Then they simply go somewhere else: The effect of hotspot policing on crime development

Introduction

In the context of policing crime hotspots, weapon ban zones (WBZ) are popular among practitioners and policymakers likewise. WBZs are clearly defined geographical areas where the carrying of certain weapons, or in some cases all weapons and potentially weaponizable objects, is legally prohibited by authorities. Their symbolic effect and their comparative ease of implementation make WBZs appealing for both practitioners and policymakers alike. Consequently, the implementation of an increasing number of WBZs has become a subject of public debate (Siegel, 2025). However, the effects of WBZs remain unclear. Moreover, there is a widespread belief that increased police activity and hotspot policing may in fact not reduce crime but only disperse it spatially (Hatten and Piza, 2022).

Analyses from a WBZ in the German city of Leipzig revealed that there was a slight decrease in severe crimes due to its introduction (Mühler et al., 2022). However, evidence for the spatial displacement of crimes was inconclusive on a higher aggregated level. Additionally, going into more detail regarding possible displacement effects was beyond the scope of the evaluation. Therefore, it remains unclear whether there was a decrease in crime rates or a spatial displacement inside and outside the WBZ. Hence, in this article, we seek to develop a more nuanced view on the case by answering the following question: Does the introduction of a WBZ lead to an actual reduction or a displacement of crime? Thereby, we seek to contribute to the body of research on WBZ effectiveness in general.

Since the works of Becker (1974), the effects of environmental change, such as the implementation of a new police policy, on crime rates have been a dominant topic of criminological research. From an empirical point of view, this assumption is still far from being certain. Several meta-analyses demonstrated displacement effects, but at the same time reveal that displacement is similarly likely compared to the so-called ‘diffusion of control benefits’ – namely a crime reduction and its diffusion in the surrounding areas (Braga, 2007; Eck, 1993; Guerette and Bowers, 2009; Hesseling, 1994). For instance, Guerette and Bowers (2009) demonstrated that displacement effects were observed in 26% of the studies they reviewed, while reduction and diffusion effects were observed in 27% of the studies. Sorg et al. (2013) showed that observed displacement effects often appear only in the short-term and are not enduring. Other studies suggest that displacement is not a probable outcome of hotspot policing (Bowers et al., 2011; Ratcliffe and Breen, 2011; Weisburd et al., 2006). Additionally, the evidence presented by Ratcliffe and Breen (2011) indicates that displacement effects can differ between types of crimes. Thus, there is still no clear picture of the actual effects of hotspot policing on crime development and spatial distribution.

We contribute to this literature in three ways. First, we provide evidence for a European context. Most studies on the effects of hotspot policing are based on data from the United States or the United Kingdom while studies from continental Europe are scarce (Guerette and Bowers, 2009; Hinkle et al., 2020; Turchan and Braga, 2024). Therefore, little is known whether the results on crime displacement are generalizable to other contexts or not.

Second, displacement is often analyzed with a quotient on an aggregated data level (Bowers and Johnson, 2003). Hence, changes in smaller units of analysis, for example, the displacement from highly trafficked streets into smaller alleys, remain unnoticed. Therefore, we provide additional insight into displacement in smaller area units by using spatial data on a micro level; namely, exact coordinates for all offenses committed 24 months before and 24 months after the introduction of the WBZ. We analyze the data by using kernel density estimation (KDE) and hotspot analysis to create a more in-depth view of the movements of crime.

Third, we argue for a differentiated view of different types of crimes. In doing so, we theoretically rely on a dual-process perspective as presented by Van Gelder (2013). So far, the theoretical approach to explain spatial displacement is mainly focused on rational behavior. However, recent approaches to sociological action theory propose the idea that the majority of human behavior is based on spontaneous and automatic cognitive processes (Evans, 2018; Tutić, 2022). These approaches call into question the idea of rational choice models as general theory of action. The findings of Abramovaite et al. (2023) provide substantiating evidence for the assertion that an augmented detection probability is associated with a decline in theft and burglary, but not violent offenses. As we will argue, it is plausible to assume differences in the underlying cognitive processes behind violent offenses and theft. Therefore, this may in part explain the ambiguous empirical results concerning spatial crime displacement.

The remainder of the paper is organized as follows. First, we provide background information on the WBZ in Leipzig in order to describe the actual features of this specific measure of hotspot policing. Second, we develop a theoretical model of the consequences of introducing a WBZ on crime development by combining classical criminological theories with approaches from a dual-process perspective. After describing the data and our analytical approach, we present and discuss our results. We close with a discussion of our findings, limitations, and directions for future research.

The WBZ in Leipzig

The WBZ in Leipzig, Germany, was established in November 2018 as a reaction to high crime rates and group conflicts in the area around the street ‘Eisenbahnstraße’. This area is a crime hotspot within the city, as evidenced by crime rates that are significantly higher than the city average (Mühler et al., 2022). These high crime rates also led to the Eisenbahnstraße gaining a reputation as a particularly dangerous area among Leipzig's citizens and authorities.

The WBZ was implemented as an open-ended, indefinite measure rather than a time-limited intervention. With the WBZ, a new police policy was established that consisted of three core aspects. First, the carrying of weapons and dangerous objects was prohibited (§1 Polizeiverordnung des Sächsischen Staatsministeriums des Innern über das Verbot des Mitführens gefährlicher Gegenstände in Leipzig and §1 Sächsische Waffenverbotszonenverordnung Leipzig). In Germany, carrying a weapon requires several valid permits. In the WBZ, individuals in possession of those permits were still prohibited from carrying weapons openly. ¹ Dangerous objects include any object capable of injuring a person, such as knives, screwdrivers or even so-called ‘selfie sticks’ for mobile phones.

Second, the police were given broad rights to conduct suspicionless stops, thus suspending the normally applicable rule of requiring at least reasonable suspicion (De Maillard et al., 2018). Unfortunately, there is no data on overall police activity in the area. Checks were carried out incidentally during routine police activity, such as patrols, as well as during dedicated control operations, in which checks were specifically planned and conducted (Mühler et al., 2022). Because of these planned control operations, it can be assumed that police activity increased with the implementation of the WBZ.

Third, big yellow signs were placed at the entry and exit points of the WBZ, clearly indicating its boundaries. This ensured that persons on site were generally aware that they had entered the zone and understood what this entailed. Furthermore, widespread public debate and media coverage of the WBZ and related police activities probably increased awareness of police activity in the area and therefore increased the perception of risk from the perspective of potential offenders.

Theoretical approach

Several theories have been presented to provide insight into the social consequences of changes in crime policy (Fedchak et al., 2024). The most prominent approaches refer to the decision-making behavior of potential offenders through rational choice models (Becker, 1974; Cornish and Clarke, 1987) or, more specifically, a routine activity model (Cohen and Felson, 1979). Rational choice approaches are based on the notion that rational individuals evaluate their environment to balance their own interests against the perceived risks of committing a crime. It has been argued that increased police presence leads to a change in perceived risk for potential offenders (e.g., Andenæs, 1974; Armour, 1986; Lin, 2009) because police forces act as a symbol of the criminal justice apparatus (Dau et al., 2023: 192). From this basic rational choice theory, it follows that the main effect of a newly established WBZ is a reduction in crime, since the latter acts as a manifest representation of law enforcement, for example, mediated by warning signs, media coverage, and increased police presence.

As an extension to basic rational choice models, representatives of the routine activity approach (Cohen and Felson, 1979) argue that local changes in the incentive structure might influence a rational actor's evaluations regarding this specific area but will not influence potential offenders’ general motivation to commit an offense. Therefore, a change in one environment might lead offenders to seek opportunities in other locations (Hatten and Piza, 2022). In addition, offenders seem to choose places that are not far from their daily routines because these places are familiar and, in contrast to more distant places, present less unknown risks (Bowers et al., 2011; Eck, 1993). Thus, the expected effect of a WBZ would not be a reduction, but a displacement of crime into surrounding areas.

While offering a parsimonious theoretical framework for social action in general and criminal behavior in particular, rational choice models are prone to fundamental criticism. Referring to both ambiguous empirical evidence and conceptual weaknesses of anthropological premises, critics emphasize the explanatory limitations of such rational choice models (Frederick, 2005; Goeree and Holt, 2001; Green and Shapiro, 1994; Nagel, 1995; Simon, 2000; Zafirovski, 2000). A body of research in behavioral and social sciences suggests that rational choice models fail to grasp the actual motivations and causality of human decision making in the real world (Gigerenzer, 2000; Gigerenzer and Gaissmaier, 2011; Kahneman and Tversky, 1979; Tutić, 2015).

Such criticism is often met with the argument that the purpose of those assumptions is not to accurately represent the real world, but rather to provide parsimonious explanations. Although parsimony is undoubtedly an important criterion of scientific theories, there is a danger that such premises will be regarded as ‘as-if anthropologies’ (Jones, 2003: 409) at the latest when thinking about the practical consequences of research results. It is therefore important to further develop these action-theoretical foundations.

Consequently, a different branch of theoretical literature considers that potential offenders are also guided by emotions and affect when choosing between deviant and compliant behavior (Loewenstein et al., 2001; Van Gelder, 2013). Van Gelder (2013) argues for a hot/cold perspective on criminal behavior and bases this idea on a dual-process perspective. The core of this approach is the notion of two distinct types of processes in human decision making (Evans, 2018; Kahneman, 2012). One type is theorized to be intuitive, implicit and fast (Type 1) while the other type is thought to be conscious, deliberate and slow (Type 2) (Tutić, 2022). Type 1 processes lead ‘automatic’ behavior that is acted out habitually or spontaneously. Type 2 processes are theorized to influence a class of actions that are grounded in explicit intentions as well as analytic and rational thinking. This behavioral model seems to have a much better empirical fit than the rational actor and therefore is increasingly considered the more suitable theory of action (Esser and Kroneberg, 2015). However, it should still be understood as an ideal typology, since real-world behavior seems to be produced in an even more complex interplay of unconscious and conscious factors (Evans and Stanovich, 2013; Gigerenzer and Gaissmaier, 2011).

Van Gelder (2013) translates these types into two modes of producing action in the context of crime. The ‘hot mode’ (Type 1) represents criminal actions that are primarily driven by emotion and affect, thus being not strategically planned but acted out spontaneously and on an immediate trigger. In contrast, in the ‘cold mode’ (Type 2) offenders evaluate risks and rewards consciously and rationally, as it is proposed by the rational choice perspective. Modifications to environmental factors that increase the probability of detection, such as the introduction of a WBZ, should not influence ‘hot’ mode crimes too much, as the detection probability should play a lesser role in a potential offender's decision-making process due to their affective state. However, for ‘cold’ mode crimes, the probability of detection plays a crucial role, as it alters the cost-benefit calculation and may lead to different behavioral choices of potential offenders. Consequently, crimes that are subject to a Type 1 process should be less affected by changes in the environment than Type 2 processes (Van Gelder and De Vries, 2014). Accordingly, crime is analyzed as an interplay of individual dispositional conditions on the one hand and situational conditions on the other (Wikström and Kroneberg, 2022).

Several studies have shown that violent crimes often occur in a state of high arousal (Van Gelder, 2013) or even under the influence of alcohol (e.g., McClelland and Teplin, 2001; Murdoch and Ross, 1990; Rossow et al., 1999). In these cases, offenders are unlikely to make rational decisions independently of their current environment. In other words, a highly emotional person, who may even be under the influence of alcohol, is not likely to weigh the increased likelihood of being caught in a WBZ against the benefits of hitting someone; it just happens.

By contrast, theft and similar offenses are less often committed in such states. While some qualitative studies show that individuals commit theft without planning (Feeney, 2014; Gill, 2000), other studies have found evidence of careful evaluation of the physical environment (Lin, 2009). Furthermore, analogous patterns emerge within the field of deterrence research. Previous research indicates that the efficacious influence of deterrence in crime reduction is contingent upon the nature of the offense, and further nuances emerge when classifying the crime as either a Type 1 or Type 2 offense (e.g., Abramovaite et al., 2023). Overall, it seems plausible that the evaluation of the environment is a minimum requirement for committing theft. Therefore, we expect stronger effects of changes in the environment for theft than for violent crimes.

The outlined approach calls for a more nuanced theoretical view on the deterrent effects of interventions on criminal offenses. Cohen and Felson (1979) have argued that changes in incentive structures do not change the motivation of potential offenders to commit offenses. However, if an offense is committed in a ‘hot mode’ (Type 1 process), rational motivation plays a lesser role in the decision between deviant and compliant behavior. In other words, if a person commits an offense in ‘the heat of the moment’, this person is unlikely to search for opportunities elsewhere beforehand. By contrast, if a person is planning an offense, opportunities away from areas of increased police presence are likely to be considered in the decision process.

Note that this argument does not contradict the notion of a moral filter (Wikström, 2014). Moreover, our argument depends on the general eagerness to commit a criminal offense. However, from the viewpoint of a dual-process perspective the influence of a person's moral code might depend on its strength and its chronic availability in a given situation (Esser and Kroneberg, 2015).

This theoretical perspective has strong implications for the expected effects of a WBZ. Since the introduction of a WBZ sets incentives that need to be rationally processed, we expect violent crimes (Type 1 offenses) to be less affected by the intervention than property crimes (Type 2 offenses). Note that we are not arguing that every theft is subject to rational decision making, or that every violent offense is based on affect. Instead, these types of offenses are at least to be seen as typical cases for the respective modes. Therefore, we expect to see differences in the effect of the WBZ on the aggregate level.

Data and methods

To test these theoretical assumptions, we analyze data on crimes recorded by the Saxon police between January 2016 and October 2020 within the WBZ and the adjacent areas in the districts of Volkmarsdorf and Neustadt-Neuschönefeld in Leipzig. The introduction of the WBZ in November 2018 falls roughly in the middle of our observation period. ² Police-recorded data in Germany, while standardized, has two main shortcomings: First, it does not include victim data, leading to a bias toward reported crimes (Kleinewiese, 2022), and second, it can be influenced by police practices like racial profiling. However, these biases remained relatively constant over the study period (Baur et al., 2020) and do not confound the testing of our hypotheses. In addition, since only one police station was located within the WBZ throughout the entire study period (see Figure 1), a confounding influence on crime patterns prior to or during the intervention is very unlikely.

OPEN IN VIEWER

Figure 1.

Division of districts in and around the WBZ.

Our analyses focus on two types of crimes: violent offenses and property crimes, such as theft. We operationalize violent offenses using the Police Crime Statistics (PCS) categories of violent crimes and offenses against personal freedom, which include acts like physical aggression and violations of individual liberty, and crimes against life, which encompass murder and manslaughter. We excluded incidents in private settings because the WBZ aims to reduce public violence. ³

Property crimes are operationalized using PCS categories for petty theft and grand theft. Unlike violent offenses, we included incidents in private spaces, like residential burglary, assuming that rational offenders would consider the increased police presence in the WBZ and the higher risk of detection. For every offense, the police recorded the exact date, time, and geographic coordinates, which form the basis for our geospatial analyses. We analyze the data using two complementary approaches: an aggregated data approach and a micro-level data approach.

Results

Violent offenses

Calculations of the Weighted Displacement Quotient for violent crimes are shown in Table 1. The metrics presented include the impact of the WBZ within the target area as well as on the adjacent buffer zone, the WDQ and its logarithmic transformation, and the Total Net Effects (TNEs). The TNE reflects the overall changes in crime across all areas (see Bowers et al., 2009 for interpretation).

OPEN IN VIEWER

Table 1.

Results of WDQ calculation for violent offenses.

Window	Success	Buffer	WDQ	WDQlog	TNE
3-month	0.41	−1.06	−0.39	−0.57	11.00
6-month	−0.29	−0.91	0.32	0.51	51.77
9-month	−0.39	−0.81	0.48	0.72	82.40
12-month	−0.22	−0.44	0.5	0.76	59.32
15-month	−0.31	−0.42	0.73	1.07	76.33
18-month	−0.25	−0.19	1.32	1.92	55.68
21-month	−0.02	−0.01	1.73	1.70	8.55
24-month	0.02	−0.03	−0.68	−0.68	3.07

WDQ: Weighted Displacement Quotient; TNE: Total Net Effect.

The WDQ analysis revealed that the introduction of the WBZ had a mixed impact on violent crime. In the first three months, a modest reduction was observed in the intervention area, though this was partly offset by crime displacement into the buffer zone, as indicated by a negative WDQ. However, by the six-month mark, the WDQ became more positive, suggesting a more effective intervention over time. The WDQ values continued to rise, peaking at 21 months, but then decreased again at 24 months, indicating that the positive effects had diminished and displacement may have returned. Thus, the WDQ calculations suggest, that a notable decline of crime in the WBZ was only temporary. Crime displacement remained an issue at various points, as fluctuating WDQ values suggest that crime patterns did not remain stable.

Figure 2 provides a more detailed analysis by displaying the exponentially smoothed ¹¹ data for violent offenses (depicted by the dark red solid-dotted line), overlaid with the forecast results from the Prophet model (depicted by the blue dashed line, accompanied by 95% confidence intervals). ¹² In line with the WDQ calculations, the decline of violence in the intervention zone (top graph) was short-lived, lasting approximately three months after the WBZ's introduction and remaining below pre-introduction levels for about six months. Forecasting that accounts for the long-term trend, as well as weekly and yearly seasonality (blue dashed line), projects a stable trajectory for violent offenses, followed by fluctuations. After the five-month mark, the actual trend largely mirrors the forecast but remains slightly below it. Notably, following this initial decline, violent offenses began to rise again, peaking in July and December 2019, despite the WBZ remaining in place.

OPEN IN VIEWER

Figure 2.

Time series and forecasting of violent offenses.

An opposite trend was observed in the buffer zone, where an initial increase in violent offenses contradicted the WDQ analysis and long-term forecasts. Despite this, the WBZ appears to show a unique dynamic, as violent crime levels in both the buffer and control zones remained elevated compared to the forecasts. However, given the preexisting differences between the intervention, buffer, and control zones, this pattern should be interpreted with caution, as it cannot be unambiguously attributed to the WBZ. The discrepancies between the aggregated WDQ analysis and the time series data highlight the limitations of relying on (highly) aggregated data alone. Therefore, we used KDE and hotspot analysis to examine micro-level spatiotemporal patterns.

Figure 3 presents the KDE results for violent crimes in six-month increments. As indicated by the dark contour lines, violent crime concentration remained highest in the intervention area, both before and after the implementation of the WBZ. However, after the intervention, a subtle shift of crime towards the east and southeast near the WBZ's borders became evident, suggesting the emergence of smaller hotspots just outside the zone. This is further supported by the overall shift of the crime distribution, as reflected by the shift of the mean center (red dot).

OPEN IN VIEWER

Figure 3.

KDE of violent offenses.

This finding is also corroborated by a separate hotspot analysis (Figure 4), which revealed a significant ‘cooling’ of the main violence hotspot within the intervention zone and the emergence of new, clustered hotspots to the north, east, and south. Most of these new hotspots were short-lived, dissolving within 18 months, with the exception of one persistent cluster to the east, in the immediate vicinity of the intervention area. Eighteen months after the intervention, the spatial shifts stabilized and became more permanent, as clustered hot and cold spots of violent crimes neither emerged nor disappeared.

OPEN IN VIEWER

Figure 4.

Hotspot analysis of violent offenses. Dark hexagons (indicating low crime surrounded by low crime, i.e., low-low) and dark red hexagons (indicating high crime surrounded by high crime, i.e., high-high) represent spatial clusters where crime counts have significantly decreased or increased, respectively. These patterns show that each hexagon is part of a broader area with similar crime trends. In contrast, light blue hexagons (low crime surrounded by high crime, i.e., low-high) and light red hexagons (high crime surrounded by low crime, i.e., high-low) - hexagons whose crime levels differ from those of their neighbors. Light blue hexagons represent unusually low-crime areas within high-crime areas, while light red hexagons show the opposite - unusually high-crime areas within low-crime surroundings (hotspots within coldspots).

Hotspot analysis revealed that despite permanently lower crime rates in the intervention zone after the WBZ's implementation, crime concentration remained high in specific micro-places. Similarly, elevated crime rates in the buffer and control zones were also driven by localized clustering of incidents. In absolute terms, however, both the crime-reducing and crime-displacing effects of the WBZ were short-lived and mediocre at best.

The micro-level analysis complements the aggregated data approach by revealing the spatio-temporal dynamics of violent offenses following the WBZ's implementation, showing that moderate crime increases and decreases were driven by localized clustering rather than a uniform spread.

Property crimes

The results for petty and grand theft show a more pronounced pattern of reduction and displacement. The WDQ calculations in Table 2 reveal that, except for the initial three months, success values remain consistently negative, indicating sustained reductions in crime within the intervention area. The buffer metrics show only negative values, suggesting a diffusion of benefits to adjacent areas. This is supported by the positive WDQ values, which indicate that crime reduction extended beyond the target area. Furthermore, the consistently positive and substantial TNE values for theft indicate significant overall reductions in theft across both the intervention and buffer zones.

OPEN IN VIEWER

Table 2.

Results of WDQ calculation for petty and grand theft.

Window	Success	Buffer	WDQ	WDQlog	TNE
3-month	0.10	−0.14	−0.70	−0.60	7.37
6-month	−0.06	−0.13	0.45	0.42	83.98
9-month	−0.07	−0.11	0.66	0.67	131.63
12-month	−0.07	−0.13	0.56	0.57	214.37
15-month	−0.03	−0.05	0.67	0.69	99.27
18-month	−0.03	−0.05	0.58	0.59	114.76
21-month	−0.05	−0.04	1.10	1.10	172.87
24-month	−0.04	−0.05	0.74	0.75	173.66

WDQ: Weighted Displacement Quotient; TNE: Total Net Effect.

However, the time series analysis in Figure 5 offers a slightly different perspective. Although property offenses in the intervention area declined sharply and remained at a lower level of approximately five offenses per week (top graph) for three months, the observed trend and forecasted values converged after six months. This suggests that the reduction of property crimes within the WBZ was only short-lived. Long-term changes appear even more limited, as the forecasted and actual trends are nearly identical – unlike violent offenses, where the actual trend remained slightly below the forecast.

OPEN IN VIEWER

Figure 5.

Time series and forecasting of petty and grand theft.

In contrast, the development of petty and grand theft in the buffer zone initially followed the observed trend but later diverged, resulting in elevated property crime levels compared to the forecasted trend – beginning in October 2019. In the control zone, both the observed and forecasted trends show a sharp decline in police-recorded theft during the first three months. Since the pre-intervention crime rate was much higher in the control zone than in the intervention zone, the relative change observed there appears larger than within the intervention zone itself. Whether this difference is driven by the WBZ's implementation or by other underlying factors cannot be conclusively determined. However, after three months, theft began to rise again – moderately in the forecasts, but more drastically in the observed data. After five months, the forecasted and actual trends converged, with the observed trend displaying more pronounced crime peaks.

WDQ and time series analysis paint a similar, though not entirely identical, picture of the development of property crimes within the WBZ. Both methods indicate a short-lived diffusion of benefits in relation to property crime. However, they diverge in their assessment of the permanence of the reduction. While the WDQ analysis suggests a mid- to long-term reduction, the time series analysis indicates that this trend is only short-term, followed by a substantial long-term increase in property crime outside the WBZ.

Using KDE, Figure 6 illustrates the spatiotemporal development of petty and grand theft at the micro level. Before the WBZ's implementation, these crimes were strongly concentrated in the intervention zone, but unlike violent crimes, they had multiple smaller crime hotspots both inside and outside the area. Six months before the intervention, the other hotspots became more pronounced without shifting the center of the overall spatial distribution, ¹³ although property crime did not concentrate as strongly as in the main hotspot within the intervention zone.

OPEN IN VIEWER

Figure 6.

KDE of petty and grand theft.

This changed in the first six months after the WBZ was introduced. While the main property crime hotspot within the intervention zone shrank – aligning with the findings from the time series analysis – the hotspot south of the WBZ became the most dominant, followed by another in the east. The spatial displacement of property crime incidences is also evident from the mean and median centers of the overall distribution, which were located well outside the intervention area six and 12 months after the intervention. After one year, the property crime hotspots outside the WBZ still remained, although their overall crime incidence had declined, and despite spatial displacement and an overall reduction in property crimes, the WBZ remained the largest crime hotspot in the long-term.

The shift in crime concentration appears to be only partially caused by a mere reduction of crime within the intervention zone. As the hotspot analysis in Figure 7 illustrates in detail, displacement contributed to this pattern. A comparison of the three-month periods before and after the WBZ's implementation shows the emergence of three new significant clustered hotspots east and southeast of the intervention area. Of these three hotspots, one remained over a longer period, while another began to emerge to the southwest in the six-month comparison. These two clusters persisted even in the long-term comparison, highlighting a clear displacement of property crimes to areas outside the WBZ. At the same time, the analysis shows that the introduction of the WBZ was associated with a permanent reduction in petty and grand theft within the intervention zone.

OPEN IN VIEWER

Figure 7.

Hotspot analysis of petty and grand theft.

Taking into account both the results of the aggregated analysis and Figure 6, the processes of property crime reduction within the WBZ and displacement to surrounding areas did not happen simultaneously. In the first six months of the intervention, petty and grand theft reduced within the WBZ as well as, in parts, outside of it, particularly in the south. However, these crimes moderately increased in clustered hotspots in the east and southeast, despite an overall crime reduction and a diffusion of benefits (see Figure 5). Additionally, the rise in property crime levels in the buffer and control zones beginning in October 2019 was driven by the clustered crime hotspots in the south and southwest, which became more pronounced and solidified 12 months after the intervention.

Discussion

The results of the data analyses fit our theoretical expectations quite well. We hypothesized that the effect of a WBZ's implementation is contingent on the type of crime. Criminal actions that rely on conscious and deliberate rational calculation (‘cold mode’, Type 2) should be more affected than those acted out in a comparatively intuitive and affective fashion (‘hot mode’, Type 1). Furthermore, we seek to contribute to answering the open question whether and how hotspot policing measures such as WBZ affect the actual crime development at all: Do they reduce criminal activity in the intervention zone and beyond (‘diffusion of benefits’) or will criminal activity just be displaced in the surrounding areas?

Table 3 gives a comparative overview of the results of the empirical analyses. In line with our theoretical assumption based on the dual-process perspective, we find stronger associations of the WBZ's introduction with theft-related offenses. There is a significant reduction in theft incidents both within and, to some extent, outside the intervention zone, along with the spatial displacement following the WBZ's introduction. This contrasts with violent offenses, where no similarly distinct patterns were observed, suggesting that such interventions mainly affect ‘cold mode’ rather than ‘hot mode’ crimes, since perpetrators of the former type, on average, seem to more consciously adapt their behavior to avoid persecution than those of the latter type.

OPEN IN VIEWER

Table 3.

Comparative overview of empirical results on WBZ success.

Method	Violent crimes (‘hot mode’ actions)	Petty and grand theft (‘cold mode’ actions)
Weighted Displacement Quotient	– Moderate short-term reduction and signs of displacement – Weak to moderate mid-term reduction and distribution of benefits – No long-term effects	– Moderate to strong short-, mid- and long-term reduction and distribution of benefits
Time series and forecasting	– Weak and short-term reduction – Return to previous level after six months – Slight long-term increase in crime outside the WBZ	– Strong short-term reduction – Return to previous level after six months – Substantial long-term increase in crime outside the WBZ
Kernel Density Estimation	– WBZ remained largest crime hotspot in short- and long-term	– New prominent hotspots in the surrounding areas in short-term – WBZ remained largest crime hotspot in long-term
Hotspot analyses	– Subtle shift of hotspots toward the surrounding zones – Formation of one small, persistent hotspot rather than a uniform spread	– Strong and sustained reduction of crime within WBZ – Clear and lasting displacement to two new hotspots in the surrounding areas, rather than a uniform spread
Overall pattern	– Only a limited short-term and virtually no long-term effect, either within or outside the intervention zone	– Strong short-term effects in crime reduction and diffusion of benefits, but also long-term crime displacement

WBZ: weapon ban zone.

Thus, interventions aimed at reducing violent, affect-driven public crimes by increasing police presence and control authority are unlikely to achieve lasting effects in terms of deterrence and diffusion of benefits. While ban zones may have a limited positive short-term effect in reducing violent and, especially, theft crime, they also come with the downside of spatial displacement. Instead of preventing crime altogether, crime partially shifts to nearby areas outside the intervention zone. Furthermore, interventions like WBZs seem to have their strongest crime-reducing effect on property offenses, while violent crimes seem much less affected. This is particularly noteworthy given that WBZs and other instruments of hotspot policing are often implemented in response to especially severe or high-profile violent incidents.

In terms of overall effectiveness, the results are quite clear: the WBZ's introduction did not result in a long-term reduction in crime incidents, neither for violent offenses nor for theft. This is even more evident for theft, as crime levels remained permanently elevated in clustered hotspots outside the intervention zone. Thus, at least the WBZ in Leipzig can be considered unsuccessful in achieving the original objectives that politically justified its implementation. Nonetheless, the WBZ produced considerable social costs by disrupting social routines and negatively affecting daily lives of many residents in the district (Mühler et al., 2022).

Given these findings, the long-term effectiveness of broad, untargeted measures like WBZs in reducing crime and improving public perception of safety remains highly questionable. At best, WBZs may be somewhat effective at least in short terms. However, even within a six-month period, they appear to affect property crimes more than violent crimes, thereby failing to precisely address their politically intended purpose – namely, the permanent reduction of violent crime.

Conclusion

Our study has provided new insights into the impact of the WBZ in Leipzig by examining whether its introduction led to a reduction or displacement of two different types of crime: violent crime and theft. We hypothesized that the WBZ's effectiveness would vary by crime type, as its deterrent effects rely on a dual-process mechanism of rational decision-making. Thus, we expected violent crimes to be less affected by the intervention than property crimes.

Our findings largely confirm this expectation. While our analyses suggest that the WBZ had some short-term crime-reducing effects on both crime types, the effect seems particularly pronounced for theft. However, its long-term effectiveness appears to be limited. Displacement was evident for both violent and property crimes, though with different spatial patterns and persistence. While some displacement occurred, patterns of violent offenses – often driven by impulse – did not show clear indications of long-term adaptation to the intervention. In contrast, perpetrators of property crimes appeared to consciously adjust their behavior, avoiding enforcement areas while continuing offenses elsewhere.

These results contribute to the literature on crime displacement in several ways. First, we extend the discussion of increased police activity and hotspot policing to a European, specifically German, context, providing empirical evidence on spatial crime displacement beyond Anglo-American regions.

Second, our study highlights the importance of distinguishing between different types of crimes, emphasizing the need for a nuanced approach. Drawing on the dual-process perspective (Van Gelder, 2013), we suggest that crime displacement is not uniform but varies based on the cognitive and situational mechanisms underlying different offenses. This has significant policy implications, casting doubt on the effectiveness of broad policing interventions like WBZs, and underscoring the need for more targeted and sustainable crime prevention strategies tailored to specific crime types.

Applying a dual-process perspective can explain results on place-based crime interventions. For instance, evaluation results from the Clear, Hold, Build framework in the United Kingdom, a place-based crime intervention, show reductions in crime only for property crimes but not for any other type of crime (England and Corcoran, 2025). Moreover, evidence from a longitudinal study in London suggests that stop and search interventions have an effect on drug offenses but not on violent crime (Tiratelli et al., 2018). Hence, evidence from the United Kingdom is in line with the results in this study.

Furthermore, our theoretical argument can guide future research on comprehending the effects of WBZs. For instance, while we argue that violent offenses are comparatively seldom driven by rational motivations, the decision to bring a weapon into the WBZ could certainly be rational. Therefore, it can be hypothesized that the introduction of a WBZ would have an impact on the severity of violent crimes, but not on their occurrence.

Third, we analyzed crime displacement using a multimethod approach, combining WDQ and time series analysis for overall trends with KDE and Local Moran's I for micro-level spatial dynamics. By integrating these methods, we were able to develop a more comprehensive and nuanced understanding of displacement effects, rather than relying solely on aggregated data analysis such as the WDQ. This approach could inform future studies with similar research objectives.

Our findings contribute to a more comprehensive understanding of spatial crime displacement and its theoretical underpinnings. However, there are limitations. First, the design of our study does not allow for strong causal inference. While we identified associations and patterns that appear closely related to the WBZ, we cannot fully attribute these outcomes to its implementation. Therefore, further research should attempt to apply a synthetic control design relying on more detailed data. However, in Germany, access to the type of detailed police data used in our analysis for Leipzig is either highly restricted or, in many cases, unavailable due to insufficient data coverage.

Second, due to limited data access, we were only able to investigate police-registered crime displacement in the immediate vicinity of the intervention zone, leaving crime trends in other districts of Leipzig and unreported crimes unaccounted for. As a result, we cannot entirely rule out the potential confounding effects of broader crime reduction. Further research would benefit from a more extensive data basis, encompassing not only more extensive areas but also further indicators of crime, such as victimization surveys or registered police calls for service. Above that, in order to implement the suggested synthetic control design, fine-grained sociodemographic data for matching and constructing a synthetic control group is required. Not least, detailed data on police presence would be crucial as a control variable.

Third, since our study focused specifically on Leipzig and a particular type of hotspot policing, it is difficult to generalize our findings to other German or European contexts. Future research should, therefore, indeed explore whether these patterns also emerge in other cities and settings, adopting a more refined, comparative and context-sensitive approach to the investigation of crime prevention and urban policy.