Attenuation during Treatment’, Journal of Applied Behavior Analysis , 28: 93– 4. Google Scholar CrossRef Search ADS PubMed Lin M. LucasJr H. C. and Shmueli G. ( 2013), ‘ Research Commentary - Too Big to Fail: Large Samples and the p-Value Problem’, Information Systems Research , 24: 906– 17. Google Scholar CrossRef Search ADS Loughran T. A. Mulvey E. P. Schubert C. A. Fagan J. Piquero A. R. and Losoya S. H. ( 2009), ‘ Estimating a Dose-Response Relationship between Length of Stay and Future Recidivism in Serious Juvenile Offenders’, Criminology , 47: 699– 740. Google Scholar CrossRef Search ADS PubMed Luce R. D. and Edwards W. ( 1958), ‘ The Derivation of Subjective Scales from Just Noticeable Differences’, Psychological Review , 65: 222– 37. Google Scholar CrossRef Search ADS PubMed Madensen T. D. and Eck J. E. ( 2012), ‘ Crime Places and Place Management’, in F. T. Cullen and P. Wilcox, eds, The Oxford Handbook of Criminological Theory , 1– 18. Oxford University Press. Mattson M. P. ( 2008), ‘ Hormesis Defined’, Ageing Research Reviews , 7: 1– 7. Google Scholar CrossRef Search ADS PubMed Mattson M. P. and Calabrese E. J. ( 2009), ‘ Hormesis: What It Is and Why It Matters’, in M. P. Mattson and E. J. Calabrese, eds, Hormesis: A Revolution in Biology, Toxicology and Medicine , 1– 14. Springer Science. Google Scholar CrossRef Search ADS McCord J. ( 2003), ‘ Cures that Harm: Unanticipated Outcomes of Crime Prevention Programs’, The Annals of the American Academy of Political and Social Science , 587: 16– 30. Google Scholar CrossRef Search ADS Meade B. Steiner B. Makarios M. and Travis L. ( 2013), ‘ Estimating a Dose–Response Relationship between Time Served in Prison and Recidivism’, Journal of Research in Crime and Delinquency , 50: 525– 50. Google Scholar CrossRef Search ADS Miller J. G. and Hoelter H. H. ( 1979), Prepared Testimony: Oversight on Scared Straight . Government Printing Office. Papachristos A. V. Smith C. M. Scherer M. L. and Fugiero M. A. ( 2011), ‘ More Coffee, Less Crime? The Relationship Between Gentrification and Neighborhood Crime Rates in Chicago, 1991 to 2005’, City & Community , 10: 215– 40. Google Scholar CrossRef Search ADS Payne B. K. and DeMichele M. ( 2011), ‘ Sex Offender Policies: Considering Unanticipated Consequences of GPS Sex Offender Monitoring’, Aggression and Violent Behavior , 16: 177– 87. Google Scholar CrossRef Search ADS Ratcliffe J. H. and Rengert G. F. ( 2008), ‘ Near-Repeat Patterns in Philadelphia Shootings’, Security Journal , 21: 58– 76. Google Scholar CrossRef Search ADS Ratcliffe J. H. Taniguchi T. Groff E. R. and Wood J. D. ( 2011), ‘ The Philadelphia Foot Patrol Experiment: A Randomized Controlled Trial of Police Patrol Effectiveness in Violent Crime Hotspots’, Criminology , 49: 795– 831. Google Scholar CrossRef Search ADS Sampson R. J. and Laub J. H. ( 1993), Crime in the Making: Pathways and Turning Points through Life . Harvard University Press. Shaw C. R. and McKay H. D. ( 1942/1969), Juvenile Delinquency in Urban Areas . University of Chicago Press. Sherman L. W. ( 1990), ‘ Police Crackdowns: Initial Residual Deterrence’, in M. Tonry and N. Morris, eds, Crime and Justice: A Review of Research , 1– 48. University of Chicago Press. Google Scholar CrossRef Search ADS Sherman L. W. ( 1993), ‘ Defiance, Deterrence, and Irrelevance: A Theory of the Criminal Sanction’, Journal of Research in Crime and Delinquency , 30: 445– 73. Google Scholar CrossRef Search ADS Sherman L. W. Schmidt J. D. Rogan D. P. Gartin P. R. Cohn E. G. Collins D. J. and Bacich A. R. ( 1991), ‘From Initial Deterrence to Long-Term Escalation: Short-Custody Arrest for Poverty Ghetto Domestic Violence’, Criminology , 29: 821– 50. Google Scholar CrossRef Search ADS Smith M. J. Clarke R. V. and Pease K. ( 2002), ‘ Anticipatory Benefits in Crime Prevention’, Crime Prevention Studies , 13: 71– 88. Summers L. and Rossmo D. K. ( 2016, June), Situational Decision-Making of Vehicle Burglars: Prevention and Displacement . Paper presented at the International Symposium on Environmental Criminology and Crime Analysis (ECCA). Tallarida R. J. and Jacob L. S. ( 1979), The Dose-Response Relation in Pharmacology . Springer-Verlag. Google Scholar CrossRef Search ADS Telep C. W. Mitchell R. J. and Weisburd D. ( 2014), ‘ How Much Time Should the Police Spend at Crime Hot Spots? Answers from a Police Agency Directed Randomized Field Trial in Sacramento, California’, Justice Quarterly , 31: 905– 33. Google Scholar CrossRef Search ADS van der Werf H. M. ( 1996), ‘ Assessing the Impact of Pesticides on the Environment’, Agriculture, Ecosystems & Environment , 60: 81– 96. Google Scholar CrossRef Search ADS Weisburd D. Lum C. M. and Yang S. M. ( 2003), ‘ When Can We Conclude that Treatments or Programs “Don’t Work”?’, The Annals of the American Academy of Political and Social Science , 587: 31– 48. Google Scholar CrossRef Search ADS Welsh B. C. Farrington D. P. and Gowar B. R. ( 2015), ‘ Benefit-Cost Analysis of Crime Prevention Programs’, in M. Tonry, ed., Crime and Justice: A Review of Research , Vol. 44, 447– 516. University of Chicago Press. Google Scholar CrossRef Search ADS Zhang Y. Shen G. Yu Y. and Zhu H. ( 2009), ‘ The Hormetic Effect of Cadmium on the Activity of Antioxidant Enzymes in the Earthworm Eisenia fetida’, Environmental Pollution , 157: 3064– 8. Google Scholar CrossRef Search ADS PubMed Zipf G. ( 1949), Human Behaviour and the Principle of Least Effort: An Introduction to Human Ecology . Addison-Wesley Press. Footnotes 1 Biphasic means having two phases. In the case of hormesis, the phenomenon requires both a stimulatory and an inhibitory phase ‘to satisfy [its] qualitative definition’ (Calabrese and Baldwin 2002: 94). 2 To be clear, the above example is not intended as a metaphor to offenders. It is used to help readers understand the concept of hormesis. 3 Some biological studies refer to these relationships as j-shaped curves as well. 4 Because criminologists tend to study changes in offending around the time of marriage (Laub et al. 1998), little focus is placed on examining offending during the earliest stages of a romantic relationship. However, some qualitative work has suggested that motivation to commit crime can often be driven by materialistic desires (see Anderson 1999), due possibly in part to impress a prospective partner. Thus, while it has not been studied, a backfire effect is certainly possible. 5 Although the difference in slope was the only positive coefficient generated in the analysis, it was not statistically significant (see Ratcliffe et al. 2011: 812). © The Author(s) 2017. Published by Oxford University Press on behalf of the Centre for Crime and Justice Studies (ISTD). All rights reserved. For permissions, please e-mail: email@example.com
The British Journal of Criminology – Oxford University Press
Published: Mar 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera